<%BANNER%>

The Acquisition of Orthographic-Phonological Correspondence Rules in L2 and L3 Portuguese

Permanent Link: http://ufdc.ufl.edu/UFE0041910/00001

Material Information

Title: The Acquisition of Orthographic-Phonological Correspondence Rules in L2 and L3 Portuguese Error Resolution, Interference, and Generalizability
Physical Description: 1 online resource (200 p.)
Language: english
Creator: Barkley, Sharon
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: correspondence, english, grapheme, l2, l3, orthography, phoneme, phonology, portuguese, spanish
Linguistics -- Dissertations, Academic -- UF
Genre: Linguistics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The current study investigates the acquisition of orthographic-phonological correspondence rules by learners of Portuguese as an L2 or L3, with English and Spanish as the previous language(s). Acquisition is examined from the perspectives of error resolution, interference, and generalizability. Target L2/L3 production is shown to increase throughout the course of the semester of data collection, as interference from the L1/L2 decreases. The results support previous research, finding transfer from both the L1 and L2, and interference based on language typology as well as language status. Word type (cognates, non-cognates and nonce words) is not found to play as great a role in acquisition as grapheme and participants? linguistic background. It is proposed that examination of the acquisition of orthographic-phonological correspondence rules may have the potential for highlighting more specifically the source(s) of interference in an L3.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Sharon Barkley.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Lord, Gillian.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041910:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041910/00001

Material Information

Title: The Acquisition of Orthographic-Phonological Correspondence Rules in L2 and L3 Portuguese Error Resolution, Interference, and Generalizability
Physical Description: 1 online resource (200 p.)
Language: english
Creator: Barkley, Sharon
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: correspondence, english, grapheme, l2, l3, orthography, phoneme, phonology, portuguese, spanish
Linguistics -- Dissertations, Academic -- UF
Genre: Linguistics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The current study investigates the acquisition of orthographic-phonological correspondence rules by learners of Portuguese as an L2 or L3, with English and Spanish as the previous language(s). Acquisition is examined from the perspectives of error resolution, interference, and generalizability. Target L2/L3 production is shown to increase throughout the course of the semester of data collection, as interference from the L1/L2 decreases. The results support previous research, finding transfer from both the L1 and L2, and interference based on language typology as well as language status. Word type (cognates, non-cognates and nonce words) is not found to play as great a role in acquisition as grapheme and participants? linguistic background. It is proposed that examination of the acquisition of orthographic-phonological correspondence rules may have the potential for highlighting more specifically the source(s) of interference in an L3.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Sharon Barkley.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Lord, Gillian.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041910:00001


This item has the following downloads:


Full Text





THE ACQUISITION OF ORTHOGRAPHIC-PHONOLOGICAL CORRESPONDENCE
RULES IN L2 AND L3 PORTUGUESE: ERROR RESOLUTION, INTERFERENCE, AND
GENERALIZABILITY



















By

SHARON BARKLEY


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2010




























2010 Sharon Barkley



































To Herman O.









ACKNOWLEDGMENTS

I thank my dissertation chair, Dr. Gillian Lord, for her enduring patience, wise counsel,

speedy feedback, and constant availability and willingness to help. I thank my committee

members, Dr. Theresa Antes, Dr. Libby Ginway, and Dr. Edith Kaan, for their interest, advice,

and dedication in reading this dissertation. I thank my raters and fellow teaching assistants,

Andrea Ferreira and Quinn Hansen, for their friendship and for the hours they spent cheerfully

listening to and transcribing the data for this project. I thank my statistics consultant, Claudio

Fuentes, for his invaluable help in the statistical analysis of the data. I thank my professors and

colleagues in the Linguistics Program and the Department of Spanish and Portuguese Studies, for

their input and support, especially Dr. Laurie Massery and Dr. Rose Smouse. I thank my

teaching supervisor, Dr. Libby Ginway, my fellow teaching assistants, and my many students for

all that I have learned from them while teaching Portuguese. I thank my family and friends for

their continued love, support, encouragement and prayers throughout this degree program,

especially Bill and Mary Barkley, Dr. Thomas Barkley, Sharon Dice, Carol Hall, Tom and

Florence Barkley, Liz Horner, Chad Preston, and the members of the "Supper Club." Finally, I

thank God for His innumerable gifts to me, and I pray that He may bless these many people who

have so patiently accompanied me on this journey.









TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S .................................................................................... .....................4

L IS T O F T A B L E S ................................ ................................................................... 8

LIST OF FIGURES ................................... .. .... .... ................... 12

L IS T O F A B B R E V IA TIO N S ............................ ................................................... ...................13

A B S T R A C T ..............................................................................................................14

CHAPTER

1 INTRODUCTION AND LITERATURE REVIEW ................................... .................15

1.1 Introduction ......................................................................... ......... 15
1.2 L2 and L 3 A acquisition ....................... ...... .................... 17
1.2.1 Acquisition of Typologically Similar Languages ...........................................28
1.2.2 The Relatedness of Spanish and Portuguese.................................................31
1.2.3 Acquisition of L2 and L3 Phonology ...........................................................36
1.3 Acquisition of L2 and L3 Orthographic-Phonological Correspondence Rules ............40
1.4 The Relevant Rules of English, Spanish and Portuguese............................................46
1.5 D different W ord T ypes ......................................................................... ....................48
1.6 T he Present Study ........................ .... ........................ .. ........... ..... .... 52

2 METHODOLOGY ................................. .. ... .... ................... 56

2 .1 Introductio n ............................................................................................................ 56
2.2 Research Questions and Hypotheses ..................................................... ............. 56
2.2.1 Error R solution ............................................. .. ....... .............. ... 57
2.2.2 Interference ................................................................... .......... 58
2.2.3 Generalizability .................. ............................ .. ..... ................. 60
2 .3 M etho d o lo g y ......................................................................... 6 1
2.3.1 P participants .................................................................... ......... 61
2.3.2 M materials ....................................................... ............ ......... 64
2 .3.3 D ata C collection ............ .......................................................... ...... .... .... 67
2 .3 .4 T ra n scrip tio n ............................................................................................... 6 9
2 .3 .5 A n a ly s is ....................................................................................................... 7 1

3 ERROR RESOLUTION RESULTS AND DISCUSSION ...........................................77

3 .1 Introd uctio n ...................................... .................................................. 77
3.2 R results and D iscussion........... ............................................................ .. .... .... ..... 78
3.2.1 A ll Production .................. ............................ ....... ................. 78
3 .2 .2 P roductio n b y G roup .............................................................. .....................78









3.2.3 Production by G raphem e........................................................ ............... 80
3.3 O their O bservations.................................................... .. .. .. ........ ..............82
3.3.1 Production by Test and Group ....................................... ....................... 82
3.3.2.1 Production by native Spanish speakers.............................................82
3.3.2.2 Production by native English speakers with no or low proficiency
in Spanish .................. ................... ............... ...... ............. 84
3.3.2.3 Production by native English speakers with high proficiency in
S p an ish .............................................................................. .... 8 5
3.3.3 Productionby Grapheme........................... ..... ........................... 86
3.3.3.1 Production for graphem e .................................. ............... 86
3.3.3.2 Production for graphem e .................................. ............... 89
3 .4 S u m m ary ............................................................................ 9 1

4 INTERFERENCE- RESULTS AND DISCUSSION .................... ......................... 98

4 .1 Introd uctio n ...................................... .................................................. 9 8
4 .2 R esu lts and D iscussio n ................................................................ .......................... 100
4.2.1 Production for G raphem e ................................................... .................. 100
4.2.1.1 Production for grapheme by production type and test...............102
4.2.1.2 Production for grapheme by production type and group ...........104
4.2.2 Production for Graphem e .................................................... 108
4.2.2.1 Production for grapheme by group and type .............................110
4.2.2.2 Production for grapheme by group, type and test....................112
4.2.3 Relationship between Spanish Production and Portuguese Production..........114
4.3 Other Observations ................. ............... ......... .............. ......... 115
4 .4 S u m m ary ................................................................................................ ............ 1 18

5 GENERALIZABILITY RESULTS AND DISCUSSION .........................................130

5 .1 In tro d u c tio n ........................................................................................................... 1 3 0
5.2 R results and D discussion ................................................ .......................................... 130
5.2.1 Production for Graphemes and .............. ................... ......... 130
5.2.2 P reduction by G raphem e...................................................... ..................... 13 1
5.2.3 C ognate < s> Item s................................................. .............................. 132
5.3 Other Observations ................. ............... ......... .............. ......... 136
5 .4 Su m m ary ...............................................................13 7

6 CONCLUSION................ ..... .. .. .... ...... ....... ........ 145

6.1 Introduction ................................ ..........................................145
6.2 Acquisition of Orthographic-Phonological Rules...................................................... 145
6.2.1 Error R solution ..................................... ......... .... .... .. ........ .... 145
6.2.2 Interference ................................. ............................ .... ....... 146
6.2.3 Generalizability .................. .......................... .... .... ................. 148
6.3 Contributions and Implications ............................................................................150
6.3.1 T heoretical Im plications....................................................... ............... 150
6.3.2 A applied Implications ......................................................... .............. 152









6.4 Limitations and Future Directions...... ........................................................... 153
6 .5 C lo sin g ............................................................................ 1 5 6

APPENDIX

A INFORM ED CON SENT FORM .......................... ..................................... ............... 158

B LANGUAGE BACKGROUND QUESTIONNAIRE................................159

C SUMMARY OF PARTICIPANT QUESTIONNAIRES .............. .... .................165

D W ORD S U SED IN READIN G TA SK S ......................................................... ...................170

E INSTRUCTION AND SAMPLE SLIDES FOR READING TASK .................................177

F TRANSCRIPTION GUIDE FOR RATERS ...................................................178

G RATER TRAN SCRIPTION SHEET ................. .......................... .....................179

H SAMPLE OF PARTICIPANT PRODUCTION DATA AND ASSIGNED VALUES
FO R A N D ITEM S ............................................................................................ 181

I PERCENTAGES OF CORRECT PRODUCTION BY PARTICIPANT AND GROUP
FOR EACH GRAPHEM E AND TEST ......................................................................182

J PRODUCTION ON SPANISH READING TASK AND AVERAGE PRODUCTION
ON PORTUGUESE READING TASKS....................................... .......................... 185

K PERCENTAGES OF PRODUCTION BY PARTICIPANT FOR EACH
PRODUCTION TYPE AND GRAPHEME.....................................................................187

L PERCENTAGES OF PRODUCTION BY PRODUCTION TYPE FOR EACH WORD ...189

M PERCENTAGES OF PRODUCTION BYW ORD ........................................ ..................191

L IS T O F R E F E R E N C E S .................................................................................. ..................... 193

B IO G R A P H IC A L SK E TC H ............................................................................. ....................200









LIST OF TABLES


Table page

1-1 Symbols examined, Portuguese sounds, and examples ........................................... 55

1-2 Symbols examined, Portuguese contexts and sounds, and examples .............................55

1-3 Symbols examined, Spanish sounds, and examples .................................. ...............55

1-4 Symbols examined, English sounds, and examples .......................................................55

2-1 Participant groups ...................................... .. .......... ............... 75

2-2 R revised particip ant group s............................................................................... ..... .... 75

2-3 Grapheme words ................................. ...... .. ............. ....... 75

2-4 G raphem e < z> w words ........................................................................................... 75

2-5 Summary of inter-rater agreement for and items (number and percentage)........76

3-1 Number and percentage of correct items for all [z], by test .....................................93

3-2 Participant groups ...................................... .. .......... ............... 93

3-3 Number and percentage of correct items for all [z], by group and test............................93

3-4 A N O V A for all production ..................................................................... .... .................93

3-5 R revised participant groups........................................................................... .............94

3-6 Number and percentage of correct items for all [z], by revised group and test ...............94

3-7 AN OVA for all production (revised groups) ........................................ .....................94

3-8 Significant differences found between tests, by group ............................................... 94

3-9 Significant differences found between groups, by test ..................................................95

3-10 Number and percentage of correct items, by grapheme and group .................................95

3-11 Production of [z] by native Spanish speakers in English reading task ...........................95

4-1 Number and percentage of items, by production type and test...............................121

4-2 ANOVA for all production types for ................................................................ 121

4-3 Post-hoc t-test results for all production types for on Test 1 .................................. 121









4-4 Post-hoc t-test results for all production types for on Test 2 .............. .................121

4-5 Post-hoc t-test results for all production types for on Test 3 ........... .................122

4-6 Number and percentage of items, by production type and participant group ...........122

4-7 Post-hoc t-test results for production type EP for items, by participant group.........122

4-8 Post-hoc t-test results for production type ES for items, by participant group.........122

4-9 Post-hoc t-test results for production type PO for items, by participant group ........122

4-10 Post-hoc t-test results for production type SP for items, by participant group.........122

4-11 Post-hoc t-test results for all production types for items for participant group El ...123

4-12 Post-hoc t-test results for all production types for items for participant group E2...123

4-13 Post-hoc t-test results for all production types for items for participant group P.....123

4-14 Post-hoc t-test results for all production types for items for participant group S .....123

4-15 Number and percentage of items, by production type and test..............................124

4-16ANOVA for all production types for ....................... ....... ........... .....................124

4-17 Post-hoc t-test results for all production types for on Test 1 .................................. 124

4-18 Post-hoc t-test results for all production types for on Test 2.............. ................ 124

4-19 Post-hoc t-test results for all production types for on Test 3 .................................. 124

4-20 Number and percentage of items, by production type and participant group...........24

4-21 Post-hoc t-test results for production type EP for items, by participant group.........125

4-22 Post-hoc t-test results for production type SP for items, by participant group.........125

4-23 Post-hoc t-test results for all production types for items for participant group El ...125

4-24 Post-hoc t-test results for all production types for items for participant group E2...125

4-25 Post-hoc t-test results for all production types for items for participant group P.....125

4-26 Post-hoc t-test results for all production types for items for participant group S.....125

4-27 Number and percentage of items, by production type, participant group and test ...126

4-28 Post-hoc t-test results for production type EP for items, by participant group and
te st ......................................... .. ................................................................................ 1 2 6
0)









4-29 Post-hoc t-test results for production type SP for items, by participant group and
te st ................................................... ........................... ................ 1 2 6

5-1 Number and percentage of correct items for all [z], by word type ...............................140

5-2 Number and percentage of correct items, by grapheme and word type........................140

5-3 Post-hoc t-test results comparing graphemes for each word type............................. 140

5-4 Post-hoc t-test results comparing word types for .......................... ............... 140

5-5 Post-hoc t-test results comparing word types for .....................................................140

5-6 Cognate words, with corresponding sound in English ........................................... 141

5-7 Number and percentage of correct items, by cognate set ......................................141

5-8 AN O V A for cognate item s .............................................. ............................. 141

5-9 Number and percentage of correct items for cognate sets, by participant group......141

5-10 Post-hoc t-test results for cognate set with [z] in English, by participant group ......142

5-11 Post-hoc t-test results for cognate set with [s] in English, by participant group ......142

5-12 Post-hoc t-test results for cognate set with [3] in English, by participant group ......142

5-13 Post-hoc t-test results for cognate sets for participant group El ...........................142

5-14 Post-hoc t-test results for cognate sets for participant group P..............................142

5-15 Post-hoc t-test results for cognate sets for participant group E2 ...........................142

5-16 Post-hoc t-test results for cognate sets for participant group S..............................142

C- Summary of Participant Questionnaires ................................................................ 165

D -1 W words used in reading tasks............................................... ......... ............................ 170

F-l Transcription guide for raters................................................. .............................. 178

H- Sample of participant production data and assigned values for and items........181

I-1 Percentages of correct production by participant and group for each grapheme and
te st ................... ................... ......................................................... .. 1 8 2

J-1 Production on Spanish reading task and average production on Portuguese reading
ta sk s ................... ............................................................ ................ 1 8 5









L-1 Percentages of Production by Production Type for Each Word ............................. 189









LIST OF FIGURES


Figure page

3-1 Percentage of correct items by test ....................................................... .................... 96

3-2 Percentage of correct items, by test for each group ..........................................................96

3-3 Percentage of correct items, by group for each test ........................................................96

3-4 Percentage of correct items, by grapheme for each group.............................................97

3-5 Percentage of correct items, by group for each grapheme......... .................................97

4-1 Percentage of items, by test for each production type ......................... ...........127

4-2 Percentage of items, by participant group for each production type .....................127

4-3 Percentage of items, by test for each production type.............................................128

4-4 Percentage of items, by participant group for each production type ........................128

4-5 Percentage of correct items on Spanish reading task and average percentage of
correct items on Portuguese reading tasks .............. .... .......................................... 129

4-6 Relationship between production on Spanish reading task and average production on
Portuguese reading tasks ........................................................... .. ............... 129

5-1 Percentage of correct items by word type .... ....................................143

5-2 Percentage of correct items, by grapheme and word type ......................................143

5-3 Percentage of correct items, by test for each cognate set ................ ................144

5-4 Percentage of correct items, by participant group for each cognate set....................144

E-1 Instruction slides and sample word slides from PowerPoint reading task....................177

G Rater Transcription Sheet Page 1 ............................................. ............................ 179

G-2 Rater Transcription Sheet Page 2......... ... .............. ......... ....................... 180









LIST OF ABBREVIATIONS

< > Indicate a grapheme

/ / Indicate a phoneme

[ ] Indicate an allophone

L1 First language

L2 Second language

L3 Third language

CLI Cross-linguistic influence

FL(A) Foreign language (acquisition)

GPCR Grapheme-phoneme correspondence rule

IL Interlanguage

OPM Ontogeny-phylogeny model

SL(A) Second language (acquisition)

T-unit Terminal unit

U Universals (part of the OPM)

UG Universal grammar

VOT Voice-onset time









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

THE ACQUISITION OF ORTHOGRAPHIC-PHONOLOGICAL CORRESPONDENCE
RULES IN L2 AND L3 PORTUGUESE: ERROR RESOLUTION, INTERFERENCE, AND
GENERALIZABILITY

By

Sharon Barkley

August 2010

Chair: Gillian Lord
Major: Linguistics

The current study investigates the acquisition of orthographic-phonological

correspondence rules by learners of Portuguese as an L2 or L3, with English and Spanish as the

previous languagess. Acquisition is examined from the perspectives of error resolution,

interference, and generalizability. Target L2/L3 production is shown to increase throughout the

course of the semester of data collection, as interference from the L1/L2 decreases. The results

support previous research, finding transfer from both the L1 and L2, and interference based on

language typology as well as language status. Word type (cognates, non-cognates and nonce

words) is not found to play as great a role in acquisition as grapheme and participants' linguistic

background. It is proposed that examination of the acquisition of orthographic-phonological

correspondence rules may have the potential for highlighting more specifically the sources) of

interference in an L3.









CHAPTER 1
INTRODUCTION AND LITERATURE REVIEW

1.1 Introduction

Language acquisition has long been a topic of interest in the field of linguistics, initially in

the area of first language (L1) acquisition (e.g., Bloomfield, 1933; Brown, 1973; Chomsky,

1959; Skinner, 1957), and later in the area of second language (L2) acquisition (e.g., Ellis, 2008;

Krashen, 1982; Selinker, 1972; White, 2003), and beyond bilingualism (e.g., Baker, 2006;

Grosjean, 2001), third language (L3) acquisition (e.g., De Angelis, 2007), and multilingualism

(e.g., Aronin & Hufeisen, 2009; Cenoz, 2000; Cenoz & Genesee, 1998). Studies in acquisition

have examined many different aspects of language, such as phonology (e.g., Hansen Edwards &

Zampini, 2008), pragmatics (e.g., Kasper & Rose, 2002), discourse analysis (e.g., Boxer &

Cohen, 2004), and so on. In examining the acquisition of an L2 or L3, the literature has often

discussed the effects of a learner's previous languages) on the new language (e.g., Cenoz, 2003;

Cenoz, Hufeisen& Jessner, 2001), and, in recent work in the area ofpsycholinguistics, the

effects of word similarities and differences between languages, including cognates and non-

cognates, have been considered (e.g., Costa, Caramazza & Sebastian-Galles, 2000; Costa,

Santesteban & Cafio, 2005). As far as Portuguese1 is concerned, the language has not had the

long research tradition that languages such as Spanish, French and German have had (Ornstein,

1942; Holton, 1954), but its importance as a world language has increased with the growing

importance of Brazil, Latin America's biggest economy (Margolis, 2009).





1 Portuguese is the official language of Angola, Brazil, Cape Verde, Guinea-Bissau, Mozambique, Portugal and SAo
Tome and Principe, as well as one of the official languages of East Timor, and one of the languages spoken in other
areas, such as Goa and Macau. The Portuguese language ranks within the world's top 10 most spoken languages (in
terms of number of speakers), with Brazil having by far the largest number of speakers, approximated at 180
million.









The present study brings together several of these areas, as it explores an aspect of

language acquisition which has not been widely discussed in the literature: orthographic-

phonological correspondence rules. This study considers the acquisition of these rules by

learners ofPortuguese2 as an L2 or L3, examining the influence of the learners' L1 (Spanish or

English), and where applicable L2, on the target language, and also considering the effect of

different word types (cognates, non-cognates, and nonce words) on that acquisition. It is worth

mentioning that much of the literature on the acquisition of Portuguese appears to deal with

impressions and observations, with relatively little empirical evidence being offered to date,

making this study an important contributor in the area of Portuguese acquisition.

As is the case with many studies in language acquisition, the current research project arose

as a result of classroom observations. In classes of students learning Portuguese as a foreign

language, it seemed that native English speakers with high proficiency in Spanish behaved more

like native Spanish speakers than native English speakers with low or no proficiency in Spanish,

when it came to the treatment of certain sound-symbol correspondences. For instance, the

intervocalic 3 in Brasil ("Brazil") was often pronounced as the voiceless alveolar fricative [s]

rather than the voiced alveolar fricative [z], the correct pronunciation in Portuguese. While this

seemed a logical approach for native Spanish speakers, whose L1 pronounces intervocalic as

[s] and where there is no phoneme /z/, it is not necessarily logical for native English speakers,

whose L1 does have the phoneme /z/, often pronounces intervocalic as [z], and has the

country name (Brazil) pronounced with [z] (as required by the orthographic-phonological

correspondence rule for in English). It appeared that the non-native but highly proficient


2 While the orthographic-phonological correspondences discussed here are the same across different varieties of
Portuguese, Brazilian Portuguese is the variety taught at the institution at which the present study was carried out.
3 <> are used for orthographic representations, [] are used for allophones, and / / are used for phonemes.

16









Spanish speakers drew on the grapheme-phoneme rules of their L2, Spanish, rather than their L1,

English, when it came to the pronunciation of L3 Portuguese words containing certain

orthographic-phonological correspondences, even when their L1 rules would be more helpful for

the target language pronunciation.

These observations led to the overarching questions that have motivated the current study.

First, to what extent are errors in foreign language pronunciation4 resolved over time? Second,

how do the grapheme-phoneme correspondence systems of the L1 and L2 influence the

acquisition of the L3 system? Third, what is the role of different word types (cognates, non-

cognates, and nonce words) in the acquisition of L3 orthographic-phonological correspondence

rules? In order to address these questions, it is necessary to consider the relevant research in

certain key areas. First, general L2 and L3 acquisition will be discussed, including the notion of

transfer, followed by a look at the acquisition of related languages, in particular Spanish and

Portuguese. Next, acquisition of L2 and L3 phonology will be considered, especially as it relates

to pronunciation. Thereafter, the acquisition of orthographic-phonological correspondence rules

will be examined, with a presentation of the pertinent rules of English, Spanish and Portuguese

for the sound-symbol correspondences considered in the current study. Finally, previous research

regarding the effect of different word types (cognates, non-cognates, and nonce words) in

acquisition will be considered.

1.2 L2 and L3 Acquisition

The fields of foreign language acquisition5 and bilingualism6 are well established, and

research in these areas has been abundant. In the literature, numerous theories and models have


4 Here, production of the orthographic-phonological correspondence rules in question.
5 For the purpose of this dissertation, the terms "foreign language acquisition" (FLA) and "second language
acquisition" (SLA or L2 acquisition) are used synonymously, when referring to a language that is learned after the
first language (LI).

17









been proposed to account for the acquisition of a second language as a whole, or the acquisition

of certain aspects of a second language (e.g., phonology), considering learners' different

developmental stages and topics such as transfer from the L1 to the L2 (and vice-versa),

fossilization of the L2, and achieving native-like status. It is beyond the scope of this project to

exhaustively review all possible SLA theories and studies, hence the discussion here will restrict

itself to those theories that are relevant to SLA insofar as L3 is concerned, or that are applicable

to the current project.

One theory of language acquisition which has received a great deal of attention for both L1

and L2 acquisition is universal grammar (UG), which proposes that there are certain grammar

principles, shared by all languages, that are said to be innate. Noam Chomsky has been an

influential figure in this area, arguing for an innate language faculty based largely on the

observed ability of L1 learners to pick up language so quickly and seemingly effortlessly, and to

produce grammatically correct language despite poor input, at least insofar as negative evidence

is concerned that is, evidence of ungrammatical production in the input (see Chomsky, 1965).

UG has been said to be made up of principles and parameters, where the principles are the

aspects of language that all languages share (a core grammar), and parameters are the language-

specific settings for these universal principles. For example, all languages have vowels but the

specific vowel systems of languages such as English, Portuguese and Spanish differ.

There has also been significant discussion about what part UG plays, if any, in the

acquisition of languages beyond the L1, particularly for adult learners (see White, 2003).

Perspectives vary, from full access to UG during L2 acquisition at one end of the spectrum to no

access to UG at the other end of the spectrum. It is beyond the scope of this dissertation to

6 Both simultaneous and consecutive learning of two languages is understood as bilingualism, for the purpose of this
project.









review all of these perspectives; it will simply be acknowledged here that UG plays some part in

L2 acquisition. Within this broad UG framework lies the Ontogeny-Phylogeny Model (OPM)

proposed relatively recently by Major (2001) to describe and account for the acquisition process

of the L2 and beyond. Major explained the terms ontogeny and phylogeny in both biological and

linguistic terms. Biologically, he described ontogeny as the "life cycle of a single organism" (p.

81), and phylogeny as "the evolutionary development of groups of organisms" (p. 81).

Linguistically, he rephrased the difference between the two with ontogeny as "the life cycle of an

individual's language" (p. 81) and phylogeny as the "life cycle of whole languages and language

types" (p. 81), where the latter included "historical change, dialectal variation, language loss, and

language contact phenomena" (p. 81).

In Major's view, a model of L2 acquisition should consider the components of a learner's

interlanguage (IL), justifying the importance of each component and describing the interaction

between the components. He assumed that the IL has as its components the L1, the L2 and what

he calls universals (U)7, given the numerous studies in the literature which indicate that learners

produce things which are neither part of the L1 grammar nor part of the L2 grammar (e.g.,

Broselow, Chen & Wang 1998). Given these three component parts of a learner's IL, Major

described the appearance of the IL at different stages of acquisition. Initially, he claimed, the

idealized learner would have an IL that is equivalent to the L1 and, finally, the IL would be

equivalent to the L2. In the interim, the L1 would decrease, the L2 would increase, and U, which

lay dormant initially, would increase and subsequently decrease. In addition, Major claimed that

the OPM could be extended to the IL of languages learned beyond the L2 (i.e., L3, L4, L5, and



Major (2001) suggests that language univeisals comprise UG, as well as the whole set of universal language
properties: leamability theory, markedness, underlying representations, rules and processes, constraints, and stylistic
variation.









so on). For an idealized L3 learner, for example, initially the IL would contain the components

L1 and L2, to varying degrees depending on their stages of completion (degrees of proficiency),

while finally the IL would be equivalent to the L3. Again, he claimed, U would increase and

subsequently decrease.

Such a model is useful for third language acquisition and multilingualism, fields which

have been drawing increasingly more attention, no doubt in part due to the realization that

multilingualism is common around the world (Cenoz & Hoffman, 2003), with bi- or

multilingualism being at least as frequent as monolingualism, if not more frequent (Ecke, 2001;

Hammarberg, 2001). The need has thus arisen to identify specifically what distinguishes third

language acquisition from second language acquisition and multilingualism from bilingualism.

Clearly, while third language acquisition has much in common with second language acquisition,

in terms of theoretical frameworks, for instance, there are traits which are specific to the

acquisition of a third language. The increased number of linguistic systems, which entail more

possibilities of combinations and interactions between linguistic systems, contribute to the

complexity of third language acquisition. As well as considering the differences between L2 and

L3 acquisition, this new subfield considers the influence ofbilingualism on third language

acquisition. Studies considering cross-linguistic phenomena, language use phenomena, and even

the question of the ideal age at which third languages should be introduced in schools are some

of the areas investigated within this sub field, where the L3 studied has almost always been

English (Cenoz and Hoffmann, 2003).

Next, some of the differences between L2 and L3 acquisition are considered, as well as

some of the effects of bilingualism on L3 acquisition. With L2 acquisition, the two languages

may be acquired either consecutively or simultaneously (i.e., L--L2, or Lx/Ly). When there are









three languages in play, four temporal possibilities arise: the three languages are acquired

simultaneously (Lx/Ly/Lz); the first two languages are acquired simultaneously before the third

language is acquired (Lx/Ly-L3); the first language is acquired before the other two are

acquired simultaneously (L-Lx/Ly); or the three languages are acquired consecutively

(L1-L2-L3) (Cenoz, 2003).

L3 learners, then, have more language experience to draw upon than L2 learners, as they

have access to two language systems instead of one (obviously depending on the temporal nature

of their acquisition process). Folk wisdom claims that it is easier for bilinguals and multilinguals

to learn a new language than monolinguals, and this has been the subject of numerous studies,

which have attempted to determine whether in fact bilinguals and multilinguals have an

advantage over monolinguals when learning a new language. The literature is divided as to this

matter. In a review by Cenoz (2003), many studies found a negative association between

bilingualism and cognitive development, prior to a study conducted by Peal and Lambert (1962),

in which they found that French-English bilingual children achieved higher scores on certain

verbal and nonverbal tests of cognitive ability, as compared with monolingual French or English

children. Cenoz (2003) reported that, since the 1960s, studies conducted to look at the effect of

bilingualism on cognitive development, metalinguistic awareness and communicative skills have

shown that bilinguals: 1) have higher scores in tests of creative thinking, 2) possess a higher

ability to reflect on language and manipulate it, and 3) show a greater sensitivity to their

interlocutors and make use of more varied communication strategies.

In reviewing the effect of bilingualism on general L3 proficiency, Cenoz (2003) reported

numerous studies which showed advantages for bilinguals and several studies which did not.

Nevertheless, the tendency was toward the association of bilingualism with advantages in L3









acquisition, when general L3 proficiency was considered. In one such study, Errasti (2003)

looked at the effect ofbilinguals' language proficiency on their L3 acquisition, studying 155

bilingual adolescents in a school in the Basque Autonomous Community in Northern Spain. Half

of the students (78), classified in the maintenance group, had Basque as their L1, and used

mainly Basque at home, at school and in social contexts, while the other half of the students (77),

classified in the immersion group, had Spanish as their L1 or used Spanish more than Basque

outside of school. (All participants had received instruction in Basque from the age of three,

while Spanish was introduced into the curriculum at the age of 3, and English at the age of 8.)

For the study, participants were required to write an informal letter and a recipe in Basque,

Spanish and English, over a period of three months, at monthly intervals, with the Basque

materials collected first, then the Spanish, and finally the English. Materials were graded using a

holistic evaluation, taking into account content, organization, language use, vocabulary, and

mechanics of writing and a T-unit8 evaluation, where fluency was calculated based on number

of words per T-unit, grammatical complexity based on the number of clauses per T-unit, lexical

complexity on the number of lexical words per T-unit, and accuracy based on the number of

errors (semantic, morphosyntactic, alphabetical, and lexical) per T-unit.

In Basque, the maintenance group performed significantly better than the immersion group

in all areas, which included overall production, fluency, grammatical complexity, lexical

complexity, and accuracy. In English, the maintenance group again performed better than the

immersion group in all areas, but this time significant differences were only found in the

categories of overall production and in fluency. In Spanish, there were no statistical differences

between groups in any of the categories. Considering the interaction of various factors, the

8 The term T-unit was coined by Hunt (1965) and is essentially the shortest grammatical unit that can stand alone (a
dominant clause and its dependent clauses). It has been used to analyze discourse and writing in both the L1 and L2.

22









author found two interesting positive associations: 1) high levels of competence in Basque and

Spanish were related to a high level of competence in L3 English; and 2) the measurements in

Basque, Spanish and English, for the maintenance group, were highly correlated, suggesting that

writing in each particular language was not an independent process. The first association

confirmed the Threshold Hypothesis (Cummins, 1979), which stated that the students who would

benefit most from their bilingualism would be those with high levels of competence in the two

languages; the second association supported the Linguistic Interdependence Hypothesis

(Cummins, 1979), which stated that knowledge in one language could be positively transferred

to another language.

The tendency toward the association ofbilingualism with advantages in L3 acquisition,

when general L3 proficiency was considered, was not necessarily found to be the case when

specific aspects ofL3 proficiency were considered (Cenoz, 2003). Some studies in phonology

(e.g., Davine, Tucker and Lambert, 1971) found no differences between bilinguals and

monolinguals, while others did (e.g., Cohen, Tucker and Lambert, 1967; Enomoto, 1994).

Studies in the area of syntax, morphology and the lexicon presented similar mixed results,

leading to the conclusion by Cenoz (2003) that bilinguals achieved more favorable results in

studies that considered general L3 proficiency rather than proficiency in very specific L3 aspects.

The general suggestion, then, is that there is no negative effect ofbilingualism on L3 acquisition,

and that in some cases the effect is positive. The present study considers the effects of learners'

previous languages on the L3 in a very specific area the acquisition of orthographic-

phonological correspondence rules so its findings will be interesting to consider in light of the

mixed results found in previous studies, insofar as advantages for bilinguals are concerned.









Another topic widely discussed in the literature on L3 acquisition is the notion of

interference or transfer from the learner's L1 and L2. While this is to be expected, given a model

like Major's (2001) OPM, where the beginning state of a learner's IL for the L3 is made up in

part by the L1 and in part by the L2, what is perhaps of most interest is the nature of interference.

Perhaps counter-intuitively, some studies have shown that during L3 production, learners often

inadvertently produced L2 forms, rather than L1 forms (Dewaele, 1998; Herwig, 2001; Selinker

& Baumgartner-Cohen, 1995). Faerch and Kasper (1986) suggested that these learner mistakes

did not appear to be used intentionally (for instance, to bridge a lexical gap), but rather they

appeared to result from the learners' inability to suppress the intrusive language.

According to a review by Murphy (2003), cross-linguistic influence9 has been approached

in different ways in the literature. Some believe there is no significant difference between the L1

and L2 acquisition processes, claiming that errors merely result from learners' testing of

hypotheses (Corder, 1967; Krashen, 1983), while others find transfer to be a natural and even

necessary part of acquisition (Selinker, 1972; Gass, 1983, 1984). Murphy (2003) considered two

sets of variables that could affect cross-linguistic influence (CLI): learner-based variables and

language-based variables. The learner-based variables included proficiency, amount of target

language exposure and use, language mode, linguistic awareness, age, educational background,

and context. The list of language-based variables included language typology, frequency, word

class, and morphological transfer.

Proficiency, for instance, was generally agreed to inversely influence language transfer;

that is, the lower the learner's proficiency in the L2, the greater the likelihood of language


9 Some prefer this term to the terms transfer and interference, since it is more encompassing, including the notions
that transfer may be positive as well as negative, and that transfer may be bidirectional (L 1-L2, and/or L2-L 1).
This preference for terms does not concern the present study, thus no discrimination is made heretofore between the
terms transfer, interference and cross-linguistic influence.

24









transfer from the L2 to the L3 (Odlin, 1989; Poulisse & Bongaerts, 1994). According to Murphy,

proficiency was a prominent feature in the literature on L3 acquisition, and it was necessary to

take into account the level of proficiency in all of a multilingual's languages. Hammarberg

(2001), on the other hand, believed that for there to be transfer from the L2 to the L3, there must

be a certain degree of proficiency in that language, while Shanon (1991) noted that the source of

transfer was often the more recently acquired L2, which was weaker than the L1 in terms of

proficiency.

Language exposure, Murphy pointed out, behaved similarly in the acquisition of L2 and

L3: in L2 acquisition, as exposure to the L2 increased, transfer from the L1 decreased; in L3

acquisition, as exposure to the L3 increased, transfer from the L2 (and presumably the LI)

decreased. This observation is consistent with Major's (2001) OPM, which has an initial IL state

for the L3 as consisting of parts of the L1, parts of the L2 and universals (U), and predicts that

the L1 and L2 will decrease, U will increase and then decrease, and the L3 will increase.

Interestingly, Dewaele (2001) claimed that transfer from the L1 for L2 speakers declined more

quickly than transfer from the L2 (and, presumably, the LI) for L3 speakers, and Murphy (2003)

suggested that this might be due to the more complex linguistic system of the multilingual than

the bilingual, again consistent with Major's model.

In terms of language mode, Grosjean (2001) proposed a continuum in which the "base"

language (L1) was always activated, while the "guest" language (L2) became increasingly more

active depending on the language mode (monolingual to bilingual). Murphy (2003) suggested

that this model might be adapted for L3 learners, whose base language (L1) would always be

active but whose guest languages (L2 and L3) would be active to varying degrees. She claimed

that the L1 would be easier to deactivate than the L2, though this claim seemed to be









contradicted somewhat by her subsequent discussion about the typological similarity between a

learner's languages. Language typology is explored later in this chapter, in section 1.2.1.

Another way of accounting for these "intrusions" from one non-native language (L2) into

another (L3) was put forwards by De Angelis (2005), who proposed a cognitive process by

which learners transfer lexical items from one non-native language to another without being

aware of it, calling such a process a "system shift." She suggested that such a shift might occur in

three stages: 1) first, the learner transfers a lexical item from one non-native language to another

(presumably due to a knowledge gap), believing that the item belongs to the source system only;

2) next, the learner associates the lexical item with the guest system, believing that it may belong

in both the guest system and the source system; and 3) the learner believes the lexical item

belongs to the guest system, no longer recognizing the source of her knowledge as being found

in the source system. The first stage has been widely discussed in the literature, and could

certainly be attested to by any foreign language instructor.

The hypothesized second and third stages were tested in three separate studies conducted

by De Angelis (2005). In the first study, a French-Canadian speaker of Italian, with previous

knowledge of English and Spanish, was interviewed in Italian about her personal experiences in

London and other foreign countries. Her interview was analyzed for instances of CLI from

French or Spanish. Six months after the initial data collection, the participant was interviewed

again, about the specific problematic lexical items. The participant was first asked if she was

familiar with the English words (read to her one at a time), and then she was asked to translate

them into Italian. Afterwards, the participant was asked if she was familiar with the correct target

Italian words. For the 12 troublesome lexical items identified, the participant produced a Spanish

or Spanish-influenced word at both interviews on eight occasions. Most of the 12 words shared









similar forms in Spanish and Italian, which might account for the overt lexical transfer from

Spanish, although in some cases the participant produced a Spanish-Italian blend, suggesting

previous familiarity with the correct Italian word. De Angelis (2005) suggested that this provided

some support for the notion that the participant believed she was producing an Italian word,

leading the researcher to question whether in some cases learners are even aware of the source of

their knowledge in the original linguistic system.

In the second study, De Angelis' (2005) participants were 10 L1 English university

students, with low proficiency in Spanish, enrolled in Italian language classes (of different

levels). The learners were given a list of English words to be translated into Spanish (all non-

cognates in English, Spanish and Italian), followed by a text in English to be summarized in

writing in Italian, followed by the same list of English words, this time to be translated into

Italian. (Participants were asked not to look at previous sections after completion, and were

carefully monitored to ensure compliance.) Results showed learners providing a Spanish word in

the Italian translation task, in some cases, but then failing to provide a translation for the same

item in the Italian translation task, and vice-versa, suggesting a lack of knowledge concerning the

original source system. In her third study, De Angelis (2005) had an L1 English speaker, with

previous knowledge of Italian and Spanish (as well as Latin, French and German), keep a diary

of those things which confused him most about learning and having to speak two non-native

languages. 10 The diary illustrated the participant's struggles in keeping the two languages

separate, with instances where he knew the words in both languages but didn't know to which

language a word belonged.



10 The researcher had been approached by the participant, who was reporting some confusion between Spanish and
Italian. He was living in Spain and studying Spanish and Italian, having previously learned Italian informally while
living in Italy.









De Angelis (2005) suggested two factors that might lead a learner to effect a system shift:

1) what she called "perception of correctness" (p. 11), where a learner quite easily perceives that

words in the native language are not correct target language words, and thus is able to block

transfer of these words, while less easily blocking transfer of words from another non-native

language which learners perceive to be correct (depending, of course, on levels of proficiency in

the non-native languages); and 2) "association of foreignness" (p. 11), where a learner comes to

associate (all) non-native languages as foreign, creating a cognitive link between the foreign

languages which does not exist between the native language and a foreign language. De Angelis

believed that these two factors might help to explain why L3 learners have been observed to

block transfer from their L1 in favor of transfer from another non-native language.

The present study considers the nature of interference (from the L1 and/or the L2) by

examining the production of orthographic-phonological correspondences in the L3 by learners

with different linguistic backgrounds (L1 English with L2 Spanish, and vice-versa), and of

differing levels of proficiency in the L2. The results will be interesting to compare with those

observations outlined above, to see where the intrusions are coming from (i.e., the L1, a more

proficient L2, or a less proficient L2). Despite differing views as to how "intrusions" end up

being part of a non-native language be they from another non-native language or from the

native language; be they due to low or high proficiency in a non-native language cross-

linguistic influence is undeniable, especially in the case of similar languages (Carvalho, 2002).

For this reason, it is important to consider now what the literature has noted with regard to

related languages.

1.2.1 Acquisition of Typologically Similar Languages

Language typology is a rather broad term which refers to language classification, and there

are many ways in which to classify languages (Fromkin, Rodman & Hyams, 2007). Syntactic

28









typology, for one, considers word order in a particular language or group of languages, such that

languages with subject-verb-object (SVO) word order (e.g., English, Spanish, and Portuguese)

would be considered typologically similar to each other but different from languages with, for

example, subject-object-verb (SOV) word order (e.g., German, Dutch, and Japanese). From a

phonological perspective, typology may refer to the syllable structure permissible within a

language: from a simple consonant-vowel (CV) and vowel (V) structure (e.g., Japanese) to more

complex structures including consonant clusters at the beginning and end of syllables (e.g.,

English, Spanish, and Portuguese). Typology may also refer to the origin of a language, or the

family to which a language belongs. For instance, Spanish and Portuguese are derived from

Latin, while English, German and Swedish are derived from an earlier form of Germanic known

as Proto-Germanic (Fromkin, Rodman & Hyams, 2007).

With regard to the effect of language typology on the acquisition of L2 and L3, it seems

that similar typology is often advantageous, being a source of facilitation or positive transfer.

Considering similarities due to language family, Orr (1987, as cited by Jarvis & Odlin, 2000)

looked at the acquisition ofprefixation in Chichewa, a Bantu language, and found an advantage

for L1 speakers ofNgoni, another Bantu language, compared with learners of Chichewa with

Gujarati as their L1, an Indo-European language that does not have the same Bantu prefixing

system.

Considering languages that were related (and unrelated) in their morphological structures,

Jarvis and Odlin (2000) examined L1 Finnish and L1 Swedish learners of English as an L2, in

their use or non-use of prepositions with the verbs take, sit and put, in a task where they were

asked to write a narrative for a short silent film they had just watched. Swedish and English were

considered typologically similar in this study, with regard to their free, prepositional









morphology, while Finnish was considered typologically dissimilar, with bound, agglutinative

morphology for spatial references. The authors found that learners with the typologically

dissimilar L1, Finnish, omitted prepositions in English in all of the spatial contexts examined,

while those learners with the typologically similar L1, Swedish, did not omit English

prepositions in any of the contexts.

Cenoz (2001) also found evidence of greater cross-linguistic influence between L2/L3

English and a typologically similar L1 or L2 than a dissimilar one. In a study of L1 Basque

and/or L1 Spanish elementary and secondary school children learning English", using a story-

telling task in English, Cenoz found greater transfer from Spanish, an Indo-European language

like English, than from Basque, a non-Indo-European language. These studies suggest more

positive cross-linguistic influence when the L1 is typologically similar to the L2.

Studies considering the effect of the typological similarity or dissimilarity of an L1 and L2

on the L3 have shown mixed results. On the one hand, facilitation has been found between an L2

and L3 when they were typologically similar. For instance, Ecke and Hall (2000, as cited in

Ecke, 2001) found that L1 Spanish speakers with L2 English and L3 German had much stronger

influence from the L2 than from the L1 in written production of the L3. In other cases, no

facilitation was found between typologically similar languages. For example, in a study of

German prepositional verbs, where participants had a variety of L1, L2, L3 and sometimes L4

backgrounds, Gibson, Hufeisen and Libben (2001) examined the effect of similar and dissimilar

previous languages on learners' production of German prepositions in a fill-in-the-blank task12

Gibson et al. found that those participants with L1 s structurally similar to German did not

1' All participants were native speakers of Basque and/or Spanish, had Basque as the language of instruction at
school, and had Spanish and English as school subjects.
12 Verbs were given and participants had to supply the following preposition. For example, gehoren zu ("belong
to"), and sprechen iiber ("talk about").









perform differently from those with structurally dissimilar LI s, nor did having English as a

foreign language help German learners, even when the German preposition would be a

translation equivalent of English.

The current study considers Spanish and Portuguese, two languages which are closely

related in terms of language family, many structural aspects and lexical items. In the case of

these two languages, most of the literature would agree that knowledge of one language

generally facilitates the acquisition of the other language, although it may also prove to be a

hindrance in some respects. To follow is a review of the literature that considers the relatedness

of these two languages.

1.2.2 The Relatedness of Spanish and Portuguese

Garrison (1979) highlighted the similarities between Spanish and Portuguese, observing

that there are few languages as closely related, and that there are fundamental similarities that

transcend even regional dialects of the two languages. In an overview of a presentation on

Portuguese given to Advanced Spanish students, he suggested that around 80% of Portuguese

words have Spanish cognates, going on to point out some of the systematic differences between

Spanish and Portuguese orthography (e.g., the grapheme <9> which exists in Portuguese but not

in Spanish), phonology (e.g., nasal vowels in Portuguese, Portuguese monophthongs for certain

Spanish diphthongs), lexical items in the form of non-cognates (e.g., "street" is calle in Spanish

and rua in Portuguese) and false cognates (e.g., apellido is Spanish for "last name" while apelido

is Portuguese for "nickname"), and grammatical items, such as gender differences on nouns (e.g.,

masculine mensaje in Spanish and feminine mensagem in Portuguese, "message"), verb

conjugations (e.g., the first person singular conjugation of the verbpoder, "to be able to," is

puedo in Spanish andposso in Portuguese), verb tenses (e.g., Spanish does not have the personal

(conjugated) infinitive that Portuguese has), and so on.

31









Holton (1954) also highlighted the relatedness of Spanish and Portuguese. In an effort to

promote the great literary works in the Portuguese language, he suggested that Spanish speakers

might easily learn to read Portuguese with minimum effort, given an understanding of certain

basic differences between the two languages. He delineated some of the orthographic and

phonological differences between Spanish and Portuguese, as well as some grammatical

differences (focusing on articles, verbs and pronouns), with the intent of showing how simple it

would be for a Spanish speaker to develop a recognition skill like reading. He clarified that such

a method would not suffice for the acquisition of an active command of the language,

highlighting that speaking and comprehension of spoken Portuguese would require a more

concerted effort. Nonetheless, the fact remains that the languages do share considerable

structural and lexical similarities.

Resnick (1945) illustrated the advantages and disadvantages of having a background in

Spanish when learning Portuguese by listing examples of words which are cognates, non-

cognates and false cognates in the two languages. He also highlighted that cognates (including

false cognates) often differ in the languages in terms of spelling and pronunciation, something

which learners must bear in mind during the acquisition process.

Azevedo (1978) observed that university students acquainted with one language (Spanish)

experienced facilitation in the acquisition of a new language (Portuguese) due to the structural

similarities between the two languages, yet at the same time, these precise similarities often

caused problems for the learners, leading to transfer of several Spanish features, such as

morphological, phonological, lexical and syntactic markers, which did not belong in Portuguese.

In terms of phonological errors, Azevedo specifically pointed out, among other things, the

feature under investigation here: the tendency of Spanish speakers to devoice the voiced alveolar









fricative [z] in Spanish-Portuguese cognates like zero ("zero") and casa ("house"). Additionally,

Azevedo highlighted the fact that although Portuguese instructors might be aware of Spanish

interference, it was difficult for them to correct the problem, "...because the diagnosis of the

problem is done, more often than not, in an impressionistic and unsystematic manner" (p. 19),

rather than being based on empirical evidence, thus highlighting the need for further research on

learners' production in order to enable targeted correction strategies.

Tarquinio (1977) also mentioned some of the difficulties with pronunciation that arise for

learners of Portuguese. As a native Brazilian Portuguese speaker and an instructor of Portuguese

and Spanish, she was particularly attuned to common pronunciation mistakes which resulted

from interference (particularly from Spanish), including the devoicing of the intervocalic ,

which is voiced in Portuguese. She went so far as to state that these interferences hurt the ears

of the native listener because they are foreign" (p. 82), and suggested that students be alerted

early on as to these common interference, by explicitly drawing their attention to them.

While much of the literature on Spanish and Portuguese highlights the mutual

intelligibility of the two languages, there are those who maintain that the spoken languages are

quite different in pronunciation, resulting in reduced understanding particularly for Spanish

speakers trying to understand Portuguese (Ellison & Andrews, 1969; Jebsen and Biel, 1986;

Timberlake, 1989). These observations about the similarity between the two languages, or lack

thereof, prompted Jensen (1989) to conduct a study looking at extent and direction of

intelligibility, as well as the effect of certain non-linguistic factors (age, sex, education, attitude

and experience with the other languages). Thirty-nine native speakers of Portuguese listened to

four recorded texts read in Spanish, while 32 native speakers of Spanish listened to four recorded









texts read in Portuguese. 13 Each group had to answer five multiple-choice comprehension

questions per text, written in their native language, and complete a short questionnaire asking

about the non-linguistic factors considered by the study. The native Portuguese speakers

performed significantly better than the native Spanish speakers on three of the four texts; the text

on which native Spanish speakers performed better than native Portuguese speakers related to

Ecuador, and it was postulated that the Spanish speakers, all of whom were from Latin America,

might have had more familiarity with the subject than the Portuguese speakers, all of whom were

from Brazil. Of the non-linguistic factors, the only one found to significantly and positively

affect participants' scores was experience with the other language, which came as no great

surprise. In this study of passive listening to recorded voices, Jensen found Spanish and

Portuguese to be mutually intelligible, but only at a level of about 50% to 60%. Additionally, the

results supported the common belief that Portuguese speakers understand spoken Spanish better

than Spanish speakers understand spoken Portuguese, although the difference was not

staggering, and might be attributable to individual factors.

Jordan (1991) described the differences between Spanish and Portuguese as mainly

morphological and phonological, as she made a case for the use of a contrastive method in the

teaching of Portuguese to speakers of Spanish (or, presumably, vice-versa). Jordan claimed that

there is little point in a contrastive course if the goal is simply for speakers to understand each

other and make themselves understood, since these two tasks are achieved relatively effortlessly

with little or no instruction in the other language (see Jensen, 1989). However, for a speaker to

reach a certain level of proficiency in writing and speaking the non-native language, Jordan

suggested the use of a contrastive method, highlighting the basic phonological and

13 Participants with "extensive experience of the other language" (p. 850) or who were of "foreign (e.g., European,
Asian or North American) background" (p. 850) were eliminated from analysis.

34









morphological differences, and particularly the few exceptions to the general rule. Given that

learners would make comparisons between the languages anyway, Jordan believed that it would

be helpful to make this process a conscious one, thereby limiting the erroneous assumptions and

generalizations that students might make. In a qualitative examination of a class of Spanish

speakers learning Portuguese, where she used a reference book designed to cover the principal

phonological and morphological differences between the two languages (making no mention of

similarities), Jordan found that students concentrated on certain differences while having

problems with other discrepancies, one of which was the intervocalic which was often not

voiced. To overcome these problems, Jordan suggested the need for increased exposure to the

non-native language, particularly in the areas of listening and reading, and also for as much

communication as possible to be in the target language in the classroom, in order that learners

might overcome the fear of making mistakes.

Carvalho (2002) discussed the inevitability of transfer, in particular from a known

language to a similar language being learned, where this transfer might be positive or negative.

In the case of Spanish and Portuguese, where there is a high degree of mutual intelligibility14

even between monolingual speakers of the two languages, a primary concern of instruction is

early fossilization. Where learners are able to achieve a high level of communicative ability early

on, Hadley (1986) stated that "fossilization is more likely to occur if learners see no reason to

improve their interim grammar and decide that it is adequate to serve its needs" (p. 268). Thus,

Carvalho (2002) claimed that instruction for speakers of a typologically similar language must

include activities which emphasize grammatical correction and promote metalinguistic



14 Jensen (1989) found this to be approximately 50-60% in a recorded listening task, but Carvalho (2 '" ,2) suggested
that this percentage might be higher in a conversational setting, where speakers could negotiate meaning and make
use of visual cues.









reflection, and she called for further studies to contribute empirical data which might reveal and

contrast different stages involved in the acquisition of Portuguese by Spanish speakers.

In this section, particular consideration has been given to the relatedness between Spanish

and Portuguese, and it has been noted that phonological differences between these two languages

often pose problems for learners. The current study considers the role that the typology of

learners' previous languages plays in the acquisition of the L3, as it examines interference from

English and Spanish in participants' production of Portuguese. Here, Spanish is considered to be

closer to Portuguese (more typologically similar) than English, given the language family,

structural aspects and lexical items that Spanish and Portuguese share. In the next section, the

acquisition ofL2 and L3 phonology will be discussed, specifically as it relates to the acquisition

of pronunciation.

1.2.3 Acquisition of L2 and L3 Phonology

In some of the earlier approaches to second and foreign language pedagogy, pronunciation

and oral skills in general were not stressed as highly as listening skills (Lazaraton, 2001). With

the advent of more communicative approaches, the importance of speaking has increased,

although focus on pronunciation has still been somewhat lacking. Blanche (2004) noted that the

Communicative Approach does not concern itself so much with accent reduction as with

"intelligibility" (p. 178), and Keys (2000) stated that this approach "encouraged a tendency to

leave pronunciation matters to one side" (p. 91). Goodwin (2001) described pronunciation as a

critical element because it is "the language feature that most readily identifies speakers as non-

native" (p. 117). Poor pronunciation, as perceived by native speakers, may result in the non-

native speaker feeling embarrassed and inferior (Goodwin, 2001), while "good" pronunciation in

a foreign language does not draw the attention of a native speaker away from what is being said

to how it is being said (Hockett, 1950).









While the place of pronunciation in pedagogy may be debatable, it is clear that

pronunciation poses problems for many learners (e.g., with relation to the acquisition of

Portuguese pronunciation, see Azevedo, 1978; Tarquinio, 1977; Jordan, 1991), and as such,

pronunciation is an area of FL acquisition which has led to many studies exploring the nature of

production. It is beyond the scope of this dissertation to review exhaustively the studies which

have looked at the acquisition of L2 and L3 phonology, since much of the work which has been

carried out in these fields is, at best, only tangentially related to the present study, which looks at

the role of orthography in the acquisition of pronunciation. (For a review of theoretical

approaches to phonology and current trends in L2 phonology see Major, 2001, 2008, and Hansen

Edwards and Zampini, 2008.) However, by way of example, a few of the studies which have

examined the influence of learners' previous languages) on their acquisition of pronunciation

will be presented here.

Major (2001) states that there has been much research done in L2 phonology looking at

claims (from Contrastive Analysis) that similar phenomena are more difficult to learn than

dissimilar phenomena, with the psycholinguistic reason for this seeming to lie in the fact that

"gross differences are more often noticed, due to perceptual saliency, [than] minimal

differences" (p. 37). Flege (1986) stated that, given the L1 as a starting point for the L2, learners

could categorize L2 phones into three categories: "identical," "similar," and "new." "New"

phones in the L2 are those which have no counterpart in the L1; "similar" phones are those

where the sounds in the L1 and L2 are acoustically similar but not exactly the same; "identical"

phones are those which share all acoustic properties in the two languages. Flege (1987)

hypothesized that "new" phones in the L2, with no counterpart in the L1, would be easier to









acquire than "similar" phones, with a counterpart in the L1 that differed phonetically in some

minimal way.

In his study, Flege (1987) examined Voice Onset Time (VOT) in the production of

"similar" phones (/u/ and /t) and a "new" phone (/y/) by bilingual English-French speakers and

monolingual English and French speakers. His participants included: 1) French and English

monolinguals (to provide an idea of phonetic "norms" in the two languages), 2) three groups of

American English speakers who had learned French as adolescents or young adults (the groups

differed in their level of proficiency in French), and 3) French speakers who were highly

proficient in English. Participants carried out two tasks, both based on lists of phrases provided

in French and English, containing /tu/ and /ti/ in English (two and TV), and /tu/ and /ty/ in French

(tous "all" and tu "you"). In the first task, participants simply read the lists of phrases; in

the second task, they were required to produce original, complete sentences based on the phrases

they had just read in isolation. Results for /t/ production showed that the greater the learners'

proficiency in L2 French, the closer their production was to that of French monolinguals, both

for English /t/ and French /t/. The group of L1 English speakers highly proficient in French

produced English /t/ intermediate to the norms of the L1 and L2. Similarly, the group ofFrench

speakers highly proficient in English produced both English and French /t/ intermediate to the

norms of English and French. These results seemed to suggest that the greater the experience in

two languages, the less difference there is between the L1 and the L2 forms. As for /u/

production, the L1 English speakers consistently produced English/u/ similarly to monolingual

English speakers. However, the L1 French speakers failed to produce French /u/ according to the

norm set by the French monolinguals. Thus, learning French did not appear to affect the English

speakers' production of L1 /u/, but learning English did affect the French speakers' production of









L1 /u/. Finally, for /y/ production, only the group of English speakers with the lowest level of

proficiency in French differed significantly from the group of French monolinguals. This result

showed that, with the "new" phone (/y/), where there was no possibility of an approximation to

an English (L1) counterpart, the learners of French were able to come close to producing an

authentic L2 sound.

Baker and Trofimovich (2005) examined groups of L1 Korean speakers learning English

as an L2 (children and adults), in order to determine how age of acquisition influenced the

organization of learners' phonetic systems. The researchers carried out a picture-naming task to

elicit six English vowels in 18 CVC monosyllabic words and five Korean vowels in 10 disyllabic

words, first collecting baseline data from monolingual Korean children and adults, and

monolingual English children and adults, then collecting data from the L1 Korean speakers

learning English as an L2, with the latter participants being divided into early bilinguals (those

who learned the L2 before the age of 15) and late bilinguals (those learning the L2 after the age

of 15). Analyzing the production data acoustically, the researchers compared the production of

the bilinguals with that of the monolinguals, as well as comparing the production of the early

bilinguals with that of the late bilinguals. Results showed that early bilinguals produced different

acoustic realizations of the English and Korean vowels, while late bilinguals produced English

vowels that were often "colored" by the L1. In fact, the more acoustically similar the vowels in

the L1 and L2 were, the more likely the "coloring" of the L2 vowels by the acoustic properties of

the L1 vowels, such that late bilinguals only produced acoustically different sounds when the Ll-

L2 pairs of sounds were very different. The researchers also observed that length of exposure to

the L2 in late bilinguals was significant, in that learners tended to produce L2 sounds as L1

sounds more in the early stages of acquisition than in later stages of acquisition.









Wade-Woolley (1999) observed that a learner's L1 phonological system constrained the

learner's ability to perceive and produce sounds in the L2, citing several studies to illustrate this

point: 1) a study by Werker and Tees (1984) showed L1 English speakers' inability to

distinguish between Hindi dental and retroflex stops, since this distinction does not exist in

English; and 2) studies by Goto (1971) and Yamada & Tohkura (1992) showed L1 Japanese

speakers learning English having problems with the /1/-/r/ contrast which exists in the L2 but not

in the L1.

It is clear from these studies that L1 phonology has a considerable role to play in L2

phonology, and transfer from a learner's previous language is an important factor to consider in

the acquisition of L2 pronunciation. Further, Muller and Muller (1968) observe that transfer is

particularly problematic when a written stimulus is used to evoke an oral response. Thus, it is

important to take the written form into account in the discussion of the acquisition of

pronunciation. The correspondence between orthography and phonology is at the heart of the

present study, and the acquisition of such correspondence rules will be discussed next.

1.3 Acquisition of L2 and L3 Orthographic-Phonological Correspondence Rules

In the field of L2 and L3 acquisition, few studies have focused on the relationship between

orthography and phonology, and more specifically how acquisition of the two is influenced by a

learner's previous languagess. This area merits further attention, however, because sound-

symbol correspondences are important not only for effective reading and writing skills, but they

can also relate to good pronunciation (Olshtain, 2001). Many studies looking at the relationship

between orthography and phonology have focused on English as an L2 or L3 and concentrated

on differences arising as a result of different L1 scripts, that is, alphabetic and non-alphabetic

scripts.









Koda (1999) described orthographic structures as varying along two dimensions: 1) the

fundamental unit of orthographic representation, and 2) the depth of representation. The basic

unit of representation in alphabets is the phoneme, while the morpheme is the basic unit in

logographies. With alphabets, readers must systematically analyze component letters and letter

clusters within a word, and reading competence requires that readers realize that written symbols

correspond to speech units. With logographies, it has been argued that readers arrive at

phonological information through whole-word lexical retrieval rather than through word-internal

analysis (Gleitman, 1985). Orthographic depth, the second dimension described by Koda, relates

to the degree of regularity found in sound-symbol correspondences. Shallow orthographies, like

Spanish and Portuguese, have a high degree of orthographic-phonological regularity, while deep

orthographies, such as English, have much less consistent correspondences, as evidenced by the

orthographically similar yet phonologically dissimilar related words anxious and anxiety.

With these differences in mind, Koda (1999) examined adult learners of L2 English, and

explored the effect of different L1 backgrounds (alphabetic and logographic) on the learners'

intraword sensitivity. The study's participants were 20 native speakers of Chinese, chosen

because of their experience with a logographic script, and 20 native speakers of Korean, selected

because of their experience with the non-Roman alphabetic script Hangul. The two groups were

comparable in terms of their length of stay in the US (less than 6 months), the type and length of

instruction they had had in English in their own countries (grammar/reading methods in high

school), and their scores on listening and reading sections of the Test of English as a Foreign

Language (TOEFL).

The participants completed two tasks: an orthographic acceptability judgment task and two

decoding tasks. For the first task, a series of 40 nonsense words were created by rearranging real









words, such as "double" and "report", to yield legal strings (according to English phonotactic

rules) like "boudel" and "troper," as well as illegal strings such as "ebdluo" and "tproer."

Participants were asked to judge the orthographic acceptability of these nonsense words, being

allowed to spend as much time as necessary on the test. 15 For the second task, two decoding

activities were used. The first required that participants read aloud 50 pseudo-English words,

with some allowance being made for certain non-native pronunciation (such as devoicing of final

/b/ by Korean speakers). The second decoding task was a homophone judgment test, where 30

real English words were presented visually paired with orthographically legal strings, with half

of the pairs being homophonic (e.g., "please"-"pleeze") and half non-homophonic (e.g.,

"dream"-"draim"). Participants indicated their judgment of these pairs, by circling S for same or

D for different. It was predicted that the Korean learners, given their experience with another

orthographic script and ensuing intraword sensitivity, would outperform the Chinese learners on

both tests. In fact, the Korean learners did outperform the Chinese learners in the orthographic

acceptability test on those nonsense words with illegal strings, suggesting that their L1 intraword

sensitivity might have been extended to their L2. However, there were no significant differences

between learners on the decoding tests, possibly resulting from the fact that English has two

representational properties for its orthography morphemes and phonemes. Koda suggested that

learners using logographic strategies might have had less success with a phonologically shallow

orthography, an observation requiring empirical testing.

Wade-Woolley (1999) also examined the orthographic and phonological effects of

different L1 scripts on word reading in the L2, in a study of L1 Russian and L1 Japanese learners

of English as an L2. Participants undertook seven tasks which included a mixture of standardized

15 Presumably, Koda used the term orthographicc acceptability" to refer to phonotactic acceptability, since the illegal
strings violate English phonotactic rules rather than, specifically, orthographic rules.

42









and experimental tests involving reading comprehension and vocabulary, reading of isolated

words (previously seen and unseen), matching of aural stimuli with a correctly spelled word,

distinguishing between visually similar (phonotactically permissible and non-permissible)

English pseudo words, repetition of aurally presented pseudo words, and phoneme deletion.

Japanese learners were found to perform better on tasks which required recognition of correct or

permissible orthography (real and pseudo words, respectively), while Russian learners performed

better on the phoneme deletion task. These results suggested that Japanese learners relied less on

phonology than orthography in reading, as the researcher had anticipated, due to the learners'

non-alphabetic L1, lending support to the idea that there are transfer effects to be found in L2

reading.

Relatively few studies have dealt specifically with the acquisition of Portuguese

pronunciation, and particularly as it relates to orthography. One such study was conducted by

Defior, Martos and Cary (2002), who examined children learning to read in their L1 Portuguese

or Spanish. Although both of these languages are considered to have shallow orthographiesl6, the

grapheme-phoneme correspondence rules (GPCRs) in Portuguese are less consistent than those

in Spanish. In their study, Defior et al. (2002) examined the extent to which differences between

the two orthographies impacted reading strategies, where these were divided into two types: 1)

phonological, sublexical, or indirect; and 2) visual, lexical, or direct. (Strategies of the first type

relied more on GPCRs to assemble the pronunciation of lexical items while strategies of the

second type required a reader to use direct access to the lexicon in order to retrieve an item's

pronunciation.) Given the greater asymmetry and complexity in Portuguese than in Spanish, the

authors hypothesized that Spanish children would perform better (i.e., faster and with greater

16 In a shallow orthography, the grapheme-phoneme correspondence rules (GPCRs) have highly consistent rules,
while the GPCRs in a deep orthography, such as English, have inconsistent and unpredictable rules.

43









accuracy) than Portuguese children on nonce words. In their experiment with 120 Spanish and 94

Portuguese children in grades one through four, participants read a list of numerals (2, 3, 4, etc.),

number words (dois, tres, quatro17, etc.), and nonce words (nois, nes, datro, etc.). Errors were

divided into two types, in order to disclose different types of reading strategies (i.e., indirect

phonological vs. direct lexical): 1) phonological errors, where deletion, substitution or some

other inversion of consonants and/or vowels led to the production of another nonce word; and 2)

lexical errors, where errors led to the production of a real word. Thus, phonological errors were

indicative of a failure in the use of the indirect phonological strategy, while lexical errors

reflected a failure in the use of the direct, lexical strategy.

Across all grades and both orthographies, participants were significantly slower with nonce

words than with numerals and number words. Additionally, results confirmed the authors'

hypothesis, as the Portuguese children were slower with nonce words than the Spanish children.

Spanish children also read number words faster than Portuguese children, but there were no

significant differences found between the two orthographies with numerals. In terms of errors,

there were no significant differences found between grades or orthographies for numerals or

number words. The pattern of errors in reading nonce words was different for the two

orthographies: Portuguese children steadily decreased in number of errors from grades one to

three, while Spanish children maintained a stable rate of errors from grade two on. Additionally,

Portuguese children made significantly more phonological errors than Spanish children (except

at grade three). Spanish children also made fewer lexical errors initially than Portuguese

children, although this distinction disappeared in grade two, only to reappear in grade four,

where Portuguese children made fewer lexical errors than Spanish children. The increase in


7 Dois ("two"), tres ("three"), quatro ("four"), etc.









lexical errors between grades three and four, found for both orthographies, might be indicative of

a change in reading strategies: in grade three, Spanish and Portuguese children might be turning

from an indirect, phonological strategy to a direct, lexical one. In summary, the participants

generally performed better on numerals and number words than nonce words, and the children

with the simpler orthography (Spanish) performed better on nonce words than those with the

more complex orthography (Portuguese).

Muller and Muller (1968) also examined the influence of orthography on the acquisition of

pronunciation in Portuguese, but with L2 learners rather than L1 learners. In their study of 140

high school students receiving instruction in Portuguese, they considered: 1) whether the

exposure of the written form interfered with the acquisition of pronunciation; and if so, 2) which

letters or letter combinations were most likely to cause interference. Participants were divided

into two groups: those who were denied access to the written form during the first four weeks of

instruction in Portuguese, and those who were allowed access to the written form but without

explicit instruction on the relationship between orthography and phonology. After four weeks of

instruction, participants were given picture cue sheets and asked to produce sentences which had

formed part of drill materials used during the instruction phase. The participant utterances were

recorded and later listened to and judged by raters who were both native speakers of Portuguese

and language instructors, using a three-point scale"1. Results indicated that the group of

participants who had had access to written materials performed significantly worse than the

group which was denied access to the written form, suggesting that the written form had indeed

interfered with the acquisition of pronunciation.


18 The criteria for judging the participants' production items (sentences) are not included by the authors, thus it is not
possible to know whether assessment was global or specific. However, given that the authors go on to postulate
about GPCRs, it would seemthat assessment of specific sounds would be necessary in order to allow an informed
discussion of those GPCRs which were found to be most problematic for their participants.

45









The researchers then considered, in abstract terms, possible combinations of grapheme-

phoneme correspondences in English and their correspondents in a target language, and

predicted that the most logical source of interference would be when a grapheme A in both

languages represented a phoneme X in one language and (an entirely different) phoneme Y in the

target language. Considering this specific permutation AX AY in their data, the researchers

found that the participants who had been denied access to the written form performed

significantly better than those who had been allowed access to the written form. This suggested

evidence of interference due to the difference in pronunciation of the same grapheme in the two

languages.

The present study explores the nature of interference in the case of this same orthographic-

phonological permutation, where a single grapheme represents different phonemes in different

languages, by considering how L1 English/L2 Spanish and L1 Spanish/L2 English speakers treat

the graphemes and in L3 Portuguese. As previously mentioned, the typological

similarity between Spanish and Portuguese is at once an asset and a liability (Resnick, 1945),

with the differences in GPCRs facilitating certain phonological combinations but leading to

continued mispronunciations by learners in other cases (Jordan, 1991). Examining the sound-

symbol correspondence systems in English, Portuguese and Spanish for the GPCRs considered

in the present study will highlight the similarities and differences between the three languages.

1.4 The Relevant Rules of English, Spanish and Portuguese

Two orthographic symbols are considered in the present study: and intervocalic 19

In Portuguese, both of these graphemes are pronounced as [z] (Table 1-1). In Portuguese, there is




19 It was originally intended that the study examine six sound-symbol correspondences -[z], intervocalic -
[z], -[3], -[3], -[h], and word-initial -[h] (Faraco & Moura, 1990) and data were collected with
this objective in mind. However, due to the complexity of the project, this dissertation will examine only the first
46










not always a one-to-one correspondence between graphemes and phonemes. In the case of the

two graphemes considered in this study, has a one-to-one grapheme-phoneme

correspondence but does not. The grapheme is always pronounced as [z]. The grapheme

has two possible pronunciations: 1) [s] word-initially, and after consonants <1, n, r>; and 2)

[z] in intervocalic position and before voiced consonants (Table 1-2). It is worth mentioning that

[s] exists between vowels, but with a change in the orthography, from to . This

orthographic change (with accompanying phonological change) is phonemic, resulting in a

change in meaning, as demonstrated by the minimal pairs asa ([aza], "wing"), and assa ([asa],

"s/he bakes"). The orthographic-phonological correspondence rule -[s] was not considered

in the present study, however, as the aim was to consider the phoneme /z/ and its two

orthographic correspondences, and intervocalic 20

The sound-symbol correspondences in English and Spanish are rather different from those

in Portuguese, for the two orthographic symbols considered in the present study. It is important

to understand these differences, in order to be able to discuss later the effect of English and

Spanish on the acquisition of the Portuguese correspondences. Where Portuguese has [z] for

and intervocalic , Spanish has a different sound from Portuguese, but the same sound for both

graphemes: [s]21 (Table 1-3). The English sound-symbol correspondences for the graphemes


two correspondences. While the chapter on methodology will make mention of the other four correspondences, the
results and discussion chapters will consider only the -[z] and -[z] correspondences.
20 It is worth noting that the phoneme /z/ is represented orthographically not only by and intervocalic , but
also by before voiced consonants (as mentioned previously), and by , in words such as exemplo ("example")
and exame ("exam"). However, is not considered in the present study due to its irregularity of pronunciation,
since it can also be pronounced as [s], as in excelente ("excellent") and excess ("excess"), and as the voiceless
palatal fricative D] in words such as caixa ("box') and xicara ("cup").
21 While this may not hold true for Castillian Spanish, where is pronounced as [0], it is true for most varieties of
Spanish. Additionally, it is noteworthy that Spanish has no voiced alveolar fricative phoneme /z/, although there is
the possibility that speakers may realize orthographic as [z] when it occurs before certain voiced consonants, as
in mismo ("same"), asno ("donkey"), and desde ("since"), although speakers may not be aware of the sounds they
produce (Schwegler & Kempff, 2007).









examined in the current study differ slightly from those in both Spanish and Portuguese. In

English, is pronounced as [z], as it is in Portuguese. The intervocalic grapheme ,

however, has three possible pronunciations in English: [z], [s], and the voiced palatal fricative [3]

(Table 1-4). In the current study, the effect of these different orthographic-phonological

correspondence rules is considered as learners produce different types of Portuguese words,

discussed in the following section.

1.5 Different Word Types

The different word types which are of interest in the present study are cognates (true or

false), non-cognates and nonce words. Typically, cognates and non-cognates have been

discussed in the literature, with respect to L2 and L3 acquisition, in one of two ways: 1) as

lexical items to be acquired by learners (e.g., Singleton, 1999; Gass & Selinker, 2001); or 2) as

items in speech production studies aimed at gaining a better understanding of how the bilingual

or multilingual brain handles lexical storage of and access to its different languages (e.g., Costa,

Santesteban and Cafio, 2005). Nonce words, or pseudo words, have typically been used in studies

examining reading skills (e.g., Defior, Martos & Cary, 2002). The present study considers these

word types from a rather different perspective, as it examines the effect of these word types on

the acquisition of sound-symbol correspondence rules in L2 or L3. Nevertheless, a brief

consideration of the literature with respect to these word types may be helpful in forming

predictions about their interaction with the sound-symbol correspondence rules examined here.

Generally, vocabulary items which are cognates in different languages are seen to be

relatively easy to acquire (e.g., with respect to Spanish and Portuguese, Chandler, 1958; Resnick,

1945; Garrison, 1979), as their orthographic (and often phonological) relatedness in the two

languages is high. There are, of course, pitfalls with cognates, if they differ slightly in spelling

and/or pronunciation in the two languages, as learners must first be aware of these differences

48









and, second, commit these differences to memory if the vocabulary items are to be produced

correctly in speech and in writing (Resnick, 1945). Another danger for learners is when cognates

turn out to be false cognates in different languages, where the orthography and/or phonology

may be similar but the meaning differs (e.g., brincar in Spanish means "to jump" while in

Portuguese it means "to play, to joke"). Again, these are words which learners must make an

effort to remember in order for them to be used correctly. Nevertheless, the high degree of

relatedness of cognates makes them easily recognizable to new learners, and therefore generally

easier to remember and use than other new vocabulary items which bear little or no orthographic

or phonological resemblance to the L1 (non-cognates).

The effect of word differences has also been considered in cognitive studies, with the

objective of understanding better how bi- and multilingual learners' organize their mental

lexicon(s). An exhaustive discussion of brain organization is beyond the scope of this

dissertation, but a brief mention of the effects of word relatedness in production is warranted

here. For more on bilingual mental organization, refer to Singleton (1999), for example.

In speech production studies using related words in picture-naming studies, response times

tend to suggest that relatedness may or may not be advantageous. In a picture-naming study with

Dutch-English bilinguals, where Hermans, Bongaerts and Schreuder (1998) used phonologically

related and unrelated words as distractors, response times were longer when the distractors were

phonologically related to the L1 translation of the L2 target word. For instance, learners shown a

picture of a mountain (berg in Dutch) were slower with the phonologically related distractor

word berm ("verge") than with the phonologically unrelated word kaars ("candle"). Although

the related (distractor) words which Hermans et al. used were not cognates, and while they were

not interested in orthography in their study, nevertheless there are aspects of their methodology









which relate to the present study. To a large extent, the related words which Hermans et al.

presented were orthographically similar to the words targeted for production, and the current

study presents words which are orthographically similar (cognates) and dissimilar (non-cognates

and nonce words) in learners' previous languages and the target language, in order to determine

whether there exists an element of "distraction" in related words which is not to be found in

unrelated words. Given the results of the Hermans et al. study, where related words had a

negative effect on the target words, it is possible to hypothesize that word type will have an

effect on production in the present study, with related words (cognates) being produced less

accurately than unrelated words (non-cognates and nonce words).

On the other hand, facilitatory effects have been found for cognates, which are related

words of a different kind from those used by Hermans et al. While their words were

phonologically and orthographically related, they were not also semantically related, as cognates

are (except in the case of false cognates). In a review of the literature, Costa, Sanstesteban and

Cafio (2005) found several production studies which showed a positive effect, insofar as speed

and accuracy were concerned, with naming cognates but not non-cognates (e.g., Costa,

Caramazza & Sebastian-Galles, 2000; Kroll, Dijkstra, Janssen & Deslauriers, 2000), even with

aphasic bilingual speakers (e.g., Roberts & Deslauriers, 1999; Kohnert, 2004). Costa et al. (2005)

likened the cognate effect to neighborhood effects, where neighborhood density plays an

important role in production. (Neighborhoods are made up of similar-sounding words, which

share an onset, such as cat, cap, and cash, or which differ in only one phoneme, like cat, scat,

and at, and dense neighborhoods are those which have many words in them. For more on

neighborhoods and their effects, see Altmann, 1997.) Costa et al. (2005) stated that picture-

naming studies (e.g., as Vitevitch 2002, 2003) showed that words with many neighbors were









named faster than words with few neighbors. Similarly, they claimed, the processing of cognates

is facilitated, where the words have phonological overlap, as do neighbors, but across languages.

Another possibility to consider is that seen words (cognates and non-cognates in the

present study) will be produced more accurately than unseen words (here, nonce words), given

learner familiarity with the former. In a study looking at the acquisition of stress patterns in L2

Spanish, Lord (2007) found that L1 English speakers, with varying proficiency levels in L2

Spanish, and L1 Spanish speakers had different accuracy rates and production times when

reading sentences containing real Spanish words versus sentences including synthetic (created)

Spanish-like words. The results showed evidence that a learner's lexicon plays an important part

in stress assignment, with known words being produced more accurately and more quickly than

unknown words.

Defior et al. (2002) claimed that "it is widely accepted that the reading of [nonce words] is

a good indicator of knowledge of the alphabetic code" (p. 146), thus making them ideal items to

include in order to demonstrate participants' ability to generalize GPCRs from seen words to

unseen words. Considering nonce words, recall that Defior et al.'s (2002) study (previously

described in detail) found that children were generally slower and less accurate when reading

them (nonce words) than when reading real words (numbers).

Extrapolating these observations and results to the current study is somewhat challenging

because this study does not look at the acquisition of the words as vocabulary items per se, nor

are response times recorded in the present study (since accuracy in production is the primary

concern here). If the target sound-symbol correspondences in cognates are produced more

accurately than those in other words (non-cognates and nonce words), that would suggest a

degree of facilitation for cognates. On the other hand, if the correspondences in cognates are









produced less accurately than those in other words, that would be indicative of some form of

"distraction" with cognates. Alternatively, if the target correspondences in seen words (cognates

and non-cognates) are produced more accurately than those in unseen words (nonce words), that

would imply that learners are relying more on direct access to the lexicon to retrieve

pronunciation, rather than on the grapheme-phoneme correspondence rules to assemble the

pronunciation of the items.

Given these different possibilities, it is difficult to predict the effect of the different word

types on the acquisition and production of the orthographic-phonological correspondences

considered in the study at hand. Nevertheless, due to the fact that some of the literature on L2

and L3 acquisition has discussed the role of cognates, non-cognates and nonce words, and given

the use of different word types in reading studies which considered orthographic-phonological

correspondences (e.g., Defior et al., 2002), it seems pertinent to consider these word types in this

study.

1.6 The Present Study

The literature reviewed in this chapter sets the stage for the current study. Recalling its

three main areas of focus error resolution, interference and generalizability this study looks

first to address the question of how well the sound-symbol correspondences in L3 Portuguese

(with L1 English and L2 Spanish, or vice-versa) are acquired. What evidence does the present

study find to support Major's (2001) OPM, which predicts an increase in L3 in a learner's IL?

In terms of the effect of learners' previous languages on their acquisition of an L3, the

literature showed conflicting results in several areas: 1) whether or not there is an advantage for

bilinguals over monolinguals in the acquisition of a new language (Cenoz, 2003); 2) whether

proficiency in the L2 has to be low or high in order for it to be the source of transfer for the L3

(Odlin, 1989; Poulisse & Bongaerts, 1994; Shanon, 1991; vs. Hammarberg, 2001); and 3)

52









whether or not facilitation occurs between typologically similar languages (e.g., Ecke and Hall,

2000; vs. Gibson, Hufeisen and Libben, 2001). Given these mixed observations, the current

study questions what the nature of interference is with respect to the sound-symbol

correspondences in question. To what extent are the GPCRs in the L3 "colored" (to extrapolate

fromFlege, 1987, and Baker & Trofimovich, 2005) by the GPCRs in the L1 and/or the L2? What

evidence is there for the effect of language status (L1/L2) versus language typology

(similar/dissimilar) in the acquisition of the L3 correspondence rules? It was also observed in the

literature reviewed that Portuguese instructors might be aware of interference from Spanish but

that this has been based more on observation than empirical evidence (Azevedo, 1978), and that

further research on the acquisition of Portuguese by Spanish speakers is necessary (Carvalho,

2002). Thus, the current study also asks what the relationship is between proficiency in Spanish

and acquisition of Portuguese.

As far as generalizability is concerned, the main question is to what extent learners are able

to generalize the GPCRs learned, from seen words (cognates and non-cognates) to unseen words

(nonce words). Thus, it is of interest to know what differences arise in production due to

different word types. What support is there for Defior et al.'s (2002) and Lord's (2007) findings,

where known words (here, cognates and non-cognates) were read faster and more accurately than

unknown words (nonce words)? Or is there evidence to suggest that there is a difference between

production of related words cognates and unrelated words non-cognates and nonce words

(to extrapolate from Hermans, Bongaerts & Schreuder, 1998, and Costa, Sanstesteban and Canfo,

2005)?

The next chapter describes the design of this study, as well as the methods used in data

collection, transcription and analysis. Chapters three, four and five present and discuss the









findings related to each of the three research questions (error resolution, interference and

generalizability, respectively). The final chapter brings the dissertation to a close, discussing

implications from the study, as well as limitations and future directions.









Table 1-1. Symbols examined, Portuguese sounds, and examples
Symbol Portuguese Sound Example
[z] zebra [zebra] ("zebra")
Intervocalic [z] casa [kaza] ("house")

Table 1-2. Symbols examined, Portuguese contexts and sounds, and examples
Symbol Portuguese Context Portuguese Sound Example
All contexts [z] zebra [zebra] ("zebra")
Intervocalic [z] casa [kaza] ("house")
Before [z] desde [dezd3i] ("since")
Before [z] vesgo [vezgu] ("cross-eyed")
Before [z] asno [aznu] ("donkey")
Before [z] mesmo [mezmu] ("same")
Word-initial [s] sapato [sapatu] ("shoe")
After <1> [s] balsa [bawsa] ("ferry")
After [s] ganso [gisu] ("goose")
After [s] urso [tusu] ("bear")

Table 1-3. Symbols examined, Spanish sounds, and examples
Symbol Spanish Sound Example
[s] caza [kasa] ("hunt")
Intervocalic [s] casa [kasa] ("house")

Table 1-4. Symbols examined, English sounds, and examples
Symbol English Sound Example
[z] zebra
Intervocalic [z] raisin
Intervocalic [s] basin
Intervocalic [3] Asian









CHAPTER 2
METHODOLOGY

2.1 Introduction

In the present study, participants enrolled in introductory Portuguese classes at the

university level were recorded at three intervals during their semester of instruction (at the

beginning, in the middle, and at the end of the semester), as they read a series of Portuguese or

Portuguese-like words including English-Spanish-Portuguese orthographic cognates, non-

cognates and nonce words containing certain grapheme-phoneme correspondences. To follow is

a presentation of the research questions that motivated the study, as well as initial hypotheses

concerning these questions. Thereafter is a detailed description of the methodology used in the

study to test these hypotheses, including participants and materials, as well as the procedures of

data collection, transcription and analysis. Concluding this chapter is an overview of subsequent

chapters.

2.2 Research Questions and Hypotheses

In the literature reviewed in the previous chapter, it was claimed that the part played by the

L3 in a learner's IL would increase over time, while the role of the L1 and L2 would decrease

(Major 2001). In terms of the nature of interference, on the one hand, there was the claim that the

lower the learner's proficiency in the L2, the greater the likelihood of language transfer from the

L2 to the L3 (Odlin, 1989; Poulisse & Bongaerts, 1994); on the other hand, there was the belief

that there must be a certain degree of proficiency in the L2 in order for it to be the basis of

transfer to the L3 (Hammarberg 2001). Language typology was suggested to be a significant

factor in transfer (Cenoz, 2001), with the typologically similar language being the source of

transfer more often than the typologically dissimilar language. As far as word effects, priming

studies showed delayed response times when phonologically related distractor words were used









(Hermans, Bongaerts & Schreuder, 1998), but also positive effects for cognates (Costa,

Sanstesteban and Cafio, 2005), and nonce words were seen to be read with less speed and

accuracy than real words (Defior et al., 2002; Lord, 2007). The current study addresses these

areas of error resolution, interference and generalizability (word type) by examining the

acquisition of L2 and L3 orthographic-phonological correspondences. Although these areas

overlap to some extent, and certain questions apply in all three cases, they are presented and

discussed separately hereafter, for the sake of clarity.

2.2.1 Error Resolution

Three basic research questions (R. Q.) arise with regard to error resolution, listed here with

hypotheses (H.) concerning participant production.

R. Q. 1. To what extent do learners acquire the Portuguese grapheme-phoneme

correspondence rules (GPCRs) under consideration, throughout the course of the semester?22

H. 1. Significant increases in correct production are expected to occur throughout the

semester of instruction, at least when all production data is considered together (both graphemes,

all word types, all participant groups), in accordance with Major's (2001) OPM, which predicts

that a learner's IL will move towards an end state of L3.

R. Q. 2. What differences in acquisition exist between participant groups23?

H. 2. It is hypothesized that there will be differences in correct production between the

language groups, with lower accuracy being found in the production of the more proficient

Spanish speakers, given the persistent problems discussed in the literature with regard to Spanish

speakers learning Portuguese pronunciation (Azevedo, 1978; Jordan, 1991; Tarquinio, 1977).

22 For the purpose of this dissertation, acquisition is measured by the correct pronunciation of the target Portuguese
GPCRs, although in some cases correct production may actually result frompositive transfer from English, as will
be discussed further below.
23 Participants and their grouping will be described further in section 2.3.1.

57









R. Q. 3. What differences in acquisition are there between the two graphemes and

intervocalic corresponding to the phoneme /z/?

H. 3. It is anticipated that there will be significant differences between tests for both

graphemes, and significant differences between the two graphemes, based on the fact that one of

them has a one-to-one grapheme-phoneme correspondence in Portuguese (-/z/), while the

other does not (-/z/ in intervocalic position and before voiced consonants, -/s/ elsewhere),

and in light of the much documented problem ofdevoicing of intervocalic (Azevedo, 1978;

Jordan, 1991; Tarquinio, 1977).

2.2.2 Interference

There are several principal questions concerning interference, presented here with

hypotheses regarding participant production.

R. Q. 1. What evidence is there of interference from the L1 and/or the L2?

H. 1. It is anticipated that there will be evidence of interference from both the L1 and the

L2, with these reducing over time, given the predictions made by Major's (2001) OPM that the

learner's initial IL consists in part of the L1 and in part of the L2, with these decreasing over

time. To extrapolate somewhat from the OPM's prediction that the universals (U) component

will increase and then decrease throughout the course of the IL, it is hypothesized that there will

be evidence in the data that participants are producing sounds which are neither consistent with

the correspondence systems of the L1/L2 nor with that of the L3. Such production is expected to

reflect universal tendencies toward unmarked features, such as voicelessness.

R. Q. 2. What differences in interference are there between participant groups?

H. 2. Given the conflicting claims in the literature with regard to the relationship between

proficiency in the L2 and the level of transfer from the L2 to the L3 (e.g., Odlin, 1989; Poulisse

& Bongaerts, 1994; vs. Hammarberg, 2001), it is difficult to predict what differences might exist

58









between participant groups in the present study. Certainly it would seem that there should be

differences between participant groups of different language backgrounds and L2 proficiency

levels. If the "weaker" L2 is the greater source of interference, then native English speakers

should show greater interference from Spanish, while native Spanish speakers should show

greater transfer from English. On the other hand, if there must b e a certain level of proficiency in

the L2 in order for it to be the basis of transfer to the L3, then L1 English speakers who are more

proficient in Spanish should show greater transfer from L2 Spanish than L1 English speakers

who are less proficient in Spanish. At the same time, proficient L2 English speakers would be

expected to show greater transfer from English than from Spanish. Alternatively, it may be that

all participants will transfer to a greater extent from the more similar language (Spanish) than

from the less similar language (English). Given the researcher's own observations in the

classroom, the hypothesis selected here is that the higher the proficiency level in the closely

related language, Spanish, the more evidence of transfer there will be from that language in the

target language production.

R. Q. 3. What differences in interference exist between the two graphemes and ?

H. 3. It is expected that there will be differences between the two graphemes, in light of the

GPCRs in English, Portuguese and Spanish, and in view Muller and Muller's (1968) prediction

about the most difficult GPCR to acquire being of the AX AY type (where language 1 has sound

X for a particular grapheme A and language 2 has sound Y for that same grapheme A). Since

grapheme always has the same correspondence rule in English and Portuguese (although not

in Spanish), but grapheme only sometimes has the same correspondence rule in English and

Portuguese (and never in Spanish), it is anticipated that there will be greater accuracy in

production of due to the possibility of positive transfer from English, and less accuracy in









production of due to the possibility of only negative transfer from Spanish and generally

negative transfer from English.

R. Q. 4. To what extent is production in Spanish related to acquisition (correct production)

ofPortuguese?

H. 4. It is anticipated, based on the researcher's own classroom observations and

impressions, as well as observations in the literature about the difficulty encountered by Spanish

speakers with Portuguese pronunciation, in particular with intervocalic (e.g., Azevedo, 1978;

Tarquinio, 1977; Jordan, 1991), that production in Spanish will be inversely related to

acquisition of the Portuguese GPCRs in question. That is, it is expected that the greater the

accuracy in Spanish production, the lower the accuracy in Portuguese production will be.

2.2.3 Generalizability

The term generalizability is used here to refer to participants' ability to apply the GPCRs

learned for seen words (cognates and non-cognates) to unseen words (nonce words). There is one

principal question to be answered here, along with a hypothesis regarding participant production.

R. Q. 1. What differences in production arise due to word type (cognates, non-cognates and

nonce-words)?

H. 1. On the one hand, given findings that real words were read faster and more accurately

than nonce words (Defior et al., 2002; Lord, 2007), it would be expected that the L2 and L3

participants in the current study would perform better on real words than on nonce words. On the

other hand, extrapolating Hermans et al.'s (1998) and Costa et al.'s (2005) findings that related

words (cognates) affected production to a greater extent than unrelated words (non-cognates and

nonce words), where the former found related words to have a negative effect, and the latter

found them to have a positive effect, it might be expected in the present study that cognates

would affect production to a greater or lesser extent than non-cognates and nonce words. Given

60









the literature, it is difficult to predict what the current study will show with regard to the

relationship between word type and the acquisition of the GPCRs in question, so once again, the

researcher's classroom observations are instrumental in hypothesis selection here. It is predicted

that the orthographic-phonological correspondences in seen words (cognates and non-cognates)

will be produced more accurately than those in unseen words (nonce words), due to learners'

greater familiarity with and exposure to the former. That is, production will be lexically driven

rather than rule driven.

In answering these research questions, a better understanding of L3 production will be

gained. A contribution to the general body of research will be made by finding evidence to

support or not support the theoretical framework and previous studies cited, with insight

added regarding the role of an L3 learner's previous languages in the acquisition process.

Pinpointing the source of interference in an L3 learner's IL is not straightforward (Dewaele,

1998), and by considering the acquisition of specific orthographic-phonological

correspondences, stronger conclusions may be drawn than have been possible thus far with

regard to the sources) of interference. With these thoughts in mind, the next section presents the

methodology used in the current study, in order to test the hypotheses and answer the research

questions.

2.3 Methodology

Consideration will be given first to the participants involved in the present study, then to

the materials and tasks developed, and finally to the methods of collecting, transcribing and

analyzing the data.

2.3.1 Participants

The participants in the current study were students in three introductory classes of

Portuguese: two Beginner 1 classes (the first semester of a two-semester course) and one

61









Introduction to Portuguese and Brazil class for fluent (native or advanced) speakers of other

Romance languages (an accelerated one-semester course). All students of the three classes were

asked to participate in the study just as they would participate in any class activity, since the

tasks were administered during normal class time, and there was no penalty for students who

failed to participate in one or more of the tasks due to absence from class on that day. Thus,

participation was considered voluntary, since no additional time commitment outside of class

was required and there was no penalty for non-participation. Before participating, the students

indicated that they had been informed about the study and its requirements by completing a

consent form (Appendix A).

Next, participants completed a questionnaire (Appendix B), giving details about the

languages they spoke, whether they were native speakers of them, how and when they had

learned these languages, and with whom and where they used them. They were also given

descriptions of four proficiency levels (1 being the lowest and 4 being the highest) for the four

skills of listening speaking reading and writing, and were asked to rate themselves for each

language by choosing from the descriptions the proficiency level that they felt most closely

reflected their own. (Participants were not asked to rate themselves for English, since it was

presumed that all were proficient, if not native, English speakers, due to their enrollment in an

English-speaking institution of higher education.)

Of the 66 students who participated in the study, data from 52 were chosen for inclusion in

the analysis. This selection was based upon two criteria: 1) attendance all three times the task

was administered; and 2) audible (and therefore usable) recordings. No participants were

eliminated based on questionnaire answers. (The questionnaire asked participants if they had

ever been diagnosed with a reading disorder, such as dyslexia. Had anyone responded









affirmatively to this question, which was not the case, their data would most likely have been

excluded.)

For the purpose of data analysis, participants were divided into groups, based on the

information they supplied in the questionnaire (summarized in Appendix C), and on the course

(beginner or accelerated) in which they were enrolled. It bears mentioning here that, at the

institution where the data were collected, enrollment in Portuguese is relatively low (at least

compared with Spanish and French, for instance), and there is no standardized placement test for

students. Those who enroll in the accelerated introductory class, designed for speakers of other

Romance languages, have generally taken classes or a placement test in the other language, are

native speakers of that language, or occasionally, are heritage speakers ofPortuguese.

Sometimes, informal interviews with the instructor or coordinator are conducted for the purpose

of assessing the student's level of proficiency in their other Romance languages) in order to

ensure that they enroll in the appropriate Portuguese class. Those who do not meet the criteria for

the accelerated class enroll in the beginner class. It is worth mentioning that, occasionally, when

scheduling conflicts occur, these enrollment criteria are not strictly adhered to, due to the limited

number of classes offered each semester and the desire to have enrollment be as high as possible.

When this happens, students who would otherwise be in the accelerated class are sometimes

permitted to enroll in the beginner class, but never vice-versa, due to the difficulty that beginner

students would encounter in the accelerated class.

The 52 participants in the present study were initially divided into five groups (Table 2-1).

After initial data analysis, where no significant differences were found between group O and any

of the other groups apart from S, the five groups were collapsed into four (Table 2-2). The results

which led to this finalized grouping are discussed in Chapter 3.









2.3.2 Materials

In order to examine participants' production at different points during the semester of

instruction, with respect to the orthographic-phonological correspondences considered in the

study24, a reading task was developed, which required participants to read 130 words in

Portuguese, 36 words in English and 20 words in Spanish (Appendix D). The English and

Spanish words were given only the first time the task was administered, to ascertain participants'

knowledge of the GPCRs of those languages for the graphemes considered in the current study.

The task was created specifically to require reading of the graphemes in all three languages and

to elicit the production of target sounds in all three languages, as much as possible25

Spontaneous oral data would not have been appropriate for this study because participants would

not have been able to produce anything at the beginning of the semester, given that most of them

had no previous exposure to or instruction in Portuguese. Even if this were not the case, it would

be difficult to ensure production of sufficient instances of the target grapheme-phoneme

correspondences by all participants in natural speech. Additionally, early learners such as these

might feel more anxious about being recorded while producing spontaneous oral data than during

a reading task such as the one developed.

The English and Spanish words were all real words, all of which contained either target

graphemes or graphemes which would elicit the target sounds examined in the study, but

whether or not they were cognates or non-cognates with each other or with Portuguese was not

taken into account, since their purpose was simply to demonstrate participants' knowledge of the

24 The study was designed to consider six sound-symbol correspondences: -[z], intervocalic -[z], -[3],
-[3], initial -[h], intervocalic -[h]. However, due to the complexity of the project, only data for the
first two of these correspondences are analyzed and discussed in this dissertation. Nevertheless, this chapter makes
mention of the sixcorrespondences since they formed part of the materials developed and used.

25 Recall that the phoneme /z/ does not occur in Spanish, although it is possible for to be pronounced as [z]
before the voiced consonants , , and , as in mismo ("same"), asno ("donkey") and desde ("since").

64









GPCRs in English and Spanish, and their physical ability to produce the sounds. The Portuguese

words consisted of:

* 48 English-Spanish-Portuguese cognates (C), such as bdsico ("basic") and horizonte
("horizon")

* 35 non-cognates (NC), such as casaco ("coat") and zangado ("angry")

* 37 nonce words (N), such asfeserel andpaimozes

* 10 filler words (F) which did not contain any of the target graphemes, such as chato
("boring") and crianqa ("child")

As far as possible, the cognates and non-cognates were taken from the textbook, Ponto de

Encontro (Klobucka, A., Jouet-Pastre, C. M. C., Moreira, M. L. de B., Sobral, P. I., &

Hutchinson, A. P., 2007) that participants used for class. While classroom materials and

instruction were not factors examined in this study, a brief note on them is warranted here.

Although the textbook itself makes no reference to pronunciation, pronunciation guides and

exercises appear throughout ancillary reference materials such as the accompanying Student

Activities Manual and online audio exercises. Students are expected to read these notes regarding

pronunciation on their own, outside of the classroom, but pronunciation is also taught and

reinforced in the classroom. At the beginning of the semester, one or two class periods in both

class levels are dedicated to explicit instruction regarding orthographic-phonological

correspondence rules, and correction is offered to a certain extent (not measured) throughout the

semester of instruction.

Because participants were enrolled in different courses that progressed through the

textbook at different rates, only words from the first three chapters were used (i.e., the chapters

covered by all classes between the first and second administrations of the task), in an effort to

ensure that all participants would have been exposed to the cognate and non-cognate words by

the time the task was administered for the second time. In some grapheme-phoneme-word type

65









categories, it was not possible to find sufficient tokens from the vocabulary found in the

textbook, so to make the categories as balanced as possible, words were added, such as manhoso

("smart," "whiny") which was added to the -[z]-NC category, and azedo ("sour") which was

added to the -[z]-NC category. Nonce words were created by listing all of the syllables found

in the cognate and non-cognate words, then using the random function in Excel to rearrange

themto yield words with two, three or four syllables. Any syllables or words which were

phonotactically impossible or unlikely to occur in Portuguese were excluded. For example,

words ending in are very rare in Portuguese so any nonce word created with a final was

excluded. The pronunciation of the target graphemes in the nonce words was agreed on by the

three raters in the study (Section 2.3.4 provides more information about the raters). Of the 120

non-filler words, 13 contained two target correspondences, such as realizado ("realized") which

contained the initial -[h] and -[z] correspondences, yielding a total of 133 target

correspondences for consideration. Of these, 53 were found in cognates, 40 in non-cognates and

40 in nonce words. Of the 133 tokens, 42 contained the target phoneme /h/ (24 and 18 ),

49 contained the target phoneme /z/ (30 and 19 ), and 42 contained the target phoneme

/3/ (22 and 20 ). 26

Considering only the and words which are analyzed and discussed in this

dissertation, the 49 words were later reduced to 48, as it was noticed that the in one of the

words in the -[z]-C category, transiqdo ("transition"), was not intervocalic. The words

are shown broken down into cognates, non-cognates and nonce words (Table 2-3), as are the

words (Table 2-4), with asterisks marking those words which were not found among the

vocabulary items listed in the textbook. It is worth mentioning here that three sets of cognate

26 Recall that only the /z/ items are analyzed here; the analysis of the other sounds is beyond the scope of this
dissertation.









words were included, due to the three possible pronunciations of in English: [z] as in raisin;

[s] as in basin, and [3] as in Asian. The Cl cognates are those where English would have [z], the

C2 cognates are those where English would have [s], and the C3 cognates are those where

English would have30.

A PowerPoint (2003) presentation was created with instructions for the participants,

followed by the words to be read in each language, with each word appearing on a separate slide.

After initial instructions, all the Portuguese words were shown, followed by a slide indicating the

end of the Portuguese words and the beginning of the English words. After the English words

came another slide indicating the end of the English words and the beginning of the Spanish

words. After the Spanish words, there was a slide which indicated the conclusion of the task

(Appendix E shows the instruction slides and examples of the words that participants saw for

each language). For all three languages, the words were randomly ordered, using the random

function in Excel, then two versions of the PowerPoint presentation were created, "Odd" and

"Even," since participants were seated in odd and even numbered recording booths in the

Language Learning Center. "Odd" had Portuguese words 1-130, English words 1-36 and Spanish

words 1-20; "Even" had Portuguese words 66-130 followed by 1-65, English words 19-36

followed by 1-18, and Spanish words 11-20 followed by 1-10. The purpose of the two versions

was to roughly control for fatigue and to minimize distractions from neighboring participants,

since all participants were reading and being recorded at the same time.

2.3.3 Data Collection

Data were collected at the university's Language Learning Center, where participants sat at

booths equipped with a computer, monitor and headphones with attached microphone. Seated at

booths numbered with odd and even numbers, participants were instructed to open the









PowerPoint version ("Odd" or "Even") respective to their booth number27. After reading the

instructions in the PowerPoint file, participants were instructed to begin speaking, at which point

recording was started by the researcher, from a central computer, using a Sanako Lab 300

system. Participants were able to go at their own pace, clicking the mouse or pressing a key on

the keyboard to move from one slide to the next. It was felt that, by allowing participants to

proceed when ready, they would not become bored by having to wait for the slides to advance

automatically, nor would they be pressured if the slides were moving too quickly for them. The

main disadvantage of having participants advance through the slides on their own was that they

could skip slides, inadvertently or on purpose. Still, it was felt that this disadvantage was

outweighed by the advantages, so it was deemed better to have participants progress at their own

rate. Once all participants had finished reading, the recording was stopped, again by the

researcher, and the recorded data was captured into individual files for each participant and

saved on an external drive.

The reading task was administered three times throughout the semester: 1) on the second

day of class, before any formal instruction on pronunciation had been given; 2) during Week 8 of

a 16-week semester; and 3) during Week 15 of the 16-week semester. Participants read the 130

Portuguese words each time the task was administered, while the English and Spanish words

were read only on the first occasion. Whether or not participants read the same version ("Odd" or

"Even") on each occasion was not deemed important since participants themselves would

likely not remember the order in which the words had been presented in previous tests and thus

was not verified.




27 Whether or not participants opened and used the correct version was not verified precisely. Observing that
alternating participants seemed to be reading different versions was deemed sufficient verification.

68









Once all the data had been collected, the files were examined for acoustic problems during

the recording process, as well as any gaps in individual participation (that is, where a participant

failed to attend one or more test dates). Then, using Adobe Audition, each file was truncated at

the beginning to remove the participant's name and at the end to remove the empty portion after

the participant had finished the reading task, in preparation for raters to listen to the recordings.

Once the recordings had been prepared for the raters, the files were identified by random

numbers to ensure anonymity.

2.3.4 Transcription

Three raters listened to all of the data. Rater 1 is a native speaker of Brazilian Portuguese

(from Rio de Janeiro) but not a linguist. Rater 2 is a highly fluent speaker of Brazilian

Portuguese and a linguist. Rater 3 (the researcher) is a native speaker of Brazilian Portuguese

(from Sao Paulo) and a linguist. All three raters were instructors of Portuguese at the university

at the time of data collection: rater 1 was the instructor of the two beginner classes; rater 2 was

the instructor of the accelerated class for speakers of other Romance languages; and rater 3 was

the instructor of other classes which did not participate in the study. While perhaps somewhat

undesirable, the use of the instructors and the researcher as the raters in the study was necessary

due to a lack of other fluent Portuguese speakers available.

Before carrying out the ratings, the researcher met with the other two raters for training on

how to transcribe and assess the recorded data. Recordings from a pilot study were used in the

training, so that the raters would be familiar with the words and target sounds to be transcribed

without actually being exposed to (and perhaps prejudiced by) any of the data from the present

study. The researcher reviewed the target Portuguese sounds for the six graphemes under

consideration in the study, as well as some of the alternate sounds which participants were

hypothesized to produce, based on their knowledge of English and Spanish pronunciation rules.

69









Raters 2 and 3, the linguists, were familiar with the IPA symbols necessary to transcribe all

possible sounds, but a guide (Appendix F) was produced especially for rater 1, the non-linguist.

The guide showed words in Portuguese, English and Spanish, with highlighted graphemes and

the symbols for their respective sounds, to facilitate transcription. Alternative sounds (due to

dialectal variation) were indicated with slash marks on the guide. Raters 1 and 2 were also

instructed to consult with the researcher raterr 3) if they were unsure how to transcribe a

particular sound28

Provided with necessary equipment, all three raters listened to all of the data on their own,

and transcribed the target sounds for each word, as had been discussed in the training session.

Including those in the filler words, the number of target graphemes transcribed per participant,

per task, was 155. To facilitate the process, transcription sheets (Appendix G) were provided for

the raters to circle the sounds heard, with four columns for each target grapheme/phoneme: the

first column contained the symbol for the sound expected for someone using English

pronunciation rules; the second column contained the symbol expected for someone using

Spanish pronunciation rules; the third column contained the symbol for the (correct) target

Portuguese sound; and the fourth column was left blank for the rater to write in the symbol for a

sound other than those in the first three columns. Raters either circled one of the symbols listed

on the sheet or wrote in a symbol if it was not listed.

All of the raters' transcriptions were entered into an Excel spreadsheet then compared in

order to calculate inter-rater agreement. In cases where one rater did not agree with the other

two, the majority rating was selected (Lord, 2007). As relates to and items only, in a

very small percentage of cases (0.52%), none of the raters agreed, so the three met to listen to

28 For example, rater 1 inquired about how to transcribe the voiced velar stop [g], which was not provided on the
transcription guide but was used by many participants for in words such asferrugem ("rust").

70









and discuss those sounds until agreement was reached, at least between two of the raters. Table

2-5 provides details on the inter-rater agreement process for and items. Overall, the

three raters agreed 85.03% of the time; the cases where only two raters agreed were fairly evenly

distributed: Raters 1 and 2 (but not rater 3) agreed 3.58% of the time; raters 1 and 3 (but not 2)

agreed in 4.49% of instances; and raters 2 and 3 (but not 1) agreed on 6.9% of the items.

2.3.5 Analysis

For the purpose of statistical analysis, the data entered into Excel were assigned different

numerical and alphabetic values for different purposes. (Appendix H shows a sample of

participant production for and tokens, as well as the numerical and alphabetic values

assigned to each sound.) In order to examine participants' accuracy with respect to Portuguese

pronunciation, numerical values were assigned: a score of 1 for a correct (target) Portuguese

sound, and 0 for anything else. Scores were then summed per participant, per grapheme, and per

word type for each test, to facilitate analysis.

Separately, the data were examined for the sounds which participants produced (such as

the projected English or Spanish sound on the rating sheet, or the target Portuguese sound). The

sounds were assigned alphabetic values as follows: 1) EN for transfer from English (based on

English pronunciation rules), when transfer from Spanish or correct pronunciation in Portuguese

would differ from English, such as the realization of the in visual ("visual") as [3]; 2) EP for

instances when English and Portuguese pronunciation rules coincided but differed from Spanish,

such as the realization of the inpresidente ("president") or the in horizonte ("horizon")

as [z]; 3) ES for cases when English and Spanish pronunciation rules coincided with each other

but differed from Portuguese, such as the realization of the in bdsico ("basic") as [s]; 4) NO

when participants produced nothing for a target phoneme (either because syllables were skipped

or transposed, or because entire words/slides were skipped); 5) OTH when a sound was produced

71









other than the sounds expected by the rules of English, Portuguese and Spanish, such as the

realization of the in visual ("visual") as [J]; 6) PO when production was correct according to

the Portuguese rule, where this differed from the English and Spanish rules, such as the

realization of the in bdsico ("basic") as [z]; and 7) SP for transfer from Spanish (based on

Spanish pronunciation rules) when the Spanish rule differed from the English and Portuguese

rules, such as the realization of the inpresidente ("president") or the in horizonte

("horizon") as [s]. Each of these alphabetic values was then summed per participant, per

grapheme and per word type for each test, to facilitate analysis. Once participant scores were

summed, they were calculated as percentages of the total number of items per category, due to

the unbalanced nature of the categories (e.g., there were 17 cognate items but only six

cognate items per test). These percentages were then imported into SAS for statistical

analyses to be carried out. The statistical tests conducted are described separately here, according

to the research question they were intended to address.

With regard to the first research question, concerning error resolution, a general linear

model (mixed effects) was used to analyze the percentages calculated from the numerical scores

assigned to each sound produced, considering as factors the 52 participants (random factor), and

as fixed factors: the five participant groups 29, three tests, two graphemes and three word types, to

yield a total of 936 observations. 30 Tukey post-hoc t-tests were conducted when necessary, to

establish where there were significant differences, and all analyses used a significance level of

a=0.0005. The output from the model and the results of the t-tests are presented and discussed in

Chapter 3.


29 These five groups were later reduced to 4, as discussed in greater detail in the next chapter.

30 Note that the same model was used to address the third research question, regarding generalizability, which is
discussed separately momentarily.









For the purpose of addressing the second research question, regarding the role of

interference in acquisition, a general linear model (mixed effects) was used to analyze the

percentages calculated from the alphabetic scores assigned to each sound produced, considering

as factors the 52 participants (in four groups), three tests, and seven production types for , to

yield a total of 1092 observations. For , the factors considered were the 52 participants (in

four groups), three tests, and four production types, to yield a total of 624 observations. Tukey

post-hoc t-tests were conducted when necessary, to establish where there were significant

differences, and all analyses used a significance level of a=0.0005. All production was

considered by group and grapheme, according to the various production types. The output from

the model and the results of the t-tests are presented and discussed in Chapter 4. In order to

address the question on the relationship between accuracy in Spanish production and accuracy in

Portuguese production, a regression model was used to analyze participants' production on the

Spanish reading task compared with average production for the three Portuguese reading tasks.

The results of the regression model are also presented and discussed in Chapter 4.

As for the third research question, regarding generalizability, the results of the general

linear model used to address the first research question were also used to address the third

question, where the results relevant to the latter question pertain to the fixed factor word type.

Tukey post-hoc t-tests were conducted when necessary, to establish where there were significant

differences, and all analyses used a significance level of a=0.0005. All production was

considered together, then production by grapheme was examined. Next, the cognate words

were divided into three sets, according to the sound in the English equivalent ([s], [T])pr [

with production of these sets being considered according to participant group. The output from

the model and the results of the t-tests are presented and discussed in Chapter 5.









In the final chapter, the results of the present study are reviewed, with respect to the

research questions it set out to answer and the hypotheses proposed, and the study's

contributions and implications are discussed, as well as its limitations and future directions.









Table 2-1. Participant groups
Group Group Description No.
El Native English speakers with no/low proficiency in Spanish (beginner class) 21
E2 Native English speakers with high proficiency in Spanish (accelerated class) 4
P Participants with previous exposure to or instruction in Portuguese 9
O Participants with exposure to or instruction in other foreign languages 9
S Native Spanish speakers 9

Table 2-2. Revised participant groups
Group Group Description No.
El Native English speakers with no/low proficiency in Spanish (beginner class) 25
E2 Native English speakers with high proficiency in Spanish (accelerated class) 5
P Participants with previous exposure to or instruction in Portuguese 9
S Native Spanish speakers 11


Table 2-3. Grapheme words
Cl C2 C3 NC N
apresentacgo basico audiovisual atencioso disudo
esquisito casos decisdo casaco feserel
museu curiosidade divis6es desenho isagio
president filosofia revisdo manhoso* lomosa
visivel generosidade televised poloneses maresa
persuasive visao preguigoso quasano
* Indicates words not in textbook.


Table 2-4. Grapheme words
C NC N
colonizacgo azedo* bazerdo
horizonte buzios paimozes
idealizaCgo dizia* pastiza
localizado gizes (pl*) prozida
organizacgo lazer trazentar
realizado rapazes zalito
zangado
* Indicates words not in textbook.









Table 2-5. Summary of inter-rater agreement for and items (number and percentage)
Test Raters 1& 2 Raters 1 & 3 Raters 2 & 3 Raters 1, 2 & 3
1 items agreed on (#) 107 126 205 2058
1 items agreed on (%) 4.29 5.05 8.21 82.45
2 items agreed on (#) 82 102 165 2147
2 items agreed on (%) 3.29 4.09 6.61 86.02
3 items agreed on (#) 79 108 147 2162
3 items agreed on (%) 3.17 4.33 5.89 86.62
All items agreed on (#) 268 336 517 6367
All items agreed on (%) 3.58 4.49 6.90 85.03









CHAPTER 3
ERROR RESOLUTION RESULTS AND DISCUSSION

3.1 Introduction

This chapter addresses the first broad research question, regarding error resolution, by

presenting and discussing the pertinent results. Before considering these, however, it is helpful to

recall the research questions asked and hypotheses stated (in section 2.2.1) concerning error

resolution. The first question asked to what extent learners acquire the Portuguese GPCRs under

consideration, throughout the course of the semester. It was hypothesized that correct production

would increase, at least when all production data were considered together (i.e., the two

graphemes, all word types, all participant groups), in accordance with Major's (2001) OPM,

which predicts that a learner's IL will move towards an end state of L3. The second question

explored the differences in acquisition which might exist between participant groups. It was

predicted that there would be differences in correct production between the language groups,

with lower accuracy being found in the production of the more proficient Spanish speakers,

given the problems discussed in the literature with regard to Spanish speakers learning

Portuguese pronunciation (Azevedo, 1978; Jordan, 1991; Tarquinio, 1977). The third question

considered the differences in acquisition between the two graphemes and intervocalic

corresponding to the phoneme /z/. It was expected that there would be significant differences

between tests for both graphemes, and significant differences between the two graphemes would

be evident, based on the one-to-one grapheme-phoneme correspondence which exists in

Portuguese for but not for (-/z/ in intervocalic position and before voiced

consonants, -/s/ elsewhere), and given the much documented problem ofdevoicing of the

intervocalic by L2/L3 learners (Azevedo, 1978; Jordan, 1991; Tarquinio, 1977). The results

will be presented and discussed as follows: 1) all production; 2) production according to









participant group; and 3) production according to grapheme.31 Variation between groups and

individuals will then be discussed, with a general summary of results to follow.

3.2 Results and Discussion

Results are presented here in terms of raw production scores, proportions of correct items

out of all possible items per category (e.g., cognate , cognate , etc.), and statistically

significant differences (output from the model), which are marked with an asterisk in the tables.

3.2.1 All Production

Considering together the production of all the participants, irrespective of their group, the

grapheme or the word type, the data showed that correct production of [z] occurred for 845 items

on Test 1, 1118 items on Test 2, and 1244 items on Test 3. Expressed as percentages of possible

correct items, these scores translate to 33.85% on Test 1, 44.79% on Test 2, and 49.84% on Test

3 (Table 3-1), shown graphically in Figure 3-1. Tukey post-hoc t-tests revealed significant

differences between Tests 1 and 2, Tests 1 and 3, and Tests 2 and 3 (p<.0001, for all three pairs

of tests). These results are encouraging: they indicate that progress in accurate sound production,

that is, acquisition, occurred throughout the semester. Next, production by participant group is

examined, which takes participants' linguistic backgrounds into consideration.

3.2.2 Production by Group

As previously discussed in section 2.3.1, participants were grouped according to their

linguistic background, based on the information that they provided on their language background

questionnaires with regard to their proficiency in English, Spanish, Portuguese and other

languages, and also based on the Portuguese course in which they were enrolled (beginner or

accelerated). With these criteria in mind, five groups emerged (Table 3-2), as was discussed


31 Production according to word type (cognate, non-cognate and nonce) will be discussed in Chapter 5, which
addresses the third broad research question, regarding generalizability.

78









previously. Due to the unbalanced number of participants in the five groups, least square means

(LS means) were used instead of arithmetic means in the statistical procedures, since LS means

are appropriate for unbalanced designs with more than one effect. (For more on LS means,

including examples, see Khuri, 2009.) The number of correct items, and their expression as

percentages of possible correct items, are shown in Table 3-3.

The ANOVA results (Table 3-4) indicated that all of the factors group, test, grapheme,

and word type were significant, as were the interactions between group and test, group and

grapheme, test and grapheme, and grapheme and word type32. Tukey post-hoc t-tests revealed

significant differences between groups El and S (p<.0001), groups E2 and S (p=0.0384), groups

O and S (p<.0001), and groups P and S (p<.0001). Because no significant differences were found

between group O and any group other than S at this point, it was decided to consolidate the

participants in group O into other groups. 33 Of the nine participants originally grouped in 0, six

were transferred to El, one to E2, and two to S, yielding four groups (Table 3-5).34

Considering these four revised groups, the number of correct items and their expression as

percentages of possible correct items were calculated (Table 3-6). Graphically, these results are

shown by group (Figure 3-2) and by test (Figure 3-3). The ANOVA results using the revised

groups (Table 3-7) indicated that all of the factors group, test, grapheme, and word type were

significant, as were the interactions between group and test, and group and grapheme.


32 The interaction between grapheme and word type will be discussed in Chapter 5, which considers the effect of
word type on acquisition.
33 It seems likely that the reason for the significant difference found between groups O and S is due to the linguistic
background of the participants in group O. Disregarding their proficiency in another language (besides English,
Spanish and Portuguese), seven of the nine participants would be classified as El or E2 participants, both of which
groups showed significant differences from group S.

34 The interaction between language group and test found no significant differences between O and groups other
than S, but there were some significant differences between other groups (El and E2, and E2 and P on Test 1),
therefore further consolidation of these groups would not have seemed a reasonable step.

79









Given that the ANOVA results indicated a significant interaction between group and test,

Tukey post-hoc t-tests were carried out to establish where significant differences existed. These

were found between tests for all groups, though not between all tests (Table 3.8). Once again, it

is encouraging to find that each group makes significant progress throughout the semester of

instruction in terms of accurate production, at least between the first and third tests.

Significant differences were also found between groups El and E2 on Test 1, between

groups El and S on all three tests, between groups E2 and P on Test 1, and between groups P and

S on all three tests (Table 3.9 tests are indicated in parentheses). These results are worth

highlighting, for they show that native English speakers with high proficiency in Spanish

behaved more similarly to native Spanish speakers than to other native English speakers (with no

or low proficiency in Spanish), at least initially. On the second and third tests, however, there

were no appreciable differences between E2 and any of the other groups. While this was not

unexpected for Test 2, it was a little unexpected for Test 3, at which point there were no

significant differences between El, E2 and P, but there were significant differences between El

and S, and between P and S. The reason for this may lie in the fact that group E2 had a low

number of participants, relative to the other groups, which led to a less conclusive statistical

result (the standard errors in the differences of LS means on Test 3 were: 5.5453 for El and S,

6.9680 forP and S, and 8.3616 for E2 and S), so that a statistical difference between E2 and S

was not observed on Test 3. Such a difference might have been found had there been a greater

number of participants in E2, a point to be remembered for future research.

3.2.3 Production by Grapheme

Now consideration will be given to the two graphemes, and . The ANOVA results

(Table 3-7 in section 3.2.2) showed a significant difference between and , but did not

show a significant interaction between test and grapheme (p=0.0547), nor was the three-way

80









interaction between test, group and grapheme found to be significant (p=0.1853), although the

interaction between group and grapheme was shown to be significant (p<.0001). The number of

correct items and their expression as percentages of possible correct items were calculated per

grapheme for each group (Table 3-10). Graphically, these results are shown per group (Figure 3-

4) and per grapheme (Figure 3-5).

Tukey post-hoc t-tests showed significant differences between and for all groups

(p<.0001). These results clearly show the major role which the grapheme plays in pronunciation,

and they would suggest that the -[z] correspondence rule in Portuguese is easier for learners

to acquire than the -[z] rule. Between groups, significant differences were found only

between groups El and S on (p<.0001), betweenP and S on (p<.0001), and betweenP

and S on (p=0.0284).

It is interesting to see that few differences emerged between groups. It was not entirely

unexpected that the group of native English speakers with low or no proficiency in Spanish

should perform significantly better on than the group of native Spanish speakers, given that

the former group has the same correspondence rule in their L1 (-[z]) as in Portuguese, while

the latter group has no phoneme /z/ in their L1. Nor was it particularly unexpected that the native

English speakers with high proficiency in Spanish should perform similarly to the native Spanish

speakers, since both groups are fluent in Spanish, a language which is so closely related to

Portuguese. It is noteworthy, though, that those participants with previous exposure to

Portuguese should achieve results which are significantly different only from the group of native

Spanish speakers. This would suggest that the previous instruction in or exposure to Portuguese

generally did not have a great impact on the learners, at least insofar as the GPCRs in question

are concerned, and at least at these early levels.









3.3 Other Observations

This section highlights observations of interest with regard to production by particular

individuals or groups. Correct production for each participant was summed for each grapheme on

each test, and calculated as percentages of possible correct items; averages were calculated for

each group (Appendix I). In this section, individual participants are referred to by group and

number, for ease of reference. For instance, participant 1, in the group of participants with

previous exposure to Portuguese, is referred to as P-1.

3.3.1 Production by Test and Group

First, it is noteworthy that none of the 52 participants produced all of the 48 tokens

correctly, on any of the three tests, indicating room for improvement for all of the participants at

the end of the semester of data collection. While this result is not unexpected for beginners of

Portuguese (enrolled in either of the two Portuguese courses), it is interesting that this is the case

even for the participants who had had exposure to or instruction in Portuguese prior to the

semester in which the data were collected.

3.3.2.1 Production by native Spanish speakers

Contrary to expectations, the participant with the highest score on Test 1, with 71%

accuracy, was a native Spanish speaker (S-39). His score went down slightly on Test 2 (63%),

and remained the same on Test 3 (63%). Similarly, the participant with the highest score by the

end of the semester was a native Spanish speaker (S-37), although this time performance

improved from 6% accuracy on Test 1 to 35% on Test 2, and 83% on Test 3. Ofthe group of

native Spanish speakers, S-19 achieved moderate accuracy on all three tests: 42%, 65% and

54%, and S-50 showed a dramatic improvement between the first two tests only: 0%, 40% and

46%. Production by these four native Spanish speakers was certainly not the norm for the group

of native Spanish speakers as a whole. As expected, most of the other participants in group S had

82









much less success in accuracy: S-2 achieved 4%, 13% and 35% on Tests 1, 2 and 3, respectively;

S-3 had 0%, 4% and 17%; S-4 achieved 8%, 10% and 6%; S-16 obtained 2%, 2% and 4%; S-43

achieved 0%, 4% and 19%; S-45 had 15%, 19% and 23%; and S-48 achieved 15%, 25% and

10%. The only native Spanish speaker to have had previous instruction in or exposure to

Portuguese (P-1) accurately produced 8%, 25% and 44% of items, at each of the three tests.

It is worth pointing out that the two participants who had the highest scores on the first and

third tests (S-39 and S-37, respectively) were the only ones who had had previous instruction in

another language other than Portuguese (French, in both cases). The exposure that these

participants had to French may have been a contributing factor to their increased success in the

tests, in comparison with the other participants in group S. It is somewhat surprising that they

performed better than the native Spanish speaking participant who had had previous instruction

in Portuguese, as it would seem more plausible to assume that previous instruction in the target

language (Portuguese) would be more advantageous for the acquisition of Portuguese GPCRs

than previous instruction in another language (French). Of course, individual factors may come

into play here, including length of exposure to and proficiency in the language in question, as

well as attention to phonological details. These are factors that will have to be left for future

studies.

It is also important to mention here that the group of native Spanish speakers varied with

respect to their exposure to and instruction in Spanish and English. Some had spent a significant

portion of their lives in Spanish-speaking countries (Colombia, Venezuela, El Salvador, Bolivia),

while others had lived largely in the U.S. These differences in upbringing and instruction, not to

mention other individual differences, no doubt also impacted their proficiency in English, with

some learning English as an L2 and others as an L1 (along with Spanish). The English reading









task, which was administered at the time of the first test, was given to assess participants'

knowledge of the English GPCRs and their ability to produce the target graphemes/phonemes.

Table 3-11 gives the results of the English task, with respect to the sound [z], for the native

Spanish speakers, along with a brief description of their background in Spanish. (Results for the

other groups, which consisted of native English speakers, are not shown or discussed, because

their production was as expected for native speakers.) There were eight English words in the task

which are pronounced with [z], containing either the grapheme (gymnasium, physics,

present, result) or the grapheme (bizarre, realization, zebra, zipper); the results in the table

show how many of these words each participant produced accurately out of the eight. Also

shown in the table is the highest level of accuracy that the participants achieved in the

Portuguese reading tasks, as mentioned previously in this section.

The purpose of considering together the highest level of accuracy in Portuguese production

and the score on the English reading task is to illustrate that accuracy in English did not

necessarily correspond to accuracy in Portuguese. Participants who managed to produce all of

the English [z] items correctly (19, 29, 45 and 48) achieved very different levels of accuracy in

Portuguese (65%, 71%, 23%, and 25%, respectively). Also, those who did not produce any of the

English [z] items accurately (2, 4, 16) achieved different levels of accuracy in Portuguese (35%,

10%, and 4%, respectively). Thus, despite obvious differences within this group, it would be

difficult to come up with clear-cut criteria for dividing the group further. Hence, the group's

production is still considered as a whole, in future discussions, but with the knowledge that its

participants are heterogeneous in many respects.

3.3.2.2 Production by native English speakers with no or low proficiency in Spanish

As for the native English speakers with no or low proficiency in Spanish (group El), a

great deal of diversity was observed in terms of accurate production. On Test 1, for instance,

84









accuracy ranged from 2% (E1-30) to 67% (E1-20). This variability may result from differences

in proficiency in Spanish35, since the former participant achieved 90.91% accuracy on the

Spanish reading task completed on the first test, while the latter participant achieved only 50%

accuracy on the Spanish reading task. Of this group of native English speakers (El), El-20

achieved the highest score on Test 1, with 67%, and also achieved the highest scores of the group

on Tests 2 and 3, with 71% and 79%, respectively. While gains for this participant may not have

been great from test to test, his scores did increase, nevertheless.

Of the other 26 participants in this group, 11 had scores that remained the same or

increased from one test to another. The other 15 participants had different combinations of

increases and decreases between tests (e.g., an increase from Test 1 to Test 2, and a decrease

from Test 2 to Test 3; a decrease from Test 1 to Test 2, followed by an increase between Tests 2

and 3; and so on.). Differences between scores for Tests 1 and 3 were both positive and negative:

El-18, for instance, increased from 19% on Test 1 to 63% on Test 3, while E1-24 decreased

from 35% on Test 1 to 21% on Test 3. These fluctuations in scores between tests are reflected in

all of the language groups. Of the 52 participants, 23 consistently improved their scores on the

three tests. The other participants showed various permutations of increases, decreases and

equivalent scores between the three tests, which is to be expected from participants from intact

classes.

3.3.2.3 Production by native English speakers with high proficiency in Spanish

The group of native English speakers with high proficiency in Spanish (E2) is interesting

to consider because of the similarities and differences within the group. Four of the five

participants in this group show similar levels of accuracy in the three tests with the level being


35 The effect of the L1 and L2 willbe discussed in detail in Chapter 4, which deals with the second research
question, regarding interference.









quite low on the first test, followed by noticeable rises on the subsequent two tests: E2-10

achieved 15%, 27% and 56%; E2-12 had 8%, 31% and 46%; E2-13 obtained 8%, 42% and 65%;

and E2-23 had 2%, 29% and 52%. The fifth participant (E2-26) shows a markedly different

pattern of approximately half of the items produced accurately on all three tests, with accuracy

levels of 52%, 54% and 52% on Tests 1, 2 and 3, respectively. An explanation for this may lie in

the level of Spanish proficiency of these five individuals. The scores of the first four participants

(E2-10, E2-12, E2-13 and E2-23) on the Spanish reading task given on Test 1 were all

considered appropriate for proficient speakers of Spanish (100%, 90.91%, 100% and 100%,

respectively), while the score for the fifth participant (E2-26) on the Spanish reading task was

considerably lower (59.09%), and not at all what would be expected from a fluent Spanish

speaker. It bears repeating here that the grouping of the native English speakers into El and E2

was based on the class in which they were enrolled beginner Portuguese and accelerated

Portuguese, respectively and not on the information they supplied on their language

background questionnaire or on their performance in the Spanish reading task on Test 1.

3.3.3 Production by Grapheme

The average accuracy on items, across all participants, for Tests 1, 2 and 3 is 60%,

73% and 80%, respectively, while the average accuracy on items is 17%, 26% and 30%.

Production of each of the graphemes will now be discussed.

3.3.3.1 Production for grapheme

It is helpful to remember that all of the participants were fluent, if not native, English

speakers, enrolled in an English speaking post-secondary institution. Also, in English is

pronounced [z], as in Portuguese, so it is understandable that participants should perform better

on items than on items (which may be realized as [z], [s] or [3] in English, and is almost

always realized as [s] in Spanish), as has already been discussed previously in this chapter.

86









However, it is still interesting to observe that not all participants achieve a high level of accuracy

on items, of which there were 19. Only six of the 52 participants achieved 100% accuracy

on these tokens on Test 1, while eight did so on Test 2, increasing to 12 participants on Test 3.

(Five participants achieved 100% accuracy on all three tests.) For the native English speakers

with no or low proficiency in Spanish (El), it might be expected that the scores would be

consistently high, since English and Portuguese have the same correspondence rule for , but

this is not necessarily the case. The average levels of accuracy for group El were 74%, 84% and

88% on Tests 1, 2 and 3, respectively. Ofthe six aforementioned participants who achieved

100% accuracy on Test 1, five belong to group El. However, there were several participants with

low accuracy on this test, including El-30 who did not produce a single token correctly. One

possible explanation for this lack of production of [z] lies in the participant's accuracy on the

Spanish reading task: 90.91%. Although the participant was enrolled in the beginner Portuguese

class, along with all of the other participants in group El, she apparently had an excellent grasp

of Spanish orthographic-phonological correspondence rules (in fact, out of all the participants in

El, she had the highest percentage of accuracy on the Spanish reading task), which seems to be

the most plausible explanation for her completely inaccurate reading of the Portuguese words on

the first test. Interestingly, the participant seems to have overcome this negative effect relatively

well by the second and third tests, when she achieves 84% and 74% accuracy, respectively.

Another anomalous participant, E1-24, had accurate production on 74% of items on Test 1,

then dropped to an alarming 21% (the second lowest level of accuracy in this group, on any of

the tests, following the aforementioned 0% on Test 1), rising to 47% in Test 3. In this instance,

however, a plausible explanation for this performance is not easily found in the participant's

linguistic background, consisting of four years of Spanish in high school (accuracy of 68.18% in









the Spanish reading task), as well as three semesters of Italian in college and time spent in Italy.

It is possible that this individual's three background languages (English, Spanish and Italian)

interacted in such a way as to cause her confusion, and in each test she may have been testing

different hypotheses with regard to the orthographic-phonological correspondence rules in

Portuguese.

The native English speaking participants with high proficiency in Spanish (group E2) had

lower averages on the first two tests than the other group of native English speakers (33% and

65% for E2, compared with 74% and 84% for El), but both groups averaged roughly the same

level of accuracy on Test 3 with 87% for E2 and 88% for El. The participants with previous

instruction in or exposure to Portuguese (group P) tended to achieve relatively high levels of

accuracy on all the tests, averaging 77%, 83% and 89% on Tests 1, 2 and 3, respectively. This

was not really unexpected, although there were two participants in this group whose performance

stood out. P-l scored low, relative to the others 21%, 37% and 68% which may be attributed

to her native Spanish status and to reportedly little instruction in Portuguese (only five months in

high school, after which the class was cancelled due to instructor illness). This participant's

production was more accurate than that of most of the participants in the native Spanish group

(S), which supports her classification in group P, although her level of accuracy in Portuguese

was quite a bit lower than that of the other participants in this group. P-42 also had relatively low

accuracy on Test 1 (37%), with close to average scores on Tests 2 and 3 (74% and 95%,

respectively). The below-average score on the first test is not easily explained, as the participant

claimed to be a native speaker of Portuguese and to speak the language with relatives and at

home with her mother. She did not claim to have previous instruction in Portuguese, however, so









it is possible that her learning of Portuguese (including pronunciation) had never been in a

formal setting where correction might be offered.36

The group of native Spanish speakers (S) had lower average accuracy than any of the other

groups for items, on all three tests (22%, 40% and 52%, on Tests 1, 2 and 3, respectively).

Two participants achieved much higher accuracy than the average on all three tests: S-19 (58%,

89% and 84%) and S-39 (95% on all three tests). There does not seemto be an immediately

apparent reason why S-19 did much better than the majority of the other participants in the

group. According to his language background questionnaire, he had lived in a Spanish-speaking

country until the age of 5 and had completed Spanish to IB level at school. It is interesting to

note that his accuracy on the English reading task on Test 1 was 100%, unlike most of his fellow

native Spanish speakers. This level of accuracy in English may have had a positive effect on his

production of items in Portuguese, especially on the first test, when data were collected

before any instruction on Portuguese pronunciation had been given in the classroom. As for

participant S-39, whose scores for items were consistent across all three tests, it is possible

that his high accuracy in English (100% on the English reading task), coupled with four years of

French in high school, positively affected his production in Portuguese.

3.3.3.2 Production for grapheme

Considering the items, the participants were generally less accurate on these than on

the items. 37 The highest accuracy for items was 55% (S-39) on Test 1, 59% on Test 2



36 After data analysis, this participant subsequently enrolled in a Portuguese class taught by the researcher, whose
opinion it is that this individual was by no means a true native speaker of the language, as claimed on her language
background questionnaire. The participant's accent, vocabulary and grammar were sorely lacking in comparison
with what would be expected of a native speaker. Perhaps a better descriptive termfor this participant would be
"heritage learner," given her familial exposure to the language and yet non-native-like production.

37 This probably has a lot to do with the fact that not only does intervocalic in Spanish have a different
phonological correspondence rule from Portuguese ([s] and [z], respectively), but in many cases in English does
as well. (Remember that, in English, intervocalic has three possible phonological correspondences: [z] as in
89









(E1-20 and P-29), and 79% on Test 3 (S-37). With this 79% accuracy rate as the highest on

items, it is clear that no participant came close to achieving full accuracy on any of the three

tests, while for items, there were 25 instances of participants achieving 100% accuracy.

There are a few interesting observations to be made regarding production on items.

Group P averaged higher accuracy on the three tests than the other three groups, and group S

tended to have very low accuracy (about a third of the scores for S were zero). S-16 had 0%

accuracy for on all three tests, S-4 achieved one correct token on the first test (3 % accuracy)

but none on the second and third tests, and S-3 had 0% accuracy on the first two tests and then

7% (2 correct items) on the third test. These low-scoring tendencies are not observed in any of

the other groups, but there are two participants whose accuracy levels of 0%, 0% and 10% (El-

27), and 0%, 3% and 17% (E1-28) are somewhat surprising. It is noteworthy that these two

participants (twin sisters) reported exposure to Spanish "all through school" and "K-IB,"

respectively, and achieved 81.82% and 86.36% accuracy on the Spanish reading task. This may

account for production which is more similar to that of the native Spanish speakers than that of

the other native English speakers (from both groups El and E2). Interestingly, these two

participants did not score particularly low on items, suggesting that their proficiency in

Spanish played a greater role with items than with items.

This closer examination of the variation in production by individuals and groups in this

section has complemented the statistical results given in the previous section. The two have

shown general progress by the participants in their production of the target Portuguese sounds,


raisin, [s] as in basin, and [3] as inAsian.) Without examining specific word types at this point, it is not possible to
identify participants' performance on the tokens found in cognate words where English would have different
phonological realizations. Hence, the present discussion will consider only overall accuracy on items, and the
more detailed discussion, taking word types into account, will follow in Chapter 5.









while also illustrating that there is still room for improvement by all participants for one or both

graphemes.

3.4 Summary

With the original research questions on error resolution in mind, the results presented in

this chapter supported Major's (2001) OPM, which predicts that the idealized learner's IL will

move towards an end state consisting solely of L3. While such an end point was not achieved by

this group of learners as a whole, given that overall accuracy at Test 3 was 50%, there was

evidence to show significant progress between tests.

In terms of differences between participant groups, the prediction that the more proficient

Spanish speakers would have greater difficulty with acquisition of the Portuguese GPCRs was

partially supported. Certainly the native Spanish speakers tended to produce the target

Portuguese sounds less accurately than the native English speakers with low proficiency in

Spanish. However, the native English speakers with high proficiency in Spanish only tended to

have less accuracy in Portuguese pronunciation initially, and by the end of the semester their

level of accuracy was not significantly different from that of any of the other groups.

With respect to the differences between the two graphemes in question, the hypothesis that

there would be significant increases between tests for both graphemes was not supported, as the

test by grapheme interaction was found not to be significant. However, the prediction that

learners would have different levels of accuracy for the two graphemes was supported, with

produced significantly more accurately than by all participant groups. Thus, not only was

evidence found to support the literature that shows the persistent problem that Spanish speakers

have in devoicing Portuguese , but evidence was found to show that English speakers also

appear to have this problem.









In the examination of individual variation, disparity in production was found within and

across all groups, as well as increases and decreases from one test to another. No participant

achieved 100% accuracy on all items for any test, although some participants did correctly

produce all items in the three tests. Items with intervocalic appeared to be more

problematic for participants, most likely due to interference from their L1 and L2, a discussion

which will be taken up in detail in the next chapter.









Table 3-1. Number and percentage of correct items for all [z], by test
Production Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 %
All [z] 845 33.85 1118 44.79 1244 49.84


Table 3-2. Participant groups
Group Group Description No.
El Native English speakers with no/low proficiency in Spanish (beginner class) 21
E2 Native English speakers with high proficiency in Spanish (accelerated class) 4
P Participants with previous exposure to or instruction in Portuguese 9
0 Participants with exposure to or instruction in other foreign languages 9
S Native Spanish speakers 9

Table 3-3. Number and percentage of correct items for all [z], by group and test
Group Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 %
El 430 42.66 518 51.39 513 50.89
E2 37 19.27 73 38.02 108 56.25
0 127 29.40 197 45.60 246 56.94
P 210 48.61 243 56.25 274 63.43
S 41 9.49 87 20.14 103 23.84


Table 3-4. ANOVA
Effect


for all production
Num D


group
test
grapheme
word type
group*test
group* grapheme
group* word type
test* grapheme
test* word type
grapheme* word type
group*test* grapheme
group*test*word type 1
test* grapheme* word type
group*test*grapheme*word type


iF F Value
4 13.71
2 93.03
1 1091.61
2 7.66
8 6.84
4 44.46
8 1.07
2 4.73
4 0.68
2 4.88
8 1.88
6 0.64
4 0.80
4 0.79


P value
*<.0001
*<.0001
*<.0001
*0.0005
*<.0001
*<.0001
0.3808
*0.0091
0.6093
*0.0078
0.0596
0.8556
0.5255
0.7477









Table 3-5. Revised participant groups
Group Group Description No.
El Native English speakers with no/low proficiency in Spanish (beginner class) 25
E2 Native English speakers with high proficiency in Spanish (accelerated class) 5
P Participants with previous exposure to or instruction in Portuguese 9
S Native Spanish speakers 11

Table 3-6. Number and percentage of correct items for all [z], by revised group and test
Group Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 %
El 516 39.81 653 50.39 667 51.47
E2 41 17.08 88 36.67 130 54.17
P 210 48.61 243 56.25 274 63.43
S 78 14.77 134 25.38 173 32.77


Table 3-7. ANOVA for all production (revised groups)
Effect Num DF F Value p-value
group 3 80.55 *<.0001
test 2 53.73 *<.0001
grapheme 1 616.02 *<.0001
word type 2 3.29 *0.0378
group*test 6 3.95 *0.0007
group*grapheme 3 33.38 *<.0001
group*word type 6 0.68 0.6622
test*grapheme 2 2.92 0.0547
test*word type 4 0.38 0.8242
grapheme*word type 2 3.01 0.0500
group*test*grapheme 6 1.47 0.1853
group*test*word type 12 0.30 0.9900
test* grapheme* word type 4 0.66 0.6171
group*test*grapheme*word type 18 0.57 0.9241

Table 3-8. Significant differences found between tests, by group
Group Tests 1 & 2 Tests 1 & 3 Tests 2 & 3
El *<.0001 *<.0001
E2 *0.0004 *<.0001 *0.0014
P *0.0003
S *0.0093 *<.0001









Table 3-9. Significant differences found between groups, by test
El E2 P S
El 0.0452 (Test 1) <.0001 (Test 1)
<.0001 (Test 2)
0.0198 (Test 3)
E2 0.0096 (Test 1)
P <.0001 (Test 1)
0.0002 (Test 2)
0.0004 (Test 3)
S

Table 3-10. Number and percentage of correct items, by grapheme and group
Group # % # %
El 1261 82 575 24
E2 176 62 83 19
P 426 83 301 38
S 237 38 148 15

Table 3-11. Production of [z] by native Spanish speakers in English reading task
No. Background in Spanish English Portuguese
1 Dominican Rep.; Miami; Span. classes in middle school and HS 7 44%
2 Colombia 0 35%
3 Up to IB in HS; relatives 6 17%
4 From Venezuela 0 10%
16 K-10; El Salvador till 15 0 4%
19 Venezuela till 5; IB Spanish 8 65%
37 K-12, Colombia 6 83%
39 4314 (advanced Spanish class at university); Mexico for 2 yrs 8 71%
43 Graduated HS in Colombia 1 19%
45 Since 5th grade 8 23%
48 Bolivia for 14 yrs 8 25%
50 Studied for 6 yrs; Colombia 2 46%











All Production


100

80

60

40

20 -

0-


Test


Figure 3-1. Percentage of correct items by test


Production per Group


100
80
60
40
20
0


54


Group El


Group E2


Group P


Group


Figure 3-2. Percentage of correct items, by test for each group


Group Production per Test


100
so -
60
40 -
20
0


1$


Figure 3-3. Percentage of correct items, by group for each test


* [z]


* Test 1
* Test 2
I Test 3


p 33

Group S


2"5 E.


33


* Group El
* Group E2
* Group P
I Group S


2 3

Test










Grapheme Differences per Group


100

80

60

40

20

0 -


U Grapheme
I Grapheme


Group


Figure 3-4. Percentage of correct items, by grapheme for each group



Group Differences per Grapheme


38








* Group El
* Group E2
" Group P
* Group S


Grapheme


Figure 3-5. Percentage of correct items, by group for each grapheme


I









CHAPTER 4
INTERFERENCE- RESULTS AND DISCUSSION

4.1 Introduction

This chapter addresses the second overarching research question, which asked about the

role of interference, or language transfer, in acquisition. Before presenting and discussing the

relevant results, it is helpful to recall the research questions asked and hypotheses stated (in

section 2.2.2) concerning interference. The first question asked what evidence there was of

interference from the L1 and/or the L2, and predicted, according to Major's (2001) OPM, that

there would be evidence of interference from both the L1 and the L2, with these reducing over

time. It was also expected, given the U component in the model, that there would be evidence of

production which was not consistent with the grammars of the L1, the L2 or the L3.

The second question considered whether there would be differences in interference

between participant groups, and it was anticipated that the different groups would show varying

degrees of transfer from English and Spanish, depending on their native language and on their

level of proficiency in Spanish. It was predicted that native English speakers with low or no

proficiency in Spanish would show greater transfer from English (at least initially), while native

English speakers with high proficiency in Spanish and native Spanish speakers would show

significant transfer from Spanish (at least initially), due to their greater experience with a

language more closely related to Portuguese than English. Thus, the prediction was that the more

typologically similar language, Spanish, would be the basis of transfer where proficiency in that

language was high, while English would be the basis of transfer where proficiency in Spanish

was lower, according to Hammerberg's (2001) belief that there must be a certain level of

proficiency in the L2 in order for it to be a basis of transfer to the L3, and according to findings

in the literature regarding the facilitation of transfer between similar languages.









The third question explored whether differences in interference existed between the two

graphemes and , and it was hypothesized that there would indeed be differences due to

the GPCRs of English, Spanish and Portuguese for the two graphemes in question. Furthermore,

it was predicted, according to Muller and Muller's (1968) observations, that the more difficult

Portuguese GPCR to acquire would be the -[z] one, since it is different from (most of) the

GPCRs in Spanish and English for intervocalic (although English does have the -[z]

correspondence rule for some lexical items), and Portuguese -[z] would be easier to acquire

since English has the same rule. Thus, evidence of positive transfer from English (or acquisition

of Portuguese) was expected for , while evidence of negative transfer from English and/or

Spanish was expected for .

The final question explored the extent to which accuracy of production in Spanish could be

said to be related to acquisition (correct production) of Portuguese. It was expected, based on the

researcher's own classroom-motivated impressions, as well as observations in the literature

regarding the difficulties encountered by Spanish speakers learning Portuguese pronunciation,

that there would be an inverse relationship between accuracy of production in Spanish and

production of the Portuguese GPCRs in question. That is, it was predicted that the greater the

proficiency in Spanish (as measured by accuracy in production in the Spanish reading task), the

lower the accuracy in production in Portuguese would be.

The results will be presented and discussed by grapheme and participant group. Following

will be an analysis of participants' production in the Spanish reading task on the first test,

compared to their production in the Portuguese reading tasks. 38 In the section with other




38 Production according to word type (cognate, non-cognate and nonce) will be discussed in Chapter 5, which
addresses the third research question, about generalizability.









observations, production of individual participants and words will be examined, and finally the

chapter's results will be summarized.

4.2 Results and Discussion

Seven production types were possible for grapheme , while only four were possible for

grapheme . Hence, the results for the two graphemes are presented and discussed separately.

4.2.1 Production for Grapheme

Results are presented in terms of raw production scores, percentages of items for each

production type (out of all items on a test), and statistically significant differences (output from

the model), which are marked with an asterisk in the tables. For , seven production types

were possible: EN, EP, ES, NO, OTH, PO and SP.

The first, coded as EN (English), represents [3], and it could occur when the English

pronunciation rule differed from the Spanish and Portuguese rules. Its occurrence was logically

limited to the six cognate words where the English equivalent would be pronounced with

[3]39, and it was taken to indicate evidence of negative transfer from English.

The second production type, coded as EP (English Portuguese), represents [z], occurring

when the English and Portuguese pronunciation rules coincided with each other but differed

from Spanish. This production type was limited to the five cognate words where the English

equivalent would be pronounced with [z]40, and it was considered to be indicative of positive

transfer from English and/or acquisition of Portuguese.

The third type of production, coded as ES (English Spanish), represents [s], occurring

when the English and Spanish pronunciation rules coincided with each other but differed from

39 Audiovisual ("audiovisual"), decisao ("decision"), divisaes ("division"), revisao ("revision"), televised
("television"), visdo ("vision").
40 Apresentagqo ("presentation"), esquisito (a false cognate meaning "odd"), museu ("museum"), president
("president"), visivel ("visible").









target Portuguese. This production type was possible for 18 words: the six cognate words

where the English equivalent would be pronounced with [s]41, the six non-cognate words

and the six nonce words42. This production type suggested negative transfer from English

and/or Spanish.

The fourth production type, coded as PO (Portuguese), represents [z], and it was possible

when the Portuguese rule differed from the English and Spanish ruless. This was possible for 24

words: the six cognate words which would be pronounced with [3] in English and with [s] in

Spanish, the six cognate words which would be pronounced with [s] in both English and

Spanish, the six non-cognate words, and the six nonce words. This production type

would indicate evidence of acquisition of Portuguese pronunciation rules, rather than

interference from English or Spanish.

The fifth production type, coded as SP (Spanish), represents [s], and it could occur when

the Spanish rule differed from the English and Portuguese ruless. This was possible for 11

words: the five cognate words which would be pronounced with [s] in Spanish but with [z]

in English and Portuguese, and the six cognate words which would be pronounced with [s]

in Spanish but with [3] in English and with [z] in Portuguese.

The final two types of production possible were when participants produced nothing for

the target grapheme (coded as NO), as a result of a misread word, a skipped syllable or a skipped

word, or when participants produced another sound (coded as O TH), which was unexpected


41 Basico ("basic"), casos ("cases"), curiosidade ("curiosity"), filosofia ("philosophy"), generosidade ("generosity"),
persuasive ("persuasive").
42 The "default" pronunciation for in English was assumed to be [s], since this is the sound that occurs in most
of the contexts where the grapheme appears, while pronunciation of as [z] or [3] in English is idiosyncratic
and must be learned. Thus, it was assumed that if participants transferred English pronunciation rules, the in the
Portuguese non-cognate and nonce words would be pronounced as [s], since these words should not have triggered
any idiosyncratic pronunciation from English as cognate words with [z] and [3] in English might have done.









given the pronunciation rules of English, Spanish and Portuguese for intervocalic . These two

production types, logically, could occur for any of the 29 items.

All the sounds produced for words were coded (according to the seven production

types), and the codes summed for each test. These sums (raw scores) were then converted to

percentages, with these adding up to 100% for each test (Table 4-1), and grouped according to

production type for graphic representation (Figure 4-1).

Numerically, slight increases were seen in correct production of as [z] (EP, PO)

throughout the semester and slight decreases in incorrect production of as [s] (ES, SP),

though there was no decrease of as [3]. Negative transfer (EN, ES and SP) accounted for

approximately 81% of the total production on the first test, 73% on the second test, and 69% on

the third test. Positive transfer and/or acquisition (EP and PO), on the other hand, accounted for

roughly 17% on Test 1, 27% on Test 2, and 30% on Test 3. (NO and OTH accounted for the rest

of the production, adding up to only approximately 1% on each of the three tests.) The ANOVA

results (Table 4-2) indicated that production type was a significant factor, and significant

interactions were found between production type and test, and between production type and

participant group.

4.2.1.1 Production for grapheme by production type and test

Tukey post-hoc t-tests revealed a number of significant differences on Test 1 (Table 4-3),

Test 2 (Table 4-4) and Test 3 (Table 4-5). Not all of the differences shown to be significant are

actually interesting to consider, given that different production types could logically occur for

different sets of words. The noteworthy comparisons, here and throughout this chapter, are

between production types EP and SP, and ES and PO (marked with+ in the tables). The

comparison between EP and SP is interesting to consider because these two production types are

essentially two sides of the same coin, so to speak. The same can be said for the comparison









between ES and PO. In the first, where English and Portuguese rules coincide (EP), they differ

from the Spanish rule (SP); in the second; where English and Spanish rules coincide (ES), they

differ from the Portuguese rule (PO). Thus, in each pair, there is a production type that

corresponds to correct production (EP and PO, respectively) and one that corresponds to

incorrect production (SP and ES, respectively), as far as target Portuguese is concerned.

There were significant differences found between production types EP and SP on all three

tests: SP (347 tokens) differed significantly from EP (113 tokens) on Test 1 (p<.0001); SP (279

tokens) occurred significantly more than EP (148 tokens) on Test 2 (p<.0001); and SP (275

tokens) differed significantly from EP (146 tokens) on Test 3 (p<.0001). On all three tests,

transfer from Spanish was greater than transfer from English/correct Portuguese production.

Additionally, there were no significant differences for EP production between tests, nor for SP

production between tests. This suggests that, for words where the Spanish rule would

produce something different from the English and/or Portuguese rules, participants were mainly

employing the Spanish rule, incorrectly producing [s], and this remained relatively constant

throughout the semester.

Examining the production types ES and PO, the picture is quite similar. Negative transfer

from English and/or Spanish (ES) was consistently higher than correct Portuguese production

(PO) for the relevant words. In the interactions on all three tests, the difference between these

two types of production was statistically significant (p<.0001). There were, however, significant

differences between tests for these two production types. ES showed a significant decrease from

Test 1 to Test 2 (p=0.0304), and between Test 1 and Test 3 (p<.0001), while there was a

significant increase for PO from Test 1 to Test 2 (p=0.0001), and between Tests 1 and 3

(p<.0001). When the correspondence rules in the learners' L1 and L2 coincided, but differed









from the L3 rule, negative transfer from the L1 and L2 abounded. Although there was a

significant decrease in the negative transfer and a significant increase in target Portuguese

production between the first two tests, no significant gains were made between Tests 2 and 3,

suggesting that learners require a good deal more exposure to Portuguese and instruction in

pronunciation to overcome the hindrances posed by the L1 and L2, or they fossilize incorrect

correspondence rules fairly early on (between Tests 2 and 3). Obviously, longer term studies are

necessary in order to discover the extent of interference from the L1 and L2.

4.2.1.2 Production for grapheme by production type and group

Considering the production type by group interaction for , those comparisons to be

discussed are between participant groups for the same production type (e.g., groups El and E2

for type EP), and between different production types for the same group (e.g., EP and SP for

group El). The production types for each language group were summed for all three tests (raw

scores) and then calculated as percentages of the total production43, such that each group's

production adds up to 100% (Table 4-6). These results are also shown graphically for the four

participant groups (Figure 4-2).

Tukey post-hoc t-tests were carried out, and a number of significant differences were

found. These results are given first according to production type for all groups. No significant

differences between groups were found for production types EN, NO or OTH; the production

types found to have significant inter-group differences were EP, ES, PO and SP. For production

type EP (Table 4-7), the only significant differences found were between El and S (p=0.0137)

and between P and S (p=0.0104). Recalling that EP is the production type that occurs when

English and Portuguese rules coincide (but differ from the Spanish rule), and remembering that


43 Raw scores and percentages are not shown for each test because the ANOVA did not find the three-way
interaction between test, production type and group to be significant.









El is the group of native English speakers with low or no proficiency in Spanish, and P is the

group of participants with previous instruction in Portuguese (most of whom would be in El

were it not for their prior exposure to Portuguese), it is perhaps to be expected that these

significant differences should exist. This seems to suggest that groups El and P are transferring

from English on these words (or showing acquisition of the Portuguese rule) more than group S

does. It is also noteworthy that there were no significant differences between group E2 and any

of the other groups. It would seem that the overall production of this group falls sufficiently

between that of groups El and P (on the high end) and group S (on the low end) to render any

differences insignificant. 44

Production type ES (Table 4-8) shows group P differing significantly from groups El

(p=0.0005) and S (p=0.0013). Given what was found for production type EP, it would not be

illogical to assume that group El participants are transferring mainly from English where ES

occurs for them, and S participants are most likely still transferring from Spanish in instances

where ES occurs. What is interesting again, is that there were no significant differences found

between E2 and the other groups. Recalling the percentages for ES production for the different

groups (Table 4-6 and Figure 4-2), it is clear that P showed the least amount of interference from

English and/or Spanish on the words where the English and Spanish rules coincided (but differed

from the Portuguese rule). S showed the greatest amount of interference (presumably mainly

from Spanish), El showed the second highest amount of interference (believed to be primarily

from English), and E2 production was between El and P in terms of interference. This seems to

suggest that previous Portuguese has indeed been a help to group P in terms of enabling correct




44 This result might be explained somewhat if test were considered, as the production of E2 more closely resembled
that of S initially, while later on E2 production was more similar to that of groups El and P.









production of Portuguese, and for some reason group E2 also seems to be at less of a

disadvantage than El and P when it comes to the hindrance caused by the L1 and/or L2.

Production type PO (Table 4-9), which is approximately the counterpart of ES, showed

significant differences between group P and the other three groups (p<.0001). This is an

interesting result, for two principal reasons. First, it validates the creation of a separate group for

those with previous instruction in Portuguese, in that these participants differed significantly

from the other groups in their production (at least of items). Second, it suggests that

instruction in Portuguese does make a difference when it comes to acquisition of pronunciation

rules. Group P significantly outperformed the other three groups (the percentage for P on this

production type was almost double the percentage of the next highest group), demonstrating

better production of the Portuguese correspondence rule for intervocalic than was found with

the other groups. (The fact that there is a statistical difference seen here for group P which was

not observed in the previous chapter, concerning error resolution, is due to the fact that here the

data are broken down further than in the previous chapter, into different possible production

types, allowing a more detailed perspective on what different participant groups are producing in

different contexts.)

Differences on the last production type, SP (Table 4-10), the approximate counterpart of

EP, showed those groups consisting primarily of native English speakers with low or no Spanish

proficiency (El and P) differing from those groups with proficient Spanish participants (E2 and

S). These results would seem to indicate that El, P and S are transferring significantly from their

L1 (English for El and P, and Spanish for S), while E2 seems to be transferring significantly

from the L2 (Spanish). This would indicate that where proficiency in the typologically more

similar language (here, Spanish) is higher, transfer from that language (be it from the L1 or L2)









to the L3 is higher, suggesting that in the acquisition of related languages, typological similarity

or relatedness is more important than language status (L1/L2).

Consideration will now be given to comparisons between different types of production

within each of the four participant groups. Those comparisons of interest (marked with + in the

tables) are EP-SP and ES-PO. For group El (Table 4-11), there were significant differences for

the two key comparisons EP-SP and ES-PO (for both, p<.0001). ES production was

significantly greater than PO production. While this finding is not particularly unexpected, given

that ES is indicative of transfer from English and/or Spanish (the L1 and/or the L2), it is

noteworthy that SP production was significantly greater than EP production, which indicates

negative interference from Spanish (this group's L2).

Group E2 (Table 4-12) also had significantly greater SP production than EP production

(p<.0001), suggestive of more negative transfer from Spanish than positive transfer from English

(and/or acquisition of the Portuguese rule). Additionally, this group showed significantly greater

production of ES than PO (p<.0001), with negative transfer from English and/or Spanish

occurring more often than target Portuguese production.

For group P (Table 4-13), there was no significant difference between EP and SP

(p=1.0000), indicating that positive transfer from English (and/or acquisition of the Portuguese

rule) and negative transfer from Spanish were roughly equivalent for the relevant words.

However, as for groups El and E2, there was a significant difference between ES and PO

(p<.0001). Although group P showed a great deal more of production type PO than did the other

groups, the fact remains that negative transfer from English and/or Spanish was still occurring

abundantly, even for the group with previous instruction in Portuguese.









Group S (Table 4-14) showed significant differences on both of the comparisons under

consideration (EP-SP and ES-PO). In both cases, S showed more negative transfer from Spanish

(SP and ES) than positive transfer from English and/or target Portuguese pronunciation (EP and

PO). Given that this is the group of native Spanish speakers, it is not unexpected that the bulk of

their production should show transfer from Spanish (or possibly English, for production types ES

and EP). It is encouraging, though, that they achieve some target Portuguese pronunciation

(evidenced clearly by production type PO).

4.2.2 Production for Grapheme

With these findings for the grapheme in mind, production for the grapheme will

be examined in this section. Results are presented in terms of raw production scores, percentages

of items for each production type (out of all items on a test), and statistically significant

differences (output from the model), which are marked with an asterisk in the tables. For this

grapheme, there was a reduced number of possible production types, since English and

Portuguese orthographic-phonological rules always coincide to produce [z], and Spanish always

differs, with its rule of-[s] (or
There were, therefore, four possible production types for this grapheme: EP, NO, OTH,

and SP. All four of these types could occur on all of the 19 words, as participants produced a

sound expected according to the English and/or Portuguese phonological systems (EP),

according to the phonological system of Spanish (SP), according to none of the three systems

(OTH), or no sound at all for the target grapheme (NO). All the sounds produced were coded

according to these four production types, and the codes summed for each test. These sums (raw




45 The voiceless interdental fricative [8] was produced for by only one participant, for three different words on
the first test; all were allocated the production code SP.









scores) were then converted to percentages, with these adding up to 100% on each test (Table 4-

15), and grouped according to production type for graphic representation (Figure 4-3).

Numerically, increases in correct production of as [z] (EP) were seen throughout the

semester, as were decreases in incorrect production of as [s] (SP), while the marginal

occurrences of NO and OTH remained roughly the same across all three tests (1-2% for each

type). Negative transfer from Spanish (SP) accounted for 37% of the total production on the first

test, 23 % on the second test, and 16% on the third test. Target Portuguese (EP), on the other

hand, accounted for 60% on Test 1, 73% on Test 2, and 80% on Test 3. (NO and OTH accounted

for the rest of the production, with both types adding up to approximately 3% on Tests 1 and 3,

and 4% on Test 2.)

The ANOVA results (Table 4-16) indicated that production type was a significant factor,

and the interaction between production type and test was significant, as was the interaction

between production type and participant group. The interaction between group and test was not

found to be significant, but the three-way interaction between group, test and production type

was significant for . Tukey post-hoc t-tests revealed a number of significant differences on

Test 1 (Table 4-17), Test 2 (Table 4-18) and Test 3 (Table 4-19). Those comparisons of interest

are between production types EP and SP (marked with + in the tables).

Despite what appears to be a sizeable numerical difference between EP and SP on the first

test, no statistically significant difference was found between these production types. However,

they did differ from each other on Tests 2 and 3 (p<.0001), with EP occurring significantly more

than SP. These results suggest that there was no statistically significant difference between

transfer from English and transfer from Spanish before instruction in Portuguese (i.e., at Test 1),









but once instruction took place, negative transfer from Spanish decreased while acquisition of

Portuguese GPCRs increased (possibly facilitated by positive transfer from English).

Between tests, there were some significant differences found for EP: between Tests 1 and 2

(p=0.0006), and between Tests 1 and 3 (p<.0001). However, there was no significant difference

found between Tests 2 and 3 (p=0.1343), suggesting that greater gains were made earlier in the

semester. The pattern was similar for SP, with significant differences between Tests 1 and 2

(p=0.0009), and between Tests 1 and 3 (p<.0001), but no significant difference between Tests 2

and 3 (p=0.1242), indicating that negative transfer decreased more markedly at the beginning of

the semester. Although instruction was not a factor that was specifically taken into account in

this study, it does seem to have had an effect.

4.2.2.1 Production for grapheme by group and type

Considering the interaction between participant group and production type, the

comparisons of interest are between groups for the same production type (e.g., groups El and E2

for type EP), and comparisons between different production types for the same group (e.g., EP

and SP for group El). The production types for each group were summed for all three tests (raw

scores) and then calculated as percentages of the total production, such that each group's

production adds up to 100% (Table 4-20). These results are also shown graphically for the four

participant groups (Figure 4-4).

Tukey post-hoc t-tests were carried out to determine where the significant differences

could be found. These results are given first according to production type for all groups. No

significant differences between groups were found for production types NO or OTH; the

production types to have significant inter-group differences were EP and SP. For production type

EP (Table 4-21), there were a number of significant differences to be considered. In fact, the

only inter-group difference which was found not to have statistical significance was the one









between El and P. Recalling that most of the participants in group P (with previous instruction in

Portuguese) would otherwise have been included in group El, it is perhaps not unexpected that

their production should be similar when considered across all three tests. It is also perhaps not

unexpected that these two groups should behave differently from the groups of speakers

proficient in Spanish (E2 and S). It is noteworthy that the two groups of Spanish speakers,

however, should differ significantly from each other. It seems that the different Lis of these two

groups may have played a role in their Portuguese production, despite the similarity they shared

in their proficiency in Spanish.

For production type SP (Table 4-22) the only inter-group difference that was not

statistically significant was that found between groups El and P. Again, it seems reasonable that

these two groups, comprised primarily of native English speakers with low or no proficiency in

Spanish, should have less negative transfer from Spanish than the other groups, comprised of

proficient Spanish speakers. It is also worth highlighting that a significant difference was found

between the two groups of proficient Spanish speakers (E2 and S), where the native Spanish

speakers showed more negative transfer from Spanish than did the non-native Spanish speakers.

Consideration will now be given to comparisons between different types of production

within each of the four participant groups. The noteworthy comparisons (EP-SP) are again

marked with+ in the tables. The EP-SP comparison showed statistically significant differences

for all of the participant groups. However, the differences were not the same for all groups. That

is, for groups El (Table 4-23), E2 (Table 4-24) and P (Table 4-25), production of EP was

significantly greater than production of SP, while for group S (Table 4-26), the reverse was true

- production type SP occurred significantly more often than production type EP. Thus, it seems

that the L1 was an important factor when it came to production: for those groups where the L1









was primarily English (El, E2 and P), target Portuguese exceeded negative transfer from Spanish

for (perhaps facilitated by the possibility of positive transfer from English), while the group

of participants whose L1 was Spanish transferred principally from Spanish for (from where

there would be no possibility of positive transfer to L3 Portuguese).

These results may be somewhat oversimplified, however, when test is not taken into

account. Thus, it is necessary to examine the three-way interaction of group, production type and

test.

4.2.2.2 Production for grapheme by group, type and test

For each participant group, raw scores were summed for each production type and test, and

percentages were calculated, with each group's production for each test adding up to 100%

(Table 4-27)46. For ease of reading tests are indicated in parentheses. As would be hoped for

language learners, there were increases in target production of [z] (EP) and decreases in incorrect

production of [s] (SP) across all groups, between all tests. As is to be expected, groups differed

in the proportions of their production, and post-hoc Tukey tests revealed those differences which

were statistically significant for EP and SP. (Test times are shown in parentheses.)

For production type EP (Table 4-28), there were a number of significant differences on the

first test. Groups El and P differed significantly from groups E2 and S, but not from each other,

nor did E2 and S differ significantly from each other. That is, the two groups of primarily native

English speakers (El and P) were seen to differ from the two groups of proficient Spanish

speakers (E2 and S), but the groups within the pairs did not differ significantly from each other.

By Test 2, the picture was slightly different: El and P still differed significantly from S, but they

no longer differed significantly from E2; and E2 continued not to differ significantly from S.


46 Production for NO and OTH is not shown, for the sake of simplicity, since there were no significance differences
between groups or tests for these production types.









This would seem to indicate that the production of E2 was somewhere in the middle in terms of

accuracy: it was not quite as accurate as El and P, but it was also not quite as inaccurate as S.

This progression continued to Test 3: El and P remained significantly different from S and there

continued to be no significant difference between El, P and E2, but at this test E2 differed

significantly from S. These results would suggest that native English speakers proficient in

Spanish (E2) behave more like native Spanish speakers (S) initially than like native English

speakers with no or low proficiency in Spanish (El). However, by the end of the semester, the

E2 participants largely overcame the negative transfer from Spanish, at least to the same extent

as the native English speakers in E2 (and P). The native Spanish speakers in S, however, did not

seem as capable of overcoming the negative transfer from Spanish as the native English speakers

proficient in Spanish in E2, suggesting that the L1 may be more involved in determining

acquisition of the Portuguese GPCR -[z] than proficiency in a typologically similar language.

For production type SP (Table 4-29), there was roughly the same significance pattern as

for EP, albeit with the raw scores and percentages decreasing for SP where they increased for

EP. The only divergence from the previous pattern was that for SP there was no significant

difference between groups E2 and S on Test 3 (although it approached significance at p=0.0752).

The reason for this non-significant difference is unclear, given that there was a significant

difference found between E2 and S on Test 3 for production type EP, but it may be related to the

relatively small number of participants in group E2.

Between tests, only four significant differences occurred. For group E2, there were

significant differences between Tests 1 and 3 for production type EP (p<.0001) and for type SP

SP (p=0.0008). For group S, there were significant differences between Tests 1 and 3 for

production type EP (p=0.0132) and for type SP (p=0.0055). In the case of both groups, EP









production increased and SP production decreased. These results show that, for , groups El

and P started out with a relatively high level of accuracy, whether due to previous Portuguese

instruction (for P) or due to the orthographic-phonological rule -[z] that is shared by English

and Portuguese, allowing for the possibility of positive transfer from English in the case of these

items. Due to this relatively high level of accuracy to begin with, it is not entirely unexpected

that there should be no great gains made by these groups for . Also, it is perhaps to be

expected that the other two groups, E2 and S, should display lower levels of accuracy initially,

given their proficiency in Spanish and the typological similarity between Spanish and

Portuguese. It is encouraging, then, to see both of the latter groups making significant gains in

accuracy by the end of the semester.

4.2.3 Relationship between Spanish Production and Portuguese Production

In this section, the results of the Spanish reading task given at Test 1 are considered in

relation to Portuguese production, in order to examine the hypothesis that accuracy of production

in Spanish is inversely related to accuracy of production in Portuguese. The results (Appendix J)

are presented in ascending order, according to level of accuracy on the Spanish reading task. A

comparison between production on the Spanish reading task and an average of the Portuguese

production on the three tests is shown graphically in Figure 4-5. (Production on the Spanish

reading task was first compared to each of the three Portuguese tests separately, but statistical

tests did not reveal any significant differences between the three comparisons, so it was decided

to calculate an average of the participants' production on the Portuguese tests and use that

average as a basis for comparison with the production on the Spanish task.)









A regression model was used to analyze these results, in order to determine the relationship

between participants' production on the Spanish reading task and production on the three

Portuguese reading tasks. The results of the regression model are shown in Figure 4-6. The slope

relating Spanish to Portuguese is negative; that is, there is a negative relationship between

accuracy in Spanish and accuracy in Portuguese. In other words, the greater the accuracy in

Spanish was, the lower the accuracy in Portuguese was. This pattern appears to suggest that

accuracy in Spanish pronunciation has a negative bearing on accuracy in Portuguese

pronunciation, an observation that is corroborated by other results seen in this chapter.

4.3 Other Observations

In this section, production is considered at the level of individual participants and words.

Production was summed per production type for each participant, and then percentages of

possible production calculated by grapheme for each participant, such that production of

grapheme adds up to 100% and production of grapheme adds up to 100% for each

participant. These percentage results were then ordered according to the four participant groups

(Appendix K).

The items of most interest to consider are those where participants produced something

other than what was expected, given the pronunciation rules for and in English, Spanish

and Portuguese. For grapheme , production type OTH could logically occur on any of the 29

words in each test, yet there was a relatively low proportion of OTH for most participants,

where production ranged from 0% to 3% across all three tests. Two participants had greater OTH

production: participant S-4, with 5%, and participant E1-14, with 8%. Participant S-4 produced

the voiceless palatal fricative [f] for two words in the first test audiovisual ("audiovisual")

and rlev'ii'\, ("television"), and for the same word in Tests 2 and 3 -presidente

("president"). Participant E1-14 produced the voiceless palatal fricative [] for six words on the









first test- museu ("museum"), visdo ("vision"),feserel (a nonce word), quasano (nonce), casaco

("coat"), preguigoso ("lazy") and [sj] for one word on the first test le i't it ("decision"). It

would seem that for the cognate words where English would have the voiced palatal fricative [3],

the participants were producing something approximating this sound with [f] and [sj]. In the

other cases, it is not clear what would have led the participants to produce the sound they did. In

fact, the voiceless palatal fricative [f] was the most common sound categorized as OTH for all of

the participants, and generally occurred on cognate words where English would have the voiced

palatal fricative [3], though not always. Although this production type accounted for a very small

proportion of the overall production, it is interesting that it is not altogether absent, supporting

claims in the literature that in acquisition of a language after the L1 (i.e., L2 and beyond),

elements often emerge that are not consistent with the grammar of the L1 nor with the L2

grammar (Broselow, Chen & Wang, 1998).

For grapheme , production type OTH could logically occur on any of the 19

words. Proportions of this production type were relatively high for , compared with . Two

participants E1-18 and E1-24 produced something other than the expected phonemes for 18%

of their total production, which equates to roughly 10 words across the three tests. Another

18 participants also registered OTH production, with percentages ranging from 2% to 12% of

their total production. These other sounds tended to be the voiced palatal fricative [3] or the

affricate [ts], neither of which would be expected, given the pronunciation rules for in

English, Spanish and Portuguese. However, again, it is interesting to observe that participants did

not produce just any sound in these OTH cases; rather, their production was generally some sort

of approximation to the target in the case where [3] was produced instead of [z], the

approximation is in place of articulation palatall instead of alveolar) rather than voicing, as was









often observed for grapheme . These observations would indicate universal tendencies

toward unmarked features.

Consideration will now be given to the production of particular words. The 48 words are

listed (Appendix L) along with their word type (C for cognate, N for nonce, and NC for non-

cognate) and the sound expected according to English pronunciation rules (the Portuguese rules

would always produce [z]; and the Spanish rules would always produce [s], or possibly [0] for

). Production was summed for each word, according to production type, and percentages

calculated, with production for each word totaling 100%.

For the five cognate words where English would have [z], target Portuguese

pronunciation (possibly due to positive English transfer) was recorded approximately half of the

time for four of the words; the fifth word, president ("president"), was produced correctly on

roughly two thirds of instances. It is possible that this word appeared more like the English form

of the cognate than the other four words in this category, which may have led to its correct

production more often. For the six cognate words where English would have [s] as well as

Spanish, there was an overwhelming production of[s] (close to 100%) on all but one word -

casos ("cases") where incorrect [s] was produced on 73% of instances. It is possible that this

word was more common in the classroom, which led to its higher correct production in

comparison with the other words in this category. In the case of the six cognate words where

English would have [3], it is interesting to find relatively low proportions of EN ([3]) on all the

words apart from audiovisual ("audiovisual"), where participants produced [3] on 35% of

occasions. Again, it is possible that this word was produced with English [3] more often than the

other words in this category due to its exact orthographic equivalence in English. It is also

noteworthy that participants produced something other than what would be expected by English,









Portuguese and Spanish rules on 8% of occasions for this word. This OTH production tended to

be [f] which approximates the English [3]. Across the words in this category, production ofSP

([s]) tended to be quite a bit higher than that of PO ([z]), except in the case of audiovisual

("audiovisual") and visdo ("vision") where production was fairly evenly split between SP and

PO. For the nonce and non-cognate words, participants generally produced inaccurate ES

([s]) roughly 85% of the time, although the non-cognate word casaco ("coat") yielded only 70%

inaccurate production. This may be due to this word being more common in the classroom than

the other non-cognate words. It is interesting that the results for the non-cognate words and for

the nonce words appear very similar. 47

Results for the words were mixed. The percentages of accuracy ranged from 55% on

the non-cognate rapazes ("young men") to 86% on the cognate horizonte ("horizon"). The target

in idealizagdo ("idealization") was skipped on 15% of occasions, by far the greatest

percentage of NO production for any word. Additionally, pastiza (nonce) and rapazes ("young

men") had relatively high OTH production, with 8% and 6%, respectively. The other sounds

produced for these words tended to be [3] and [ts], which are not attributable to Spanish or

English.

4.4 Summary

With the original research questions on interference in mind, the results presented in this

chapter support Major's (2001) OPM, which predicts interference from the L1 and/or the L2 and

this interference reducing over time as the L3 increases. Evidence was found for interference

mainly from the L1 for groups El and S, while group E2 had interference from the L2. All

groups tended to have increases in target Portuguese production throughout the semester,


47 This will be discussed in greater detail in Chapter 5.









indicative of a reduction of the role of the L1/L2 and an increase in the L3 in the learners' IL.

Additionally, there were occurrences of some items which were not consistent with the

orthographic-phonological correspondence systems of the L1, L2 or L3. These "other" sounds

(production type OTH) tended to be approximations of the sounds expected by the L1, L2 or L3

correspondence rule systems. For example, where English would have the [s]-[3] rule (i.e., on

words like visual, "visual"), the voiceless counterpartJ] was often produced. Rather thanjust

being "rogue" sounds in the production, these approximations suggest evidence for the universal

elements (U) predicted by the OPM, with less marked (e.g., voiceless) sounds appearing in the

place of more marked (e.g., voiced) sounds.

Hammarberg (2001) proposed that for an L2 to be the basis of interference, proficiency in

it would have to be relatively high. The results in this chapter support that claim to a certain

extent. E2, the group of L1 English speakers highly proficient in Spanish, tended to show more

interference from the L2 (at least initially) than El, the group of L1 English speakers with no or

low proficiency in Spanish. Group S, the L1 Spanish speakers, showed little evidence of clear

transfer from the L2, however; their transfer tended to be from the L1, even though their L2 was,

presumably, every bit as strong as the L2 of group E2. Thus, it seems that the L1 was the

principal basis of transfer for groups El, P and S. For group E2, it appears that the more

typologically similar L2 was the basis of transfer initially (before instruction in Portuguese

pronunciation), but transfer from this language was reduced later on, in favor of positive transfer

from the L1 and/or target L3 production.

Evidence was also found to support Muller and Muller's (1968) prediction that the more

difficult GPCR to acquire would be the one which differed from the L1 and L2 (i.e., the -[z]

rule, in this case). In general, participants produced items correctly more often than









items, perhaps due in part to the possibility of positive transfer from English in the case of ,

but not (except in the case of those cognate items where English would have [z]).

A negative relationship was also found between participants' production in the Spanish

reading task and in the Portuguese reading tasks: the greater the accuracy in the Spanish reading

task was, the lower the accuracy in Portuguese tended to be. This inverse relationship supports

observations and studies in the literature reviewed above, which found persistent difficulties for

Spanish speakers learning Portuguese pronunciation.

Finally, the section on variation showed some of the other sounds which participants

produced, and displayed some of the disparity found in the production of different words. All of

the words showed evidence of negative transfer: a cognate item, horizonte ("horizon"),

suffered the least negative transfer (on only 14% of its production), while a cognate item

where both English and Spanish would have [s], generosidade ("generosity"), suffered the most

negative transfer (on 99% of its production). It is interesting that these two ends of the spectrum,

in terms of transfer (negative and possibly positive, too), should be found on cognate words. The

discussion of the effect of word type on production, and the question regarding participants'

ability to generalize the pronunciation rules from seen to unseen words, is taken up in more

detail in the next chapter.









Table 4-1. Number and percentage of items, by production type and test
Type Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 %
EN 21 1 34 2 24 2
EP 113 7 148 10 146 10
ES 860 57 780 52 741 49
NO 7 0 6 0 6 0
OTH 16 1 10 1 11 1
PO 144 10 251 17 305 20
SP 347 23 279 19 275 18


Table 4-2. ANOVA for all production types for
Effect Num DF F Value P Value
test 2 0.00 1.0000
group 3 0.00 1.0000
production type 6 697.13 *<.0001
test*productiontype 12 9.12 *<.0001
group*production type 18 13.25 *<.0001
group*test 6 0.00 1.0000
group*test*production type 36 0.74 0.8681


Table 4-3. Post-hoc t-test results for all production types for
Type EN EP ES NO
EN *0.0187 *<.0001 1.0000
EP *<.0001 *0.0033
ES *<.0001
NO
OTH
PO
SP


on Test 1
OTH
1.0000
*0.0147
*<.0001
1.0000


Table 4-4. Post-hoc t-test results for all production types for on Test 2
Type EN EP ES NO OTH PO SP
EN 0.0003 K<.0001 1.0000 1.0000 <.0001 <.0001
EP <.0001 <.0001 <.0001 <.0001 <.0001
ES '<.0001 '<.0001 *<.0001 '<.0001
NO 1.0000 <.0001 <.0001
OTH <.0001 <.0001
PO 0.9473
SP


PO
*<.0001
0.8903
<.0001
<.0001
<.0001


SP
<.0001
<.0001
<.0001
<.0001
<.0001
<.0001









Table 4-5. Post-hoc t-test results for all production types for on Test 3
Type EN EP ES NO OTH PO SP
EN <.0001 <.0001 1.0000 1.0000 <.0001 <.0001
EP <.0001 <.0001 <.0001 <.0001 + <.0001
ES <.0001 <.0001 + <.0001 <.0001
NO 1.0000 <.0001 <.0001
OTH <.0001 <.0001
PO 1.0000
SP

Table 4-6. Number and percentage of items, by production type and participant group
EN EN EP EP ES ES NO NO OTH OTH PO PO SP SP
# % # % # % # % # % # % # %
El 53 2 242 10 1267 54 15 1 26 1 333 14 413 18
E2 0 0 31 7 228 52 1 0 3 1 52 12 120 28
P 23 3 94 12 361 46 2 0 4 1 207 26 92 12
S 3 0 40 4 525 55 1 0 4 0 108 11 276 29


Table 4-7. Post-hoc t-test results for production type EP for <
Group El E2 P S
El 0.9992 1.0000 *0.0137
E2 0.9517 1.0000
P *0.0104
S

Table 4-8. Post-hoc t-test results for production type ES for <
Group El E2 P S
El 1.0000 *0.0005 1.0000
E2 0.5945 1.0000
P *0.0013
S

Table 4-9. Post-hoc t-test results for production type PO for
Group El E2 P S
El 1.0000 *<.0001 0.9813
E2 *<.0001 1.0000
P *<.0001
S

Table 4-10. Post-hoc t-test results for production type SP for
Group El E2 P S
El *0.0004 0.0713 *<.0001
E2 *<.0001 1.0000
P *<.0001
S


:s> items, by participant group


:s> items, by participant group







items, by participant group







items, by participant group









Table 4-11. Post-hoc t-test results for all production types for items for participant group El
Type EN EP ES NO OTH PO SP
EN *<.0001 *<.0001 0.9999 1.0000 <.0001 *<.0001
EP *<.0001 *<.0001 *<.0001 0.1370 +*<.0001
ES *<.0001 *<.0001 *<.0001 *<.0001
NO 1.0000 *<.0001 *<.0001
OTH *<.0001 *<.0001
PO 0.3669
SP

Table 4-12. Post-hoc t-test results for all production types for items for participant group E2
Type EN EP ES NO OTH PO SP
EN 0.6029 *<.0001 1.0000 1.0000 *0.0024 +*<.0001
EP *<.0001 0.6724 0.7977 0.9913 *<.0001
ES *<.0001 *<.0001 *<.0001 *<.0001
NO 1.0000 *0.0034 *<.0001
OTH *0.0071 *<.0001
PO *<.0001
SP

Table 4-13. Post-hoc t-test results for all production types for items for participant group P
Type EN EP ES NO OTH PO SP
EN *0.0016 *<.0001 0.9999 1.0000 *<.0001 *0.0029
EP *<.0001 *<.0001 *<.0001 *<.0001 1.0000
ES *<.0001 *<.0001 *<.0001 *<.0001
NO 1.0000 *<.0001 *<.0001
OTH *<.0001 *<.0001
PO *<.0001
SP

Table 4-14. Post-hoc t-test results for all production types for items for participant group S
Type EN EP ES NO OTH PO SP
EN 0.9289 *<.0001 1.0000 1.0000 *<.0001 *<.0001
EP *<.0001 0.8812 0.9469 *0.0206 *<.0001
ES *<.0001 *<.0001 *<.0001 *<.0001
NO 1.0000 *<.0001 *<.0001
OTH *<.0001 *<.0001
PO *<.0001
SP









Table 4-15. Number and percentage of items, by production type and test
Type Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 %
EP 588 60 719 73 793 80
NO 13 1 17 2 8 1
OTH 17 2 21 2 24 2
SP 370 37 231 23 163 16


Table 4-16. ANOVA for all production types for
Effect Num DF F Value P value
test 2 0.00 1.0000
group 3 0.00 1.0000
production type 3 405.24 *<.0001
test*production type 6 18.33 *<.0001
group*production type 9 46.20 *<.0001
group*test 6 0.00 1.0000
group*test*production type 18 1.83 *0.0196


Table 4-17. Post-hoc t-test results for all production types for on Test 1
Type EP NO OTH SP
EP *<.0001 *<.0001 +0.9280
NO 1.0000 *<.0001
OTH *<.0001
SP

Table 4-18. Post-hoc t-test results for all production types for on Test 2
Type EP NO OTH SP
EP *<.0001 *<.0001 +*<.0001
NO 1.0000 *<.0001
OTH *<.0001
SP

Table 4-19. Post-hoc t-test results for all production types for on Test 3
Type EP NO OTH SP
EP *<.0001 *<.0001 *<.0001
NO 1.0000 *0.0003
OTH *0.0014
SP

Table 4-20. Number and percentage of items, by production type and participant group
EP# EP% NO# NO% O OTH OTH% SP# SP %
El 1261 82 25 2 40 3 213 14
E2 176 62 4 1 3 1 102 36
P 426 83 3 1 10 2 74 14
S 237 38 6 1 9 1 375 60









Table 4-21. Post-hoc t-test results for production type EP for
Type El E2 P S
El *0.0009 1.0000 *<.0001
E2 *0.0042 *0.0002
P *<.0001
S

Table 4-22. Post-hoc t-test results for production type SP for
Type El E2 P S
El *0.0002 1.0000 *<.0001
E2 *0.0039 *0.0002
P *<.0001
S


items, by participant group







items, by participant group


Table 4-23. Post-hoc t-test results for all production types for items for participant group El
Type EP NO OTH SP
EP *<.0001 *<.0001 +*<.0001
NO 1.0000 *0.0002
OTH *0.0010
SP


Table 4-24. Post-hoc t-test results for all production types for items for participant group E2
Type EP NO OTH SP
EP *<.0001 *<.0001 *0.0011
NO 1.0000 *<.0001
OTH *<.0001
SP

Table 4-25. Post-hoc t-test results for all production types for items for participant group P
Type EP NO OTH SP
EP *<.0001 *<.0001 *<.0001
NO 1.0000 0.1070
OTH 0.2339
SP

Table 4-26. Post-hoc t-test results for all production types for items for participant group S
Type EP NO OTH SP
EP *<.0001 *<.0001 *<.0001
NO 1.0000 *<.0001
OTH *<.0001
SP









Table 4-27. Number and percentage of items, by production type, participant group and test


EP#
El 380 (T 1)
432 (T 2)
449 (T 3)
E2 31 (T1)
62 (T 2)
83 (T 3)
P 131 (T1)
142 (T2)
153 (T 3)
S 46 (T 1)
83 (T 2)
108 (T 3)


EP%
74 (T 1)
84 (T 2)
88 (T 3)
33 (T 1)
65 (T 2)
87 (T 3)
77 (T 1)
83 (T 2)
89 (T 3)
22 (T 1)
40 (T 2)
52 (T 3)


SP #
115 (T1)
52 (T 2)
46 (T 3)
59 (T 1)
32 (T 2)
11 (T 3)
37 (T 1)
25 (T 2)
12 (T 3)
159 (T 1)
122 (T2)
94 (T 3)


SP %
22 (T 1)
10 (T 2)
9(T3)
62 (T 1)
34 (T 2)
12 (T 3)
22 (T 1)
15 (T 2)
7 (T3)
76 (T 1)
58 (T 2)
45 (T 3)


Table 4-28. Post-hoc t-test results for production type EP for items, by participant group
and test
Type El E2 P S
El *0.0002 (T 1) 1.0000 (T 1) *<.0001 (T 1)
0.9530 (T 2) 1.0000 (T 2) *<.0001 (T2)
1.0000 (T 3) 1.0000 (T 3) *<.0001 (T 3)
E2 0.0011 (T 1) 1.0000 (T 1)
0.9986 (T 2) 0.6390 (T 2)
1.0000 (T 3) *0.0298 (T 3)
P *<.0001 (T 1)
*<.0001 (T 2)
*0.0002 (T 3)
S

Table 4-29. Post-hoc t-test results for production type SP for items, by participant group
and test
Type El E2 P S
El *0.0005 (T 1) 1.0000 (T 1) *<.0001 (T 1)
0.5856 (T 2) 1.0000 (T 2) *<.0001 (T2)
1.0000 (T 3) 1.0000 (T 3) *<.0001 (T 3)
E2 *0.0061 (T 1) 1.0000 (T 1)
0.9944 (T 2) 0.7211 (T 2)
1.0000 (T 3) 0.0752 (T 3)
P *<.0001 (T 1)
*<.0001 (T 2)
*0.0002 (T 3)
S












All Production per Type for


q'4


1 2 2

EN EP
EN EP


120C 2.3197 I
i1i


100

80

60

40

20

0


PO SP


Figure 4-1. Percentage of items, by test for each production type


Production by Type and Group for


SEl

* E2

P

MS


Figure 4-2. Percentage of items, by participant group for each production type


ES NO

Production Types


* Test 1

* Test 2

* Test 3


EN EP ES NO OTH PO SP

Production Types


57,












All Production per Type for


S1


1 2 1


2 2 2


Production Types


Figure 4-3. Percentage of items, by test for each production type




Production by Type and Group for


100

80

60

40

20

0


82 83


m El

SE2


MS


Figure 4-4. Percentage of items, by participant group for each production type


100

80

60

40

20

0


* Test 1

* Test 2

* Test 3


EP NO OTH SP

Production Types


I














Spanish Reading Task Production and Average Portuguese

Production


---Spanish %

-Average Portuguese %


Figure 4-5. Percentage of correct items on Spanish reading task and average percentage of

correct items on Portuguese reading tasks


Participant numbers are shown on the horizontal axis, ordered according to accuracy of

production on the Spanish reading task. For the sake ofreadability, only some of the participant
numbers are shown (one in four).


prt Ll
83-


Eo






4D


= 64 CI -0Q 3Bgs3ari sh


i i i i
0 10 3) 30 4) 5)
^^cm dnl


OD ?0 :) 90 -iO


Figure 4-6. Relationship between production on Spanish reading task and average production on

Portuguese reading tasks


N

aJq
Q43w


a 1179


+ +




4 J4
+ +
+ -- +
+ +


4 4
J ++ +
4- "--- 4








+ +
44-




4 4
+ +
++









CHAPTER 5
GENERALIZABILITY RESULTS AND DISCUSSION

5.1 Introduction

This chapter addresses the third broad research question, regarding generalizability, by

presenting and discussing the relevant results. Before considering these, it is useful to recall the

research questions asked and hypotheses stated (in section 2.2.3) concerning generalizability.

The term generalizability was used to refer to participants' ability to apply the GPCRs that they

had learned for seen words to unseen words. The principal question was what differences in

production might arise due to word type: cognates, non-cognates and nonce-words. In light of

Defior et al.'s (2002) and Lord's (2007) findings that real words were read faster and more

accurately than nonce words, it was postulated that the orthographic-phonological

correspondences in seen words (cognates and non-cognates) would be produced more accurately

than those in unseen words (nonce words), due to learners' greater familiarity with and exposure

to the former. The results will be presented and discussed first for all production, then production

by grapheme. Next, production of the cognate items will be examined separately, in view of

the three sounds that occur for the English equivalents of these items. Variation between groups

and individuals will then be discussed, with a general summary of results to follow.

5.2 Results and Discussion

Results are presented in terms of raw production scores, proportions of correct items out of

all possible items per word type, and statistically significant differences (output from the model),

which are marked with an asterisk in the tables.

5.2.1 Production for Graphe mes and

Considering together the production of all the participants, irrespective of group and

grapheme, the data showed correct production of [z] on 1437 (out of3588) cognate items, 842









(out of 1872) nonce items, and 928 (out of 2028) non-cognate items. Expressed as percentages of

possible correct items, these raw scores are equivalent to 40%, 45% and 46%, respectively

(Table 5-1). When all production (both graphemes and all three tests) was considered, correct

production appeared to be roughly the same across all three word types (Figure 5-1).

The ANOVA (refer back to Table 3-7) showed that the interaction between grapheme and

word type was significant. Other interactions which were shown to be significant by the

ANOVA did not involve word type, which is of principal concern in this chapter, so only the

significant interaction with word type is discussed here. Other significant interactions have been

discussed in previous chapters. Since the ANOVA results showed the grapheme by word type

interaction to be significant, it is necessary to consider the graphemes separately.

5.2.2 Production by Grapheme

Correct production was summed (raw scores) for each word type, by grapheme, and then

calculated as a percentage of all possible production for that word type and grapheme (Table 5-

2). The percentage of correct tokens is also shown graphically for each word type and grapheme

(Figure 5-2). Post-hoc Tukey t-tests were carried out to determine statistically significant

differences between graphemes for each word type, and between word types for each grapheme.

Results of the t-tests revealed significant differences between and on all three word

types, with produced correctly significantly more often than (Table 5-3). This result

suggests that accuracy in production has less to do with word type and more to do with

orthography, an observation that appears to be confirmed by the fact that no significant

differences were found between word types for items (Table 5-4).

Since participants produced items with relatively the same accuracy across all word

types, it would appear that there were no facilitatory or inhibitory effects for cognates as

compared with words which were not cognates (i.e., non-cognates and nonce words).









Additionally, the lack of significant differences observed between seen words (cognates and non-

cognates) and unseen words (nonce words) does not permit any conclusion to be drawn

concerning whether production is lexically driven or rule driven.

In the case of tokens, however, there appeared to be a facilitatory effect for cognates

compared with non-cognates and nonce words (Table 5-5). Cognate items were accurately

produced significantly more often than non-cognate and nonce items. As with ,

though, no significant difference was observed between non-cognate and nonce items, which

again makes it impossible to draw any conclusion about learners' ability to generalize from seen

to unseen words. The ostensibly facilitatory effect found for cognates bears further

investigation, considering that this group of tokens included words which would have different

sounds in English for .

5.2.3 Cognate Items

The group of cognate items had an unusual characteristic: one orthographic-

phonological correspondence rule in Portuguese (-[z]), and one rule, albeit a different one, in

Spanish (-[s]), but three rules in English: five words pronounced with [z], six words

pronounced with [s], and six words pronounced with [3] (Table 5-6). This group of items

allows a more in-depth exploration of orthographic-phonological correspondences. Till now, this

dissertation has examined only differences between graphemes ( and ), and between word

types (cognate, non-cognate and nonce); now, the consideration is of items within a single word

type with one grapheme. What makes this analysis interesting is that the three sets of English

sounds for these cognates fall into different production categories with respect to the two other

languages under consideration, namely Portuguese and Spanish. That is, those words where

English would have [z] share a correspondence rule with Portuguese but not with Spanish; those

words where English would have [s] share a correspondence rule with Spanish but not with









Portuguese; and those words where English would have [3] do not share a correspondence rule

with either Portuguese or Spanish. Thus, it would seem plausible to infer that the English

correspondence rules play a part where differences are observed between the three sets of

cognate tokens.

Correct production for these three sets of words (where the set name is the sound in

English) was summed (raw scores) and then calculated as a percentage of possible items (Table

5-7). Results are shown graphically for the three sets of cognates for each test (Figure 5-3).

Numerically, scores tend to increase between tests, for the three sound groups. Tokens with [z] in

English are seen to be produced correctly more often than tokens with [3] in English, which in

turn are observed to be produced accurately more regularly than tokens with [s] in English.

The ANOVA (Table 5-8) shows that each of the individual factors group, test and

English sound is significant, as is the interaction between group and sound; thus it is necessary

to examine the production of cognate words according to the four participant groups. Correct

production for the four language groups was summed (raw scores) and then calculated as a

percentage of possible production (Table 5-9). Results are shown graphically by participant

group for each of the three sets of cognate items (Figure 5-4). A number of statistically

significant differences were found, both between and within participant groups.

For cognate tokens where English would have [z] in the pronunciation (Table 5-10),

groups El and P produced correctly as [z] (according to the target Portuguese rules)

significantly more often than group S. There were no significant differences between L1 English

speakers with low or no Spanish (El and P, mostly) and the L1 English speakers proficient in

Spanish (E2), nor was there a significant difference between the L1 English speakers proficient

in Spanish (E2) and the native Spanish speakers (S). In other words, for this set of items, group









E2 fell in the middle (between groups El and P at the higher end, and group S at the lower end)

with production that proved not to be significantly different from any other group. When it came

to the cognate tokens where English would have [s] (Table 5-11), no significant differences

were found between groups. That is, L1 English speakers, irrespective of proficiency in Spanish,

L1 Spanish speakers, and those with previous exposure to Portuguese all produced these tokens

with relatively the same (low) degree of accuracy. For those cognate items where English

would have [3] (Table 5-12), there were only significant differences between groups P and E2,

and between P and S. In both comparisons, group P participants produced the items with greater

accuracy. For this set of words, group El fell in the middle (between group P at the higher end,

and groups E2 and S at the lower end) with production that proved not to differ significantly

from any other group.

Considering each group separately, it can be seen that El production differed significantly

for the three sets of cognate tokens (Table 5-13), as did production by group P (Table 5-14).

For both groups, those tokens where English would have [z] were produced accurately most

often, followed by those tokens where English would have [3], followed by those tokens where

English would have [s]. For group E2 (Table 5-15), those tokens where English would have [z]

were produced accurately significantly more often than the other tokens, but no significant

differences were found between the other two sets of tokens (where English would have [s] or

[3]). Finally, group S showed no significant differences between any of the three sets of cognate

tokens (Table 5-16).

The results for these three sets of cognate tokens are noteworthy. It bears repeating

that all of these words are cognates containing the grapheme , but that is clearly not a

sufficient description. Since the orthographic-phonological correspondence rules for these words









remain constant in Portuguese (-[z]) and in Spanish (-[s]), it seems fair to state that the

differences which were observed between the three sets of words arose as a result of the different

English correspondence rules for the three sets. As seen in previous chapters, here again it was

observed that linguistic background played a part in production. For the native Spanish speakers

(group S), who had a consistent correspondence rule in their L1, there were no statistical

differences observed across the three sets of cognate words. For the native English speakers

(groups El, E2 and most ofP), who had different correspondence rules in their L1, there were

statistical differences in production between the sets. For all of the L1 English speakers, accurate

production, according to the target Portuguese rule, was significantly greater when the L1

correspondence rule coincided with the target rule (-[z]).

When the L1 and L2 rules coincided with each other (-[s]) but differed from the L3

Portuguese rule (-[z]), no differences were observed between groups, corroborating what has

been discussed in previous chapters: when two similar rules in a learner's previous languages

coincide, there is greater inaccuracy in production according to a different rule in the new

language (L3). When the three languages had different correspondence rules, it appeared that

differences in production were due primarily to previous instruction in the new language (group

P significantly outperformed groups E2 and S), rather than due to the L1 (groups El and E2 did

not differ significantly from each other), or due to proficiency in Spanish (groups E2 and S did

not differ significantly from each other). Consideration will now be given to some of the

variation found for particular words and word categories.









5.3 Other Observations

Here, production of individual words is discussed. The percentages of correct production

by word (Appendix M) relate to overall production, that is, by all participants on all tests.48 It is

interesting to observe tendencies within groups of words. For cognate words where English

would have [z], accuracy ranged from 46% for visivel ("visible") to 68% forpresidente

("president"). For cognate words where English would have [s], accuracy ranged from 1%

for generosidade ("generosity") to 27% for casos ("cases"). For cognate words where English

would have [3], accuracy ranged from 13% for di'i\i'\ ("divisions") to 48% for visdo ("vision").

For nonce words, accuracy ranged from 11% for lomosa to 24% for isdgio; while for non-

cognate words, accuracy ranged from 6% for manhoso ("whiney") to 29% for casaco

("coat"). For cognate words, accuracy ranged from 62% for idealizaqdo ("idealization") to

86% for horizonte ("horizon"). For nonce words, accuracy ranged from 53% forpaimozes to

83% for trazentar. Finally, for non-cognate words, accuracy ranged from 55% for rapazes

("young men") to 79% for azedo ("sour").

Clearly, there was a great deal of variation between the high and low points of the ranges,

yet the distance between these points was relatively similar across the different word types and

graphemes. These tendencies are interesting to consider as they illustrate the variability that

occurs, in terms of accurate production, within a single word category for a single grapheme.

Why there is such variability is difficult to pinpoint, but possible reasons may include how often

the participants were exposed to the words in class (for seen words), word length, closeness in





48 Although certain differences in production were found between groups in the previous section, here production by
all participants is considered together, for the sake of brevity and clarity, since the purpose of this section is merely
to present some of the salient and interesting differences between words.









form to known words, or a combination of such factors. While the effect of these things was not

controlled for or looked at in this study, future studies addressing these issues may be beneficial.

It is noteworthy that the most accurately produced word was the cognate word

horizonte ("horizon"), where English would have [z], and the least accurately produced word

was the cognate word generosidade ("generosity"), where English and Spanish would have

[s]. With the former being produced correctly on 86% of instances, and the latter being produced

correctly on only 1% of instances, there is clearly still room for improvement on all items, which

is to be expected for learners in the early stages of language acquisition. What seems almost

incomprehensible is that the least accurately produced word, generosidade ("generosity") was

produced correctly by only one participant on one test (that is, once out of a possible 156 times).

This level of inaccuracy is difficult to explain, though it might be due to a combination of the

aforementioned factors: it is a relatively long word (six syllables), its use in the classroom may

have been relatively low, and its closeness to known words (generous, generosity in English, for

example) is high. Additionally, it falls into the category of words where the L1 and L2 share a

correspondence rule which differs from the rule in the target L3, a case which is potentially

subject to negative transfer from both of a learner's previous languages. These observations

about particular words and groups of words reinforce yet again the effect of the grapheme on

accurate production, as well as the role of the learners' previous languages in the acquisition of

orthographic-phonological correspondence rules in L3.

5.4 Summary

In this chapter, word type (cognate, non-cognate, or nonce) was considered as a factor

which might affect production. Output from the statistical model showed a significant interaction

between word type and grapheme, revealing that accurate production of grapheme items

occurred significantly more than accurate production of grapheme items, across all word









types. Differences between word types were not found to be statistically significant for , and

only showed statistically significant differences between cognates and other words (non-

cognates and nonce). Thus, evidence was not found to support the hypothesis that participants

would do better on seen words than unseen words, based on the findings from the study by

Defior et al. (2002) and Lord (2007) where real words were read faster and more accurately than

nonce words.

A closer examination of the cognate words, which contained three sets of tokens,

according to different grapheme-phoneme correspondence rules in English, revealed an

interaction between participant groups and sets of cognate words with different sounds in

English ([z], [s] and [3]). It was found that accuracy of production of these sets of cognate

words varied, depending largely on the L1 of the participants. Native Spanish speakers (group

S), whose L1 correspondence rule would have been the same for all three sets of cognate

tokens (-[s]), showed no significant differences in production across the three sets. L1

English speakers (groups El, E2 and most ofP) displayed greater accuracy on words where

English would have [z], similar to the target Portuguese rule. When the L1 and L2

correspondence rules coincided (-[s]), differing from the target L3 rule (-[z]), no

significant differences were found between groups.

All of these results appear to indicate that word type did not play as important a part in

production in this study as the factors discussed in previous chapters. Nevertheless, the closer

examination of the cognate tokens with different English correspondence rules was

revealing, as it shed light on the crucial role that the Li's system of correspondence rules played

in L3 production. This analysis of the three sets of cognate items is useful because it may

help to clarify the origin of transfer (English or Spanish) for non-cognate and nonce items,









where the rules for English and Spanish pronunciation would coincide (-[z]). For the results

discussed in previous chapters, it wasn't possible to tease apart whether incorrect production of

[s] for grapheme items was suggestive of negative transfer from English or from Spanish.

Now, however, it would appear that the groups of native English speakers (El, E2 and most ofP)

were producing sounds according to the rules of their L1, given that there were significant

differences between the sets of cognates for these groups; while the native Spanish speakers

(S) were producing sounds according to the rules of their L1, since there were no significant

differences between the cognate sets for this group. Thus, it seems plausible to propose that

the groups producing [s] on the non-cognate and nonce items would also be transferring

largely from their L1. This suggestion bears further investigation, something which will be

addressed further in the next chapter's concluding remarks.









Table 5-1. Number and percentage of correct items for all [z], by word type
Word Type Raw Scores Possible Percentages
Cognate 1437 3588 40
Nonce 842 1872 45
Non-cognate 928 2028 46

Table 5-2. Number and percentage of correct items, by grapheme and word type
Word Type by Grapheme Raw Scores Possible Percentages
Cognate 777 2652 29
Nonce 166 936 18
Non-cognate 164 936 18
Cognate 660 936 85
Nonce 676 936 72
Non-cognate 764 1092 82

Table 5-3. Post-hoc t-test results comparing graphemes for each word type
compared to
Cognate *<.0001
Nonce *<.0001
Non-cognate *<.0001

Table 5-4. Post-hoc t-test results comparing word types for
Cognate Nonce Non-cognate
Cognate 0.9618 0.9998
Nonce 0.8824
Non-cognate

Table 5-5. Post-hoc t-test results comparing word types for
Cognate Nonce Non-cognate
Cognate *<.0001 *<.0001
Nonce 1.0000
Non-cognate









Table 5-6. Cognate words, with corresponding sound in English
Cognate Word English
apre',eutiiti, ("presentation") [z]
esquisito ("odd") [z]
museu ("museum") [z]
president ("president") [z]
visivel ("visible") [z]
bdsico ("basic") [s]
casos ("cases") [s]
curiosidade ("curiosity") [s]
filosofia ("philosophy") [s]
generosidade ("generosity") [s]
persuasive ("persuasive") [s]
audiovisual ("audiovisual") [3]
(IeL i,,t ("decision") [31
lii'ieV ("division") [3]
revisdo ("revision") [3]
i'le'i\i7, ("television") [3]
visdo ("vision") [3]


Number and percentage
Test 1 Test 1
# %


of correct items, by cognate set
Test 2 Test 2 Test 3 Test 3
# % # %


English [z] 113 43 148 57 146 56 407 52
English [s] 11 4 29 9 37 12 77 8
English [3] 76 24 101 32 116 37 293 31


Table 5-8. ANOVA for cognate items
Effect Num DF F Value P Value
group 3 4.60 *0.0036
test 2 12.11 *<.0001
sound 2 102.92 *<.0001
group*test 6 0.65 0.6882
group*sound 6 12.53 *<.0001
test*sound 4 0.59 0.6704
group*test*sound 12 0.58 0.8600


Table 5-9. Number and percentage of correct items for cognate sets, by participant group
[z]# [z]% [s]# [s]% [3]# [31%
El 242 60 26 5 160 33
E2 31 41 14 16 10 11
P 94 70 18 11 84 52
S 40 24 19 10 39 20


Table 5-7.
Cognate



Total
#


Total
%









Table 5-10. Post-hoc t-test results for cognate set with [z]
-English [z] El E2 P S
El 0.7097 0.9733 *<.0001
E2 0.2561 0.8872
P *<.0001
S

Table 5-11. Post-hoc t-test results for cognate set with [s]
-English [s] El E2 P S
El 0.9948 0.9998 1.0000
E2 0.9948 1.0000 1.0000
P 0.9998 1.0000 1.0000
S 1.0000 1.0000 1.0000

Table 5-12. Post-hoc t-test results for cognate set with [3]
-English [3] El E2 P S
El 0.4503 0.3014 0.7393
E2 *0.0086 0.9996
P *0.0114
S


Table 5-13. Post-hoc t-test results for cognate
Group El [z] [s] [3]
[z] *<.0001 *<.0001
[s] *<.0001
[3]


in English, by participant group






in English, by participant group






in English, by participant group


sets for participant group El


Table 5-14. Post-hoc t-test results for cognate sets for participant group P
Group P [z] [s] [3]
[z] *<.0001 *0.0425
[s] *<.0001
[3]

Table 5-15. Post-hoc t-test results for cognate sets for participant group E2
Group E2 [z] [s] [3]
[z] *0.0177 *0.0017
[s] 1.0000
[3]

Table 5-16. Post-hoc t-test results for cognate sets for participant group S
Group S [z] [s] [3]
[z] 0.1003 0.9986
[s] 0.6242
[3]











All Production per Word Type


100

80

60

40

20

0


* Cognate
* Nonce
* Non-cognate


Word Type
Figure 5-1. Percentage of correct items by word type


Production by Grapheme and Word Type

100
85 82
80 7

4iur C5u r, of t
; 60
U Nonce
40 Nonce
29
20 18 18 Non-cognate


0
Grapheme/Word Type


Figure 5-2. Percentage of correct items, by grapheme and word type











Cognate Production


100

00
Ca
60
U 43
0 40 -
" I
S.- n -^


57 56

i


E-igl,.h [z] E-,igl.h [s] ErFRl,-,'i [3]

Cognate Type/Test

Figure 5-3. Percentage of correct items, by test for each cognate set


Cognate Sets by Group


Test 1
Test 2
Test 3


100

80 -
4 60
4 60 -
41
5 40 -
40

^ 20

n


rtl ,
162
2.1 --1


[z] [s] [3]
Cognate Sets


Figure 5-4. Percentage of correct items, by participant group for each cognate set


SEl
S E2

E SP
0 NP


2
211


I n


I


.37
I-









CHAPTER 6
CONCLUSION

6.1 Introduction

The purpose of the current study was to investigate the acquisition of orthographic-

phonological correspondence rules by learners of Portuguese as an L2 or L3, with L1 English

and, where applicable, L2 Spanish, or L1 Spanish and L2 English. Acquisition was examined

from the perspectives of error resolution, interference, and generalizability, in an effort to answer

the study's overarching research questions regarding these three areas. In the next section, these

questions and the initial hypotheses proposed (in section 2.2) are restated in summarized form,

with the hypotheses confirmed or rejected, in light of the results found. Again, each of the

questions is presented separately, for sake of clarity.

6.2 Acquisition of Orthographic-Phonological Rules

6.2.1 Error Resolution

R. Q. 1. To what extent do learners acquire the Portuguese grapheme-phoneme

correspondence rules (GPCRs) under consideration, throughout the course of the semester?

H. 1. Significant increases in correct production were expected to occur throughout the

semester of instruction, at least when all production data were considered together (both

graphemes, all word types, all participant groups).

This hypothesis was confirmed, since significant progress was indeed found between tests,

although for this group of learners as a whole, there was still much room for improvement at the

end of the semester, in terms of accuracy in production.

R. Q. 2. What differences in acquisition exist between participant groups?









H. 2. It was hypothesized that there would be differences in correct production between the

language groups, with lower accuracy being found in the production of the more proficient

Spanish speakers.

This hypothesis was partially supported. Certainly the native Spanish speakers tended to

produce the target Portuguese sounds less accurately than the native English speakers with low

or no proficiency in Spanish. However, the native English speakers with high proficiency in

Spanish tended to have less accuracy in Portuguese pronunciation only initially, and by the end

of the semester their accuracy was not significantly different from that of any other group.

R. Q. 3. What differences in acquisition are there between the two graphemes and

intervocalic corresponding to the phoneme /z/?

H.3. It was anticipated that there would be significant differences between tests for both

graphemes, and significant differences between the two graphemes.

This hypothesis was partially confirmed. While there were no significant differences found

between tests for each grapheme (the ANOVA result for the test by grapheme interaction was

not found to be significant), a significant difference was found between the two graphemes, with

produced significantly more accurately than by all participant groups.

6.2.2 Interference

R. Q. 1. What evidence is there of interference from the L1 and/or the L2?

H. 1. It was anticipated that there would be evidence of interference from both the L1 and

the L2, with these reducing over time. It was also hypothesized that there would be evidence in

the data that participants were producing sounds which were neither consistent with the

correspondence systems of the L1/L2 nor with that of the L3.

These hypotheses were confirmed. As a whole, participants tended to show evidence of

transfer from the L1 and L2 (to varying degrees), with the influence of these reducing over time









as the role of the L3 increased in the learners' IL, evidenced by increased accuracy in target

production. Additionally, participants produced sounds which were not consistent with the

correspondence systems of the L1, L2 nor L3. Such production was almost always an

approximation of a sound in one of the languages (e.g., a voiceless palatal fricative was produced

where the English cognate would have the voiced palatal fricative), rather than a "rogue" sound,

suggesting a universal tendency toward some unmarked feature (voicelessness, in the case of the

example given).

R. Q. 2. What differences in interference are there between participant groups?

H. 2. It was hypothesized that the higher the proficiency level in the closely related

language, Spanish, the more evidence of transfer there would be from that language in the target

language production.

This hypothesis was supported to a certain extent. The proficient Spanish speakers (both

native and non-native) showed significantly greater interference from Spanish than the native

English speakers with low or no proficiency in Spanish, at least initially. By the end of the

semester, the native English speakers with high proficiency in Spanish were not found to differ

significantly from any other group. It is interesting to note that the native English speakers with

low proficiency in Spanish also showed evidence of transfer from Spanish, their L2, although to

a lesser extent than the participants proficient in Spanish. Additionally, it is worth mentioning

that the native Spanish speakers, proficient in their L2, English, showed more evidence of

transfer from the L1 (the more similar language to the target language) than the L2 (the less

similar language). Thus, while it is evident that both language status (L1/L2) and language

relatedness come into play in transfer, the results for the group of native Spanish speakers would

suggest that language relatedness trumps language status.









R. Q. 3. What differences in interference exist between the two graphemes and ?

H. 3. It was expected that there would be differences between the two graphemes, in light

of the GPCRs in English, Portuguese and Spanish: it was anticipated that there would be greater

accuracy in production of due to the possibility of positive transfer from English, and less

accuracy in production of due to the possibility of only negative transfer from Spanish and

generally negative transfer from English.

This hypothesis was supported, since participants as a whole produced items correctly

more often than items, perhaps in part due to the possibility of positive transfer from English

in the case of, but not (except in the case of those cognate items where English

would have [z]).

R. Q. 4. To what extent is production in Spanish related to acquisition (correct production)

of Portuguese?

H. 4. It was anticipated that production in Spanish would be inversely related to acquisition

of the Portuguese GPCRs in question. That is, it was expected that the greater the accuracy in

Spanish production, the lower the accuracy in Portuguese production would tend to be.

Again, this hypothesis was confirmed, with a negative relationship being found between

participants' production in the Spanish reading task and in the Portuguese reading tasks: students

with higher accuracy in Spanish tended to have lower accuracy in Portuguese.

6.2.3 Generalizability

R. Q. 1 What differences in production arise due to word type (cognates, non-cognates and

nonce-words)?

H. 1. It was predicted that the orthographic-phonological correspondences in seen words

(cognates and non-cognates) would be produced more accurately than those in unseen words









(nonce words), due to learners' greater familiarity with and exposure to the former. That is,

production would generally be lexically driven rather than rule driven.

This hypothesis was mostly rejected, since there were no significant differences between

word types for grapheme , and significant differences for grapheme were only found

between cognates and non-cognates, and between cognates and nonce words, with cognates

produced more accurately than the other words. The results for grapheme would seem to

suggest that all production was rule driven, while the results for grapheme would appear to

indicate that there was a facilitatory effect for cognates, but this positive effect was found to be

limited to the set of words where English would have the same correspondence rule as the target

language, Portuguese (-[z]), and was only found for native English speakers, suggesting that

for grapheme items, production is lexically driven, and more specifically, driven by the

lexicon of the L1.

In summary, a general trend was found toward increased accuracy in production and

decreased interference throughout the course of the semester of data collection. Some differences

in production were observed between participant groups, graphemes, and word types. Proficient

Spanish speakers, native and non-native, showed greater interference from Spanish than did non-

proficient Spanish speakers, at least initially. This finding would suggest two things: 1)

proficiency in a language has to be relatively high in order for it to be a significant source of

transfer; and 2) structural relatedness (i.e., between Spanish and Portuguese here) appears to play

a greater role in interference than language status (i.e., L1/L2), at least in the beginning.

Significantly greater accuracy was found on grapheme items than on grapheme items,

demonstrating that acquisition of pronunciation is closely related to orthography. Differences

between word types were only found on items, where cognates were produced more









accurately than other words. A closer examination of these cognate items revealed that

groups with L1 English were significantly more accurate on those items where their L1 had the

same correspondence rule as the target L3 rule (-[z]), while no such difference was found for

L1 Spanish speakers.

The findings of the present study corroborate the complexity of a multilingual's linguistic

systemss, as discussed previously in the literature. Both the L1 and the L2 play a role in the

acquisition of an L3, to varying extents, depending on such factors as proficiency, language

status, and language typology. The specific contributions and implications of the current study

are considered further in the next section.

6.3 Contributions and Implications

This study makes a contribution in several areas, and has theoretical and applied

implications. It contributes to the fields ofL2 acquisition, L3 acquisition, and phonology, by

specifically addressing the area of orthographic-phonological correspondence rules and systems,

something which has not received considerable attention to date. This study also contributes to

the existing literature on the acquisition of Portuguese, a less commonly taught and researched

language.

6.3.1 Theoretical Implications

The results from the present study lend support to previous work done in several areas.

Evidence was found to corroborate Muller and Muller's (1968) prediction that the more difficult

correspondence rule to acquire in an L3 is that which has the same grapheme representing

different phonemes in the L1/L2 and the L3. The findings here also support Major's (2001)

OPM, showing an initial IL consisting of the L1 and L2, with these decreasing over time as the

L3 increases, and with language universals (U) also playing a part in the IL.









Some of the conflicting results found in the literature for instance, with regard to the

effects ofbilingualism on L3 acquisition, transfer from the L2 to the L3, the role of language

status in transfer, the effects of a typologically similar language, etc. are also reflected in this

study's results. On the one hand, bilingualism at first seems to be disadvantageous for L1

English speakers proficient in Spanish, since they appear to transfer negatively from Spanish

initially; but later production is at least as accurate as that of native English speakers not

proficient in Spanish, suggesting that bilingualism may have positive effects (or maybe no effect)

later on. On the other hand, bilingualism does not appear to be of much help to native Spanish

speakers fluent in English, since their production seems to generally reflect negative transfer

from their L1, even when positive transfer from the L2 would be possible.

As to whether proficiency in the L2 has to be low or high for it to be the basis of transfer in

the L3, the present study shows greater influence from the L2 when proficiency is high, but only

in the case of a more typologically similar language, since L1 Spanish speakers with high

proficiency in L2 English tended not to transfer as much from their L2 as from their L1. Thus,

the evidence found in this study indicates that language relatedness (similar vs. dissimilar) is

more important than language status (L1 vs. L2), but only when proficiency in the related

language is high. Given these findings, this study proposes that examination of orthographic-

phonological correspondence rules, while uncommon, may perhaps be a clearer way to to

pinpoint more specifically the source of interference in anL3.

As for the role of different word types in the acquisition of grapheme-phoneme

correspondence rules, this study did not generally support previous work which found facilitation

or inhibition with cognates. Additionally it is hard to determine whether production here was

lexically driven or rule driven, given that there was generally no difference between production









for seen words and for unseen words. The implication of these inconclusive results here is that

additional research is required in this area, something which will be discussed in the section on

future directions.

The findings of the current study not only support previous observations in the literature

with regard to the persistent problem that Spanish speakers have in devoicing Portuguese

intervocalic , but also show that English speakers have the same problem. The fact that

English has the same GPCR for intervocalic as Spanish (in most cases) is not discussed in

the literature, but is important to remember, in light of the fact that many Portuguese learners are

also speakers of English, albeit not always as a native language. This point is discussed further in

the next section, which considers some of the applied implications of the present study.

6.3.2 Applied Implications

This study further corroborates the complexity of orthographic-phonological

correspondence systems, making salient the fact that acquisition of pronunciation in a new

language involves more than simply learning new sounds, stress or intonation patterns;

orthography also has a significant part to play in the acquisition of pronunciation. In order to

assist learners in acquiring new or different orthographic-phonological correspondence rules, it

may be helpful to have them participate in self-analysis projects, which would allow them to

record, examine and monitor their own progress in acquisition (Lord, 2005).

It is also worth pointing out that, often, it is the contrasts between languages that are

highlighted to learners, when maybe in this case explicit instruction regarding the similarities

between a known language, English, and the target language, Portuguese, would be more

advantageous to learners. For example, by illustrating to learners that English shares a rule for

intervocalic with Portuguese, and by providing specific examples of this, students may be









able to make helpful analogies in the L3, enabling them to extend that specific English rule for

intervocalic to all instances of intervocalic in Portuguese.

The fact that the learners in this study did not appear to generalize the GPCRs to nonce

words suggests that they are using lexical knowledge or stored exemplars to arrive at

pronunciation, indicating that they may not have learned the rules for the realization of

Portuguese grapheme-phoneme correspondences. Again, this highlights the need for explicit

instruction regarding these rules and specific activities to practise their implementation, in order

to assist in the acquisition of pronunciation.

While the findings of the present study have been inconclusive in certain respects, it has

raised interesting questions regarding the acquisition of L3 orthographic-phonological

correspondence rules, and the results firmly indicate how complex these rules and systems are,

which must prompt further studies in this field, a point which will be taken up in the next section.

6.4 Limitations and Future Directions

Although the contributions are noteworthy, as with any study there are also several

limitations to be addressed here. The method of participant selection resulted in a number of

potential complications. Participants were selected from students enrolled in introductory

Portuguese classes at the time of the study. While this form of participant selection was

convenient, it made factors such as age, gender, and length of instruction in or exposure to

previous languages impossible to control for, something which future studies might do well to

take into account. Furthermore, it was difficult to find a good way to classify native English

speakers according to their proficiency in Spanish. Grouping them according to the class in

which they were enrolled (beginner or accelerated) seemed the most objective way to divide the

participants, but some participants may have been assigned to one group when it might have

been better to assign them to another group. The grouping criteria may have obfuscated some of









the results, and future studies might achieve different results if different criteria for grouping

were used, say grouping according to some kind of pronunciation or proficiency test.

Additionally, having students participate in the study as they would in any normal class activity,

without penalty or reward as far as their grade was concerned, may have resulted in some

participants rushing through the activities, without giving due care and attention to the task.

Perhaps offering some form of incentive (grade or monetary) would address this limitation,

though it may prove difficult to quantify the effect of an incentive (or lack thereof).

The materials and tasks developed are not without their limitations. In order to obtain data

for the graphemes targeted by the present study, production was elicited rather than spontaneous.

While elicited production may have its limitations (being contrived, it may not reflect what

learners would do in natural speech), it would not have been possible for students in introductory

classes of Portuguese to produce spontaneous data prior to instruction in the language, as was

discussed in an earlier chapter. Thus, the materials and tasks were specifically developed for the

present study in order to elicit production by all participants on the same words containing the

particular GPCRs under investigation. In creating the materials, words found in the initial

chapters of the textbook were used to ensure all participants had equal opportunities for exposure

to them, with the additional inclusion of some words not found in the textbook in order to

balance word categories as much as possible. Due to the limited number of words in the textbook

containing the target grapheme, it was not possible to control for word length or frequency, nor

for similarity or dissimilarity to words which participants might have known in their other

languagess. Such factors might have affected the results (recall how the target grapheme in the

longer word idealizag(o, "idealization", was often not produced), and it would be good if future

studies could limit some of these differences.









In terms of the reading task itself, having participants move through the words (slides) at

their own pace, rather than controlling the speed at which words were presented, may have

resulted in some participants skipping over words inadvertently or on purpose in an effort to

finish the activity more quickly. On the other hand, allowing participants to go at their own pace

may have achieved two things: 1) it may have minimized the tedium for those participants who

were fast readers and who would otherwise have spent a good deal of time waiting for words to

appear automatically on the screen; and 2) it may have minimized anxiety for those participants

who were not fast readers and who would otherwise not have had sufficient time to read words

carefully. It would be interesting to have post-hoc feedback from participants as to whether they

perceived this aspect of the study design to be positive or negative, something which another

study might consider incorporating in order to inform future research.

As well as the method of data collection not being as controlled as possible, the equipment

used was certainly not the most precise available, but realistically and logistically speaking, it

was more accessible, easy to use, and inexpensive than experimental psycholinguistic equipment.

Future work might consider using more precise equipment to record and analyze the data, rather

than relying on rater judgments. Additionally, it would be interesting to consider response times

along with accuracy in production. Response times might enable a difference to be seen in

different word types, something which was not possible in the current study's examination of

accuracy alone.

Another factor which was not controlled for in the present study was the effect of

instruction. This clearly had an impact, although the extent to which instruction played a role is

not known, nor what type of instruction was used or found more helpful (i.e., classroom vs.

supplementary materials). Future studies can try to isolate this factor and see whether and how it









helps, by incorporating specific treatments (instruction), and by having a control group (with no

explicit instruction).

As mentioned in the previous section, examining the acquisition of orthographic-

phonological correspondence rules has real potential for pinpointing more precisely the sources)

of interference in an L3. Thus, it would be interesting to pursue this line of investigation with

other correspondence rules and participants. It would be worth looking at graphemes or

phonemes that would be new to participants with different linguistic backgrounds, such as the

<9>-[s] correspondence rule in Portuguese, which does not exist in English or Spanish. It might

also be worthwhile to compare elicited (reading) and spontaneous (spoken) data from more

advanced learners to establish the effect of written stimuli in later stages of acquisition.

Longitudinal studies with students majoring or minoring in Portuguese might also be of interest,

to determine how acquisition progresses over a longer period; and studies looking at the effect of

study abroad programs could also be considered. In sum, there are numerous ways to explore the

potential of this particular area of L3 phonological acquisition in an effort to determine its

viability as an indicator of acquisition of pronunciation and of interference from other languages.

6.5 Closing

The current study examined the role of the grapheme in the acquisition of pronunciation, in

light of the fact that different languages have different GPCRs. This aspect of L2 and L3

acquisition has not received a great deal of attention in the literature to date, yet has been shown

here to have a significant impact on accurate production in the target language. Further studies in

this area will provide more empirical evidence with regard to common pronunciation mistakes

and may provide teachers with useful means of correction, based on learners' linguistic

differences.









In addition, it is proposed here that real potential exists for this type of examination in

order to determine more precisely the nature of interference from an L3 learner's previous

languages. Through future studies examining the role of the L1 and L2 in the acquisition of

orthographic-phonological correspondence rules in the L3, it maybe possible to gain new

insights into how bilingual and multilingual learners organize not only the GPCR systems of

their different languages but perhaps also their mental lexicon(s), which would be of great

interest to researchers, and potentially very helpful to instructors and learners alike.









APPENDIX A
INFORMED CONSENT FORM

Protocol Title: Developing Pronunciation in Portuguese as a Foreign Language

Please read this consent document carefully before you decide to participate in this study.

Purpose of the research study: The purpose of this study is to investigate development in
pronunciation in a foreign language class.

What you will be asked to do in the study: You will be asked to read lists of words in English,
Spanish and Portuguese, while being audio-recorded.

Time required: No additional time outside of class.

Risks and Benefits: After the study is concluded, you may benefit by receiving feedback
concerning your development in pronunciation in Portuguese. There are no risks involved.

Compensation: None.

Confidentiality: Your identity will be kept confidential to the extent provided by law. Your
information will be assigned a code number. The list connecting your name to this number will
be kept in a locked office. When the study is completed and the data have been analyzed, the list
will be destroyed. Your name will not be used in any report.

Voluntary participation: Your participation in this study is completely voluntary. There is no
penalty for not participating and participation will not affect your course grade in any way.

Right to withdraw from the study: You have the right to withdraw from the study at any time
without consequence.

Whom to contact if you have questions about the study:
Sharon Barkley, Graduate Student, Linguistics Program, sbarkley@ufl.edu
(Supervisor: Dr. Gillian Lord, Romance Language and Literatures Department,
glord@rll.ufl.edu)

Whom to contact about your rights as a research participant in the study:
UFIRB Office, Box 112250, University of Florida, Gainesville, FL 32611-2250; Phone (352)
392-0433

Agreement:
I have read the procedure described above. I voluntarily agree to participate in the procedure and
I have received a copy of this description.

Participant: Date:
Principal Investigator: Date:









APPENDIX B
LANGUAGE BACKGROUND QUESTIONNAIRE


Name:


Class: POR


Time:


The information you provide here will help to inform us about your language experience,
enabling us to place you in the right class. Please answer all questions as fully as possible. For
the sections on rating your language skills, please refer to the descriptions at the end of this
questionnaire.

Sex: Age:

Have you ever been diagnosed with a reading disorder (e.g., dyslexia) or speech deficit (e.g.,
stuttering)? If so, please explain.


Language 1: English

1. Are you a native speaker of English?



2. If not, where did you learn English (at school, somewhere else)? In what grade did you start
taking classes in English? How long have you studied English?



3. Where do you currently speak this language (at home, at school, at work)?



4. With whom do you speak this language (parents, grandparents, siblings, spouse, children, co-
workers, friends)?



Language 2: Spanish

5. Are you a native speaker of Spanish?









6. Have you had formal instruction in Spanish? In what grade did you start taking classes? How
long have you studied Spanish? Have you taken any college classes in Spanish? If so, what
level was the last class you took (1000, 2000, 3000 or 4000)?




7. Have you lived in a country or community where Spanish is spoken? If so, explain.





8. Where do you currently speak Spanish (at home, at school, at work)?



9. With whom do you speak Spanish (parents, grandparents, siblings, spouse, children, co-
workers, friends)?



10. Please rate your communication skills in Spanish, referring to the descriptions at the end of
this questionnaire. Check only ONE number, corresponding to the description which you feel
best matches your abilities.
Listening 1 2 3 4
Speaking 1 2 3 4
Reading 1 2 3 4
Writing 1 2 3 4

Language 3: Portuguese

11. Are you a native speaker of Portuguese?



12. Have you had formal instruction in Portuguese? In what grade did you start taking classes?
How long have you studied Portuguese? Have you taken any college classes in Portuguese?
If so, what level was the last class you took (1000, 2000, 3000 or 4000)?









13. Have you lived in a country or community where Portuguese is spoken? If so, explain.


14. Where do you currently speak Portuguese (at home, at school, at work)?



15. With whom do you speak Portuguese (parents, grandparents, siblings, spouse, children, co-
workers, friends)?



16. Please rate your communication skills in Portuguese, referring to the descriptions at the end
of this questionnaire. Check only ONE number, corresponding to the description which you
feel best matches your abilities. (The descriptions are at the end of this document.)
Listening 1 2 3 4
Speaking 1 2 3 4
Reading 1 2 3 4
Writing 1 2 3 4

Language 4 (please specify):

17. Are you a native speaker of this language?



18. Have you had formal instruction in this language? In what grade did you start taking classes?
How long have you studied this language? Have you taken any college classes in this
language? If so, what level was the last class you took (1000, 2000, 3000 or 4000)?




19. Have you lived in a country or community where this language is spoken? If so, explain.





20. Where do you currently speak this language (at home, at school, at work)?









21. With whom do you speak this language (parents, grandparents, siblings, spouse, children, co-
workers, friends)?



22. Please rate your communication skills in this language, referring to the descriptions at the
end of this questionnaire. Check only ONE number, corresponding to the description which
you feel best matches your abilities.
Listening 1 2 3 4
Speaking 1 2 3 4
Reading 1 2 3 4
Writing 1 2 3 4

Language 5 (please specify):

23. Are you a native speaker of this language?



24. Have you had formal instruction in this language? In what grade did you start taking classes?
How long have you studied this language? Have you taken any college classes in this
language? If so, what level was the last class you took (1000, 2000, 3000 or 4000)?




25. Have you lived in a country or community where this language is spoken? If so, explain.





26. Where do you currently speak this language (at home, at school, at work)?



27. With whom do you speak this language (parents, grandparents, siblings, spouse, children, co-
workers, friends)?









28. Please rate your communication skills in this language, referring to the descriptions at the
end of this questionnaire. Check only ONE number, corresponding to the description which
you feel best matches your abilities.
Listening 1 2 3 4
Speaking 1 2 3 4
Reading 1 2 3 4
Writing 1 2 3 4









Use the following descriptions to rate your language skills for listening speaking, reading and
writing. For each skill, you should choose ONLY the description which you feel best describes
your abilities.

Listening
1. I understand only a few isolated words and phrases.
2. I can recognize a number of phrases and can understand simple questions about myself, my
activities, my likes and dislikes, as well as other people's conversations about such things. I
understand people when they speak slowly and use simple past, present and future tenses
3. I can recognize a number of grammatical structures, including conditional and subjunctive,
and I understand vocabulary on a variety of topics, such as the weather, work, studies, films,
books and so on. I can understand and follow most directions, I can understand native
speakers in a restaurant or store situation.
4. I have an extensive understanding of grammar and vocabulary, and I can understand in-depth
conversations on most topics. I can understand a native speaker of any age or social
background, and I can understand discussions on topics such as sports, politics and religion.
Speaking
1. I know a few words and phrases but I cannot have a conversation.
2. I know a number of phrases and can have a simple conversation. I can ask people questions
about themselves, their likes and dislikes, their usual activities, and I can talk about the same
things for myself I can use simple past, present and future tenses.
3. I know quite a bit of vocabulary and grammar, and I can carry on a conversation on a variety
of topics. I can ask for directions and give directions, I can order a meal in a restaurant, I can
give my opinion and have a simple discussion on common topics, such as the weather, work,
studies, films and so on. I can use a variety of tenses, including conditional and subjunctive.
4. I have extensive knowledge of grammar and vocabulary, and I can carry on in-depth
conversations on most topics. I can talk with a native speaker of any age or social
background, and I can debate topics such as sports, politics and religion. I can use most
tenses with little concern.
Reading
1. I can recognize a few words and phrases but cannot read or understand simple texts without
looking up most words in a dictionary.
2. I can get the gist of a simple text, but I need a dictionary to understand all of the words in the
text.
3. I can understand a variety of texts with occasional use of a dictionary.
4. I can understand most texts with little use of a dictionary.
Writing
1. I can write a few words and phrases but cannot put them into a grammatically correct
sentence.
2. I can write sentences containing simple grammatical structures, on a limited number of topics
(descriptions of people, places, activities, and so on).
3. I can write sentences containing complex grammatical structures (including subordinate and
relative clauses), in a variety of registers (e-mails, informal letters, compositions).
I can write business letters and academic papers, using a variety of vocabulary and grammatical
structures.









APPENDIX C
SUMMARY OF PARTICIPANT QUESTIONNAIRES


Table C-1. Summary of Participant Questionnaires
No Class_ Sex Age Read Eng Spn Spn Origin Spn Prt Prt Origin Prt Oth Oth Oth
Per Dis L/S/ Instr L/S/ Lang Origin L/S/
R/W R/W R/W
1 1130_3 F 18 N NNS NS DR; Miami; 4/4/2/2 Yes SryrHS 5 2/1/1/1 French 2 yrs 2/1/1/1
cl in elem & ms; teacher HS


mid sch


2 3010 F 22 N
3 1130_3 F 18 N


4 1130 3 F 22 N

5 3010 F 19 N


sick; cl
cancel


NNS NS Colombia 4/4/4/4 No
NS NS Up to IB in 3/4/4/3 No


relatives
NNS NS Fr


4/4/4/4 No


Venezuela
NS NNS Grades 2-5 3/1/3/2 Yes POR 1130; 4/2/3/2 French HS
parents fr
Portugal


3/1/2/2


6 1130 4 F 28 N
7 1130 3 M 21 N
8 1130_3 F 30 N



9 1130 4 M 18 N
10 3010 F 24 N
11 1130_4 M 18 N

12 3010 F 21 N


NS NNS 2 yrs HS
NS NNS 2 yrs HS
NS None


2/2/2/2 No
1/1/1/1 No
Yes Priv lessons; 1.5/1/1/ French 2 yrs
2 ms 1 HS
Curitiba Urb
Reg


NS None No
NS NNS SPN 3300 3/3/3/3 No
NS NNS 3 yrs HS 3/3/3/3 Yes Lev 1
Rosetta Sto
NS NNS 4000 level 4/4/3/3 No
in Mexico


1/2/1/1


French 2000
level


2/1/1/1


2/1/2/1









Table C-1. Continued


No Class Sex Age Read Eng Spn Spn Origin Spn Prt Prt Origin Prt Oth Oth Oth
Per Dis L/S/ Instr L/S/ Lang Origin L/S/
R/W R/W R/W


13 3010 M
14 1130 4 F
15 1130 3 M

16 1130 4 F

17 1130_4 F

18 1130_4 M

19 1130 3 M


20 1130 4 M
21 1130_3 F
22 1130 3 M


NS NNS 3301
NS NNS 3 yrs HS
NS NNS 1 coll
course


22 N NNS NS K-10;El
Salv till 15


3/3/3/4 No
2/2/3/2 No
2/1/2/1 No


Haitian NS


Brazilian
fiance


4/4/4/4 No


19 N NS NNS Chile, 1-4- 3/2/1/1 No
y-old
18 N NS None No

18 N NNS NS Venezuela 4/4/4/4 No
till 5; IB
Spanish
18 N NS NNS 2yrsHS 1/1/2/1 No
19 N NS NNS 3 yrs HS 2/2/2/2 No
19 N NS NNS 3 yrsHS 2/2/3/2 No


French 2 yrs
HS
French 4 yrs
HS


23 3010 F 23 N NS NNS Coll major; 4/4/4/3 No
7+ yrs
study
24 1130_4 F 21 N NS NNS 4yrsHS 2/2/2/2 No


Italian 3 coll
sem;
Italy


25 1130 4 M


21 N NS NNS 10 days in 2/2/2/1 No
Peru, 10 in
Ecuador


26 3010 F 19 N NS NNS 2240


4/4/4/3


2/2/1/1

3/3/3/3


3/3/3/3


3/2/4/3 No









Table C-1. Continued
No Class Sex Age Read Eng Spn Spn Origin Spn Prt Prt Origin Prt Oth Oth Oth
Per Dis L/S/ Instr L/S/ Lang Origin L/S/
R/W R/W R/W


27 1130 3 F


17 N NS NNS All through 3/3/3/3
school


28 1130 3 F 17
29 1130 3 M 24


30 1130 4 F 18
31 1130 4 M 18
32 1130 3 F 19
33 3010 M 20


NS NNS K-IB
NS None


NNS 2clHS
NNS 3 yrs HS
NNS 2 yrs HS
NNS 9-12 HS


3/3/3/3 No
No


1/1/1/1
1/1/1/1
1/1/1/1
2/1/2/1


No
Yes


Born in
Brazil;
moved to US
at 4


POR 1130;
1/2 family
fluent in Port


2/2/1/1


3/2/3/3 French 2 yrs
HS


34 3010 F 30 N NS None


Yes Private tutor 2/2/2/2
3 yrs; 8 ms
in Brazil


35 1130 3 F
36 1130 4 F


NS NNS 6-12 HS
NS NNS 7-10HS


37 3010 F 20 N NNS NS


K-12,
Colombia


3/3/3/3
1/2/2/1


No
No French 2 yrs
HS


4/4/4/4 No


French 2241


38 1130_3 F 19
39 3010 M 21


N NS NNS 3 yrsHS
N NNS NS 4314;
Mexico for
2 yrs
N NS NNS 9-10 HS


2/2/3/2
4/4/4/4


Visited
Brazil a few
times


2/1/2/1 French Lvls 1- 3/2/3/2
4 HS


40 11304 M 17 2/1/2/1 No


2/1/2/2


2/2/2/2

3/3/3/2


40 1130 4 M 17


2/1/2/1 No









C-1. Continued


No Class_ Sex Age Read Eng Spn Spn Origin Spn Prt Prt Origin Prt Oth Oth Oth
Per Dis L/S/ Instr L/S/ Lang Origin L/S/
R/W R/W R/W
41 1130 4 M 20 N NS NNS 7-9 HS 1/1/1/1 No Born in Br; 1/1/1/1
moved to US
at 4; parents
only Eng


42 3010 F 18 N NS NNS 6 yrs



43 3010 F 22 N NNS NS Graduated
HS in
Colombia


1130
1130

1130
1130


NS
NS

NS
NS


None
NS


Since 5th
grade


NNS 8-12HS
None


4/3/4/3 No Speaks at
home with
mom
4/4/4/4 No


3/3/1/1


No
4/3/4/3 No

2/2/2/2 No


French 7-10
HS


48 1130 4 M 19 N NNS NS Bolivia for
14 yrs


1130 3
3010


NS
NNS


None
NS


Studied for
6 yrs;
Colombia


4/4/4/4 No

No
4/4/4/4 No


51 1130_3 F 19 N NS NNS 10-11 HS 3/2/3/2 No G'parents
used to live
in Br; never
taught Port


3/2/2/2


Table









Table C-1. Continued
No Class Sex Age Read Eng Spn Spn Spn L/S/ Prt Prt Prt Oth Oth Oth
Per Dis Origin R/W Instr Origin L/S/ Lang Origin L/S/
R/W R/W
52 1130 4 M 22 N NS NNS 9-10HS 2/2/1/1 No










APPENDIX D
WORDS USED IN READING TASKS


Table D-1. Words used in reading tasks


Word Gloss Category


C


Language
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese


Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25


agenda
apiterrar
apresentaCgo
arranjar*
arrogante
atencioso
audiovisual
azedo*
barreida
basico
bazerdo
buzios
canja
casaco
casos
chato
chegar
chuberrdo
ciberrar
colonizaCdo
compromisso
conjugar
corresponder
corrimao*
crianga


agenda
N/A
presentation
to arrange
arrogant
attentive
audiovisual
sour
N/A
basic
N/A
shells
chicken soup
coat
cases
boring
to arrive
N/A
N/A
colonization
commitment
to conjugate
to correspond
banister
child


Symbol 1

g
rr
s
rr
rr
s
s
z
rr
s
z
z
ji
s
s
N/A
N/A
rr
rr
z
NITA
S


















j






rr
rr
N/A
N/A

IT
IT
N/A


C
NC
C
NC
C
NC
N
C
N
NC
NC
NC
C
F
F
N
N
C
F
C
C
NC
F


Sound 1

3
h
z
h
h
z
z
z
h
z
z
z
3
z
z
N/A
N/A
h
h
z
N/A
3
h
h
N/A


Symbol 2
N/A
N/A
N/A
j
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A


Sound 2
N/A
N/A
N/A
3
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A










Table D-1. Continued


Language
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese


Number
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52


Word
curiosidade
curricular (sg*)
dangar
decisdo
desenho
digital
disudo
divis6es
duzia*
enderego
esquisito
fechar
feijoada
ferrete
ferrugem*
feserel
filosofia
garrafa*
gemotal
general
generosidade
geral
gigante*
givido
gizes (pl*)
guilajo
horizonte


Gloss
curiosity
curricular
to dance
decision
drawing
digital
N/A
divisions
dozen
address
weird
to close
bean stew
N/A
rust
N/A
philosophy
bottle
N/A
general (mil)
generosity
(in) general
giant
N/A
chalks
N/A
horizon


Category
C
C
F
C
NC
C
N
C
NC
F
C
F
NC
N
NC
N
C
NC
N
C
C
NC
C
N
NC
N
C


Symbol 1
s
rr
N/A
s
s
g
s
s
z
N/A
s
N/A
ji
rr
rr
s
s
rr
g
g
g
g
g
g
g
ji
z
z


Sound 1
z
h
N/A
z
z
3
z
z
z
N/A
z
N/A
3
h
h
z
z
h
3
3
3
3
3
3
3
3
z


Symbol 2
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
g
N/A
N/A
N/A
N/A
N/A
s
N/A
N/A
N/A
z
N/A
N/A


Sound 2
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
3
N/A
N/A
N/A
N/A
N/A
z
N/A
N/A
N/A
z
N/A
N/A










Table D-1. Continued


Language
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese


Number
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81


Word
idealizaCgo
igreja
isagio
jago
jade*
janela
jantar
jeletida
jilio
joaninha
jofiro
journal
junho
justificagdo
laginho
lazer
localizado
lomosa
manhoso*
maresa
moga
mongendo
museu
narrar
nernej a
nirruto
objeto
organizagao
origem


Gloss
idealization
church
N/A
N/A
jade
window
dinner
N/A
N/A
ladybug
N/A
newspaper
June
justification
N/A
leisure
localized
N/A
smart; whiny
N/A
young lady
N/A
museum
to narrate
N/A
N/A
object
organization
origin


Category
C
NC
N
N
C
NC
NC
N
N
NC
N
C
C
C
N
NC
C
N
NC
N
F
N
C
C
N
N
C
C
C


Symbol 1
z
ji
s












j
j

z
z
s
s
s
N/A

s
rr
ji
J















rr
ji

g
z
z














g


Sound 1
z
3
z
3
3
3
3
3
3
3
3
3
3
3
3
z
z
z
z
z
N/A
3
z
h
3
h
3
z
3


Symbol 2
N/A
N/A
g
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A


Sound 2
N/A
N/A
3
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A











Table D-1. Continued


Language
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese


Number
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109


Word
paimozes
pastiza
persuasive
pessoa
pingente*
pirralho*
poloneses
preguicoso
prenegem
president
press
prozida
quasano
rab icho
racial
ragaria
rapazes
reagir
real
realizado
reciro
refeicgo
regiao
regico
regional
relat6rio
renalda
revisao


Gloss
N/A
N/A
persuasive
person
pendant
child (slang)
Polish (plural)
lazy
N/A
president
hurry
N/A
N/A
N/A
racial
N/A
boys
to react
real
realized
N/A
meal
region
N/A
regional
report
N/A
revision


Category
N
N
C
F
NC
NC
NC
NC
N
C
F
N
N
N
C
N
NC
NC
C
C
N
NC
C
N
C
NC
N
C


Symbol 1
z
z
s
N/A
g
rr
s
s
g
s
N/A
z
s
r
r
r
r
g
r
r
r
r
r
r
r
r
r
r
IT




























r


Sound 1
z
z
z
N/A
3
h
z
z
3
z
N/A
z
z
h
h
h
h
3
h
h
h
h
h
h
h
h
h
h


Symbol 2
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
z
r
N/A
z
N/A
N/A
g
g
g
N/A
N/A
s


Sound 2
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
z
h
N/A
z
N/A
N/A
3
3
3
N/A
N/A
z










Table D-1. Continued


Language
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
Portuguese
English
English
English
English
English
English
English
English


Number
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
1
2
3
4
5
6
7
8


Word
rigor*
rinagem
riscar*
rodumar
romintico
roxo
rua
rulica
rural
sergipano*
sorridente*
televised
territ6rio
terrivel
transicao
trazentar
vagem*
visao
visivel
zalito
zangado
ahead
behalf
bizarre
cohort
decision
decisive
engineer
fusion


Gloss
rigor
N/A
to risk
N/A
romantic
purple
street
N/A
rural
from Sergipe
smiling
television
territory
terrible
transition
N/A
green bean
vision
visible
N/A
angry
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A


Category
C
N
NC
N
C
NC
NC
N
C
NC
NC
C
C
C
C
N
NC
C
C
N
NC
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A


Symbol 1
r
r
r
r
r
r
r
r
r
g
rr
s
rr
rr
s
z
ji
s
s
z
z
h
h
z
h
s
s
g
s


Sound 1
h
h
h
h
h
h
h
h
h
3
h
z
h
h
z
z
3
z
z
z
z
h
h
z
h
3
s
d3
3


Symbol 2
N/A
g
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A


Sound 2
N/A
3
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A










Table D-1.
Language
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
English
Spanish


Continued
Number
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
1


Word
garage
gender
generosity
gymnasium
happy
heavy
history
hungry
jacket
jelly
journey
judge
leisure
philosophy
physics
pleasure
present
realization
reclusive
regenerate
region
reheat
result
television
treasure
visual
zebra
zipper
antropologia


Gloss
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
anthropology


Category
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A


Symbol 1
g(e)
g
g
g
h
h
h
h
j
j
j
j
s
s
s
s
s
r
r
r
r
r
r
s
s
s
z
z
g


Sound 1
3/d3
d3
d3
d3
h
h
h
h
d3
d3
d3
d3
d3
s
z
3
z
I
I
I
I
I
I
3
3
3

z
z
h


Symbol 2
N/A
N/A
s
s
N/A
N/A
N/A
N/A
N/A
N/A
N/A
dg
N/A
N/A
N/A
N/A
N/A
z
s
g
g
h
s
N/A
N/A
N/A
N/A
N/A
N/A
N/A


Sound 2
N/A
N/A
s
z
N/A
N/A
N/A
N/A
N/A
N/A
N/A
d3
N/A
N/A
N/A
N/A
N/A
z
s
d3
d3
h
z
N/A
N/A
N/A
N/A
N/A
N/A










Table D-1.
Language
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish
Spanish


Continued
Number
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20


Word
asno
caja
calle
caza
desde
girasol
jardin
jefe
Ilamar
lleno
mismo
refrigerador
rezar
rojo
rollo
zapato
zarpar
zocato
zumbar


Gloss
donkey
box
street
hunt
since
sunflower
garden
boss
to call
full
same
refrigerator
to pray
red
roll
shoe
to set sail
overripe
to hum


Category
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A


Symbol 1
s
j
11
z
s
g
j
j
11
11
s
r
r
r
r
z
z
z
z


Sound 1
s/z
h
j/3/d3
s
s/z
h
h
h
j/3/d3
j/3/d3
s/z
r
r
r
r
s
s
s
s


Symbol 2
N/A
N/A
N/A
N/A
N/A
s
N/A
N/A
N/A
N/A
N/A
g
z
j
11
N/A
N/A
N/A
N/A


Sound 2
N/A
N/A
N/A
N/A
N/A
s
N/A
N/A
N/A
N/A
N/A
h
s
h
j/3/d3
N/A
N/A
N/A
N/A













INSTRUCTION


APPENDIX E
AND SAMPLE SLIDES FOR READING TASK


Thank you for taking part in this
activity.

You will need to move through a
number of slides in order to do
so, press the space bar or Enter
key, or click the left mouse
button.


All of the Portuguese words will
be presented first, then the
English words, and finally the
Spanish words. You will be told
when the list of words for each
language will begin.


In this activity, you will be reading
words in Portuguese, English and
Spanish. (Even though you may
never have taken Portuguese or
Spanish, please do not skip
those parts of the activity.)


Before you start reading, you will
be asked to state your name.

When you finish reading, you
may leave quietly, please! It is
important that you do not close
any windows on the computer.

The activity should take
approximately 10 minutes in total.


Most of the Portuguese words
are real words, but some have
been made up for the purpose of
this activity. All of the English
and Spanish words are real
words. Please attempt to read all
ot the words euen if some ol-
them are unfamiliar.


Please STOP here, until you are
instructed to continue.


Please state your full name-


Wel done! You have finished the
words in Portuguese.

Now you will see a number of
words in Engisn You should
pronounce them however you
think is best.


caza


You will now see a number of
words in Portuguese. You should
pronounce them however you
think is best. Please speak as
clearly as possible.


reheat


fechar


Great! You have finished the
words in English.

In tne last section, you will see a
number of words in Spanish. You
should pronounce them however
you think is best.


Congratulations! You have now
finished the activity. You may
leave the room quietly, but
PLEASE DO NOT CLOSE any
windows on the computer.


Thank you!!



Figure E-1. Instruction slides and sample word slides from PowerPoint reading task









APPENDIX F
TRANSCRIPTION GUIDE FOR RATERS

Table F-1. Transcription guide for raters
Language Grapheme Symbol for Sound
Portuguese sapo/calca s
cas a/zebra z
acho f
tchau
jacare 3
diabo d3
rato/carro h
English sad s
design/zebra z
show S
church f
visual 3
jacket d3
hospital h
react J
Spanish zapato/casa s
mis mo s/z
calle j/f/3/d3
chico
jefe h/x/X
rollo/perro r
Recall that the original data set was much larger and included more sounds, but only the sounds
[s] and [z] are analysed in this dissertation.












APPENDIX G
RATER TRANSCRIPTION SHEET


Participant Number
Word IE OSP0OiE


PESSOA
persuasive
riscar
jago
generosidade
general
jantar
paimozes
basico
desenho
rulica
cumrcular
racial
dangar
cornrnao
nerneja
president
ferrugem
regico


narrar j
duzia z
feserel s
rigor
romAntico j
sergipano d;
ciberrar j
chato tJ
televisao 3
relatorio J
isagio s
divisoes
zangado I
prozida z


s Is I'
s z
ir hi
d3 13
d3 h '3
d3 h 3
d3 h 13
z s z
s sI
s 's ,z
J h
S Ir h
I r


7 r h
d3 h '3
z s z.
i r Ih
j r h l


r K
s -

_
r h



r Ji
s z
r f


s z i


S P 0 Word
ferrete
audiovisual
geral
k s trazentar
s z temtono
moca
rabicho


d3 h 33
d3 h 3


organizaQ&o z s_ z _l k s
conjugar d3 h 3 i


manhoso s
chuberrio ti
decis&o 3_
pingente d3
press s_
barrelda I
jade d3
igreja d3
jofiro d3
arrogance j
gigante d3
reagir .1


Figure G-l1.


s l


h 1 -
h l<


rh
h 3
r h


I h


disudo
horizonte


journal
vagem
guilajo
regional
gizes
pastiza
visivel
roxo
reciro
FECHAR


E -SP'OE IS P 0


r h
sz
h 3
S z
r L1h _
k Is i
r h t

s z
82S
s Ir


d3 h 3
d3 h3


justificagao d3
idealizagao z
rapazes j
givido d3
agenda d3
feijoada d3
mongendo jd3


Casos
lazer


rinagem
realizado


corresponder IJ
compromisso s

bazerio z
esquisito z
garrafa
laginho d3
objeto dd3
refeio jo


- -


apiterrar
localizado
Iransiqgo
buzios
jilo;5
crian;a
real
pirralho


h jI3
rh h
h 37


s Iz
5 |z I
r ib
([
r h i
tl I

s z
rh
h 53

h 3
h 3

r i
r h
r lb


r h


5 17
S 2z


h 3
h 13
r h
r h
Is z


z 3s
z s

k k
i r
. ir


t tJ J


ks ks J


d3 h
z s z__










k k s




----


Rater Transcription Sheet Page 1


Words in capital letters (PESSOA and FECHAR) indicate the start point of the two different
PowerPoint versions created for the task.


i i


I I


i'


d3 lh 3
z s z


v I


rihi


-e~-tt~---t-


-













Word
renalda
digital


iE ISIPI IE ISIP OWord


J r h
d3 h 3


canja d
quasano s
junho d3
museu z
arranjar j
rodumar
reglAo j
azedo z
chegar tf
jeletida d3


h 13
s3 z
b 3


d3 h





r ii.


coUonizaqoII U z S5 J K l
poloneses i s z
lomosa s s z
viso s s Z
cunosidade s s z
somdente a r h
gemolal d3 h 3
terrivel I r h
janela 'd3 h 3
zalito z s z
apresenlarao z s z J k s
enderego k k s
casaco s s z
ragaria h rh
rural J rh
fllosofia s s 2
joaninha d3 h 3
nirruto i r h
revisao a rh 3 s z
rua r h r
maresa s s z
ongem d3h 3
pregui;oso k ks s s z
prenegem _d3 h 1

behalf h +
zipper zs
cohort h
ahead h _
generosity d3 s
television 3
present z s
visual 3 __
decision 3 3
realization J_-L z Is


history
gender
happy
jacket
decisive


jelly
engineer


E S P

d3
h I


d31
d3
d i


garage 1d3 5
reheat
physics z s
result
region
leisure .3
fusion i3
heavy ,h
bizarre ,s
journey d3
reclusive 'j
regenerate
gymnasium d3
philosophy s
judge d5
zebra z s
hungry h
treasure 3
pleasure 3


zocato
mismo
zumbar
desde
asno
girasol
antropologia
jardin
zapato
calle
caza
rollo
refrigerator
Ilamar
Ileno
rezar
zarpar
caja
jefe
rojo


z [s z
s is Z
z s z
s Z z
S 'Z Z
d3 h 3
d3 h 3
03 h 3


z 3 Z
. rhh
j rh


r Ih
z si
d3 h 3
d3 h 3
a rTh


Figure G-2. Rater Transcription Sheet Page 2


English and Spanish words were read only on the first test; for tests 2 and 3, only the Portuguese
words appeared on the second page of the transcription sheet.








180


O|E


S P'O


4 ~ 'L + 4 ,


z IS


I _I
_03 h 3







d[ h 3


I I


I I


d3 h 15-









APPENDIX H
SAMPLE OF PARTICIPANT PRODUCTION DATA AND ASSIGNED VALUES FOR


AND ITEMS


Table H-l. Sample of participant production data and assigned values for and items
Word Cl Cl Cl Cl Cl Cl Cl Cl Cl
Type
Grapheme s s s s s s s s s
Value SYMBOL TYPE VALUE SYMBOL TYPE VALUE SYMBOL TYPE VALUE
No apresentaCgo apresentaCgo apresentagao esquisito esquisito esquisito museu museu museu
1 s SP 0 s SP 0 s SP 0
2 s SP 0 s SP 0 s SP 0
3 s SP 0 s SP 0 s SP 0
4 s SP 0 s SP 0 s SP 0
5 z EP 1 z EP 1 s SP 0
6 z EP 1 z EP 1 s SP 0
7 z EP 1 z EP 1 z EP 1
8 z EP 1 z EP 1 z EP 1
9 z EP 1 z EP 1 s SP 0
10 z EP 1 s SP 0 s SP 0
11 s SP 0 s SP 0 z EP 1
12 s SP 0 s SP 0 s SP 0
13 s SP 0 s SP 0 s SP 0
14 z EP 1 s SP 0 f OTH 0
15 s SP 0 z EP 1 z EP 1
16 s SP 0 s SP 0 s SP 0
17 s SP 0 s SP 0 s SP 0
18 s SP 0 z EP 1 s SP 0
19 s SP 0 z EP 1 z EP 1
20 z EP 1 z EP 1 z EP 1
21 z EP 1 z EP 1 z EP 1











APPENDIX I
PERCENTAGES OF CORRECT PRODUCTION BY PARTICIPANT AND GROUP FOR EACH GRAPHEME AND TEST

Table I-1. Percentages of correct production by participant and group for each grapheme and test
Part Group Test 1 Test 2 Test 3 Test 1 Test 2 Test 3 Test 1 Test 2 Test 3 All All All
% % % % % % All% All% All% % % %
6 El 24 34 38 42 74 84 31 50 56 67 32 46
7 El 21 21 24 100 100 100 52 52 54 100 22 53
9 El 14 14 17 84 79 79 42 40 42 81 15 41
14 El 7 17 34 32 95 89 17 48 56 72 20 40
17 El 3 17 24 53 79 84 23 42 48 72 15 38
18 El 7 38 45 37 84 89 19 56 63 70 30 46
20 El 48 59 66 95 89 100 67 71 79 95 57 72
21 El 14 28 24 100 100 100 48 56 54 100 22 53
22 El 17 34 21 68 79 95 38 52 50 81 24 47
24 El 10 7 3 74 21 47 35 13 21 47 7 23
25 El 38 28 41 100 100 100 63 56 65 100 36 61
27 El 0 0 10 42 74 68 17 29 33 61 3 26
28 El 0 3 17 68 79 100 27 33 50 82 7 37
30 El 3 45 24 0 84 74 2 60 44 53 24 35
31 El 7 14 17 79 68 74 35 35 40 74 13 37
32 El 17 31 24 89 89 95 46 54 52 91 24 51
35 El 24 10 14 100 84 89 54 40 44 91 16 46
36 El 10 45 45 47 89 89 25 63 63 75 33 50
38 El 14 24 21 95 95 89 46 52 48 93 20 49
40 El 21 55 38 74 79 84 42 65 56 79 38 54
41 El 3 14 17 95 95 74 40 46 40 88 11 42
44 El 14 41 34 89 89 95 44 60 58 91 30 54
46 El 34 31 34 89 95 100 56 56 60 95 33 58
47 El 41 48 52 89 79 100 60 60 71 89 47 64









Table I-1. Continued
Part Group Test 1 Test 2 Test 3 Test 1 Test 2 Test 3 Test 1 Test 2 Test 3 All All All
% % % <> % <> % % % %
49 El 31 52 38 100 100 100 58 71 63 100 40 64
51 El 28 34 17 84 84 74 50 54 40 81 26 48
52 El 17 17 10 74 89 89 40 46 42 84 15 42
El Average 17 28 28 74 84 88 40 50 51 82 24 47
10 E2 3 14 38 32 47 84 15 27 56 54 18 33
12 E2 0 3 24 21 74 79 8 31 46 58 9 28
13 E2 3 31 55 16 58 79 8 42 65 51 30 38
23 E2 0 14 24 5 53 95 2 29 52 51 13 28
26 E2 28 28 21 89 95 100 52 54 52 95 25 53
E2 Average 7 18 32 33 65 87 17 37 54 62 19 36
1 P 0 17 28 21 37 68 8 25 44 42 15 26
5 P 31 24 38 89 100 89 54 54 58 93 31 56
8 P 28 24 21 95 89 84 54 50 46 89 24 50
11 P 24 24 31 84 68 74 48 42 48 75 26 46
15 P 31 48 59 95 100 100 56 69 75 98 46 67
29 P 45 59 69 100 100 100 67 75 81 100 57 74
33 P 41 52 55 74 79 95 54 63 71 82 49 63
34 P 52 55 66 95 100 100 69 73 79 98 57 74
42 P 21 45 52 37 74 95 27 56 69 68 39 51
PAverage 30 39 46 77 83 89 49 56 63 83 38 56
2 S 3 3 14 5 26 68 4 13 35 33 7 17
3 S 0 0 7 0 11 32 0 4 17 14 2 7
4 S 3 0 0 16 26 16 8 10 6 19 1 8
16 S 0 0 0 5 5 11 2 2 4 7 0 3
19 S 31 48 34 58 89 84 42 65 54 77 38 53
37 S 7 21 79 5 58 89 6 35 83 51 36 42
39 S 55 41 41 95 95 95 71 63 63 95 46 65
43 S 0 3 7 0 5 37 0 4 19 14 3 8












Table I-1. Continued
Part Group Test 1 Test 2 Test 3 Test 1 Test 2 Test 3 Test 1 Test 2 Test 3 All All All
% % % % % % All% All% All% % % %
45 S 0 7 3 37 37 53 15 19 23 42 3 19
48 S 10 28 10 21 21 11 15 25 10 18 16 17
50 S 0 24 28 0 63 74 0 40 46 46 17 28
S Average 10 16 20 22 40 52 15 25 33 38 15 24









APPENDIX J
PRODUCTION ON SPANISH READING TASK AND AVERAGE PRODUCTION ON
PORTUGUESE READING TASKS

Table J-1. Production on Spanish reading task and average production on Portuguese reading
tasks
Part Group Accuracy on Average Accuracy
Spanish on Portuguese
Items % Items %









Table J-1. Continued
Part Group Accuracy on Average Accuracy
Spanish on Portuguese
Items % Items %
45 S 68 19
27 El 82 26
11 P 86 46
28 El 86 37
48 S 86 17
12 E2 91 28
30 El 91 35
50 S 95 28
1 P 100 26
2 S 100 17
3 S 100 7
10 E2 100 33
13 E2 100 38
16 S 100 3
19 S 100 53
23 E2 100 28
37 S 100 42
39 S 100 65
43 S 100 8
The results are ordered according to accuracy on the Spanish reading task.










APPENDIX K
PERCENTAGES OF PRODUCTION BY PARTICIPANT FOR EACH PRODUCTION TYPE
AND GRAPHEME

Table K-1. Percentages of Production by Participant for Each Production Type and Grapheme


Part Group


EN



EP ES NO OTH PO SP EP NO OTH SP



6 El 0 10 51 1 2 22 14 67 4 0 30
7 El 3 16 59 0 0 6 16 100 0 0 0
9 El 1 10 61 0 0 5 23 81 2 12 5
14 El 2 5 52 0 8 15 18 72 2 2 25
17 El 1 7 54 1 2 8 26 72 4 4 21
18 El 1 10 53 0 1 20 15 70 0 18 12
20 E1 3 17 36 0 0 40 3 95 0 2 4
21 El 1 14 60 0 0 8 17 100 0 0 0
22 El 1 15 55 0 0 9 20 81 5 0 14
24 El 0 7 62 0 1 0 30 47 2 18 33
25 El 3 16 54 0 0 20 7 100 0 0 0
27 El 0 0 59 0 0 3 38 61 2 0 37
28 El 0 0 54 2 0 7 37 82 4 0 14
30 El 1 6 54 0 0 18 21 53 0 0 47
31 El 7 8 60 0 1 5 20 74 4 7 16
32 El 3 11 57 0 0 13 15 91 0 0 9
35 El 1 9 56 0 1 7 25 91 5 0 4
36 El 3 9 47 0 0 24 16 75 4 2 19
38 El 1 11 57 0 1 8 21 93 0 2 5
40 E1 2 16 51 0 0 22 9 79 0 0 21
41 El 10 6 57 6 2 6 13 88 2 2 9
44 El 1 10 47 5 3 20 14 91 7 0 2
46 El 3 10 54 1 2 23 6 95 0 0 5
47 El 2 17 44 0 0 30 7 89 0 2 9
49 El 0 15 53 0 0 25 7 100 0 0 0
51 El 3 14 53 0 3 13 14 81 0 2 18
52 El 2 7 57 1 0 8 24 84 0 0 16
10 E2 0 8 52 0 3 10 26 54 0 0 46
12 E2 0 1 54 0 0 8 37 58 0 0 42
13 E2 0 9 43 0 0 21 28 51 4 0 46
23 E2 0 5 55 0 0 8 32 51 4 5 40
26 E2 0 13 59 1 0 13 15 95 0 0 5
1 P 0 2 55 0 2 13 28 42 0 4 54
5 P 3 11 49 1 0 20 15 93 2 0 5
8 P 10 16 61 0 0 8 5 89 0 5 5










TableK-1. Continued
Part Group EN EP ES NO OTH PO SP EP NO OTH SP

11 P 3 7 49 0 0 20 21 75 4 7 14
15 P 7 16 40 1 0 30 6 98 0 0 2
29 P 0 14 36 0 0 44 7 100 0 0 0
33 P 0 15 47 0 1 34 2 82 0 2 16
34 P 0 14 32 0 0 44 10 98 0 0 2
42 P 2 13 45 0 1 26 13 68 0 0 32
2 S 0 0 56 0 0 7 37 33 0 2 65
3 S 0 1 61 0 0 1 37 14 2 0 84
4 S 1 0 62 1 5 1 30 19 0 12 68
16 S 0 0 62 0 0 0 38 7 0 0 93
19 S 0 9 47 0 0 29 15 77 0 0 23
37 S 0 11 45 0 0 24 20 51 7 0 42
39 S 1 13 39 0 0 33 14 95 0 0 5
43 S 0 0 59 0 0 3 38 14 0 0 86
45 S 1 2 61 0 0 1 34 42 2 2 54
48 S 0 6 56 0 0 10 28 18 0 0 82
50 S 0 3 55 0 0 14 28 46 0 0 54









APPENDIX L
PERCENTAGES OF PRODUCTION BY PRODUCTION TYPE FOR EACH WORD

Table L-1. Percentages of Production by Production Type for Each Word
Word Type English sound EN % EP % ES % NO% OTH% PO % SP %
apresentaCgo C [z] 0 53 0 0 0 0 47
esquisito C [z] 0 47 0 3 1 0 49
museu C [z] 0 47 0 0 1 0 53
president C [z] 0 68 0 0 1 0 31
visivel C [z] 0 46 0 2 0 0 53
basico C [s] 0 0 92 0 1 8 0
casos C [s] 0 0 73 0 0 27 0
curiosidade C [s] 0 0 93 1 0 6 0
filosofia C [s] 0 0 97 1 0 2 0
generosidade C [s] 0 0 99 0 0 1 0
persuasive C [s] 0 0 92 3 0 6 0
audiovisual C [3] 35 0 0 0 8 28 30
decisdo C [3] 3 0 0 1 4 27 66
divis6es C [3] 1 0 0 0 1 13 85
revisdo C [3] 2 0 0 0 0 35 63
televised C [3] 5 0 0 0 1 37 57
visdo C [3] 5 0 0 0 1 48 46
disudo N 0 0 87 0 0 13 0
feserel N 0 0 80 0 1 19 0
isagio N 0 0 76 0 0 24 0
lomosa N 0 0 89 0 0 11 0
maresa N 0 0 83 0 0 17 0
quasano N 0 0 77 0 1 22 0
atencioso NC 0 0 87 1 1 10 0
casaco NC 0 0 70 1 1 29 0
desenho NC 0 0 72 0 1 27 0
manhoso NC 0 0 94 0 0 6 0
poloneses NC 0 0 81 0 0 19 0
preguigoso NC 0 0 84 1 2 13 0
idealizaCgo C 0 62 0 15 0 0 23
localizado C 0 67 0 0 1 0 33
organizaCgo C 0 71 0 4 1 0 24
realizado C 0 71 0 0 2 0 27
colonizaCgo C 0 67 0 4 1 0 28
horizonte C 0 86 0 0 0 0 14
bazerdo N 0 71 0 0 0 0 29









Table L-1. Continued
Word Type English sound EN % EP % ES % NO% OTH% PO % SP %
paimozes N 0 53 0 0 3 0 44
prozida N 0 77 0 0 2 0 21
trazentar N 0 83 0 0 1 0 15
pastiza N 0 70 0 1 8 0 21
zalito N 0 79 0 0 1 0 20
azedo NC 0 79 0 0 1 0 20
buzios NC 0 66 0 0 4 0 29
duzia NC 0 73 0 1 4 0 22
gizes NC 0 65 0 0 2 0 33
lazer NC 0 77 0 0 1 0 22
rapazes NC 0 55 0 0 6 0 39
zangado NC 0 74 0 0 1 0 25










APPENDIX M
PERCENTAGES OF PRODUCTION BYWORD


Table M-1. Percentages of Production by Word
Word Grapheme Word English Cognate


apresentadao
esquisito
museu
president
visivel
basico
casos
curiosidade
filosofia
generosidade
persuasive
audiovisual
decisao
divis6es
revisao
televisao
visao
disudo
feserel
isagio
lomosa
maresa
quasano
atencioso
casaco
desenho
manhoso
poloneses
preguigoso
idealizaCdo
localizado
organizagao
realizado
colonizagao
horizonte







































Sound


Type
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
N
N
N
N
N
N
NC
NC
NC
NC
NC
NC
C
C
C
C
C
C


Accuracy
(Raw Scores)
83
74
73
106
71
12
42
10
3
1
9
43
42
20
55
58
75
20
30
38
17
26
35
16
45
42
10
30
21
97
104
110
111
104
134


Accuracy
(Percentage)
53
47
47
68
46
8
27
6
2
1
6
28
27
13
35
37
48
13
19
24
11
17
22
10
29
27
6
19
13
62
67
71
71
67
86










Word Grapheme Word English Cognate Accuracy Accuracy
Type Sound (Raw Scores) (Percentage)
bazerdo N 110 71
paimozes N 83 53
prozida N 120 77
trazentar N 130 83
pastiza N 109 70
zalito N 124 79
azedo NC 123 79
buzios NC 103 66
duzia NC 114 73
gizes NC 102 65
lazer NC 120 77
rapazes NC 86 55
zangado NC 116 74


Table M-1.


Continued









LIST OF REFERENCES

Altmann, G. T. M. (1997). The ascent of Babel: An exploration of language, mind and
understanding. Oxford, UK: Oxford University Press.

Aronin, L., & Hufeisen, B. (eds). (2009). The exploration of multilingualism: Development of
research on L3, multilingualism and multiple language acquisition. Philadelphia, PA: John
Benjamins.

Azevedo, M. M. (1978). Identifying Spanish interference in the speech of learners of Portuguese.
The Modern Language Journal, 62, 18-23.

Baker, C. (2006). Foundations of bilingual education and bilingualism. Clevedon, UK:
Multilingual Matters.

Baker, W., & Trofimovich, P. (2005). Interaction of native- and second-language vowel
systems) in early and late bilinguals. Language and Speech, 48, 1-27.

Blanche, P. (2004). Using dictation to teach pronunciation. Linguagem & Ensino, 7, 175-191.

Bloomfield, L. (1933). Language. New York, NY: Holt, Rinehart and Winston.

Boxer, D., & Cohen, A. (Eds.) (2004). Studying speaking to inform second language learning.
Clevedon, UK: Multilingual Matters.

Broselow, E., Chen, S., & Wang, C. (1998). The emergence of the unmarked in second language
phonology. Studies in Second Language Acquisition, 20, 261-280.

Brown, R (1973). A first language: The early stages. Cambridge, MA: Harvard University Press.

Carvalho, A. M. (2002). Portugues para falantes de espanhol: Perspectivas de um campo de
pesquisa. Hispania, 85, 597-608.

Cenoz, J. (2000). Research on multilingual acquisition. In J. Cenoz & U. Jessner (Eds.), English
in Europe: The acquisition of a third language (pp. 39-53). Clevedon, UK: Multilingual
Matters.

Cenoz, J. (2001). The effect of linguistic distance, L2 status and age on cross-linguistic influence
in third language acquisition. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross-
linguistic influence in third language acquisition: Psycholinguistic perspectives (pp. 8-20).
Clevedon, UK: Multilingual Matters.

Cenoz, J. (2003). The additive effect ofbilingualism on third language acquisition: A review.
The International Journal of Bilingualism, 7, 71-89.

Cenoz, J., & Genesee, F. (eds.). (1998). Beyond bilingualism: Multilingualism and multilingual
education. Clevedon, UK: Multilingual Matters.









Cenoz, J., & Hoffmann, C. (2003). Acquiring a third language: What role does bilingualism
play? The International Journal of Bilingualism, 7, 1-6.

Cenoz, J., Hufeisen, B., & Jessner, U. (eds). (2001). Cross-linguistic influence in third language
acquisition: Psycholinguistic perspectives. Clevedon, UK: Multilingual Matters.

Chandler, R. E. (1958). Spanish-Portuguese transfer. The Modern Language Journal, 42, 329-
336.

Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.

Chomsky, N. (1959). Review of verbal behavior. Language, 35, 26-58.

Cohen, S. P., Tucker, R., & Lambert, W. E. (1967). The comparative skills of monolinguals and
bilinguals in perceiving phoneme sequences. Language and Speech, 10, 159-168.

Corder, S. (1967). A role for the mother tongue. In S. Gass & L. Selinker (Eds.), Language
transfer in language learning (pp. 85-97). Rowley, MA: Newbury House.

Costa, A., Caramazza, A., & Sebastian-Galles, N. (2000). The cognate facilitation effect:
Implications for models of lexical access. Journal of Experimental Psychology: Learning,
Memory and Cognition, 26, 1283-1296.

Costa, A., Santesteban, M, & Cafio, A. (2005). On the facilitatory effects of cognate words in
bilingual speech production. Brain and Language, 94, 94-103.

Cummins, J. (1979). Linguistic interdependence and the educational development of bilingual
children. Review of Educational Research, 49, 221-251.

Davine, M., Tucker, R., & Lambert, W. E. (1971). The perception of phoneme sequences by
monolingual and bilingual elementary school children. Canadian Journal of Behavioral
Science, 3, 72-76.

De Angelis, G. (2005). Multilingualism and non-native lexical transfer: An identification
problem. International Journal of Multilingualism, 2, 1-25.

De Angelis, G. (2007). Third or additional language acquisition. Clevedon, UK: Multilingual
Matters.

Defior, S., Martos, F., & Cary, L. (2002). Differences in reading acquisition development in two
shallow orthographies: Portuguese and Spanish. Applied Psycholinguistics, 23, 135-148.

Dewaele, J. (1998). Lexical inventions: French interlanguage as L2 versus L3. Applied
Linguistics, 19, 471-490.









Dewaele, J. (2001). Activation or inhibition? The interaction of L1, L2 and L3 on the language
mode continuum. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross-linguistic influence
in third language acquisition: Psycholinguistic perspectives (pp. 69-89). Clevedon, UK:
Multilingual Matters.

Ecke, P. (2001). Lexical retrieval in a third language: Evidence from errors and tip-of-the-tongue
states. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross-linguistic influence in third
language acquisition: Psycholinguistic perspectives (pp. 90-114). Clevedon, UK:
Multilingual Matters.

Ecke, P., & Hall, C. J. (2000). Lexicalische Fehler in Deutsch als Drittsprache:
Translexikalischer Einfluss auf3 Ebenen der mentalen Reprasentation. Deutsch als
Fremdsprache, 1, 31-37.

Ellis, R. (2008). The study of second language acquisition (2nd ed.). Oxford, UK: Oxford
University Press.

Ellison, F. P., & Andrews, N. (1969). Portuguese in the high schools. InD. D. Walsh (Ed.), A
handbook for teachers of Spanish and Portuguese (pp. 250-264). Lexington, MA: D. C.
Heath.

Enomoto, K. (1994). L2 perceptual acquisition: The effect of multilingual linguistic experience
on the perception of a "less novel" contrast. Edinburgh Working Papers in Applied
Linguistics, 5, 15-29.

Errasti, M. P. S. (2003). Acquiring writing skills in a third language: The positive effects of
bilingualism. The International Journal of Bilingualism, 7, 27-42.

Faerch, C., & Kasper, G. (1986). Cognitive dimensions of language transfer. InM. Sharwood
Smith & E. Kellerman (Eds.), Crosslinguistic influence in second language acquisition (pp.
49-65). Oxford, UK:Pergamon Press.

Faraco, C. E., & Moura, F. M. (1990). Gramatica. Sdo Paulo, Brazil: Editora Atica.

Flege, J. E. (1986). The production and perception of foreign language speech sounds. In H.
Wintz (Ed.), Human Communication and its Disorders, Vol. II (224-401). Norwood, NJ:
Ablex Publishing.

Flege, J. E. (1987). The production of "new" and "similar" phones in a foreign language:
evidence for the effect of equivalence classification. Journal ofPhonetics, 15, 47-65.

Fromkin, V., Rodman, R., & Hyams, N. (2007). An introduction to language. Boston, MA:
Thomson Wadsworth.

Garrison, D. L. (1979). Teaching the relatedness of Spanish and Portuguese. The Modern
Language Journal, 63, 8-12.









Gass, S. (1983). Language transfer and universal grammatical relations. In S. Gass & L. Selinker
(Eds.), Language transfer in language learning (pp. 69-82). Rowley, MA: Newbury House.

Gass, S. (1984). A review ofinterlanguage syntax: Language transfer and language universals.
Language Learning, 34, 115-131.

Gass, S., & Selinker, L. (2001). Second language acquisition: An introductory course. Mahwah,
NJ: Lawrence Erlbaum Associates.

Gibson, M, Hufeisen, B., & Libben, G (2001). Learners of German as an L3 and their
production of German prepositional verbs. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.),
Cross-linguistic influence in third language acquisition: Psycholinguistic perspectives (pp.
138-148). Clevedon, UK: Multilingual Matters.

Gleitman, L. R. (1985). Orthographic resources affect reading acquisition- If they are used.
Remedial and Special Education, 6, 24-36.

Goodwin, J. (2001). Teaching pronunciation. In M. Celce-Murcia (Ed.), Teaching English as a
second or foreign language (3rd ed.) (pp. 117-133). Boston, MA: Heinle & Heinle.

Goto, H. (1971). Auditory perception by normal Japanese adults of the sounds "1" and "r."
Neuropsychologia, 9, 317-323.

Grosjean, F. (2001). The bilingual's language modes. In J. Nicol (Ed.), One mind, two
languages: Bilingual language processing (pp. 1-22). Oxford, UK: Blackwell.

Hadley, A. O. (1986). Teaching language in context: Proficiency-oriented instruction. Boston,
MA: Heinle & Heinle.

Hammarberg, B. (2001). Roles of L1 and L2 in L3 production and acquisition. In J. Cenoz, B.
Hufeisen, & U. Jessner (Eds.), Cross-linguistic influence in third language acquisition:
Psycholinguistic perspectives (pp. 21-41). Clevedon, UK: Multilingual Matters.

Hansen Edwards, J. G. H. & Zampini, M. L. (eds.). (2008). Phonology and second language
acquisition. Philadelphia, PA: John Benjamins.

Hermans, D., Bongaerts, T., De Bot, K., & Schreuder, R. (1998). Producing words in a foreign
language: can speakers prevent interference from their first language? Bilingualism:
Language and Cognition, 1, 213-230.

Herwig, A. (2001). Plurilingual lexical organization: evidence from lexical processing in L1-L2-
L3-L4 translation. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross-linguistic influence
in third language acquisition: Psycholinguistic perspectives (pp. 115-137). Clevedon, UK:
Multilingual Matters.

Hockett, C. (1950). Learning pronunciation. The Modern Language Journal, 34, 261-269.

Holton, J. S. (1954). Portuguese for Spanish speakers. Hispania, 37, 446-452.









Hunt, K. (1965). Grammatical structures written at three grade levels. NCTE Research Report, 3.
Champaign, IL: NCTE.

Jarvis, S., & Odlin, T. (2000). Morphological type, spatial reference, and language transfer.
Studies in Second Language Acquisition, 22, 535-556.

Jebsen, H., & Biel, S. H (Eds.). (1986). Let's go: The budget guide to Spain, Portugal and
Morocco. New York, NY: St. Martin.

Jensen, J. B. (1989). On the mutual intelligibility of Spanish and Portuguese. Hispania, 72, 848-
852.

Jordan, I. (1991). Portuguese for Spanish speakers: A case for contrastive analysis. Hispania, 74,
788-792.

Kasper, G., & Rose, K. R. (2002). Pragmatic development in a second language. Oxford, UK:
Blackwell.

Keys, K. G. (2000). Discourse level phonology in the language curriculum: A review of current
thinking in teaching pronunciation in EFL courses. Linguagem & Ensino, 3, 89-105.

Khuri, A.I. (2009). Linear model methodology. London, UK: Chapman& Hall/CRC.

Klobucka, A., Jouet-Pastre, C. M. C., Moreira, M. L. de B., Sobral, P. I., & Hutchinson, A. P.
(2007). Ponto de encontro: Portuguese as a world language. Upper Saddle River, NJ:
Pearson Prentice Hall.

Koda, K. (1999). Development of L2 intraword orthographic sensitivity and decoding skills. The
Modern Language Journal, 83, 51-64.

Kohnert, K. (2004). Cognitive and cognate-based treatments for bilingual aphasia: A case study.
Brain and Language, 91, 294-302.

Krashen, S. (1982). Principles and practice in second language acquisition. London, UK:
Pergamon.

Krashen, S. (1983). Newmark' s 'Ignorance Hypothesis' and current SLA theory. In S. Gass & L.
Selinker (Eds.), Language transfer in language learning (pp. 135-153). Rowley, MA:
Newbury House.

Kroll, J. F., Dijkstra, A., Janssen, N., & Schriefers, H. (2000). Selecting the language in which to
speak: Experiments on lexical access in bilingual production. Paper presented at the 41st
annual meeting of the psychonomic society, New Orleans, LA.

Lazaraton, A. (2001). Teaching Oral Skills. InM. Celce-Murcia (Ed.), Teaching English as a
Second or Foreign Language (3rd ed.) (pp. 103-115). Boston, MA: Heinle & Heinle.









Lord, G. (2005). (How) Can we teach foreign language pronunciation? On the effects of a
Spanish phonetics course. Hispania, 88, 557-567.

Lord, G. (2007). The role of the lexicon in learning second language stress patterns. Applied
Language Learning, 17, 1-14.

Major, R. C. (2001). Foreign accent: The ontogeny and phylogeny of second language
phonology. Mahwah, NJ: Lawrence Erlbaum Associates.

Margolis, M. (2009). Brazil's Oil Rush. Newsweek, Oct. 2.

Muller, D. H., & Muller, T. V. (1968). The problem of interference in beginning Portuguese. The
Modern Language Journal, 52, 201-205.

Murphy, S. (2003). Second language transfer during third language acquisition. Columbia
University Working Papers in TESOL and Applied Linguistics, 3, 1-21.

Odlin, T. (1989). Language transfer. Cambridge, UK: Cambridge University Press.

Olshtain, E. (2001). Functional tasks for mastering the mechanics of writing and going just
beyond. In M. Celce-Murcia (Ed.), Teaching English as a second or foreign language (pp.
207-217). Boston, MA: Heinle & Heinle.

Omstein, J. (1942). Problems in the teaching of Portuguese. The Modern Language Journal, 26,
512-516.

Orr, G. (1987). Aspects of the second language acquisition of Chichewa noun class morphology.
Unpublished doctoral dissertation, University of California, Los Angeles.

Peal, E., & Lambert, W. E. (1962). The relationship ofbilingualism to intelligence.
Psychological Monographs, 76, 1-23.

Poulisse, N., & Bongaerts, T. (1994). First language use in second language production. Applied
Linguistics, 15, 36-57.

Resnick, S. (1945). Pitfalls in Spanish-Portuguese homonyms and cognates. Hispania, 28, 233-
235.

Roberts, P., & Deslauriers, L. (1999). Picture naming of cognate and non-cognate nouns in
bilingual aphasia. Journal of Communication Disorders, 32, 1-23.

Schwegler, A., & Kempff J. (2007). Fonetica y fonologia espafiolas (3rd ed). Hoboken, NJ: John
Wiley & Sons.

Selinker, L. (1972). Interlanguage. International Review of Applied Linguistics, 10, 209-231.

Selinker, L., & Baumgartner-Cohen, B. (1995). Multiple language acquisition: 'Damn it, why
can't I keep these two languages apart?'. Language, Culture and Curriculum, 8, Specia
Issue: Multilingualism and Language Learning, 115-123.


198









Shanon, B. (1991). Faulty language selection in polyglots. Language and Cognitive Processes, 6,
339-350.

Singleton, D. (1999). Exploring the second language mental lexicon. Cambridge, UK:
Cambridge University Press.

Skinner, B. F. (1957). Verbal behavior. New York, NY: Appleton Century Crofts.

Tarquinio, L. T. L (1977). The interference of Spanish in beginning Portuguese classes.
Hispania, 60, 82-87.

Timberlake, C. (1989). Portugal still wary of closest neighbor: Report from Portugal. The Miami
Herald, May 31, 2A.

Vitevitch, M. S. (2002). The influence of phonological similarity neighborhoods on speech
production. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28,
735-747.

Vitevitch, M. S. (2003). The influence of sublexical and lexical representations on the processing
of spoken words in English. Clinical Linguistics and Phonetics, 17, 487-499.

Wade-Woolley, L. (1999). First language influences on second language word reading: All roads
lead to Rome. Language Learning, 49, 447-471.

Werker, J., & Tees, R. (1984). Phonemic and phonetic factors in the adult cross-language speech
perception. Journal of the Acoustical Society of America, 75, 1866-1878.

White, L. (2003). Second language acquisition and universal grammar. Cambridge, UK:
Cambridge University Press.

Yamada, R., & Tohkura, Y. (1992). Perception of American English /r/ and /1/ by native speakers
of Japanese. In E. Tohkura et al. (Eds.), Speech perception, production and linguistic
structure (pp. 155-174). Tokyo, Japan: Ohmsha.









BIOGRAPHICAL SKETCH

Sharon Barkley earned a Joint-Honours Bachelor of Arts in Linguistics and Portuguese

from the University of Leeds, Leeds, U.K., in 1993. From the University of Florida, she earned a

Certificate in Teaching English as a Second Language in 2006, and her PhD in Linguistics in

2010. Sharon's research interests include second and third language acquisition, phonology (in

particular, pronunciation), and Portuguese pedagogy.





PAGE 1

1 THE ACQUISITION OF ORTHOGRAPHIC PHONOLOGICAL CORRESPONDENCE RULE S IN L2 AND L3 PORTUGUESE : ERROR RESOLUTION, INTERFERENCE AND GENERALIZABILITY By SHARON BARKLEY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVER SITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Sharon Barkley

PAGE 3

3 To Herman O.

PAGE 4

4 ACKNOWLEDGMENTS I thank my dissertation chair, Dr. Gillian Lord, for her enduri ng patience, wise counsel, speedy feedback, and constant availability and willingness to help. I thank my committee members, Dr. Theresa Antes, Dr. Libby Ginway, and Dr. Edith Kaan, for their interest, advice, and dedication in reading this dissertation. I thank my raters and fellow teaching assistants, Andrea Ferreira and Quinn Hansen, for their friendship and for the hours they spent cheerfully listening to and transcribing the data for this project. I thank my statistics consultant, Claudio Fuentes, for his invaluable help in the statistical analysis of the data. I thank my professors and colleagues in the Linguistics Program and the Department of Spanish and Portuguese Studies, for their input and support, especially Dr. Laurie Massery and Dr. Rose Smous e. I thank my teaching supervisor, Dr. Libby Ginway, my fellow teaching assistants, and my many students for all that I have learned from them while teaching Portuguese. I thank my family and friends for their continued love, support, encouragement and pra yers throughout this degree program, especially Bill and Mary Barkley, Dr. Thomas Barkley, Sharon Dice, Carol Hall, Tom and Florence Barkley, Liz Horner, Chad Preston, and the members of the Supper Club. Finally, I thank God for His innumerable gifts to me and I pray that He may bless these many people who have so patiently accompanied me on this journey

PAGE 5

5 TABLE OF CONTENTS ACKNOWLEDGMENTS ...............................................................................................................4 page LIST OF T ABLES ...........................................................................................................................8 LIST OF FIGURES .......................................................................................................................12 LIST OF ABBREVIATIONS ........................................................................................................13 ABSTRACT ...................................................................................................................................14 CHAPTER 1 INTRODUCTION AND LITERATURE REVIEW ..............................................................15 1.1 Introduction ...................................................................................................................15 1.2 L2 and L3 Acquisition...................................................................................................17 1.2.1 Acquisition of Typologically Similar Languages .............................................28 1.2.2 The Relatedness of Spanish and Po rtuguese .....................................................31 1.2.3 Acquisition of L2 and L3 Phonology ................................................................36 1.3 Acquisition of L2 and L3 Orthographic Phonological Correspondence Ru les ............40 1.4 The Relevant Rules of English, Spanish and Portuguese ..............................................46 1.5 Different Word Types ...................................................................................................48 1.6 The Present Study ..........................................................................................................52 2 METHODOLOGY .................................................................................................................56 2.1 Introduction ...................................................................................................................56 2.2 Research Questions and Hypotheses .............................................................................56 2.2.1 Error Resolution ................................................................................................57 2.2.2 Interference ........................................................................................................58 2.2.3 Generalizability .................................................................................................60 2.3 Methodology .................................................................................................................61 2.3.1 Participants ........................................................................................................61 2.3.2 Materials ............................................................................................................64 2.3.3 Data Collection ..................................................................................................67 2.3.4 Transcription .....................................................................................................69 2.3.5 Analysis .............................................................................................................71 3 ERROR RESOLUTION RESULTS AND DISCUSSION .................................................77 3.1 Introduction ...................................................................................................................77 3.2 Results and Discussion ..................................................................................................78 3.2.1 All Production ...................................................................................................78 3.2.2 Production by Group .........................................................................................78

PAGE 6

6 3.2.3 Production by Grapheme ...................................................................................80 3.3 Other Observations ........................................................................................................82 3.3.1 Production by Test and Group ..........................................................................82 3.3.2.1 Production by native Spanish speakers ...............................................82 3.3.2.2 Pr oduction by native English speakers with no or low proficiency in Spanish ............................................................................................84 3.3.2.3 Production by native English speakers with high proficiency in Spanish................................................................................................85 3.3.3 Production by Grapheme ...................................................................................86 3.3.3.1 Production for grapheme .............................................................86 3.3.3.2 Production for gra pheme .............................................................89 3.4 Summary .......................................................................................................................91 4 INTERFERENCE RESULTS AND DISCUSSION ...........................................................98 4.1 Introduction ...................................................................................................................98 4.2 Results and Discussion ................................................................................................100 4.2.1 Production for Grapheme .........................................................................100 4.2.1.1 Production for grapheme by production type and test ...............102 4.2.1.2 Production for grapheme by production type and group ...........104 4.2.2 Production for Grapheme .........................................................................108 4.2.2.1 Production for grapheme by group and type .............................110 4.2.2.2 Production for grapheme by group, type and test ......................112 4.2.3 Relationship between Spanish Production and Portuguese Production ..........114 4.3 Other Observations ......................................................................................................115 4.4 Summary .....................................................................................................................118 5 GENERALIZABILITY RESULTS AND DISCUSSION ................................................130 5.1 Introduction .................................................................................................................130 5.2 Results and Discussion ................................................................................................130 5.2.1 Production for Graphem es and .........................................................130 5.2.2 Production by Grapheme .................................................................................131 5.2.3 Cognate Items ...........................................................................................132 5.3 Other Observations ......................................................................................................136 5.4 Summary .....................................................................................................................137 6 CONCLUSION .....................................................................................................................145 6.1 Introduction .................................................................................................................145 6.2 Acquisition of Orthographic Phonological Rules .......................................................145 6.2.1 Error Resolution ..............................................................................................145 6.2.2 Interference ......................................................................................................146 6.2.3 Generalizability ...............................................................................................148 6.3 Contributions and Implications ...................................................................................150 6.3.1 Theoretical Implications ..................................................................................150 6.3.2 Applied Implications .......................................................................................152

PAGE 7

7 6.4 Limitations and Future Directions ...............................................................................153 6.5 Closing ........................................................................................................................156 APPENDIX A INFORMED CONSENT FORM ..........................................................................................158 B LANGUAGE BACKGROUND QUESTIONNAIRE ..........................................................159 C SUMMARY OF PARTICIPANT QUESTIONNAIRES .....................................................165 D WORDS USED IN READING TASKS ...............................................................................170 E INSTRUCTION AND SAMPLE SLIDES FOR READING TASK ...................................177 F TRANSCRIPTION GUIDE FOR RATERS ........................................................................178 G RATER TRANSCRIPTION SHEET ...................................................................................179 H SAMPLE OF PARTICIPANT PRODUCTION DATA AND ASSIGNED VALUES FOR AND ITEMS ..................................................................................................181 I PERCENTAGES OF CORRECT PRODUCTION BY PARTICIPANT AND GROUP FOR EACH GRAPHEME AND TEST ...............................................................................182 J PRODUCTION ON SPANISH READING TASK AND AVERAGE PRODUCTION ON PORTUGUESE READING TASKS .............................................................................185 K PERCENTAGES OF PRODUCTION BY PARTICIPANT FOR EACH PRODUCTION TYPE AND GRAPHEME .........................................................................187 L PERCENTAGES OF PRODUCTION BY PRODUCTION TYPE FOR EACH WORD ...189 M PERCENTAGES OF PRODUCTION BY WORD .............................................................191 LIST OF REFERENCES .............................................................................................................193 BIOGRAPHICAL SKETCH .......................................................................................................200

PAGE 8

8 LIST OF TABLES Table page 11 Symbols examined, Portuguese sounds, and examples .....................................................55 12 Symbols examined, Portuguese contexts and sounds, and examples ................................55 13 Symbols examined, Spanish sounds, and examples ..........................................................55 14 Symbols examined, English sounds, and examples ...........................................................55 21 Participant groups ..............................................................................................................75 22 Revised participant groups .................................................................................................75 23 Grapheme words .........................................................................................................75 24 Grapheme words .........................................................................................................75 25 Summary of inter rater agreement for and items (number and percentage) ........76 31 Number and percentage of correct items for all [z], by test ...............................................93 32 Participant groups ..............................................................................................................93 33 Number and percentage of correct items for all [z], by group and test ..............................93 34 ANOVA for all production ................................................................................................93 35 Revised participant groups .................................................................................................94 36 Num ber and percentage of correct items for all [z], by revised group and test .................94 37 ANOVA for all production (revised groups) .....................................................................94 38 Significant differences found between tests, by group ......................................................94 39 Significant differences found between groups, by test ......................................................95 310 Nu mber and percentage of correct items, by grapheme and group ...................................95 311 Production of [z] by native Spanish speakers in English reading task ..............................95 41 Number and percentage of items, by production type and test .................................121 42 ANOVA for all production types for ........................................................................121 43 Post hoc t test results for all production types for on Test 1 ....................................121

PAGE 9

9 44 Post hoc t test results for all production types for on Test 2 ....................................121 45 Post hoc t test results for all production types for on Test 3 ....................................122 46 Number and percentage of items, by production type and participant group ...........122 47 Post hoc t test results for production type EP for items, by participant group .........122 48 Post hoc t test results for production t ype ES for items, by participant group .........122 49 Post hoc t test results for production type PO for items, by participant group ........122 410 Post hoc t test results for production type SP for items, by participant group .........122 411 Post hoc t test results for all production types for items for participant group E1 ...123 412 Post hoc t test results for all production types for items for participant group E2 ...123 413 Post hoc t test res ults for all production types for items for participant group P .....123 414 Post hoc t test results for all production types for items for participant group S .....123 415 Number and percentage of items, by production type and test .................................124 416ANOVA for all production types for ............................................................................124 417 Post hoc t test results for all production types for on Test 1 ....................................124 418 Post hoc t test results for all production types for on Test 2 ....................................124 419 Post hoc t test results for all production types for on Test 3 ....................................124 420 Number and percentage of items, by production type and participant group ...........124 421 Post hoc t test results for production type EP for items, by participant group .........125 422 Post hoc t test results for production type SP for items, by participant group .........125 423 Post hoc t test results for all production types for items for participant group E1 ...125 424 Post hoc t test results for all production types for items for participant group E2 ...125 425 Post hoc t test results for all product ion types for items for participant group P .....125 426 Post hoc t test results for all production types for items for participant group S .....125 427 Number and percentage of items, by production type, participant group and test ...126 428 Post hoc t test results for production type EP for items, by participan t group and test ....................................................................................................................................126

PAGE 10

10 429 Post hoc t test results for production type SP for items, by participant group and test ....................................................................................................................................126 51 Nu mber and percentage of correct items for all [z], by word type ..................................140 52 Number and percentage of correct items, by grapheme and word type ...........................140 53 Post hoc t test results comparing graphemes for each word type ....................................140 54 Post hoc t test results comparing word types for .....................................................140 55 Post hoc t test results comparing word types for ......................................................140 56 Cognate words, with corresponding sound in English..............................................141 57 Number and percentage of correct items, by cognate set .........................................141 58 ANOVA for cognate items .......................................................................................141 59 Number a nd percentage of correct items for cognate sets, by participant group ......141 510 Post hoc t test results for cognate set with [z] in English, by participant group ......142 511 Post hoc t test results for cognate set with [s] in English, by participant group ......142 512 Post hoc t test results for cognate set wit h [ ......142 513 Post hoc t test results for cognate sets for participant group E1 ..............................142 514 P ost hoc t test results for cognate sets for participant group P .................................142 515 Post hoc t test results for cognate sets for participant group E2 ..............................142 516 Post hoc t test results for cognate sets for participant group S .................................142 C 1 Summary of Participant Questionnaires ..........................................................................165 D 1 Words used in reading tasks .............................................................................................170 F 1 Transcription guide for raters ...........................................................................................178 H 1 Sample of participant producti on data and assigned values for and items ........181 I 1 Percentages of correct production by participant and group for each grapheme and test ....................................................................................................................................182 J 1 Production on Spanish reading task and average production on Portuguese reading tasks ..................................................................................................................................185

PAGE 11

11 L 1 Percentages of Production by Production Type for Each Word ......................................189

PAGE 12

12 LIST OF FIGURES Figure page 31 Percentage of correct items by test ....................................................................................96 32 Percentage of correct items, by test for each group ...........................................................96 33 Percentage of correct items, by group for each test ...........................................................96 34 Percentage of correct items, by grapheme for each group .................................................97 35 Percentage of correct items, by group for each grapheme .................................................97 41 Percentage of items, by test for each production type ..............................................127 42 Percentage of items, by participant group for each production type ........................127 43 Percentage of items, by test for each production type ..............................................128 44 Percentage of items, by participant group for each production type ........................128 45 Percentage of correct items on Spanish reading task and average percentage of correct items on Portuguese reading tasks .......................................................................129 46 Relationship between production on Spanish reading task and ave rage production on Portuguese reading tasks ..................................................................................................129 51 Percentage of correct items by word type ........................................................................143 52 Percentage of correct items by grapheme and word type ...............................................143 53 Percentage of correct items, by test for each cognate set .........................................144 54 Percentage of correct items, by participant group for each cognate set ....................144 E 1 Instruction slides and sample word slides from PowerPoint reading task .......................177 G 1 Rater Transcription Sheet Page 1 .....................................................................................179 G 2 Rater Transcription Sheet Page 2 .....................................................................................180

PAGE 13

13 LIST OF ABBREVIATION S < > Indicate a grapheme / / Indicate a phoneme [ ] Indicate an allophone L1 First language L2 Second language L3 Third language CLI Cross linguistic influence FL(A) Foreign language (acquisition) GPCR Grapheme phoneme correspondence rule IL Interlanguage OPM Ontogeny phylogeny model SL(A) Second langua ge (acquisition) T unit Terminal unit U Universals (part of the OPM) UG Universal grammar VOT Voice onset time

PAGE 14

14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirement s for the Degree of Doctor of Philosophy THE ACQUISITION OF ORTHOGRAPHIC PHONOLOGICAL CORRESPONDENCE RULE S IN L2 AND L3 PORTUGUESE : ERROR RESOLUTION, INTERFERENCE AND GENERALIZABILITY By Sharon Barkley August 2010 Chair: Gillian Lord Major: Linguistics The current study investigate s the acquisition of orthographic phonological correspondence rules by learners of Portuguese as an L2 or L3, with English and Spanish as the previous language(s) Acquisition is examined from the perspectives of error resolu tion, interference, and generalizability Target L2/ L3 production is shown to increase throughout the course of the semester of data collection, as interference from the L1/L2 decrease s The results support previous research finding transfer from both the L1 and L2, and interference based on language typology as well as language status. Word type (cognates, non cognates and nonce words) is not found to play as great a role in acquisition as grapheme and participant s linguistic background. It is proposed t hat examination of the acquisition of orthographic phonological correspondence rules may have the potential for highlighting more specifically the source(s) of interference in an L3.

PAGE 15

15 CHAPTER 1 INTRODUCTION AND LIT ERATURE REVIEW 1.1 Introduction Language acquisition has long been a topic of interest in the field of linguistics, initially in the area of first language (L1) acquisition (e.g., Bloomfield, 1933; Brown, 1973; Chomsky, 1959; Skinner, 1957), and later in the area of second language (L2) acquisiti on (e.g., Ellis, 2008; Krashen, 1982; Selinker, 1972; White, 2003), and beyond bilingualism (e.g., Baker, 2006 ; Grosjean, 2001) third language (L3) acquisition (e.g., De Angelis, 2007) and multilingualism (e.g., Aronin & Hufeisen, 2009; Cenoz, 2000; Ce noz & Genesee, 1998) Studies in acquisition have examined many different aspects of language such as phonology ( e.g. Hansen Edwards & Zampini, 2008) pragmatics (e.g., Kasper & Rose, 2002) discourse analysis (e.g., Boxer & Cohen, 2004), and so on. In e xamining the acquisition of an L2 or L3, the literature has often discussed the effects of a learners previous language(s) on the new language (e.g., Cenoz, 2003; Cenoz, Hufeisen & Jessner, 2001), and, in recent work in the area of psycholinguistics the effects of word similarities and differences between languages, including cognates and non cognates, have been considered (e.g., Costa, Caramazza & Sebasti n Galles, 2000; Costa, Santesteban & Cao, 2005) As far as Portuguese1 1 Portuguese is the official language of Angola, Brazil, Cape Verde, Guinea Bissau, Mozambique, Portugal and So Tom and Prncipe, as well as one of the off icial languages of East Timor, and one of the languages spoken in other areas, such as Goa and Macau. The Portuguese language ranks within the worlds top 10 most spoken languages (in terms of number of speakers), with Brazil having by far the largest numb er of speakers approximated at 180 million. is concerned, the language h as not had the long research tradition that languages such as Spanish, French and German have had (Ornstein, 1942; Holton, 1954), but its importance as a world language has increased with the growing importance of Brazil, Latin Americas biggest economy (M argolis, 2009).

PAGE 16

16 The present study brings together several of these areas, as it explores an aspect of language acquisition which has not been widely discussed in the literature: orthographic phonological correspondence rules. This study considers the acqui sition of these rules by learners of Portuguese2As is the case with many studies in language acquisition, the current research project arose as a result of classroom observations. In classes of students learning Portuguese as a foreign language, it seemed that native English speakers with high proficiency in Spanish behaved more like native Spanish speakers than native English speakers with low or no proficiency in Spanish, when it came to the treatment of certain sound symbol correspon dences. For instance, the intervocalic as an L2 or L3, examining the influence of the learners L1 (Spanish or English) and where applicable L2 on the target language and also considering the effect of different word types (cognates, non cogna tes, and nonce words) on that acquisition. It is worth mentioning that much of the literature on the acquisition of Portuguese appears to deal with impressions and observations, with relatively little empirical evidence being offered to date, making this s tudy an important contributor in the area of Portuguese acquisition. 3 2 While the orthographic phonological correspondences discussed here are the same across different varieties of Portuguese, Brazilian Portuguese is the variety taught at the institution at which the present stud y was carried out. in Brasil (Brazil) was often pronounced as the voiceless alveolar fricative [s] rather than the voiced alveolar fricative [z], the correct pronunciation in Portuguese. While this seemed a logical approach for n ative Spanish speakers, whose L1 pronounces intervocalic as [s] and where there is no phoneme /z/ it is not necessarily logical for nat ive English speakers, whose L1 does have the phoneme /z/, often pronounces intervocalic as [z] and has the country name (Brazil) pronounced with [z] (as required by the orthographic phonological correspondence rule for in English). It appeared that the non native but highly proficient 3 < > are used for orthographic representations, [ ] are used for allophones, and / / are used for phonemes.

PAGE 17

17 Spanish speakers drew on the grapheme phoneme rules of their L2, Spanish, rat her than their L1, English, when it came to the pronunciation of L3 Portuguese words containing certain orthographic phonological correspondences even when their L1 rules would be more helpful for the target language pronunciation These observations led to the overarching questions that have motivated the current study. First, to what extent are errors in foreign language pronunciation41.2 L2 and L3 A cquisition resolved over time? Second, how do the grapheme phoneme correspondence systems of the L1 and L2 influence the acquisiti on of the L3 system? Third, what is the role of different word types (cognates, non cognates, and nonce words) in the acquisition of L3 orthographic phonological correspondence rules? In order to address these questions, it is necessary to consider the re levant research in certain key areas. First, general L2 and L3 acquisition will be discussed including the notion of transfer, followed by a look at the acquisition of related languages, in particular Spanish and Portuguese. Next, acquisition of L2 and L3 phonology will be considered especially as it relates to pronunciation Thereafter, the acquisition of orthographic phonological correspondence rules will be examined with a presentation of the pertinent rules of English, Spanish and Portuguese for the sound symbol correspondences considered in the current study. Finally, previous research regarding the effect of different word types (cognates, non cognates, and nonce words) in acquisition will be considered. The fields of fore ign language acquisition5 and bilingualism6 4 Here, production of the orthographic phonological correspondence rules in question. are well established, and research in these areas has been abundant. In the literature, numerous theories and models have 5 For the purpose of this dissertation, th e terms foreign language acquisition (FLA) and second language acquisition (SLA or L2 acquisition) are used synonymously, when referring to a language that is learned after the first language (L1).

PAGE 18

18 been proposed to account for the acquisition of a second language as a whole, or the acqui sition of certain aspects of a second language (e.g., phonology), considering learners different developmental stages and topics such as transfer from the L1 to the L2 (and vice versa), fossilization of the L2, and achieving native like status. It is beyo nd the scope of this project to exhaustively review all possible SLA theories and studies hence the discussion here will restrict itself to those theories that are relevant to SLA insofar as L3 is concerned, or that are applicable to the current project. One theory of language acquisition which has received a great deal of attention for both L1 and L2 acquisition is universal g rammar (UG) which proposes that there are certain grammar principles, shared by all languages, that are said to be innate. Noam Ch omsky has been an influential figure in this area, arguing for an innate language faculty based largely on the observed ability of L1 learners to pick up language so quickly and seemingly effortlessly, and to produce grammatically correct language despite poor input, at least insofar as negative evidence is concerned that is, evidence of ungrammatical production in the input (see Chomsky, 1965). UG has been said to be made up of principles and parameters, where the principles are the aspects of language t hat all languages share (a core grammar), and parameters are the language specific settings for these universal principles. For example, all languages have vowels but the specific vowel systems of languages such as English, Portuguese and Spanish differ. T here has also been significant discussion about what part UG plays if any, in the acquisition of languages beyond the L1, particularly for adult learners (see White, 2003). Perspectives vary, from full access to UG during L2 acquisition at one end of the spectrum to no access to UG at the other end of the spectrum. It is beyond the scope of this dissertation to 6 Both simultaneous and consecutive learning of two la nguages is understood as bilingualism, for the purpose of this project.

PAGE 19

19 review all of these perspectives; it will simply be acknowledged here that UG play s some part in L2 acquisition. Within this broad UG framework li es the Ontogeny Phylogeny Model (OPM) proposed relatively recently by Major (2001) to describe and account for the acquisition process of the L2 and beyond Major explain ed the terms ontogeny and phylogeny in both biological and linguistic terms. Biologica lly, he describe d ontogeny as the life cycle of a single organism (p. 81), and phylogeny a s the evolutionary development of groups of organisms (p. 81). Linguistically, he rephrase d the difference between the two with ontogeny as the life cycle of an individuals language (p. 81) and phylogeny as the life cycle of whole languages and language types (p. 81), where the latter include d historical change, dialectal variation, language loss, and language contact phenomena (p. 81). In Majors view, a mo del of L2 acquisition should consider the components of a learners interlanguage (IL), justifying the importance of each component and describing the interaction between the components. He assume d that the IL has as its components the L1, the L2 and what he calls universals (U)7 7Major (2001) suggests that language universals comprise UG, as well as the whole set of universal language properties: learnability theory, markedness, underlying representations, rul es and processes, constraints, and stylistic variation. given the numerous studies in the literature which indicate that learners produce things which are neither part of the L1 grammar nor part of the L2 grammar (e.g., Broselow, Chen & Wang, 1998) Given these three component parts of a learners IL, Major describe d the appearance of the IL at different stages of acquisition. Initially, he claimed the idealized learner would have an IL that is equivalent to the L1 and, finally, the IL would be equivalent to the L2. In the interim, the L1 would decrease, the L2 would increase, and U, which l ay dormant initially, would increase and subsequently decrease. In addition, Major claimed that the OPM could be extended to the IL of languages learned beyond the L2 (i.e., L3, L4, L5, and

PAGE 20

20 so on). For an idealized L3 learner, for example, initially the IL w ould contain the components L1 and L2, to varying degrees depending on their stages of completion (degree s of proficiency), while finally the IL would be equivalent to the L3. Again, he claimed U would incre ase and subsequently decrease. Such a model is useful for third language acquisition and multilingualism, fields which have been drawing increasingly more attention, no doubt in part due to the realization that multilingualism is common around the world (Cenoz & Hoffman, 2003), with bi or multilingualism being at least as frequent as monolingualism, if not more frequent (Ecke, 2001; Hammarberg, 2001). The need has thus arisen to identify specifically what distinguishes third language acquisitio n from second language acquisition and multilingualism from bilingualism. Clearly, while third language acquisition has much in common with second language acquisition, in terms of theoretical frameworks for instance, there are traits which are specific t o the acquisition of a third language. The increased number of linguistic systems, which entail more possibilities of combinations and interactions between linguistic systems contribute to the complexity of third language acquisition. As well as consideri ng the differences between L2 and L3 acquisition, this new subfield considers the influence of bilingualism on third language acquisition. Studies considering cross linguistic phenomena, language use phenomena, and even the question of the ideal age at whi ch third languages should be introduced in schools are some of the areas investigated within this subfield, where the L3 studied has almost always been English (Cenoz and Hoffmann, 2003). Next, some of the differences between L2 and L3 acquisition are cons idered, as well as some of the effects of bilingualism on L3 acquisition. With L2 acquisition, the two languages may be acquired either consecutivel y or simultaneously (i.e., L1 When there are

PAGE 21

21 three languages in play, four temporal possibilities arise: the three languages are acquired simultaneously (Lx/Ly/Lz); the first two languages are acquired simultaneously before the third languag e is acquired (Lx/Ly acquired simultaneously (L1 (L1 L3 learners, then, have more language experience to draw upon than L2 learners as they have access to two language systems instead of one (obviously d epending on the temporal nature of the ir acquisition process ) Folk wisdom claims that it is easier for bilinguals and multilinguals to learn a new language than monol inguals, and this has been the subject of numerous studies, which have attempted to determine whether in fact bilinguals and multilinguals have an advantage over monolinguals when learning a new language. The literature is divided as to this matter. In a r eview by Cenoz (2003), many studies found a negative association between bilingualism and cognitive development, prior to a study conducted by Peal and Lambert (1962), in which they found that French English bilingual children achieved higher scores on cer tain verbal and nonverbal tests of cognitive ability, as compared with monolingual French or English children. Cenoz (2003) report ed that since the 1960s studies conducted to look at the effect of bilingualism on cognitive development, metalinguistic awa reness and communicative skills have shown that bilinguals: 1) have higher scores in tests of creative thinking, 2) possess a higher ability to reflect on language and manipulate it, and 3) show a greater sen sitivity to their interlocutors and make use of more va ried communication strategies. In reviewing the effect of bilingualism on general L3 proficiency, Cenoz (2003) report ed numerous studies which show ed advantages for bilinguals and several studies which d id not. Nevertheless, the tendency was toward the association of bilingualism with advantages in L3

PAGE 22

22 acquisition, when general L3 proficiency was considered. In one such study, Errasti (2003) looked at the effect of bilinguals language proficiency on their L3 acquisition, studying 155 bilingual adoles cents in a school in the Basque Autonomous Community in Northern Spain. Half of the students (78), classified in the maintenance group, had Basque as their L1, and us ed mainly Basque at home, at school and in social contexts while the other half of the st udents (77), classified in the immersion group, ha d Spanish as their L1 or us ed Spanish more than Basque outside of school (A ll participants had received instruction in Basque from the age of three while Spanish was introduced into the curriculum at the age of 3, and English at the age of 8 .) For the study, participants were required to write an informal letter and a recipe in Basque, Spanish and English, over a period of three months, at monthly intervals, with the Basque materials collected first, then the Spanish, and finally the English. Materials were gra ded using a holistic evaluation taking into account content, organization, language use, vocabulary, and mechanics of writing, and a T unit8In Basque, the maintenance group performed significantly better than the immersion group in all areas, which included overall production, fluency, grammatical complexity, lexical complexity, and accuracy. In English, the maintenance group again performe d better than the immersion group in all areas, but this time significant differences were only found in the categories of overall production and in fluency. In Spanish, there were no statistical differences between groups in any of the categories. Conside ring the interaction of various factors, the evaluation, where fluency was calculated based on number of words per T unit, grammatical complexity based on the number of clauses per T unit, lexical complexity on the number of lexical words per T unit, and accuracy based on the number of errors (semantic, morphosyntactic, alphabetical, and lexical) per T unit 8 The term T unit was coined by Hunt (1965) and is essentially the shortest grammatical unit that can stand alone (a dominant clause and its dependent clauses). It has been used to analyze discourse a nd writing in both the L1 and L2.

PAGE 23

23 author found two interesting positive associations: 1) high levels of competence in Basque and Spanish were related to a high level of competence in L3 English; and 2) the measurements in Basque, Spanish and English, for the maintenance group, were highly correlated, suggesting that writing in each particular language was not an independent process. The first association confirm ed the Threshold Hypothesis (Cummins, 1979), which state d that the students who would benefit most from their bilingualism would be those with high levels of competence in the two languages; the second association support ed the Linguistic Interdependence Hypothesis (Cummins, 1979), which state d that knowledge in one language could be positi vely transferred to another language. T he tendency toward the association of bilingualism with advantages in L3 acquisition, when general L3 proficiency was considered, was not necessarily found to be the case when specific aspects of L3 proficiency were c onsidered (Cenoz, 2003). Some studies in phon ology (e.g., Davine, Tucker and Lambert, 1971) found no differences between bilinguals and monolinguals, while others did (e.g., Cohen, Tucker and Lambert, 1967; Enomoto, 1994). Studies in the area of syntax, mo rphology and the lexicon presented similar mixed results, leading to the conclusion by Cenoz (2003) that bilinguals achieved more favorable results in studies that considered general L3 proficiency rather than proficiency in very specific L3 aspects. The g eneral suggestion, then, is that there is no negative effect of bilingualism on L3 acquisition, and that in som e cases the effect is positive. The present study considers the effects of learners previous languages on the L3 in a very specific area the a cquisition of orthographic phonological correspondence rules so its findings will be interesting to consider in light of the mixed results found in previous studies, insofar as advantages for bilinguals are concerned.

PAGE 24

24 Another topic widely discussed in th e literature on L3 acquisition is the notion of interference or transfer from the learners L1 and L2. While this is to be expected, given a model like Majors (2001) OPM, where the beginning state of a learners IL for the L3 is made up in part by the L1 and in part by the L2, what is perhaps of most interest is the nature of interference. Perhaps counter intuitively, some studies have shown that during L3 production, learners often inadvertently produce d L2 forms, rather than L1 forms (Dewaele, 1998; Herw ig, 2001; Selinker & Baumgartner Cohen, 1995). Faerch and Kasper (1986) suggest ed that these learner mistakes did not appear to be used intentionally (for instance, to bridge a lexical gap), but rather they appear ed to result from the learners inability to suppress the intrusive language. According to a review by Murphy (2003), cross linguistic influence9Proficiency, for instance, was generally agreed to inversely influence language transfer; that is, the low er the learners proficiency in the L2, the greater the likelihood of language has been approached in different ways in the literature. Some believe there is no significant difference between the L1 and L2 acquisition processes, cl aiming that errors merely result from learners testing of hypotheses (Corder, 1967; Krashen, 1983), while others find transfer to be a natural and even necessary part of acquisition (Selinker, 1972; Gass, 1983, 1984). Murphy (2003) considered two sets of variables that could affect cross linguistic influence (CLI): learner based variables and language based variables. The learner based variables include d proficiency, amount of target language exposure and use, language mode, linguistic awareness, age, educ ational background, and context. The list of language based variables included language typology, frequency, word class, and morphological transfer. 9 Some prefer this term to the terms transfer and interference, since it is more encompassing, including the notions that transfer may be positive as well as negative, and that transfer may be bidirectional (L1 This preference for terms does not concern the present study, thus no discrimination is made heretofore between the terms transfer, interference and cross linguistic influence.

PAGE 25

25 transfer from the L2 to the L3 (Odlin, 1989; Poulisse & Bongaerts, 1994). According to Murphy, proficiency was a prominent feature in the literature on L3 acquisition, and it wa s necessary to take into account the level of proficiency in all of a multilinguals languages. Hammarberg (2001) on the other hand, believe d that for there to be transfer from the L2 to the L3, there must be a certain degree of proficiency in that langua ge, while Shanon (1991) note d that the source of transfer was often the more recently acquired L2, which was weaker than the L1 in terms of proficiency. Language exposure, Murphy point ed out, behave d similarly in the acquisition of L2 and L3: in L2 acquis ition, as exposure to the L2 increase d, transfer from the L1 decrease d; in L3 acquisition, as exposure to the L3 increase d, transfer from the L2 (and presumably the L1) decrease d. This observation is consistent with Majors (2001) OPM, which has an initial IL state for the L3 as consisting of parts of the L1, parts of the L2 and universals (U), and predicts that the L1 and L2 will decrease, U will increase and then decrease, and the L3 will increase. Interestingly, Dewaele (2001) claim ed that transfer from the L1 for L2 speakers decline d more quickly than transfer from the L2 (and, presumably, the L1) for L3 speakers, and Murphy (2003) suggest ed that this might be due to the more complex linguistic system of the multilingual than the bilingual, again consistent with Majors model. In terms of language mode, Grosjean (2001) propose d a continuum in which the base language (L1) was always activated, while the guest language (L2) bec ame increasingly more active depending on the language mode (monolingual to bilingual). Murphy (2003) suggest ed that this model might be adapted for L3 learners, whose base language (L1) would always be active but whose guest languages (L2 and L3) would be active to varying degrees She claimed that the L1 would be easier to deact ivate than the L2, though this claim seem ed to be

PAGE 26

26 contradicted somewhat by her subsequent discussion about the typological similarity between a learners languages. Language typology is explored later in this chapter, in section 1.2.1. Another way of acco unting for these intrusions from one no n native language (L2) into another ( L3 ) was put forwards by De Angelis (2005), who propose d a cognitive process by which learners transfer lexical items from one non native language to another without being aware o f it, calling such a process a system shift. She suggest ed that such a shift m ight occur in three stages: 1) first, the learner transfers a lexical item from one non native language to another (presumably due to a knowledge gap), believing that the item belongs to the source system only; 2) next, the learner associates the lexical item with the guest system, believing that it may belong in both the guest system and the source system; and 3) the learner believes the lexical item belongs to the guest system no longer recognizing the source of her knowledge as being found in the source system. The first stage has been widely discussed in the literature, and c ould certainly be attested to by any foreign language instructor. The hypothesized second and third stages were tested in three separate studies conducted by De Angelis (2005). In the first study, a French Canadian speaker of Italian, with previous knowledge of English and Spanish, was interviewed in Italian about her personal experiences in London and o ther foreign countries. Her interview was analyzed for instances of CLI from French or Spanish. Six months after the initial data collection, the participant was interviewed again, about the specific problematic lexical items. The participant was first ask ed if she was familiar with the English words (read to her one at a time), and then she was asked to translate them into Italian. Afterwards, the participant was asked if she was familiar with the correct target Italian words. For the 12 troublesome lexica l items identified, the participant produced a Spanish or Spanish influenced word at both interviews on eight occasions. Most of the 12 words share d

PAGE 27

27 similar forms in Spanish and Italian, which might account for the overt lexical transfer from Spanish, alth ough in some cases the participant produced a Spanish Italian blend, suggesting previous familiarity with the correct Italian word. De Angelis (2005) suggest ed that this provide d some support for the notion that the participant believed she was producing a n Italian word, leading the researcher to question whether in some cases learners are even aware of the source of their knowledge in the original linguistic system. In the second study, De Angelis (2005) participants were 10 L1 English university students with low proficiency in Spanish, enrolled in Italian language classes (of different levels) The learners were given a list of English words to be translated into Spanish (all non cognates in English, Spanish and Italian), followed by a text in English to be summarized in writing in Italian, followed by the same list of English words, this time to be translated into Italian. (Participants were asked not to look at previous sections after completion, and were carefully monitored to ensure compliance.) Resu lts showed learners providing a Spanish word in the Italian translation task, in some cases, but then failing to provide a translation for the same item in the Italian translation task, and vice versa, suggesting a lack of knowledge concerning the original source system. In her third study, De Angelis (2005) had an L1 English speaker, with previous knowledge of Italian and Spanish (as well as Latin, French and German), keep a diary of those things which confused him most about learning and having to speak t wo non native languages.10 10 The researcher had been approached by the participant, who was reporting some confusion between Spanish and Italian. He was living in Spain and studying Spanish and Italian, having previously learned Italian informally while living in Italy. The diary illustrated the participants struggles in keeping the two languages separate, with instances where he knew the words in both languages but didnt know to which language a word belonged.

PAGE 28

28 De Angelis (2005) suggest ed two f actors that might lead a learner to effect a system shift: 1) what she call ed perception of correctness (p. 11), where a learner quite easily perceives that words in the native language are not correct target language words, and thus is able to block tra nsfer of these words, while less easily blocking transfer of words from another non native language which learners perceive to be correct (depending, of course, on levels of proficiency in the non native languages); and 2) association of foreignness (p. 11), where a learner comes to associate (all) non native languages as foreign, creating a cognitive link between the foreign languages which does not exist between the native language and a foreign language. De Angelis believe d that these two factors might help to explain why L3 learners have been observed to block transfer from their L1 in favor of transfer from another non native language. The present study considers the nature of interference (from the L1 and/or the L2) by examining the production of ort hographic phonological correspondences in the L3 by learners with different linguistic backgrounds (L1 English with L2 Spanish, and vice versa), and of differing levels of proficiency in the L2. The results will be interesting to compare with those observa tions outlined above to see where the intrusions are coming from (i.e., the L1, a more proficient L2, or a less proficient L2) Despit e differing views as to how intrusions end up being part of a non native language be they from another non native lan guage or from the native language ; be they due to low or high proficiency in a non native language cross linguistic influence is undeniable, especially in the case of similar languages (Carvalho, 2002). For this reason, it is important to consider now wh at the literature has noted with regard to related languages. 1.2.1 Acquisition of T ypologically S imilar L anguages Language typology is a rather broad term which refers to language classification, and there are many ways in which to classify languages (Fro mkin, Rodman & Hyams, 2007). Syntactic

PAGE 29

29 typology, for one, considers word order in a particular language or group of languages, such that languages with subject verb object (SVO) word order (e.g., English, Spanish, and Portuguese) would be considered typolo gically similar to each other but different from languages with, for example, subject object verb (SOV) word order (e.g., German, Dutch, and Japanese). From a phonological perspective, typology may refer to the syllable structure permissible within a langu age: from a simple consonant vowel (CV) and vowel (V) structure (e.g., Japanese) to more complex structures including consonant clusters at the beginning and end of syllables (e.g., English, Spanish, and Portuguese). Typology may also refer to the origin o f a language, or the family to which a language belongs. For instance, Spanish and Portuguese are derived from Latin, while English, German and Swedish are derived from an earlier form of Germanic known as Proto Germanic (Fromkin, Rodman & Hyams, 2007). W ith regard to the effect of language typology on the acquisition of L2 and L3, it seems that similar typology is often advantageous, being a source of facilitation or positive transfer Considering similarities due to language family, Orr (1987, as cited b y Jarvis & Odlin, 2000) looked at the acquisition of prefixation in Chichewa, a Bantu language, and found an advantage for L1 speakers of Ngoni, a nother Bantu language, compared with learners of Chichewa with Gujarati as their L1, an Indo European language that does not have the same Bantu prefixing system. Considering languages that were related (and unrelated) in their morphological structures, Jarvis and Odlin (2000) examined L1 Finnish and L1 Swedish learners of English as an L2, in their use or non us e of prepositions with the verbs take sit and put in a task where they were asked to write a narrative for a short silent film they had just watched. Swedish and English were considered typologically similar in this study, with regard to their free, prepositional

PAGE 30

30 morphology, while Finnish was considered typologically dissimilar, with bound, agglutinative morphology for spatial references. The authors found that learners with the typologically dissimilar L1, Finnish, omitted prepositions in English in all of the spatial contexts examined, while those learners with the typologically similar L1, Swedish, did not omit English prepositions in any of the contexts. Cenoz (2001) also found evidence of greater cross linguistic influence between L2/L3 English and a typologically similar L1 or L2 than a dissimilar one. In a study of L1 Basque and/or L1 Spanish elementary and secondary school children learning English11Studies considering the effect of the typological similarity or dissimilari ty of an L1 and L2 on the L3 have shown mixed results. On the one hand, facilitation has been found between an L2 and L3 when they were typologically similar. For instance, Ecke and Hall (2000, as cited in Ecke, 2001) found that L1 Spanish speakers with L2 English and L3 German had much stronger influence from the L2 than from the L1 in written production of the L3. In other cases, no facilitation was found between typologically similar languages. For example, in a study of German prepositional verbs, where participants had a variety of L1, L2, L3 and sometimes L4 backgrounds, Gibson, Hufeisen and Libben (2001) examined the effect of similar and dissimilar previous languages on learners production of German prepositions in a fill in the blank task using a story telling task in English, Cenoz found greater transfer from Spanish, an IndoEuropean l anguage like English, than from Basque, a non Indo European language. These studies suggest more positive cross linguistic influence when the L1 is typologically similar to the L2. 12 11 All participants were native speakers of Basque and/or Spanish, had Basque as the language of instruction at school, and had Spanish and English as school subjects. Gibson et al. found that those participants with L1s structurally similar to German did not 12 Verbs were given and participants had to supply the following preposition. For example gehren zu (belong to ), and sprechen ber (talk about ).

PAGE 31

31 perform differently from those with structurally dissimilar L1s, nor did having English as a foreign language help German learners, even when the German preposition would be a translation equivalent of English. The current study considers Spanish and Portuguese, two languages which are closely related in terms of language family, many structural aspects and lexical items. In the case of these two languages, most of the literature would agree that knowledge of one language generally facilitates the acquisition of the other language, although it may also prove to be a hindrance in some respects. To follow is a review of the literature that considers the relatedness of these t wo languages. 1.2.2 T he Relatedness of Spanish and Portuguese Garrison (1979) highlight ed the similarities between Spanish and Portuguese, observing that there are few languages as closely related, and that there are fundamental similarities that transcend even regional dialects of the two languages. In an overview of a presentation on Portuguese given to Advanced Spanish students, he suggest ed that around 80% of Portuguese words have Spanish cognates, going on to point out some of the systematic difference s between Spanish and Portuguese orthography (e.g., the grapheme <> which exists in Portuguese but not in Spanish ), phonology (e.g., nasal vowels in Portuguese Portuguese monophthongs for certain Spanish diphthongs) lexical items in the form of non cognates (e.g., street is calle in Spanish and rua in Portuguese) and false cognates (e.g., apellido is Spanish for last name while apelido is Portuguese for nickname), and grammatical items, such as gender differences on nouns (e.g., masculine mensaje i n Spanish and feminine mensagem in Portuguese, message) verb conjugations (e.g., the first person singular conjugation of the verb poder to be able to, is puedo in Spanish and posso in Portuguese) verb tenses (e.g., Spanish does not have the persona l (conjugated) infinitive that Portuguese has) and so on.

PAGE 32

32 Holton (1954) also highlighted the relatedness of Spanish and Portuguese. In an effort to promote the great literary works in the Portuguese language, he suggest ed that Spanish speakers might easily learn to read Portuguese with minimum effort, given an understanding of certain basic differences between the two languages. He delineate d some of the orthographic and phonological differences between Spanish and Portuguese, as well as some grammatical differences (focusing on articles, verbs and pronouns), with the intent of showing how simple it would be for a Spanish speaker to develop a recognition skill like reading. He clarifie d that such a method would not suffice for the acquisition of an active c ommand of the language, highlighting that speaking and comprehension of spoken Portuguese would require a more concerted effort. Nonetheless, the fact remains that the languages do share considerable structural and lexical similarities. Resnick (1945) illustrate d the advantages and disadvantages of having a background in Spanish when learning Portuguese by listing examples of words which are cognates, non cognates and false cognates in the two languages. He also highlight ed that cognates ( including false co gnates) often differ in the languages in terms of spelling and pronunciation, something which learners must bear in mind during the acquisition process. Azevedo (1978) observe d that university students acquainted with one language (Spanish) experience d fa cilitation in the acquisition of a new language (Portuguese) due to the structural similarities between the two languages, yet at the same time these precise similarities often cause d problems for the learners, leading to transfer of several Spanish featu res, such as morphological, phonological, lexical and syntactic markers, which d id not belong in Portuguese. In terms of phonological errors, Azevedo specifically pointed out, among other things, the feature under investigation here: the tendency of Spanis h speakers to devoice the voiced alveolar

PAGE 33

33 fricative [ z ] in Spanish Portuguese cognates like zero (zero) and casa (house). Additionally, Azevedo highlight ed the fact that although Portuguese instructors might be aware of Spanish interference, it was dif ficult for them to correct the problem, because the diagnosis of the problem is done, more often than not, in an impressionistic and unsystematic manner (p. 19), rather than being based on empirical evidence, thus highlighting the need for further resea rch on learners production in order to enable targeted correction strategies. Tarquinio (1977) also mentioned some of the difficulties with pronunciation that arise for learners of Portuguese. As a native Brazilian Portuguese speaker and an instructor of Portuguese and Spanish, she was particularly attuned to common pronunciation mistakes which result ed from interference (particularly from Spanish), including the devoicing of the intervocalic , which is voiced in Portuguese. She went so far as to state that these interferences hurt the ears of the native listener because they are foreign (p. 82), and suggest ed that students be alerted early on as to these common interferences by explicitly drawing their attention to them While much of the literature on Spanish and Portuguese highlights the mutual intelligibility of the two languages, there are those who maintain that the spoken languages are quite different in pronunciation, resulting in reduced understanding, particularly for Spanish speakers trying to understand Portuguese (Ellison & Andrews, 1969; Jebsen and Biel, 1986; Timberlake, 1989). These observations about the similarity between the two languages, or lack thereof, prompted Jensen (1989) to conduct a study looking at extent and direction of i ntelligibility, as well as the effect of certain non linguistic factors (age, sex, education, attitude and experience with the other languages). Thirty nine native speakers of Portuguese listened to four recorded texts read in Spanish, while 32 native spea kers of Spanish listened to four recorded

PAGE 34

34 texts read in Portuguese.13Jordan (1991) describe d the differences between Spanish and Portuguese as mainly morphological and phonological, as she ma de a case for the use of a contrastive method in the teaching of Portuguese to speakers of Spanish ( or, presumably, vice versa). Jordan claim ed that there is little point in a contrastive course if the goal is simply for speakers to understand each other and make themselves understood, since these two tasks are achieved relatively effortlessly with littl e or no instruction in the other language ( see Jensen, 1989). However, for a speaker to reach a certain level of proficiency in writing and speaking the non native language, Jordan suggest ed the use of a contrastive method, highlighting the basic phonologi cal and Each group had to answer five multiple choice comprehension questions per text, written in their native language, and complete a short questionnaire asking about the non linguistic factor s considered by the study. The native Portuguese speakers performed significantly better than the native Spanish speakers on three of the four texts; the text on which native Spanish speakers performed better than native Portuguese speakers related to Ecua dor, and it was postulated that the Spanish speakers, all of whom were from Latin America, might have had more familiarity with the subject than the Portuguese speakers, all of whom were from Brazil. Of the non linguistic factors, the only one found to significantly and positively affect participants scores was experience with the other language, which came as no great surprise. In this study of passive listening to recorded voices, Jensen found Spanish and Portuguese to be mutually intelligible, but only at a level of about 50% to 60%. Additionally, the results support ed the common belief that Portuguese speakers understand spoken Spanish better than Spanish speakers understand spoken Portuguese, although the difference was not staggering, and m ight be att ributable to individual factors. 13 Participants with extensive experience of the other language (p. 850) or who were of foreign (e.g., European, Asian or North American) background (p. 850) were eliminated from analysis.

PAGE 35

35 morphological differences, and particularly the few exceptions to the general rule. Given that learners w ould make comparisons between the languages anyway, Jordan believe d that it would be helpful to make this process a conscious one, thereby limi ting the erroneous assumptions and generalizations that students might make. In a qualitative examination of a class of Spanish speakers learning Portuguese, where she used a reference book designed to cover the principal phonological and morphological dif ferences between the two languages (making no mention of similarities), Jordan found that students concentrated on certain differences while having problems with other discrepancies, one of which was the intervocalic which was often not voiced. To over come these problems, Jordan suggested the need for increased exposure to the non native language, particularly in the areas of listening and reading, and also for as much communication as possible to be in the target language in the classroom, in order tha t learners might overcome the fear of making mistakes. Carvalho (2002) discusse d the inevitability of transfer, in particular from a known language to a similar language being learned, where this transfer m ight be positive or negative. In the case of Span ish and Portuguese, where there is a high degree of mutual intelligibility14 14 Jensen (1989) found this to be approximately 50 60% in a recorded listening task, but Carvalho (2002) suggested that this percentage might be higher in a conversational setting, where speakers could negotiate meaning and make use of visual cues. even between monolingual speakers of the two languages, a primary concern of instruction is early fossilization. Where learners are able to achieve a high level of communicative ab ilit y early on, Hadley (1986) stated that fossilization is more likely to occur if learners see no reason to improve their interim grammar and decide that it is adequate to serve its needs (p. 268). Thus, Carvalho (2002) claim ed that instruction for spea kers of a typologically similar language must include activities which emphasize grammatical correction and promote metalinguistic

PAGE 36

36 reflection, and she call ed for further studies to contribute empirical data which might reveal and contrast different stages involved in the acquisition of Portuguese by Spanish speakers. In this section, particular consideration has been given to the relatedness between Spanish and Portuguese and it has been noted that phonological differences between the se two languages often pose problems for learners. The current study considers the role that the typology of learners previous languages plays in the acquisition of the L3, as it examines interference from English and Spanish in participants production of Portuguese. Here, Sp anish is considered to be closer to Portuguese (more typologically similar) than English given the language family, structural aspects and lexical items that Spanish and Portuguese share In the next section, the acquisition of L2 and L3 phonology will be discussed, specifically as it relates to the acquisition of pronunciation 1.2.3 Acquisition of L2 and L3 P honology In some of the earlier approaches to second and foreign l anguage pedagogy, pronunciation and oral skills in general were not stressed as hi ghly as listening skills (Lazaraton, 2001). With the advent of more communicative approaches, the importance of speaking has increased although focus on pronunciation has still been somewhat lacking Blanche (2004) noted that the Communicative Approach does not concern itself so much with accent reduction as with intelligibility (p. 178), and Keys (2000) stated that this approach encouraged a tendency to leave pronunciation matters to one side (p. 91). Goodwin (2001) describe d pronunciation as a critic al element because it is the language feature that most readily identifies speakers as non native (p. 117). Poor pronunciation, as perceived by native speakers, may result in the non native speaker feeling embarrassed and inferior (Goodwin, 2001), while good pronunciation in a foreign language does not draw the attention of a native speaker away from what is being said to how it is being said (Hockett, 1950).

PAGE 37

37 While the place of pronunciation in pedagogy may be debatable it is clear that pronunciation poses problems for many learners (e.g., with relation to the acquisition of Portuguese pronunciation, see Azevedo, 1978 ; Tarquinio, 1977; Jordan, 1991) and as such, pronunciation is an area of FL acquisition which has led to many studies exploring the nat ure of production I t is beyond the scope of this dissertation to review exhaustively the studies which have looked at the acquisition of L2 and L3 phonology, since much of the work which has been carried out in these fields is, at best, only tangentially related to the present study, which looks at the role of orthography in the acquisition of pronunciation. (For a review of theo retical approaches to phonology and current trends in L2 phonology see Major, 2001, 2008, and Hansen Edwards and Zampini, 2008.) However, by way of example, a few of the studies which have examined the influence of learners previous language(s) on their acquisition of pronunciation will be presented here. Major (2001) states that there has been much research done in L2 phonology l ooking at claims (from Contrastive Analysis) that similar phenomena are more difficult to learn than dissimilar phenomena, with the psycholinguistic reason for this seeming to lie in the fact that gross differences are more often noticed, due to perceptua l saliency, [than] minimal differences (p. 37). Flege (1986) state d that, given the L1 as a starting point for the L2, learners could categorize L2 phones into three categories: identical, similar, and new. New phones in the L2 are those which hav e no counterpart in the L1; similar phones are those where the sounds in the L1 and L2 are acoustically similar but not exactly the same; identical phones are those which share all acoustic properties in the two languages. Flege (1987) hypothesized tha t new phones in the L2, with no counterpart in the L1, would be easier to

PAGE 38

38 acquire than similar phones, with a counterpart in the L1 that differ ed phonetically in some minimal way. In his study, Flege (1987) examined Voice Onset Time (VOT) in the production of similar phones (/u/ and /t/) and a new phone (/y/) by bilingual English French speakers and monolingual English and French speakers. His participants included: 1) French and English monolinguals (to provide an idea of phonetic norms in the t wo languages), 2) three groups of American English speakers who had learned French as adolescents or young adults (the groups differed in their level of proficiency in French), and 3) French speakers who were highly proficient in English. Participants carr ied out two tasks, both based on lists of phrases provided in French and English, containing /tu/ and /ti/ in English ( two and TV ), and /tu/ and /ty/ in French ( tous all and tu you). In the first task, participants simply read the lists of phrase s; in the second task, they were required to produce original, complete sentences based on the phrases they had just read in isolation. Results for /t/ production showed that the greater the learners proficiency in L2 French, the closer their production w as to that of French monolinguals, both for English /t/ and French /t/. The group of L1 English speakers highly proficient in French produced English /t/ intermediate to the norms of the L1 and L2. Similarly, the group of French speakers highly proficient in English produced both English and French /t/ intermediate to the norms of English and French. These results seem ed to suggest that the greater the experience in two languages, the less difference there is between the L1 and the L2 forms. As for /u/ production, the L1 English speakers consistently produced English /u/ similarly to monolingual English speakers. However, the L1 French speakers failed to produce French /u/ according to the norm set by the French monolinguals. Thus, learning French did not ap pear to affect the English speakers production of L1 /u/, but learning English did affect the French speakers production of

PAGE 39

39 L1 /u/. Finally, for /y/ production, only the group of English speakers with the lowest level of proficiency in French differed si gnificantly from the group of French monolinguals. This result showed that, with the new phone (/y/), where there was no possibility of an approximation to an English (L1) counterpart, the learners of French were able to come close to producing an authen tic L2 sound. Baker and Trofimovich (2005) examined groups of L1 Korean speakers learning English as an L2 (children and adults), in order to determine how age of acquisition influence d the organization of learners phonetic systems. The researchers carrie d out a picture naming task to elicit six English vowels in 18 CVC monosyllabic words and five Korean vowels in 10 disyllabic words, first collecting baseline data from monolingual Korean children and adults, and monolingual English children and adults, th en collecting data from the L1 Korean speakers learning English as an L2, with the latter participants being divided into early bilinguals (those who learned the L2 before the age of 15) and late bilinguals (those learning the L2 after the age of 15). Anal yzing the production data acoustically, the researchers compared the production of the bilinguals with that of the monolinguals, as well as comparing the production of the early bilinguals with that of the late bilinguals. Results showed that early bilingu als produced different acoustic realizations of the English and Korean vowels, while late bilinguals produced English vowels that were often colored by the L1. In fact, the more acoustically similar the vowels in the L1 and L2 were, the more likely the coloring of the L2 vowels by the acoustic properties of the L1 vowels, such that late bilinguals only produced acoustically different sounds when the L1L2 pairs of sounds were very different. The researchers also observed that length of exposure to the L 2 in late bilinguals was significant, in that learners tended to produce L2 sounds as L1 sounds more in the early stages of acquisition than in later stages of acquisition.

PAGE 40

40 Wade Woolley (1999) observe d that a learners L1 phonological system constrain ed th e learners ability to perceive and produce sounds in the L2, citing several studies to illustrate this point: 1) a study by Werker and Tees (1984) showed L1 English speakers inability to distinguish between Hindi dental and retroflex stops, since this di stinction does not exist in English; and 2) studies by Goto (1971) and Yamada & Tohkura (1992) show ed L1 Japanese speakers learning English having problems with the /l/ /r/ contrast which exists in the L2 but not in the L1. It is clear from the se studies t hat L1 phonology has a considerable role to play in L2 phonology, and transfer from a learners previous language is an important factor to consider in the acquisition of L2 pronunciation. Further, Muller and Muller (1968) observe that transfer is particul arly problematic when a written stimulus is used to evoke an oral response. Th us, it is important to take the written form into account in the discussion of the acquisition of pronunciation. Th e correspondence between orthography and phonology is at the he art of the present study, and the acquisition of such correspondence rules will be discussed next. 1.3 Acquisition of L2 and L3 Orthographic Phonological Correspondence Rules In the field of L2 and L3 acquisition, few studies have focused on the relations hip between orthography and phonology, and more specifically how acquisition of the two is influenced by a learners previous language(s) This area merits further attention however, because s oundsymbol correspondences are important not only for effectiv e reading and writing skills, but they can also relate to good pronunciation (Olshtain, 2001). Many studies looking at the relationship between orthography and phonology have focused on English as an L2 or L3 and concentrated on differences arising as a re sult of different L1 scripts, that is, alphabetic and non alphabetic scripts.

PAGE 41

41 Koda (1999) describe d orthographic structures as varying along two dimensions: 1) the fundamental unit of orthographic representation, and 2) the depth of representation. The ba sic unit of representation in alphabets is the phoneme, while the morpheme is the basic unit in logographies. With alphabets, readers must systematically analyze component letters and letter clusters within a word, and reading competence requires that read ers realize that written symbols correspond to speech units. With logographies, it has been argued that readers arrive at phonological information through whole word lexical retrieval rather than through word internal analysis (Gleitman, 1985). Orthographi c depth, the second dimension described by Koda, relates to the degree of regularity found in soundsymbol correspondences. Shallow orthographies, like Spanish and Portuguese, have a high degree of orthographic phonological regularity, while deep orthograp hies, such as English, have much less consistent correspondences, as evidenced by the orthographically similar yet phonologically dissimilar related words anxious and anxiety With these differences in mind, Koda (1999) examined adult learners of L2 English, and explored the effect of different L1 backgrounds (alphabetic and logographic) on the learners intraword sensitivity. The studys participants were 20 native speakers of Chinese, chosen because of their experience with a logographic script, and 20 n ative speakers of Korean, selected because of their experience with the non Roman alphabetic script Hangul. The two groups were comparable in terms of their length of stay in the US (less than 6 months), the type and length of instruction they had had in E nglish in their own countries (grammar/reading methods in high school), and their scores on listening and reading sections of the Test of English as a Foreign Language (TOEFL). The participants completed two tasks: an orthographic acceptability judgment t ask and two decoding t asks. For the first task, a series of 40 nonsense words were created by rearranging real

PAGE 42

42 words, such as double and report, to yield legal strings (according to English phonotactic rules) like boudel and troper, as well as ille gal strings such as ebdluo and tproer. Participants were asked to judge the orthographic acceptability of these nonsense words, being allowed to spend as much time as necessary on the test.15Wade Woolley (1999) also examine d the orthographic and phonological effects of different L1 scripts on word reading in the L2, in a study of L1 Russian and L1 Japanese lear ners of English as an L2. Participants undertook seven tasks which included a mixture of standardized For the second task, two decoding activities were used. The f irst required that participants read aloud 50 pseudoEnglish words, with some allowance being made for certain non native pronunciation (such as devoicing of final /b/ by Korean speakers). The second decoding t ask was a homophone judgment test, where 30 re al English words were presented visually paired with orthographically legal strings, with half of the pairs being homophonic (e.g., please pleeze) and half non homophonic (e.g., dream- draim). Participants indicated their judgment of these pairs, by circling S for same or D for different. It was predicted that the Korean learners, given their experience with another orthographic script and ensuing intraword sensitivity, would outperform the Chinese learners on both tests. In fact, the Korean learners did outperform the Chinese learners in the orthographic acceptability test on those nonsense words with illegal strings, suggesting that their L1 intraword sensitivity might have been extended to their L2. However, there were no significant differences be tween learners on the decoding tests, possibly resulting from the fact that English has two representational properties for its orthography morphemes and phonemes. Koda suggest ed that learners using logographic strategies might have had less success with a phonologically shallow orthography, an observation requiring empirical testing. 15 Presuma bly, Koda used the term orthographic acceptability to refer to phonotactic acceptability, since the illegal strings violate English phonotactic rules rather than, specifically, orthographic rules.

PAGE 43

43 and experimental tests involving reading comprehension and vocabulary, reading of isolated words (previously seen and unseen), matching of aural stimuli w ith a correctly spelled word, distinguishing between visually similar (phonotactically permissible and non permissible) English pseudo words repetition of aurally presented pseudo words and phoneme deletion Japanese learners were found to perform better on tasks which required recognition of correct or permissible orthography (real and pseudo words respectively), while Russian learners performed better on the phoneme deletion task. These results suggest ed that Japanese learners rel ied less on phonology than orthography in reading, as the researcher had anticipated, due to the learners non alphabetic L1, lending support to the idea that there are transfer effects to be found in L2 reading. Relatively few studies have dealt specifically with the acquisition of Portuguese pronunciation, and particularly as it relates to orthography. One such study was conducted by Defior, Martos and Cary (2002) who examined children learning to read in their L1 Portuguese or Spanish. Although both of these languages are co nsidered to have shallow orthographies16 16 In a shallow orthography, the grapheme phoneme correspo ndence rules (GPCRs) have highly consistent rules, while the GPCRs in a deep orthography, such as English, have inconsistent and unpredictable rules. the grapheme phoneme correspondence rules (GPCRs) in Portuguese are less consistent than those in Spanish. In their study, Defior et al. (2002) examine d the extent to which differences between the two orthographies impact ed reading strategies, where these we re divided into two types: 1) phonological, sublexical, or indirect; and 2) visual, lexical, or direct. (Strategies of the first type rel ied more on GPCRs to assemble the pronunciation of lexical items while strategies of the second type require d a reader to use direct access to the lexicon in order to retrieve an items pronunciation.) Given the greater asymmetry and complexity in Portuguese than in Spanish, the authors hypothesized that Spanish children would per form better (i.e., faster and with greater

PAGE 44

44 accuracy) than Portuguese children on nonce words. In their experiment with 120 Spanish and 94 Portuguese children in grades one through four, participants read a list of numerals (2, 3, 4, etc.), number words ( do is trs, quatro17Across all grades and both orthographies, participants were significantly slower with nonce words than with numerals and number words. Addit ionally, results confirmed the authors hypothesis, as the Portuguese children were slower with nonce words than the Spanish children. Spanish children also read number words faster than Portuguese children, but there were no significant differences found between the two orthographies with numerals. In terms of errors, there were no significant differences found between grades or orthographies for numerals or number words. The pattern of errors in reading nonce words was different for the two orthographies: Portuguese children steadily decrease d in number of errors from grades one to three, while Spanish children maintain ed a stable rate of errors from grade two on. Additionally, Portuguese children made significantly more phonological errors than Spanish children (except at grade three). Spanish children also made fewer lexical errors initially than Portuguese children, although this distinction disappear ed in grade two, only to reappear in grade four, where Portuguese children made fewer lexical errors than Spanish children. The increase in etc.), and nonce words ( nois ns datro, etc.). Errors were divided into two types, in order to disclose different types of reading strategies (i.e., indirect phonological vs. direct lexical): 1) phonological errors, where deletion, sub stitution or some other inversion of consonants and/or vowels led to the production of another nonce word; and 2) lexical errors, where errors led to the production of a real word. Thus, phonological errors were indicative of a failure in the use of the in direct phonological strategy, while lexical errors reflect ed a failure in the use of the direct, lexical strategy. 17 Dois (two), trs (three), quatro (four), etc.

PAGE 45

45 lexical errors between grades three and four, found for both orthographies, might be indicative of a change in reading strategies: in grade three, Spanish and Portuguese children might be turning from an indirect, phonolo gical strategy to a direct, lexical one In summary, the participants generally performed better on numerals and number words than nonce words, and the children with the simpler orthography (Spanish) performed better on nonce words than those with the more complex orthography (Portuguese). Muller and Muller (1968) also examined the influence of orthography on the acquisition of pronunciation in Portuguese, but with L2 learners rather than L1 learners In their study of 140 high school students receiving ins truction in Portuguese, they consider ed : 1) whether the exposure of the written form interfere d with the acquisition of pronunciation; and if so, 2) which letters or letter combinations we re most likely to cause interference. Participants were divided into two groups: those who were denied access to the written form during the first four weeks of instruction in Portuguese, and those who were allowed access to the written form but without explicit instruction on the relationship between orthography and phono logy. After four weeks of instruction, participants were given picture cue sheets and asked to produce sentences which had formed part of drill materials used during the instruction phase. The participant utterances were recorded and later listened to and judged by raters who were both native speakers of Portuguese and language instructors using a three point scale18 18 The criteria for judging the participants product ion items (sentences) are not included by the authors, thus it is not possible to know whether assessment was global or specific. However, given that the authors go on to postulate about GPCRs, it would seem that assessment of specific sounds would be nece ssary in order to allow an informed discussion of those GPCRs which were found to be most problematic for their participants. Results indicated that the group of participants who had had access to written materials performed significantly worse than the group which w as denied access to the written form, suggesting that the written form had indeed interfered with the acquisition of pronunciation.

PAGE 46

46 The researchers then considered, in abstract terms, possible combinations of grapheme phoneme correspondences in English an d their correspondents in a target language, and predicted that the most logical source of interference would be when a grapheme A in both languages represented a phoneme X in one language and (an entirely different) phoneme Y in the target language. Consi dering this specific permutation AX AY in their data, the researchers found that the participants who had been denied access to the written form performed significantly better than those who had been allowed access to the written form. This suggest ed e vidence of interference due to the difference in pronunciation of the same grapheme in the two languages. The present study explores the nature of interference in the case of this same orthographic phonological permutation, where a single grapheme represe nts different phonemes in different languages, by considering how L1 English/L2 Spanish and L1 Spanish/L2 English speakers treat the graphemes and in L3 Portuguese. As previously mentioned, the typological similarity between Spanish and Portuguese is at once an asset and a liability (Resnick, 1945), with the differences in GPCRs facilitating certain phonological combinations but leading to continued mispronunciations by learners in other cases (Jordan, 1991). Examining the sound symbol correspondenc e systems in English, Portuguese and Spanish for the GPCRs considered in the present study will highlight the similarities and differences between the three languages. 1.4 The R elevant R ules of English, Spanish and Portuguese Two orthographic symbols are considered in the present study: and intervocalic 19 19 It was originally intended that the study examine six sound symbol correspondences [z], intervocalic [z], [ ], [ ] [h], and word initial [h] (Faraco & Moura, 1990) and data were collected with this objective in mind. However, due to the complexity of the project, this dissertation will examine only the first In Portuguese, both of these graphemes are pronounced as [z] (Table 1 1). In Portuguese, there is

PAGE 47

47 not always a one to one correspondence between graphemes and phonemes. In the case of the two grap hemes considered in this study, has a one to one grapheme phoneme correspondence but does not. The grapheme is always pronounced as [z]. The grapheme has two possible pronunciations: 1) [s] wordinitially, and after consonants ; an d 2) [z] in intervocalic position and before voiced consonants (Table 1 2). It is worth mentioning that [s] exists between vowels, but with a change in the orthography, from to . This orthographic change (with accompanying phonological change) is p honemic, resulting in a change in meaning, as demonstrated by the minimal pairs asa ( [aza], wing), and assa ([asa], s/he bakes). The orthographic phonological correspondence rule [s] was not considered in the present study, however, as the aim was to consider the phoneme /z/ and its two orthographic correspondences, and intervocalic 20The sound symbol correspondences in English and Spanish are rather different from those in Portuguese, for the two orthographic symbols considered in the pres ent study. It is important to understand these differences, in order to be able to discuss later the effect of English and Spanish on the acquisition of the Portuguese correspondences. Where Portuguese has [z] for and intervocalic , Spanish has a di fferent sound from Portuguese, but the same sound for both graphemes: [s] 21 two correspondences. While the ch apter on methodology will make mention of the other four correspondences, the results and discussion chapters will consider only the [z] and [z] correspondences. (Table 1 3). The English soundsymbol correspondences for the graphemes 20 It is worth noting that the phoneme /z/ is represented orthographically not only by and intervocalic , but also by before voiced consonants (as mentioned previously), and by , in words such as exemplo (example) and exa me (exam). However, is not considered in the present study due to its irregularity of pronunciation, since it can also be pronounced as [s], as in excelente (excellent) and excesso (excess), and as the voiceless palatal fricative [ caixa (box) and xcara (cup). 21 While this may not hold true for Castillian Spanish, where is Spanish. Additionally, it is noteworthy that Spanish has no voiced alveolar fricative phoneme /z/, although there is the possibility that speakers may realize orthographic as [z] when it occurs befor e certain voiced consonants, as in mismo (same), asno (donkey), and desde (since), although speakers may not be aware of the sounds they produce (Schwegler & Kempff, 2007).

PAGE 48

48 examined in the current study differ slightly from those in both Spanish and Portuguese. In English is p ronounced as [z] as it is in Portuguese The intervocalic grapheme , however, has three possible pronunciations in English: [z], [s], and the voiced palatal fricative [ ] (Table 1 4). In the current study, the effect of these different orthographic pho nological correspondence rules is considered as learners produce different types of Portuguese words, discussed in the following section. 1.5 Different W ord T ypes The different word types which are of interest in the present study are cognates ( true or fal se), non cognates and nonce words. Typically, cognates and non cognates have been discussed in the literature, with respect to L2 and L3 acquisition, in one of two ways: 1) as lexical items to be acquired by learners ( e.g., Singleton, 1999; Gass & Selinker 2001) ; or 2) as items in speech production studies aimed at gaining a better understanding of how the bilingual or multilingual brain handles lexical storage of and access to its different languages (e.g., Costa, Santesteban and Ca o, 2005 ) Nonce words, or pseudo words have typically been used in studies examining reading skills (e.g., Defior, Martos & Cary, 2002). The present study considers these word types from a rather different perspective, as it examines the effect of these word types on the acqui sition of sound symbol correspondence rules in L2 or L3. Nevertheless, a brief consideration of the literature with respect to these word types may be helpful in forming predictions about their interaction with the sound symbol correspondence rules examine d here Generally, vocabulary items which are cognates in different languages are seen to be relatively easy to acquire (e.g., with respect to Spanish and Portuguese, Chandler, 1958; Resnick, 1945; Garrison, 1979), as their orthographic (and often phonological) relatedness in the two languages is high. There are, of course, pitfalls with cognates, if they differ slightly in spelling and/or pronunciation in the two languages, as learners must first be aware of these differences

PAGE 49

49 and, second, commit these diff erences to memory if the vocabulary items are to be produced correctly in speech and in writing (Resnick, 1945). Another danger for learners is when cognates turn out to be false cognates in different languages, where the orthography and/or phonology may b e similar but the meaning differs (e.g., brincar in Spanish means to jump while in Portuguese it means to play, to joke). Again, these are words which learners must make an effort to remember in order for them to be used correctly. Nevertheless, the hi gh degree of relatedness of cognates makes them easily recognizable to new learners, and therefore generally easier to remember and use than other new vocabulary items which bear little or no orthographic or phonological resemblance to the L1 (non cognates ). The effect of word differences has also been considered in cognitive studies with the objective of understanding better how bi and multilingual learners organize their mental lexicon(s) An exhaustive discussion of brain organization is beyond the sc ope of this dissertation, but a brief mention of the effects of word relatedness in production is warranted here. For more on bilingual mental organization, refer to Singleton ( 1999), for example In speech production studies using related words in picture naming studies, response times tend to suggest that relatedness may or may not be advantageous. In a picture naming study with Dutch English bilinguals, where Hermans, Bongaerts and Schreuder (1998) used phonologically related and unrelated words as distr actors, response times were longer when the distractors were phonologically related to the L1 translation of the L2 target word. For instance, learners shown a picture of a mountain ( berg in Dutch) were slower with the phonologically related distractor wor d berm (verge) than with the phonologically unrelated word kaars (candle). Although the related (distractor) words which Hermans et al. used were not cognates, and while they were not interested in orthography in their study, nevertheless there are asp ects of their methodology

PAGE 50

50 which relate to the present study. To a large extent, the related words which Hermans et al. presented were orthographically similar to the words targeted for production, and the current study presents words which are orthographic ally similar (cognates) and dissimilar (non cognates and nonce word s) in learners previous languages and the target language, in order to determine whether there exists an element of distraction in related words which is not to be found in unrelated wor ds. Given the results of the Hermans et al. study, where related words had a negative effect on the target words, it is possible to hypothesize that word type will have an effect on production in the present study, with related words (cognates) being produ ced less accurately than unrelated words (non cognates and nonce words) On the other hand, facilitatory effects have been found for cognates, which are related words of a different kind from those used by Hermans et al. While their words were phonological ly and orthographically related, they were not also semantically related, as cognates are (except in the case of false cognates) In a review of the literature, Costa, Sanstesteban and Ca o (2005) found several production studies which showed a positive ef fect, insofar as speed and accuracy were concerned, with naming cognates but not non cognates (e.g., Costa, Caramazza & Sebastin Galles, 2000; Kroll, Dijkstra, Janssen & Deslau r iers, 2000), even with aphasic bilingual speakers (e.g., Roberts & Deslauriers 1999; Kohnert, 2004). C osta et al. (2005) likened the cognate effect to neighborhood effects, where neighborhood density plays an important role in production. ( Neighborhoods are made up of similar sounding words, which share an onset such as cat cap and cash or which differ in only one phoneme like cat scat and at and dense neighborhoods are those w hich have many words in them. For more on neighborhoods and their effects, see Altmann 1997.) Costa et al. (2005) stated that picture naming studies (e.g., as Vitevitch 2002, 2003) showed that words with many neighbors were

PAGE 51

51 named faster than words with few neighbors. Similarly, they claimed, the processing of cognates is facilitated, where the words have phonological overlap, as do neighbors, but acros s languages. Another possibility to consider is that seen words (cognates and non cognates in the present study) will be produced more accurately than unseen words (here, nonce words) given learner familiarity with the former In a study looking at the a cquisition of stress patterns in L2 Spanish, Lord ( 2007) found that L1 English speakers, with varying proficiency levels in L2 Spanish, and L1 Spanish speakers had different accuracy rates and production times when reading sentences containing real Spanish words versus sentences including synthetic (created) Spanish like words. The results showed evidence that a learners lexicon plays an important part in stress assignment, with known words being produced more accurately and more quickly than unknown words Defior et al. (2002) claim ed that it is widely accepted that the reading of [ nonce words] is a good indicator of knowledge of the alphabetic code (p. 146), thus making them ideal items to include in order to demonstrate participants ability to general ize GPCRs from seen words to unseen words. C onsidering nonce words, recall that Defior et al.s (2002) study (previously described in detail) found that children were generally slower and less accurate when reading them (nonce words) than when reading real words (numbers) Extrapolating the se observations and results to the current study is somewhat challenging, because this study does not look at the acquisition of the words as vocabulary items per se, nor are response times recorded in the present study (since accuracy in production is the primary concern here ) If the target sound symbol correspondences in cognates are produced more accurately than those in other words (non cognates and nonce words), that would suggest a degree of facilitation for cognates. On the other hand, if the correspondences in cognates are

PAGE 52

52 produced less accurately than those in other words, that would be indicative of some form of distraction with cognates. Alternatively, if the target correspondences in seen words (cognates and non cognates) are produced more accurately than those in unseen words (nonce words), that would imply that learners are relying more on direct access to the lexicon to retrieve pronunciation rather than on the grapheme phoneme correspondence rules to ass emble the pronunciation of the items. Given these different possibilities, it is difficult to predict the effect of the different word types on the acquisition and production of the orthographic phonological correspondences considered in the study at hand Nevertheless, due to the fact that some of the literature on L2 and L3 acquisition has discussed the role of cognates, non cognates and nonce words, and given the use of different word types in reading studies which considered orthographic phonological co rrespondences (e.g., Defior et al., 2002), it seems pertinent to c onsider these word types in this study. 1.6 The Present Study The literature reviewed in this chapter sets the stage for the current study. Recalling its three main areas of focus error resolution, interference and generalizability th is study looks first to address the question of how well the sound symbol correspondences in L3 Portuguese (with L1 English and L2 Spanish or vice versa) are acquired. What evidence does the present study f ind to support Majors (2001) OPM, which predicts an increase in L3 in a learners IL? In terms of the effect of learners previous languages on their acquisition of an L3, the literature showed conflicting results in several areas: 1) whether or not there is an advantage for bilinguals over monolinguals in the acquisition of a new language (Cenoz, 2003); 2) whether proficiency in the L2 has to be low or high in order for it to be the source of transfer for the L3 ( Odlin, 1989; Poulisse & Bongaerts, 1994; S hanon, 1991; vs. Hammarberg, 2001); and 3)

PAGE 53

53 whether or not facilitation occurs between typologically similar languages (e.g., Ecke and Hall, 2000; vs. Gibson, Hufeisen and Libben, 2001). Given these mixed observations, the current study questions what the n ature of interference is with respect to the sound symbol correspondences in question. To what extent are the GPCRs in the L3 colored (to extrapolate from Flege, 1987, and Baker & Trofimovich, 2005) by the GPCRs in the L1 and/or the L2? What evidence is there for the effect of language status (L1/L2) versus language typology (similar/dissimilar) in the acquisition of the L3 correspondence rules? I t was also observed in the literature reviewed that Portuguese instructors might be aware of interference from Spanish but that this has been based more on observation than empirical evidence (Azevedo, 1978), and that further research on the acquisition of Portuguese by Spanish speakers is necessary (Carvalho, 2002). Thus, the current study also asks what the rela tionship is between proficiency in Spanish and acquisition of Portuguese. As far as generalizability is concerned, the main question is to what extent learners are able to generalize the GPCRs learned, from seen words (cognates and non cognates) to unseen words (nonce words). Thus, it is of interest to know what differences arise in production due to different word types. What support is there for Defior et al.s (2002) and Lords (2007) finding s, where known words (here, cognates and non cognates) were rea d faster and more accurately than unknown words (nonce words) ? Or is there evidence to suggest that there is a difference between production of related words cognates and unrelated words non cognates and nonce words (to extrapolate from H ermans, Bongaerts & Schreuder, 1998, and Costa, Sanstesteban and Ca o 2005)? The next chapter describes the design of this study, as well as the methods used in data collection, transcription and analysis. Chapters three, four and five present and discuss the

PAGE 54

54 finding s related to each of the three research questions (error resolution, interference and generalizability, respectively). The final chapter brings the dissertation to a close, discussing implications from the study, as well as limitations and future direction s.

PAGE 55

55 Table 1 1. Symbols examined, Portuguese sound s and examples Symbol Portuguese Sound Example [z] zebra [ zeb a ] (zebra) Intervocalic [z] casa [kaza] (house) Table 1 2. Symbols examined, Portuguese contexts and sounds, and examples Symb ol Portuguese Context Portuguese Sound Example All contexts [z] zebra [zeb a] (zebra) Intervocalic [z] casa [kaza] (house) Before [z] desde [dez i] (since) Before [z] vesgo [vezgu] (cross eyed) Before [z] asno [ aznu] (donkey) Before [z] mesmo [mezmu] (same) Word initial [s] sapato [sapatu] (shoe) After [s] balsa [bawsa] (ferry) After [s] ganso [gsu] (goose) After [s] urso [u su] (bear) Table 1 3. Symbols e xamined, Spanish sounds, and examples Symbol Spanish Sound Example [s] caza [kasa] (hunt) Intervocalic [s] casa [kasa] (house) Table 1 4. Symbols examined, English sounds and e xamples Symbol English Sound Example [z] zebra Intervoca lic [z] raisin Intervocalic [s] basin Intervocalic [ ] Asian

PAGE 56

56 CHAPTER 2 METHODOLOGY 2.1 Introduction In the present study, participants enrolled in introductory Portuguese classes at the university level were recorded at three intervals during their semester of instruction (at the beginning, in the middle, and at the end of the semester), as they read a series of Portuguese or Portuguese like words including English Spanish Portuguese orthographic cognates, non cognates and nonce words containing certain grapheme phoneme correspondences. To follow is a pres entation of the research questions that motivated the study, as well as initial hypotheses concerning these questions. Thereafter is a detailed desc ription of the methodology used in the study to test these hypotheses, including participants and materials, as well as the procedures of data collection, transcription and analysis. Concluding this chapter is an overview of subsequent chapters. 2.2 Research Questions and Hypotheses In the literature reviewed in the previous chapter, it was claimed that the part played by the L3 in a learners IL would increase over time, while the role of the L1 and L2 would decrease (Major 2001). In terms of the nature o f interference, on the one hand, there was the claim that the lower the learners proficiency in the L2, the greater the likelihood of language transfer from the L2 to the L3 (Odlin, 1989; Poulisse & Bongaerts, 1994); on the other hand, there was the belief that there must be a certain degree of proficiency in the L2 in order for it to be the basis of transfer to the L3 (Hammarberg, 2001). Language typology was suggested to be a significant factor in transfer (Cenoz, 2001), with the typologically similar language being the source of transfer more often than the typologically dissimilar language. As far as word eff ects, priming studies showed delayed response times when phonologically related distractor words were used

PAGE 57

57 (Hermans, Bongaerts & Schreuder, 1998) but also positive effects for cognates (Costa, Sanstesteban and Cao, 2005), and nonce words were seen to be read with less speed and accuracy than real words (Defior et al., 2002; Lord, 2007 ). The current study addresses these areas of error resolution, interference and generalizability (word type) by examining the acquisition of L2 and L3 orthographic phonologi cal correspondences. Although these areas overlap to some extent, and certain questions apply in all three cases, they are presented and discussed separately hereafter, for the sake of clarity. 2.2.1 Error Resolution Th ree basic research questions (R. Q.) arise with regard to error resolution, listed here with hypotheses (H.) concerning participant production. R. Q. 1. To what extent do learners acquire the Portuguese grapheme phoneme correspondence rules (GPCRs) under consideration, throughout the course of the semester?22H. 1. Significant increases in correct production are expected to occur throughout the semester of instruction, at least when all production data is considered together (both graphemes, all word types, all participant groups), in accordan ce with Majors (2001) OPM, which predicts that a learners IL will move towards an end state of L3. R. Q. 2. What differences in acquisition exist between participant groups23H. 2. It is hypothesized that there will be differences in correct production between the language groups, with lower accuracy being found in the production of the more proficient Spanish speakers, given the persistent problems discussed in the literature with regard to Spanish speakers learning Portuguese pronunciation (Azevedo, 1978; Jordan, 1991; Tarquinio, 1977). ? 22 For the purpose of this dissertation, acquisition is measured by the correc t pronunciation of the target Portuguese GPCRs, although in some cases correct production may actually result from positive transfer from English, as will be discussed further below. 23 Participants and their grouping will be described further in section 2. 3.1.

PAGE 58

58 R. Q. 3. What differences in acquisition are there between the two graphemes and intervocalic corresponding to the phoneme /z/? H. 3. It is anticipated that there will be significant differences between tests fo r both graphemes, and significant differences between the two graphemes, based on the fact that one of them has a one to one grapheme phoneme correspondence in Portuguese ( /z/), while the other does not ( /z/ in intervocalic position and before voic ed consonants, /s/ elsewhere), and in light of the much documented problem of devoicing of intervocalic (Azevedo, 1978; Jordan, 1991; Tarquinio, 1977). 2.2.2 Interference There are several principal questions concerning interference, presented her e with hypotheses regarding participant production. R. Q. 1. What evidence is there of interference from the L1 and/or the L2? H. 1. It is anticipated that there will be evidence of interference from both the L1 and the L2, with these reducing over time, given the predictions made by Majors (2001) OPM that the learners initial IL consists in part of the L1 and in part of the L2, with these decreasing over time. To extrapolate somewhat from the OPMs prediction that the universals (U) component will incr ease and then decrease throughout the course of the IL, it is hypothesized that there will be evidence in the data that participants are producing sounds which are neither consistent with the correspondence systems of the L1/L2 nor with th at of the L3. Suc h production is expected to reflect universal tendencies toward unmarked features, such as voicelessness R. Q. 2. What differences in interference are there between participant groups? H. 2. Given the conflicting claims in the literature with regard to t he relationship between proficiency in the L2 and the level of transfer from the L2 to the L3 (e.g., Odlin, 1989; Poulisse & Bongaerts, 1994; vs. Hammarberg, 2001), it is difficult to predict what differences might exist

PAGE 59

59 between participant groups in the p resent study. Certainly it would seem that there should be differences between participant groups of different language backgrounds and L2 proficiency levels. If the weaker L2 is the greater source of interference, then native English speakers should sho w greater interference from Spanish, while native Spanish speakers should show greater transfer from English. On the other hand, if there must be a certain level of proficiency in the L2 in order for it to be the basis of transfer to the L3, then L1 Englis h speakers who are more proficient in Spanish should show greater transfer from L2 Spanish than L1 English speakers who are less proficient in Spanish. At the same time, proficient L2 English speakers would be expected to show greater transfer from English than from Spanish. Alternatively, it may be that all participants will transfer to a greater extent from the more similar language (Spanish) than from the less similar language (English). Given the researchers own observations in the classroom, the hypot hesis selected here is that the higher the proficiency level in the closely related language, Spanish the more evidence of transfer there will be from that language in the target language production. R. Q. 3. What differences in interference exist between the two graphemes and ? H. 3. It is expected that there will be differences between the two graphemes, in light of the GPCRs in English, Portuguese and Spanish, and in view Muller and Mullers (1968) prediction about the most difficult GPCR to acqu ire being of the AX AY type (where language 1 has sound X for a particular grapheme A and language 2 has sound Y for that same grapheme A). Since grapheme always has the same correspondence rule in English and Portuguese (although not in Spanish), but grapheme only sometimes has the same correspondence rule in English and Portuguese (and never in Spanish), it is anticipated that there will be greater accuracy in production of due to the possibility of positive transfer from English, and less acc uracy in

PAGE 60

60 production of due to the possibility of only negative transfe r from Spanish and generally negative transfer from English. R. Q. 4. To what extent is production in Spanish related to acquisition (correct production) of Portuguese? H. 4. It is a nticipated, based on the researchers own classroom observations and impressions, as well as observations in the literature about the difficulty encountered by Spanish speakers with Portuguese pronunciation, in particular with intervocalic (e.g., Azeve do, 1978; Tarquinio, 1977; Jordan, 1991), that production in Spanish will be inversely related to acquisition of the Portuguese GPCRs in question. That is, it is expected that the greater the accuracy in Spanish production, the lower the accuracy in Portug uese production will be. 2.2.3 Generalizability The term generalizability is used here to refer to participants ability to apply the GPCRs learned for seen words (cognates and non cognates) to unseen words (nonce words). There is one principal question to be answered here, along with a hypothesis regarding participant production. R. Q. 1. What differences in production arise due to word type (cognates, non cognates and nonce words)? H. 1. On the one hand, given findings that real words were read faster an d more accurately than nonce words ( Def ior et al., 2002 ; Lord, 2007), it would be expected that the L2 and L3 participants in the current study would perform better on real words than on nonce words. On the other hand, extrapolating Hermans et al.s (1998) and Costa et al.s (2005) findings that related words (cognates) affected production to a greater extent than unrelated words (non cognates and nonce words) where the former found related words to have a negative effect, and the latter found them to have a positive effect it might be expected in the present study that cognates would affect production to a greater or lesser extent than non cognates and nonce words. Given

PAGE 61

61 the literature, it is difficult to predict what the current study will show with rega rd to the relationship between word type and the acquisition of the GPCRs in question, so once again, the researchers classroom observations are instrumental in hypothesis selection here. It is predicted that the orthographic phonological correspondences in seen words (cognates and non cognates) will be produced more accurately than those in unseen words (nonce words), due to learners greater familiarity with and exposure to the former. That is, production will be lexically driven rather than rule driven. In answering these research questions, a better understanding of L3 production will be gained. A contribution to the general body of research will be made by finding evidence to support or not support the theoretical framework and previous studies ci ted, with insight added regarding the role of an L3 learners previous languages in the acquisition process. Pinpointing the source of interference in an L3 learners IL is not straightforward (Dewaele, 1998), and by considering the acquisition of specific orthographic phonological correspondences, stronger conclusions may be drawn than have been possible thus far with regard to the source(s) of interference. With these thoughts in mind, the n ext section presents the methodology used in the current study, i n order to test the hypotheses and answer the research questions. 2.3 Method ology Consideration will be given first to the participants involved in the present study, then to the materials and tasks developed, and finally to the methods of collecting, tran scribing and analyzing the data. 2.3.1 Participants The participants in the current study were students in three introductory classes of Portuguese: two Beginner 1 classes (the first semester of a twosemester course) and one

PAGE 62

62 Introduction to Portuguese an d Brazil class for fluent (native or advanced) speakers of other Romance languages (an accelerated one semester course). All students of the three classes were asked to participate in the study just as they would participate in any class activity, since th e tasks were administered during normal class time, and there was no penalty for students who failed to participate in one or more of the tasks due to absence from class on that day. Thus, participation was considered voluntary, since no additional time co mmitment outside of class was required and there was no penalty for non participation. Before participating, the students indicated that they had been informed about the study and its requirements by completing a consent form (Appendix A). Next, participants completed a questionnaire (Appendix B), giving details about the languages they spoke, whether they were native speakers of them, how and when they had learned these languages, and with whom and where they used them. They were also given descriptions of four proficiency levels (1 being the lowest and 4 being the highest) for the four skills of listening, speaking, reading, and writing, and were asked to rate themselves for each language by choosing from the descriptions the proficiency level that they fe lt most closely reflected their own. (Participants were not asked to rate themselves for English, since it was presumed that all were proficient if not native, English speakers, due to their enrollment in an English speaking institution of higher educatio n.) Of the 66 students who participated in the study, data from 52 were chosen for inclusion in the analysis. This selection was based upon two criteria: 1) attendance all three times the task was administered; and 2) audible (and therefore usable) record ings. No participants were eliminated based on questionnaire answers. (The questionnaire asked participants if they had ever been diagnosed with a reading disorder, such as dyslexia. Had anyone responded

PAGE 63

63 affirmatively to this question, which was not the ca se, their data would most likely have been excluded.) For the purpose of data analysis, participants were divided into groups, based on the information they supplied in the questionnaire (summarized in Appendix C), and on the course (beginner or accelerate d) in which they were enrolled It bears mentioning here that, at the institution where the data were collected, enrollment in Portuguese is relatively low ( at least compared with Spanish and French, for instance), and there is no standardized placement te st for students. Those who enrol l in the accelerated introductory class, designed for speakers of other Romance languages, have generally taken classes or a placement test in the other language, are native speakers of that language, or occasionally, are he ritage speakers of Portuguese. Sometimes, informal interviews with the instructor or coordinator are conducted for the purpose of assessing the students level of proficiency in the ir other Romance language(s) in order to ensure that they enrol l in the app ropriate Portuguese class. Those who do not meet the criteria for the accelerated class enrol l in the beginner class. It is worth mentioning that, occasionally, when scheduling conflicts occur, these enrol l ment criteria are not strictly adhered to, due to the limited number of classes offered each semester and the desire to have enrol l ment be as high as possible. When this happens, students who would otherwise be in the accelerated class are sometimes permitted to enroll in the beginner class, but never vic e versa, due to the difficulty that beginner students would encounter in the accelerated class. T he 52 participants in the present study were initially divided into five groups (Table 2 1) After initial data analysis, where no significant differences were found between group O and any of the other groups apart from S, the five groups were collapsed into four (Table 2 2). The results which led to this finalized grouping are discussed in Chapter 3.

PAGE 64

64 2.3.2 Materials In order to examine participants production at different points during the semester of instruction, with respect to the orthographic phonological correspondences considered in the study24, a reading task was developed, which required participants to read 130 words in Portuguese, 36 words in English and 20 words in Spanish (Appendix D). The English and Spanish words were given only the first time the task was administered, to ascertain participants knowledge of the GPCRs of those languages for the graphemes considered in the current study. The task w as created specifically to require reading of the graphemes in all three languages and to elicit the production of target sounds in all three languages, as much as possible25The English and Spanish words were all real words, all of which contained either target graphemes or graphemes which would elicit the target sounds examined in the study, but whether or not they were cognates or non cognates with eac h other or with Portuguese was not taken into account, since their purpose was simply to demonstrate participants knowledge of the Spontaneous oral data would not have been appropriate for this study because part icipants would not have been able to produce anything at the beginning of the semester, given that most of them had no previous exposure to or instruction in Portuguese. Even if this were not the case, it would be difficult to ensure production of sufficie nt instances of the target grapheme phoneme correspondences by all participants in natural speech. Additionally, early learners such as these might feel more anxious about being recorded while producing spontaneous oral data than during a reading task such as the one developed. 24 The study was designed to consider six sound symbol correspondences: [z], intervocalic [z], [ [ [h], intervocalic [h]. However, due to the complexity of the project, only data for the first two of these cor respondences are analyzed and discussed in this dissertation. Nevertheless, this chapter makes mention of the six correspondences since they formed part of the materials developed and used. 25 Recall that the phoneme /z/ does not occur in Spanish, although it is possible for to be pronounced as [z] before the voiced consonants , , and , as in mismo (same), asno (donkey) and desde (since).

PAGE 65

65 GPCRs in English and Spanish and their physical ability to produce the sounds The Portuguese words consisted of: 48 Engli sh Spanish Portuguese cognates (C), such as bsico (basic) and horizonte ( horizon) 35 non cognates (NC), such as casaco (coat) and zangado (angry) 37 nonce words (N), such as feserel and paimozes 10 filler words (F) which did not contain any of th e target graphemes, such as chato (boring) and criana (child) As far as possible, the cognates and non cognat es were taken from the textbook, Ponto de Encontro (Klobucka, A., Jout Pastr, C. M. C., Moreira, M. L. de B., Sobral, P. I., & Hutchinson, A. P., 2007) that participants used for class. While classroom materials and instruction were not factors examined in this study, a brief note on them is warranted here. Although the textbook itself makes no reference to pronunciation, pronunciation guides and exercises appear throughout ancillary reference materials such as t he accompanying Student Activities Manual and online audio exercises Students are expected to read these notes regarding pronunciation on their own, outside of the classroom, but pron unciation is also taught and reinforced in the classroom. At the beginning of the semester, one or two class periods in both class levels are dedicated to explicit instruction regarding orthographic phonological correspondence rules, and correction is offe red to a certain extent (not measured) throughout the semester of instruction. Because participants were enrolled in different courses that progressed through the textbook at different rates, only words from the first three chapters were used (i.e., the ch apters covered by all classes between the first and second administration s of the task), in an effort to ensure that all participants would have been exposed to the cognate and non cognate words by the time the task was administered for the second time. In some grapheme phoneme word type

PAGE 66

66 categories, it was not possible to find sufficient tokens from the vocabulary found in the textbook, so to make the categories as balanced as possible, words were added, such as manhoso (smart, whiny) which was added to the [z] NC category, and azedo (sour) which was added to the [z] NC category. Nonce words were created by listing all of the syllables found in the cognate and non cognate words, then using the random function in Excel to rearrange them to yield words with two, three or four syllables. Any syllables or words which were phonotactically impossible or unlikely to occur in Portuguese were excluded. For example, words ending in are very rare in Portuguese so any nonce word created with a final was excluded. The pronunciation of the target graphemes in the nonce words was agreed on by the three raters in the study (Section 2.3.4 provides more information about the raters) Of the 120 non filler words, 13 contained two target correspondences, suc h as realizado (realized) which contained the initial [h] and [z] correspondences, yielding a total of 133 target correspondences for consideration. Of these, 53 were found in cognates, 40 in non cognates and 40 in nonce words. Of the 133 tokens, 42 contained the target phoneme /h/ (24 and 18 ), 49 contained the target phoneme /z/ (30 and 19 ), and 42 contained the target phoneme / 26Considering only the and words which are analyzed and discussed in this dissertation, the 49 words were later reduced to 48, as it was noticed that the in one of the words in the [z] C category, transio (transit ion), was not intervocalic. The words are shown broken down into cognates, non cognates and nonce words (Table 2 3), as are the words (Table 2 4), with asterisks marking those words which were not found among the vocabulary items listed in the tex tbook. It is worth mentioning here that three sets of cognate 26 Recall that only the /z/ items are analyzed here; the analysis of the other sounds is beyond the sco pe of this dissertation.

PAGE 67

67 words were included, due to the three possible pronunciations of in English: [z] as in raisin ; [s] as in basin, and [ Asian The C1 cognates are those where English would have [z], the C2 cognates are those where English would have [s], and the C3 cognates are those where English would have [ A PowerPoint (2003) presentation was created with instruction s for the participants, followed by the words to be read in each language, with each word appearing on a separate slide. After initial instructions, all the Portuguese words were shown, followed by a slide indicating the end of the Portuguese words and the beginning of the English words. After the English words came another slide indicating the end of the English words and the beginning of the Spanish words. After the Spanish words, there was a slide which indicated the conclusion of the task (Appendix E sh ows the instruction slides and examples of the words that participants saw for each language). For all three languages, the words were randomly ordered, using the random function in Excel, then two versions of the PowerPoint presentation were created, Odd and Even, since participants were seated in odd and even numbered recording booths in the Language Learning Center. Odd had Portuguese words 1130, English words 136 and Spanish words 120; Even had Portuguese words 66 130 followed by 165, Englis h words 1936 followed by 1 18, and Spanish words 1120 followed by 110. The purpose of the two versions was to roughly control for fatigue and to minimize distractions from neighboring participants, since all participants were reading and being recorded at the same time. 2.3.3 Data Collection Data were collected at the universitys Language Learning Center, where participants sat at booths equipped with a computer, monitor and headphones with attached microphone. Seated at booths numbered with odd and even numbers, participants were instructed to open the

PAGE 68

68 PowerPoint version (Odd or Even) respective to their booth number27The reading task was administered three times throughout the semester: 1 ) on the second day of class, before any formal instruction on pronunciation had been given; 2) during Week 8 of a 16 week semester; and 3) during Week 15 of the 16week semester. Participants read the 130 Portuguese words each time the task was administer ed, while the English and Spanish words were read only on the first occasion. Whether or not participants read the same version (Odd or Even) on each occasion was not deemed important since participants themselves would likely not remember the order in which the words had been presented in previous tests and thus was not verified. After reading the instructions in the PowerPoint file, participants were instructed to begin speaking, at which point recording was started by the researcher, from a central computer, using a Sanako Lab 300 system. Participants were able to go at their own pace, clicking the mouse or pressing a key on the keyboard to move from one slide to the next. It was felt that, by allowing participants to proceed when ready, they would not become bored by having to wait for the slides to advance automatically, nor would they be pressured if the slides were moving too quickly for them. The main disadvantage of having participants advance through t he slides on their own was that they could skip slides, inadvertently or on purpose. Still, it was felt that this disadvantage was outweighed by the advantages, so it was deemed better to have participants progress at their own rate. Once all participants had finished reading, the recording was stopped, again by the researcher, and the recorded data was captured into individual files for each participant and saved on an external drive. 27 Whether or not participants opened and used the correct version was not verified precisely. Observing that alternating participants seemed to be reading different versions was deemed sufficient verification.

PAGE 69

69 Once all the data had been collected, the files were examined for acoustic problems during the recording process, as well as any gaps in individual participation (that is, where a participant failed to attend one or more test dates). Then, using Adobe Audition, each file was truncated at the beginning to remove the participants name and at the end to remove the empty portion after the participant had finished the reading t ask, in preparation for raters to listen to the recordings. Once the recordings had been prepared for the raters, the files were identified by random numbers to ensure anonymity. 2.3.4 Transcription Three raters listened to all of the data. Rater 1 is a na tive speaker of Brazilian Portuguese (from Rio de Janeiro) but not a linguist. Rater 2 is a highly fluent speaker of Brazilian Portuguese and a linguist. Rater 3 (the researcher) is a native speaker of Brazilian Portuguese (from So Paulo) and a linguist. All three raters were instructors of Portuguese at the university at the time of data collection: rater 1 was the instructor of the two beginner classes; rater 2 was the instructor of the accelerated class for speakers of other Romance languages; and rater 3 was the instructor of other classes which did not participate in the study. While perhaps somewhat undesirable, the use of the instructors and the researcher as the raters in the study was necessary due to a lack of other fluent Portuguese speakers avai lable. Before carrying out the ratings, the researcher met with the other two raters for training on how to transcribe and assess the recorded data. Recordings from a pilot study were used in the training, so that the raters would be familiar with the words and target sounds to be transcribed without actually being exposed to (and perhaps prejudiced by) any of the data from the present study. The researcher reviewed the target Portuguese sounds for the six graphemes under consideration in the study, as well as some of the alternate sounds which participants were hypothesized to produce, based on their knowledge of English and Spanish pronunciation rules.

PAGE 70

70 Raters 2 and 3, the linguists, were familiar with the IPA symbols necessary to transcribe all possible so unds, but a guide (Appendix F) was produced especially for rater 1, the non linguist. The guide showed words in Portuguese, English and Spanish, with highlighted graphemes and the symbols for their respective sounds, to facilitate transcription. Alternativ e sounds (due to dialectal variation) were indicated with slash marks on the guide. Raters 1 and 2 were also instructed to consult with the researcher (rater 3) if they were unsure how to transcribe a particular sound28Provided with necessary equipment, a ll three raters listened to all of the data on their own, and transcribed the target sounds for each word, as had been discussed in the training session. Including those in the filler words, the number of target graphemes transcribed per participant, per t ask, was 155. To facilitate the process, transcription sheets (Appendix G) were provided for the raters to circle the sounds heard, with four columns for each target grapheme /phoneme : the first column contained the symbol for the sound expected for someone using English pronunciation rules; the second column contained the symbol expected for someone using Spanish pronunciation rules; the third column contained the symbol for the (correct) target Portuguese sound; and the fourth column was left blank for the rater to write in the symbol for a sound other than those in the first three columns. Raters either circled one of the symbols listed on the sheet or wrote in a symbol if it was not listed. All of the raters transcriptions were entered into an Excel spre adsheet then compared in order to calculate inter rater agreement. In cases where one rater did not agree with the other two, the majority rating was selected (Lord, 2007) As relates to and items only, in a very small percentage of cases (0.52%), none of the raters agreed, so the three met to listen to 28 For example, rater 1 inquired about how to transcribe the voiced velar stop [g], which was not provided on the transcription guide but was used by many participants for in words such as ferrugem (rust).

PAGE 71

71 and discuss those sounds until agreement was reached, at least between two of the raters. Table 25 provides details on the inter rater agreement process for and items. Overall, the three rat ers agreed 85.03% of the time; the cases where only two raters agreed were fairly evenly distributed: Raters 1 and 2 (but not rater 3) agreed 3.58% of the time; raters 1 and 3 (but not 2) agreed in 4.49% of instances; and raters 2 and 3 (but not 1) agreed on 6.9% of the items. 2.3.5 Analysis For the purpose of statistical analysis, the data entered into Excel were assigned different numerical and alphabetic values for different purposes. ( Appendix H shows a sample of participant production for and tokens, as well as the numerical and alphabetic values assigned to each sound.) In order to examine participants accuracy with respect to Portuguese pronunciation, numerical values were assigned: a score of 1 for a correct (target) Portuguese sound, and 0 for anything else. Scores were then summed per participant, per grapheme, and per word type for each test, to facilitate analysis. Separately, the data were examined for the sounds which participants produced (such as the projected English or Spanish sound on the rating sheet, or the target Portuguese sound). The sounds were assigned alphabetic values as follows: 1) EN for transfer from English (based on English pronunciation rules), when transfer from Spanish or correct pronunciation in Portuguese would differ from English, such as the realization of the in visual (visual) as [ ] ; 2) EP for instances when English and Portuguese pronunciation rules coincided but differed from Spanish, such as the realization of the in presidente (president) or the in horizonte (horizon) as [z]; 3) ES for cases when English and Spanish pronunciation rules coincided with each other but differed from Portuguese, such as the realization of the in bsico (basic) as [s]; 4) NO when participants produced nothing for a target phoneme (either because syllables were skipped or transposed, or because entire words/slides were skipped); 5) OTH when a sound was produced

PAGE 72

72 other than the sounds expected by the rules of English, Portuguese and Spanish, such as the realization of the in visual (visual) as [ ]; 6) PO when production was correct according to the Portuguese rule, where this differed from the English and Spanish rules, such as the realization of the in bsico (basic) as [z]; and 7) SP for transfer from Spanish (based on Spanish pronunciati on rules) when the Spanish rule differed from the English and Portuguese rules, such as the realization of the in presidente (president) or the in horizonte (horizon) as [s]. Each of these alphabetic values was then summed per participant, per grapheme and per word type for each test, to facilitate analysis. Once participant scores were summed, they were calculated as percentages of the total number of items per category, due to the unbalanced nature of the categories ( e.g. there were 17 cognat e items but only six cognate items per test). These percentages were then imported into SAS for statistical analyses to be carried out. The statistical tests conducted are described separately here, according to the research question they were inte nded to address. With regard to the first research question, concerning error resolution a general linear model (mixed effects) was used to analyze the percentages calculated from the numerical scores assigned to each sound produced, considering as factor s the 52 participants (random factor) and as fixed factors: the five participant groups29, three tests, two graphemes and three word types, to yield a total of 936 observations.30 29 These five groups were later reduced to 4, as discussed in great er detail in the next chapter. Tukey post hoc t tests were conducted when necessary, to establish where ther e were significant differences and all analyses used a significance level of =0.0005. The output from the model and the results of the t tests are presented and discussed in Chapter 3. 30 Note that the same model was used to address the third research question, regarding generalizability, which is discussed separately momentarily.

PAGE 73

73 For the purpose of addressing the second research question, regarding the role of interference in acquisition, a general linear model (mixed effects) was used to analyze the percentages calculated from the alphabetic scores assigned to each sound produced, considering as factors the 52 participants (in four groups), three tests, and seven production types for , to yield a total of 1092 observation s. For , the factors considered were the 52 participants (in four groups), three tests, and four production types, to yield a total of 624 observations. Tukey post hoc t tests were conducted when necessary, to establish where there were significant diff erences and all analyses used a significance level of =0.0005. All production was considered by group and grapheme, according to the various production types. The output from the model and the results of the t tests are presented and discussed in Chapter 4. In order to address the question on the relationship between accuracy in Spanish production and accuracy in Portuguese production, a regression model was used to analyze participants production on the Spanish reading task compared with average product ion for the three Portuguese reading tasks. The results of the regression model are also presented and discussed in Chapter 4. As for the third research question, regarding generalizability, t he results of the general linear model used to address the first research question were also used to address the third question, where the results relevant to the latter question pertain to the fixed factor word type. Tukey post hoc t tests were conducted when necessary, to establish where there were significant differ ences and all analyses used a significance level of =0.0005. All production was considered together, then production by grapheme was examined. Next, the cognate words were divided into three sets, according to the sound in the English equivalent ([s], [z] or [ with production of these sets being co nsidered according to participant group. The output from the model and the results of the t tests are presented and discussed in Chapter 5.

PAGE 74

74 In the final chapter, the results of the present study are reviewed, with respect to the research questions it set out to answer and the hypotheses proposed, and the studys contributions and implications are discussed as well as its limitations and future directions.

PAGE 75

75 Table 2 1. Participant groups Group Group Description N o. E1 N ative English speakers with no / low proficiency in Spanish (beginner class) 21 E2 Native English speakers with high proficiency in Spanish (accelerated class) 4 P Participants with previous exposure to or instruction in Portuguese 9 O Participants with exposure to or instruction in other foreign languages 9 S Native Spanish speakers 9 Table 2 2. Revised participant groups Group Group Description N o. E1 N ative English speakers with no /low proficiency in Spanish (beginner class) 25 E2 Native English speakers with high proficiency in S panish (accelerated class) 5 P Participants with previous exposure to or instruction in Portuguese 9 S Native Spanish speakers 11 Table 2 3. Grapheme words C1 C2 C3 NC N apresentao bsico audiovisual atencioso disudo e squis ito casos deciso casaco feserel m useu curiosidade divises desenho isgio presidente filosofia reviso manhoso* lomosa v isvel generosidade televiso poloneses maresa persuasivo viso preguioso quasano Indicates words not in textbook. Table 2 4. Grapheme words C NC N colonizao azedo* bazero horizonte bzios paimozes idealizao dzia* pastiza localizado gizes (pl*) prozida organizao lazer trazentar realizado rapazes zalito zangado Indicates words not in textbook.

PAGE 76

76 Table 2 5. Summary of inter rater agreement for and items (number and percentage) Test Raters 1 & 2 Raters 1 & 3 Raters 2 & 3 Raters 1, 2 & 3 1 items agreed on (#) 107 126 205 2058 1 items agreed on (%) 4.29 5.05 8.21 82.45 2 items agreed on (#) 82 102 165 2147 2 items agreed on (%) 3.29 4.09 6.61 86.02 3 items agreed on (#) 79 108 147 2162 3 items agreed on (%) 3.17 4.33 5.89 86.62 All items agreed on (#) 268 336 517 6367 All items agreed on (%) 3.58 4.49 6.90 85.03

PAGE 77

77 CHAPTER 3 ERROR R ESOLUTION RESULTS AND DISCUSSI ON 3.1 Introduction This chapter addresses the first broad research question, regarding error resolution, by presenting and discussing the pertinent results. Before considering these, however, it is helpful to recall the res earch questions asked and hypotheses stated (in section 2.2.1) concerning error resolution. The first question asked to what extent learners acquire the Portuguese GPCRs under consideration, throughout the course of the semester. It was hypothesized that c orrect production would increase, at least when all production data were considered together (i.e., the two graphemes, all word types, all participant groups), in accordance with Majors (2001) OPM, which predicts that a learners IL will move towards an e nd state of L3. The second question explored the differences in acquisition which might exist between participant groups. It was predicted that there would be differences in correct production between the language groups, with lower accuracy being found in the production of the more proficient Spanish speakers, given the problems discussed in the literature with regard to Spanish speakers learning Portuguese pronunciation (Azevedo, 1978; Jordan, 1991; Tarquinio, 1977). The third question considered the diff erences in acquisition between the two graphemes and intervocalic corresponding to the phoneme /z/. It was expected that there would be significant differences between tests for both graphemes, and significant differences between the two graphemes would be evident, based on the one to one grapheme phoneme correspondence which exists in Portuguese for but not for ( /z/ in intervocalic position and before voiced consonants, /s/ elsewhere), and given the much documented problem of devoic ing of the intervocalic by L2/L3 learners (Azevedo, 1978; J ordan, 1991; Tarquinio, 1977). The results will be presented and discussed as follows: 1) all production; 2) production according to

PAGE 78

78 participant group; and 3) production according to grapheme.313.2 Results and Discussion Variation between groups and individuals will then be discussed, with a general summary of results to follow. Results are presented here in terms of raw production scores, proportions of correct items out of all possible items p er category (e.g., cognate , cognate , etc.), and statistically significant differences (output from the model) which are marked with an asterisk in the tables 3.2.1 All Production Considering together the production of all the participants, irresp ective of their group, the grapheme or the word type, the data showed that correct production of [z] occurred for 845 items on Test 1, 1118 items on Test 2, and 1244 items on Test 3. Expressed as percentages of possible correct items, these scores translat e to 33.85% on Test 1, 44.79% on Test 2, and 49.84% on Test 3 (Table 3 1), shown graphically in Figure 31. Tukey post hoc t tests revealed significant differences between Tests 1 and 2, Tests 1 and 3, and Tests 2 and 3 (p<.0001, for all three pairs of tes ts). These results are encouraging: they indicate that progress in accurate sound production that is, acquisition occurred throughout the semester. Next, production by participant group is examined, which takes participants linguistic backgrounds into c onsideration. 3.2.2 Production by Group As previously discussed in section 2.3.1, participants were grouped according to their linguistic background, based on the information that they provided on their language background questionnaires with regard to the ir proficiency in English, Spanish, Portuguese and other languages, and also based on the Portuguese course in which they were enrolled (beginner or accelerated). With these criteria in mind, five groups emerged (Table 3 2) as was discussed 31 Production according to word type (cognate, non cognate and nonce) will be d iscussed in Chapter 5, which addresses the third broad research question, regarding generalizability.

PAGE 79

79 previously D ue to the unbalanced number of participants in the five groups, least square means (LS means) were used instead of arithmetic means in the statistical procedures, since LS means are appropriate for unbalanced designs with more than one effect. (For more on LS means, including examples, see Khuri, 2009.) The number of correct items, and their expression as percentages of possible correct items, are shown in Table 3 3. The ANOVA results (Table 3 4) indicated that all of the factors group, test, grapheme, a nd word type were significant, as were the interactions between group and test, group and grapheme, test and grapheme, and grapheme and word type32. Tukey post hoc t tests revealed significant differences between groups E1 and S (p<.0001), groups E2 and S (p=0.0384), groups O and S (p<.0001), and groups P and S (p<.0001). Because no significant differences were found between group O and any group other than S at this point, it was decided to consolidate the participants in group O into other groups.33 Of th e nine participants originally grouped in O, six were transferred to E1, one to E2, and two to S, yielding four groups (Table 3 5).34Considering these four revised groups, the number of correct items and their expression as percentages of possible correct items were calculated (Table 3 6). Graphically, these results are shown by group (Figure 3 2) and by t est (Figure 3 3). The ANOVA results using the revised groups (Table 3 7) indicated that all of the factors group, test, grapheme, and word type were significant, as were the interactions between group and test, and group and grapheme. 32 The interaction between grapheme and word type will be discussed in Chapter 5, which considers the effect of word type on acquisition. 33 It seems likely that the reason for the significant difference found between groups O and S is due to the linguistic background of the participants in group O. Disregarding their proficiency in another language (besides English, Spanish and Portuguese), seven of the nine participants would be classified as E1 or E2 participants, both of which groups showed significant differences from group S. 34 The interaction between language group and test found no significant differences between O and groups other than S, but there w ere some significant differences between other groups (E1 and E2, and E2 and P on Test 1), therefore further consolidation of these groups would not have seemed a reasonable step.

PAGE 80

80 Given that the ANOVA results indicated a significant interaction between group and test, Tukey post hoc t tests were carried out to establish where significant difference s existed. These were found between tests for all groups, though not between all tests (Table 3.8). Once again, it is encouraging to find that each group makes significant progress throughout the semester of instruction in terms of accurate production, at least between the first and third tests. Significant differences were also found between groups E1 and E2 on Test 1, between groups E1 and S on all three tests, between groups E2 and P on Test 1, and between groups P and S on all three tests (Table 3.9 tests are indicated in parentheses). These results are worth highlighting, for they show that native English speakers with high proficiency in Spanish behaved more similarly to native Spanish speakers than to other native English speakers (with no or low p roficiency in Spanish), at least initially. On the second and third tests, however, there were no appreciable differences between E2 and any of the other groups. While this was not unexpected for Test 2, it was a little unexpected for Test 3, at which point there were no significant differences between E1, E2 and P, but there were significant differences between E1 and S, and between P and S. The reason for this may lie in the fact that group E2 had a low number of participants, relative to the other groups which led to a less conclusive statistical result (the standard errors in the differences of LS means on Test 3 were: 5.5453 for E1 and S, 6.9680 for P and S, and 8.3616 for E2 and S), so that a statistical difference between E2 and S was not observed on Test 3. Such a difference might have been found had there been a greater number of participants in E2, a point to be remembered for future research. 3.2.3 Production by Grapheme Now consideration will be given to the two graphemes, and . The ANOVA results (Table 3 7 in section 3.2.2) showed a significant difference between and , but did not show a significant interaction between test and grapheme (p=0. 0547), nor was the three way

PAGE 81

81 interaction between test, group and grapheme found to be signif icant (p=0.1853), although the interaction between group and grapheme was shown to be significant (p<.0001). The number of correct items and their expression as percentages of possible correct items were calculated per grapheme for each group (Table 3 10). Graphically, these results are shown per group (Figure 3 4) and per grapheme (Figure 3 5). Tukey post hoc t tests showed significant differences between and for all groups (p<.0001). These results clearly show the major role which the grapheme pla ys in pronunciation, and they would suggest that the [z] correspondence rule in Portuguese is easier for learners to acquire than the [z] rule. Between groups, significant differences were found only between groups E1 and S on (p<.0001), betwee n P and S on (p<.0001), and between P and S on (p=0.0284). It is interesting to see that few differences emerged between groups. It was not entirely unexpected that the group of native English speakers with low or no proficiency in Spanish should perform significantly better on than the group of native Spanish speakers, given that the former group has the same correspondence rule in their L1 ( [z]) as in Portuguese, while the latter group has no phoneme /z/ in their L1. Nor was it particular ly unexpected that the native English speakers with high proficiency in Spanish should perform similarly to the native Spanish speakers, since both groups are fluent in Spanish, a language which is so closely related to Portuguese. It is noteworthy though that those participants with previous exposure to Portuguese should achieve results which are significantly different only from the group of native Spanish speakers. This would suggest that the previous instruction in or exposure to Portuguese generally did not have a great impact on the learners, at least insofar as the GPCRs in question are concerned, and at least at these early levels.

PAGE 82

82 3.3 Other Observations This section highlights observations of interest with regard to production by particular individuals or groups. Correct production for each participant was summed for each grapheme on each test, and calculated as percentages of possible correct items; averages were calculated for each group (Appendix I ). In this section, individual participants are referred to by group and number, for ease of reference. For instance, participant 1, in the group of participants with previous exposure to Portuguese, is referred to as P 1. 3.3.1 Production by Test and Group First, it is noteworthy that none of the 52 participants produced all of the 48 tokens correctly, on any of the three tests, indicating room for improvement for all of the participants at the end of the semester of data collection. While this result is not unexpected for beginners of Portuguese (enr olled in either of the two Portuguese courses), it is interesting that this is the case even for the participants who had had exposure to or instruction in Portuguese prior to the semester in which the data were collected. 3.3.2.1 Production by native Span ish speakers Contrary to expectations, the participant with the highest score on Test 1, with 71% accuracy, was a native Spanish speaker (S 39). His score went down slightly on Test 2 (63%), and remained the same on Test 3 (63%). Similarly, the participant with the highest score by the end of the semester was a native Spanish speaker (S 37), although this time performance improved from 6% accuracy on Test 1 to 35% on Test 2, and 83% on Test 3. Of the group of native Spanish speakers, S 19 achieved moderate accuracy on all three tests: 42%, 65% and 54%, and S 50 showed a dramatic improvement between the first two tests only: 0%, 40% and 46%. Production by these four native Spanish speakers was certainly not the norm for the group of native Spanish speakers as a whole. As expected, most of the other participants in group S had

PAGE 83

83 much less success in accuracy: S 2 achieved 4%, 13% and 35% on Tests 1, 2 and 3, respectively; S 3 had 0%, 4% and 17%; S 4 achieved 8%, 10% and 6%; S 16 obtained 2%, 2% and 4%; S 43 achie ved 0%, 4% and 19%; S 45 had 15%, 19% and 23%; and S 48 achieved 15%, 25% and 10%. The only native Spanish speaker to have had previous instruction in or exposure to Portuguese (P 1) accurately produced 8%, 25% and 44% of items at each of the three tests It is worth pointing out that the two participants who had the highest scores on the first and third tests (S 39 and S 37, respectively) were the only ones who had had previous instruction in another language other than Portuguese ( French, in both cases ) The exposure that th ese participants had to French may have been a contributing factor to their increased success in the tests, in comparison with the other participants in group S. It is somewhat surprising that they performed better than the native Spa nish speaking participant who had had previous instruction in Portuguese, as it would seem more plausible to assume that previous instruction in the target language (Portuguese) would be more advantageous for the acquisition of Portuguese GPCRs than previo us instruction in another language (French). Of course, individual factors may come into play here, including length of exposure to and proficiency in the language in question, as well as attention to phonological details. These are factors that will have to be left for future studies. It is also important to mention here that the group of native Spanish speakers varied with respec t to their exposure to and instruction in Spanish and English. Some had spent a significant portion of their lives in Spanish sp eaking countries (Colombia, Venezuela, El Salvador, Bolivia ), while others had lived largely in the U.S. These differences in upbringing and instruction, not to mention other individual differences, no doubt also impacted their proficiency in English, with some learning English as an L2 and others as an L1 (along with Spanish). The English reading

PAGE 84

84 task, which was administered at the time of the first test, was given to assess participants knowledge of the English GPCRs and their ability to produce the targ et graphemes/phonemes Table 3 11 gives the results of the English task, with respect to the sound [z], for the native Spanish speakers, along with a brief description of their background in Spanish (Results for the other groups, which consisted of native English speakers, are not shown or discussed, because their production was as expected for native speakers.) There were eight English words in the task which are pronounced with [z], containing either the grapheme ( gymnasium physics present result ) or the grapheme ( bizarre realization, zebra zipper ); the results in the table show how many of these words each participant produced accurately out of the eight Also shown in the table is the highest level of accuracy that the participants achieved in the Portuguese reading tasks, as mentioned previously in this section. The purpose of considering together the highest level of accuracy in Portuguese production and the score on the English reading task is to illustrate that accuracy in English did no t necessarily correspond to accuracy in Portuguese. Participants who managed to produce all of the English [z] items correctly (19, 29, 45 and 48) achieved very different levels of accuracy in Portuguese ( 65% 71%, 23% and 25% respectively). Also, those who did not produce any of the English [z] items accurately (2, 4, 16) achieved different levels of accuracy in Portuguese (35%, 10%, and 4%, respectively). Thus, despite obvious differences within this group, it would be difficult to come up with clear cu t criteria for dividing the group further. Hence, the groups production is still considered as a whole, in future discussions, but with the knowledge that its participants are heterogeneous in many respects. 3.3.2.2 Production by native English speakers w ith no or low proficiency in Spanish As for the native English speakers with no or low proficiency in Spanish (group E1), a great deal of diversity was observed in terms of accurate production. On Test 1, for instance,

PAGE 85

85 accuracy ranged from 2% (E1 30) to 67% (E1 20). This variability may result from differences in proficiency in Spanish35Of the other 26 participants in this group, 11 had scores that remained the same or increased from one test to another. The other 15 participants had different combination s of increases and decreases between tests (e.g., an increase from Test 1 to Test 2, and a decrease from Test 2 to Test 3; a decrease from Test 1 to Test 2, followed by an increase between Tests 2 and 3; and so on.). Differences between scores for Tests 1 and 3 were both positive and negative: E1 18, for instance, increased from 19% on Test 1 to 63% on Test 3, while E124 decreased from 35% on Test 1 to 21% on Test 3. These fluctuations in scores between tests are reflected in all of the language groups. Of the 52 participants, 23 consistently improved their scores on the three tests. The other participants showed various permutations of increases, decreases and equivalent scores between the three tests, which is to be expected from participants from intact classes. since the former participant achieved 90.91% accuracy on the Spanish reading task completed on the first test, while the latter participant achieved only 50% accuracy on th e Spanish reading task. Of this group of native English speakers (E1), E120 achieved the highest score on Test 1, with 67%, and also achieved the highest scores of the group on Tests 2 and 3, with 71% and 79%, respectively. While gains for this participant may not have been great from test to test, his scores did increase, nevertheless. 3.3.2.3 Production by native English speakers with high proficiency in Spanish The group of native English speakers with high proficiency in Spanish (E2) is interesting to consider because of the similarities and differences within the group. Fou r of the five participants in this group show similar levels of accuracy in the three tests with the level being 35 The effect of the L1 and L2 will be discussed in detail in Chapter 4, whic h deals with the second research question, regarding interference.

PAGE 86

86 quite low on the first test, followed by noticeable rises on the subsequent two tests: E2 10 achieved 15%, 27% and 56%; E212 had 8%, 31% and 46%; E2 13 obtained 8%, 42% and 65%; and E2 23 had 2%, 29% and 52%. The fifth participant (E2 26) shows a markedly different pattern of approximately half of the items produced accurately on all three tests, with accuracy levels of 52%, 54% and 52% on Tests 1, 2 and 3, respectively. An explanation for this may lie in the level of Spanish proficiency of these five individuals. The scores of the first four participants (E2 10, E212, E213 and E223) on the Spanish reading task given on Test 1 were all conside red appropriate for proficient speakers of Spanish (100%, 90.91%, 100% and 100%, respectively), while the score for the fifth participant (E226) on the Spanish reading task was considerably lower (59.09%), and not at all what would be expected from a flue nt Spanish speaker. It bears repeating here that the grouping of the native English speakers into E1 and E2 was based on the class in which they were enrolled beginner Portuguese and accelerated Portuguese, respectively and not on the information they supplied on their language background questionnaire or on their performance in the Spanish reading task on Test 1. 3.3.3 Production by Grapheme The average accuracy on items, across all participants, for Tests 1, 2 and 3 is 60%, 73% and 80%, respective ly, while the average accuracy on items is 17%, 26% and 30%. Production of each of the graphemes will now be discussed. 3.3.3.1 Production for grapheme It is helpful to remember that all of the participants were fluent, if not native, English spea kers, enrolled in an English speaking post secondary institution. Also, in English is pronounced [z], as in Portuguese, so it is understandable that participants should perform better on items than on items (which may be realized as [z], [s] or [ ] in English, and is almost always realized as [s] in Spanish), as has already been discussed previously in this chapter.

PAGE 87

87 However, it is still interesting to observe that not all participants achieve a high level of accuracy on items, of which there were 19. Only six of the 52 participants achieved 100% accuracy on these tokens on Test 1, while eight did so on Test 2, increasing to 12 participants on Test 3. (Five participants achieved 100% accuracy on all three tests.) For the native English speaker s with no or low proficiency in Spanish (E1), it might be expected that the scores would be consistently high, since English and Portuguese have the same correspondence rule for , but this is not necessarily the case. The average levels of accuracy for group E1 were 74%, 84% and 88% on Tests 1, 2 and 3, respectively. Of the six aforementioned participants who achieved 100% accuracy on Test 1, five belong to group E1. However, there were several participants with low accuracy on this test, including E130 who did not produce a single token correctly. One possible explanation for this lack of production of [z] lies in the participants accuracy on the Spanish reading task : 90.91%. Although the participant was enrolled in the beginner Portuguese class, a long with all of the other participants in group E1, she apparently had an excellent grasp of Spanish orthographic phonological correspondence rules (in fact, out of all the participants in E1, she had the highest percentage of accuracy on the Spanish read ing task), which seems to be the most plausible explanation for her completely inaccurate reading of the Portuguese words on the first test. Interestingly, the participant seems to have overcome this negative effect relatively well by the second and third tests, when she achieves 84% and 74% accuracy, respectively. Another anomalous participant, E1 24, had accurate production on 74% of items on Test 1, then dropped to an alarming 21% (the second lowest level of accuracy in this group, on any of the tests, following the aforementioned 0% on Test 1), rising to 47% in Test 3. In this instance, however, a plausible explanation for this performance is not easily found in the participants linguistic background, consisting of four years of Spanish in high school (accuracy of 68.18% in

PAGE 88

88 the Spanish reading task), as well as three semesters of Italian in college and time spent in Italy. It is possible that this individuals three background languages (English, Spanish and Italian) interacted in such a way as to ca use her confusion, and in each test she may have been testing different hypotheses with regard to the orthographic phonological correspondence rules in Portuguese. The native English speaking participants with high proficiency in Spanish (group E2) had low er averages on the first two tests than the other group of native English speakers (33% and 65% for E2, compared with 74% and 84% for E1), but both groups averaged roughly the same level of accuracy on Test 3 with 87% for E2 and 88% for E1. The participant s with previous instruction in or exposure to Portuguese (group P) tended to achieve relatively high levels of accuracy on all the tests, averaging 77%, 83% and 89% on Tests 1, 2 and 3, respectively. This was not really unexpected, although there were two participants in this group whose performance stood out. P 1 scored low, relative to the others 21%, 37% and 68% which may be attributed to her native Spanish status and to reportedly little instruction in Portuguese (only five months in high school, af ter which the class was cancelled due to instructor illness). This participants production was more accurate than that of most of the participants in the native Spanish group (S), which supports her classification in group P, although her level of accurac y in Portuguese was quite a bit lower than that of the other participants in this group. P 42 also had relatively low accuracy on Test 1 (37%), with close to average scores on Tests 2 and 3 (74% and 95%, respectively). The below average score on the first test is not easily explained, as the participant claimed to be a native speaker of Portuguese and to speak the language with relatives and at home with her mother. She did not claim to have previous instruction in Portuguese, however, so

PAGE 89

89 it is possible tha t her learning of Portuguese (including pronunciation) had never been in a formal setting where correction might be offered.36The group of native Spanish speakers (S) had lower average accuracy than any of the other groups for items, on all three tests (22%, 40% and 52%, on Tests 1, 2 and 3, respectively). Two participants achieved much higher accuracy than the average on all three tests: S 19 (58%, 89% and 84%) and S 39 (95% on all three tests). There does not seem to be an immediately apparent reason why S 19 did much better than the majority of the other participants in the group. According to his language background questionnaire, he had lived in a Spanish speaking country until the age of 5 and had completed Spanish to IB level at school. It is inte resting to note that his accuracy on the English reading task on Test 1 was 100%, unlike most of his fellow native Spanish speakers. This level of accuracy in English may have had a positive effect on his production of items in Portuguese, especially o n the first test, when data were collected before any instruction on Portuguese pronunciation had been given in the classroom. As for participant S 39, whose scores for items were consistent across all three tests, it is possible that his high accuracy in English (100% on the English reading task), coupled with four years of French in high school, positively affected his production in Portuguese. 3.3.3.2 Production for grapheme Considering the items, the participants were generally less accurate on these than on the items.37 36 After data analysis, this participant subsequently enrolled in a Portuguese class taught by the researcher, whose opinion it is that this individual was by no means a true native speaker of the language, as claimed on her language background questionnaire. The participants accent, vocabulary and grammar were sorely lacking in comparison with what would be expected of a native speaker. Perhaps a better descriptive term for this participant would be heritage learner, given her familial exposure to the language and yet non native like production. The highest accuracy for items was 55% (S 39) on Test 1, 59% on Test 2 37 This probably has a lot to do with the fact that not only does intervocalic in Spanish have a different phonological correspondence rule fr om Portuguese ([s] and [z], respectively), but in many cases in English does as well. (Remember that, in English, intervocalic has three possible phonological correspondences: [z] as in

PAGE 90

90 (E1 20 and P 29), and 79% on Test 3 (S 37). With this 79% accuracy rate as the highest on items, it is clear that no participant came close to achieving full accuracy on any of the three tests, while for items there were 25 instances of participants achieving 100% accuracy. There are a few interesting observations to be made regarding production on items. Group P averaged higher accuracy on the three tests than the other three groups, and group S tended to have very low accuracy (about a third of the scores for S were zero). S 16 had 0% accuracy for on all three tests, S 4 achieved one correct token on the first test (3% accuracy) but none on the second and third tests, and S 3 had 0% accuracy on the first two tests and then 7% (2 correct items) on the third test. These low scoring tendencies are not observed in any of the other groups, but there are two participants whose accuracy lev els of 0%, 0% and 10% (E127), and 0%, 3% and 17% (E128) are somewhat surprising. It is noteworthy that these two participants (twin sisters) reported exposure to Spanish all through school and K IB, respectively, and achieved 81.82% and 86.36% accura cy on the Spanish reading task. This may account for production which is more similar to that of the native Spanish speakers than that of the other native English speakers (from both groups E1 and E2). Interestingly, these two participants did not score pa rticularly low on items, suggesting that their proficiency in Spanish played a greater role with items than with items. This closer examination of the variation in production by individuals and groups in this section has complemented the statis tical results given in the previous section. The two have shown general progress by the participants in their production of the target Portuguese sounds, raisin [s] as in basin and [] as in Asian .) Without examini ng specific word types at this point, it is not possible to identify participants performance on the tokens found in cognate words where English would have different phonological realizations. Hence, the present discussion will consider only overall a ccuracy on items, and the more detailed discussion, taking word types into account, will follow in Chapter 5.

PAGE 91

91 while also illustrating that there is still room for improvement by all participants for one or both g raphemes. 3.4 Summary With the original research questions on error resolution in mind, the results presented in this chapter supported Majors (2001) OPM, which predicts that the idealized learners IL will move towards an end state consisting solely of L 3. While such an end point was not achieved by this group of learners as a whole, given that overall accuracy at Test 3 was 50%, there was evidence to show significant progress between tests. In terms of differences between participant groups, the predict ion that the more proficient Spanish speakers would have greater difficulty with acquisition of the Portuguese GPCRs was partially supported. Certainly the native Spanish speakers tended to produce the target Portuguese sounds less accurately than the nati ve English speakers with low proficiency in Spanish. However, the native English speakers with high proficiency in Spanish only tended to have less accuracy in Portuguese pronunciation initially, and by the end of the semester their level of accuracy was n ot significantly different from that of any of the other groups. With respect to the differences between the two graphemes in question, the hypothesis that there would be significant increases between tests for both graphemes was not supported, as the test by grapheme interaction was found not to be significant. However, the prediction that learners would have different levels of accuracy for the two graphemes was supported, with produced significantly more accurately than by all participant groups. Thus, not only was evidence found to support the literature that shows the persistent problem that Spanish speakers have in devoicing Portuguese , but evidence was found to show that English speakers also appear to have this problem.

PAGE 92

92 In the examination of individual variation, disparity in production was found within and across all groups, as well as increases and decreases from one test to another. No participant achieved 100% accuracy on all items for any test, although some participants did correctly produce all items in the three tests. Items with intervocalic appeared to be more problematic for participants, most likely due to interference from their L1 and L2, a discussion which will be taken up in detail in the next chapter.

PAGE 93

93 Table 3 1. N umber and percentage of correct items for all [z], by test Production Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 % All [z] 845 33.85 1118 44.79 1244 49.84 Table 32. Participant groups Group Group Description No. E1 N ative English speakers w ith no / low proficiency in Spanish (beginner class) 21 E2 Native English speakers with high proficiency in Spanish (accelerated class) 4 P Participants with previous exposure to or instruction in Portuguese 9 O Participants with exposure to or instructio n in other foreign languages 9 S Native Spanish speakers 9 Table 3 3. Number and percentage of correct items for all [z], by group and test Group Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 % E1 430 42.66 518 51.39 513 50.89 E2 37 19.27 73 38 .02 108 56.25 O 127 29.40 197 45.60 246 56.94 P 210 48.61 243 56.25 274 63.43 S 41 9.49 87 20.14 103 23.84 Table 3 4. ANOVA for all production Effect Num DF F Value P value group 4 13.71 <.0001 test 2 93.03 <.0001 grapheme 1 1091.61 <.0001 wor d type 2 7.66 0.0005 group *test 8 6.84 <.0001 group grapheme 4 44.46 <.0001 group word type 8 1.07 0.3808 test* grapheme 2 4.73 0.0091 test* word type 4 0.68 0.6093 grapheme word type 2 4.88 0.0078 group *test* grapheme 8 1.88 0.0596 group *test* wo rd type 16 0.64 0.8556 test* grapheme word type 4 0.80 0.5255 group *test* grapheme word type 24 0.79 0.7477

PAGE 94

94 Table 35. Revised participant groups Group Group Description No. E1 N ative English speakers with no /low proficiency in Spanish (beginner clas s) 25 E2 Native English speakers with high proficiency in Spanish (accelerated class) 5 P Participants with previous exposure to or instruction in Portuguese 9 S Native Spanish speakers 11 Table 3 6. Number and percentage of correct items for all [z] by revised group and test Group Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 % E1 516 39.81 653 50.39 667 51.47 E2 41 17.08 88 36.67 130 54.17 P 210 48.61 243 56.25 274 63.43 S 78 14.77 134 25.38 173 32.77 Table 3 7. ANOVA for all productio n (revised groups) Effect Num DF F Value p value group 3 80.55 <.0001 test 2 53.73 <.0001 grapheme 1 616.02 <.0001 word type 2 3.29 0.0378 group *test 6 3.95 0.0007 group grapheme 3 33.38 <.0001 group word type 6 0.68 0.6622 test* grapheme 2 2. 92 0.0547 test* word type 4 0.38 0.8242 grapheme word type 2 3.01 0.0500 group *test* grapheme 6 1.47 0.1853 group *test* word type 12 0.30 0.9900 test* grapheme word type 4 0.66 0.6171 group *test* grapheme word type 18 0.57 0.9241 Table 3 8. Significant differences found between tests, by group Group Tests 1 & 2 Tests 1 & 3 Tests 2 & 3 E1 <.0001 <.0001 E2 0.0004 <.0001 0.0014 P 0.0003 S 0.0093 <.0001

PAGE 95

95 Table 3 9. Significant differences found between groups by test E1 E2 P S E1 0.045 2 (Test 1) <.0001 (Test 1) <.0001 (Test 2) 0.0198 (Test 3) E2 0.0096 (Test 1) P <.0001 (Test 1) 0.0002 (Test 2) 0.0004 (Test 3) S Table 3 10. Number and percentage of correct items by grapheme and group Group # % # % E1 1 261 82 575 24 E2 176 62 83 19 P 426 83 301 38 S 237 38 148 15 Table 3 11. Production of [z] by native Spanish speakers in English reading task No. Background in Spanish English Portuguese 1 Dominican Rep.; Miami; Span. classes in middle school and H S 7 44% 2 Colombia 0 35 % 3 Up to IB in HS; relatives 6 17% 4 From Venezuela 0 10% 16 K 10; El Salvador till 15 0 4% 19 Venezuela till 5; IB Spanish 8 65% 37 K 12, Colombia 6 83% 39 4314 (advanced Spanish class at university); Mexico for 2 yrs 8 71% 43 Graduated HS in Colombia 1 19% 45 Since 5th grade 8 23% 48 Bolivia for 14 yrs 8 25% 50 Studied for 6 yrs; Colombia 2 46 %

PAGE 96

96 Figure 31. Percentage of correct items by test Figure 32. Percentage of correct items, by test for each group Figure 33. Percentage of correct items, by group for each test

PAGE 97

97 Figure 34. Percentage of correct items, by grapheme for each group Figure 35. Percentage of correct items, by group for each grapheme

PAGE 98

98 CHAPTER 4 INTERFERENCE RESULTS AND DISCUSSI ON 4.1 Introduction This chapter addresses the second overarching research question, which asked about the role of interference, or language transfer, in acquisition. Before presenting and discussing the relevant results, it is helpful to recall the research que stions asked and hypotheses stated (in section 2.2.2) concerning interference. The first question asked what evidence there was of interference from the L1 and/or the L2, and predicted, according to Majors (2001) OPM, that there would be evidence of inter ference from both the L1 and the L2, with these reducing over time. It was also expected, given the U component in the model, that there would be evidence of production which was not consistent with the grammars of the L1, the L2 or the L3. The second que stion considered whether there would be differences in interference between participant groups, and it was anticipated that the different groups would show varying degrees of transfer from English and Spanish, depending on their native language and on thei r level of proficiency in Spanish. It was predicted that native English speakers with low or no proficiency in Spanish would show greater transfer from English (at least initially), while native English speakers with high proficiency in Spanish and native Spanish speakers would show significant transfer from Spanish (at least initially), due to their greater experience with a language more closely related to Portuguese than English. Thus, the prediction was that the more typologically similar language, Span ish, would be the basis of transfer where proficiency in that language was high, while English would be the basis of transfer where proficiency in Spanish was lower, according to Hammerbergs (2001) belief that there must be a certain level of proficiency in the L2 in order for it to be a basis of transfer to the L3 and according to findings in the literature regarding the facilitation of transfer between similar languages.

PAGE 99

99 The third question explored whether differences in interference existed between the two graphemes and , and it was hypothesized that there would indeed be differences due to the GPCRs of English, Spanish and Portuguese for the two graphemes in question. Furthermore, it was predicted, according to Muller and Mullers (1968) observa tions, that the more difficult Portuguese GPCR to acquire would be the [z] one, since it is different from (most of) the GPCRs in Spanish and English for intervocalic (although English does have the [z] correspondence rule for some lexical item s), and Portuguese [z] would be easier to acquire since English has the same rule. Thus, evidence of positive transfer from English (or acquisition of Portuguese) was expected for , while evidence of negative transfer from English and/or Spanish was expected for . The final question explored the extent to which accuracy of production in Spanish could be said to be related to acquisition (cor rect production) of Portuguese. I t was expected, based on the researchers own classroom motivated impressi ons, as well as observations in the literature regarding the difficulties encountered by Spanish speakers learning Portuguese pronunciation, that there would be an inverse relationship between accuracy of production in Spanish and production of the Portugu ese GPCRs in question. That is, it was predicted that the greater the proficiency in Spanish (as measured by accuracy in production in the Spanish reading task) the lower the accuracy in production in Portuguese would be. The results will be presented and discussed by grapheme and participant group. Following will be an analysis of participants production in the Spanish reading task on the first test, compared to their production in the Portuguese reading tasks.38 38 Production according to word type (cognate, non cognate and nonce) will be discussed in Chapter 5, which addresses the third research quest ion, about generalizability. In the section with other

PAGE 100

100 observations pr oduction of individual participants and words will be examined, and finally the chapters results will be summarized. 4.2 Results and Discussion Seven production types were possible for grapheme , while only four were possible for grapheme . Hence, the results for the two graphemes are presented and discussed separately. 4.2.1 Production for Grapheme Results are presented in terms of raw production scores, percentages of items for each production type (out of all items on a test), and statistically significant differences (output from the model) which are marked with an asterisk in the tables For , seven production types were possible: EN, EP, ES, NO, OTH, PO and SP. The first, coded as EN (English), represents [ ], and it could occur when t he English pronunciation rule differed from the Spanish and Portuguese rules. Its occurrence was logically limited to the six cognate words where the English equivalent would be pronounced with [ ]39The second production type, coded as EP (English Portuguese), represents [z], occurring when the English and Portuguese pronunciation rules coincided with each other but differed from Spanish. This production type was limited to the f ive cognate words where the English equivalent would be pronounced with [z] and it was taken to indicate evidence of negative tr ansfer from English. 40The third type of production, coded as ES (English Spanish), repre sents [s], occurring when the English and Spanish pronunciation rules coincided with each other but differed from and it was considered to be indicative of positive transfer from English and/or acquisition of Portuguese. 39 Audiovisual (audiovisual), deciso (decision), divises (division), reviso (revision), televiso (television), viso (vision). 40 Apresentao (presentation), esquisito (a false cognate meaning odd), museu (m useum), presidente (president), visvel (visible).

PAGE 101

101 target Portuguese. This production type was possible for 18 words: the six cognate words where the English equivalent would be pronounced with [s]41, the six non cognate words and the six nonce words42The fourth production type, coded as PO (Portuguese), represents [z], and it was possible when the Portugue se rule differed from the English and Spanish rule(s). This was possible for 24 words: the six cognate words which would be pronounced with [ ] in English and with [s] in Spanish, the six cognate words which would be pronounced with [s] in both English and Spanish, the six non cognate words, and the six nonce words. This production type would indicate evidence of acquisition of Portuguese pronunciation rules, rather than interference from English or Spanish. This production type suggested negative transfer from English and/or Spanish. The fifth production type, coded as SP (Spanish), represents [s], and it could occur when the Spanish rule differed from the English and Portuguese rule(s). This was possible for 11 words: the five cognate words which would be pronounced with [s] in Spanish but with [z] in English and Portuguese, and the six cognate words which would be pronounced with [s] in Spanish but with [ ] in English and with [z] in Portuguese. The final two types of production possible were when participants produced nothing for the target grapheme (coded a s NO), as a result of a misread word, a skipped syllable or a skipped word, or when participants produced another sound (coded as OTH), which was unexpected 41 Bsico (basic), casos (cases), curiosidade (curiosity), filosofia (philosophy), generosidade (generosity), persuasivo (persuasive). 42 The default pronunciation for in English was a ssumed to be [s], since this is the sound that occurs in most of the contexts where the grapheme appears, while pronunciation of as [z] or [ ] in English is idiosyncratic and must be learned. Thus, it was assumed that if participants transferred En glish pronunciation rules, the in the Portuguese non cognate and nonce words would be pronounced as [s], since these words should not have triggered any idiosyncratic pronunciation from English as cognate words with [z] and [ ] in English might hav e done.

PAGE 102

102 given the pronunciation rules of English, Spanish and Portuguese for intervocalic . These two pr oduction types, logically, could occur for any of the 29 items. All the sounds produced for words were coded (according to the seven production types), and the codes summed for each test. These sums (raw scores) were then converted to percentages, with these adding up to 100% for each test (Table 41), and grouped according to production type for graphic representation (Figure 4 1). Numerically, slight increases were seen in correct production of as [z] (EP, PO) throughout the semester and sligh t decreases in incorrect production of as [s] (ES, SP), though there was no decrease of as [ ]. Negative transfer (EN, ES and SP) accounted for approximately 81% of the total production on the first test, 73% on the second test, and 69% on the third test. Positive transfer and/or acquisition (EP and PO), on the other hand, accounted for roughly 17% on Test 1, 27% on Test 2, and 30% on Test 3. (NO and OTH accounted for the rest of the production, adding up to only approximately 1% on each of the three tests.) The ANOVA results (Table 4 2) indicated that production type was a significant factor, and significant interactions were found between production type and test, and between production type and participant group. 4.2.1.1 Production for grapheme by production type and test Tukey post hoc t tests revealed a number of significant differences on Test 1 (Table 4 3), Test 2 (Table 4 4) and Test 3 (Table 4 5). Not all of the differences shown to be significant are actually interesting to consider, given that different production types could logically occur for different sets of words. The noteworthy comparisons, here and throughout this chapter, are between production types EP and SP, and ES and PO (marked with + in the tables). The comparison between EP and SP is interesting to consider because these two production types are essentially two sides of the same coin, so to speak. The same can be said for the comparison

PAGE 103

103 between ES and PO. In the first, where English and Portuguese rules coincide (EP), they differ from the Spanish rule (SP); in the second; where English and Spanish rules coincide (ES), th ey differ from the Portuguese rule (PO). Thus, in each pair, there is a production type that corresponds to correct production (EP and PO, respectively) and one that corresponds to incorrect production (SP and ES, respectively), as far as target Portuguese is concerned. There were significant differences found between production types EP and SP on all three tests: SP (347 tokens) differed significantly from EP (113 tokens) on Test 1 (p<.0001); SP (279 tokens) occurred significantly more than EP (148 tokens ) on Test 2 (p<.0001); and SP (275 tokens) differed significantly from EP (146 tokens) on Test 3 (p<.0001). On all three tests, transfer from Spanish was greater than transfer from English/correct Portuguese production. Additionally, there were no signific ant differences for EP production between tests, nor for SP production between tests. This suggests that, for words where the Spanish rule would produce something different from the English and/or Portuguese rules, participants were mainly employing th e Spanish rule, incorrectly producing [s], and this remained relatively constant throughout the semester. Examining the production types ES and PO, the picture is quite similar. Negative transfer from English and/or Spanish (ES) was consistently higher tha n correct Portuguese production (PO) for the relevant words. In the interactions on all three tests, the difference between these two types of production was statistically significant (p<.0001). There were, however, significant differences between tests fo r these two production types. ES showed a significant decrease from Test 1 to Test 2 (p=0.0304), and between Test 1 and Test 3 (p<.0001), while there was a significant increase for PO from Test 1 to Test 2 (p=0.0001), and between Tests 1 and 3 (p<.0001). W hen the correspondence rules in the learners L1 and L2 coincided, but differed

PAGE 104

104 from the L3 rule, negative transfer from the L1 and L2 abounded. Although there was a significant decrease in the negative transfer and a significant increase in target Portugu ese production between the first two tests, no significant gains were made between Tests 2 and 3, suggesting that learners require a good deal more exposure to Portuguese and instruction in pronunciation to overcome the hindrances posed by the L1 and L2, or they fossilize incorrect correspondence rules fairly early on (between Tests 2 and 3). Obviously, longer term studies are necessary in order to discover the extent of interference from the L1 and L2. 4.2.1.2 Production for grapheme by production typ e and group Considering the production type by group interaction for , those comparisons to be discussed are between participant groups for the same production type (e.g., groups E1 and E2 for type EP), and between different production types for the sam e group (e.g., EP and SP for group E1). The production types for each language group were summed for all three tests (raw scores) and then calculated as percentages of the total production43 Tukey post hoc t tests were carried out, and a number of significant differences were found. These results are given first according to production type for all groups. N o significant differences between groups were found for production types EN, NO or OTH; the production types found to have significant inter group differences were EP, ES, PO and SP. For production type EP (Table 4 7), the only significant differences foun d were between E1 and S (p=0.0137) and between P and S (p=0.0104). Recalling that EP is the production type that occurs when English and Portuguese rules coincide (but differ from the Spanish rule), and remembering that such that each groups production adds up to 100% (Table 46). Th ese results are also shown graphically for the four participant groups (Figure 42). 43 Raw scores and percentages are not shown for each test because the ANOVA did not find the three way interaction between test, production type and group to be significant.

PAGE 105

105 E1 is the group of native English sp eakers with low or no proficiency in Spanish, and P is the group of participants with previous instruction in Portuguese (most of whom would be in E1 were it not for their prior exposure to Portuguese), it is perhaps to be expected that these significant d ifferences should exist. This seems to suggest that groups E1 and P are transferring from English on these words (or showing acquisition of the Portuguese rule) more than group S does. It is also noteworthy that there were no significant differences betwee n group E2 and any of the other groups. It would seem that the overall production of this group falls sufficiently between that of groups E1 and P (on the high end) and group S (on the low end) to render any differences insignificant.44Production type ES ( Table 4 8) shows group P differing significantly from groups E1 (p=0.0005) and S (p=0.0013). Given what was found for production type EP, it would not be illogical to assume that group E1 participants are transferring mainly from English where ES occurs fo r them, and S participants are most likely still transferring from Spanish in instances where ES occurs. What is interesting, again, is that there were no significant differences found between E2 and the other groups. Recalling the percentages for ES production for the different groups (Table 4 6 and Figure 42), it is clear that P showed the least amount of interference from English and/or Spanish on the words where the English and Spanish rules coincided (but differed from the Portuguese rule). S showed t he greatest amount of interference (presumably mainly from Spanish), E1 showed the second highest amount of interference (believed to be primarily from English), and E2 production was between E1 and P in terms of interference. This seems to suggest that pr evious Portuguese has indeed been a help to group P in terms of enabling correct 44 This result might be explained somewhat if test were considered, as the production of E2 more closely resembled that of S initially, while later on E2 production was more similar to that of groups E1 and P.

PAGE 106

106 production of Portuguese, and for some reason group E2 also seems to be at less of a disadvantage than E1 and P when it comes to the hindrance caused by the L1 and/or L2. Prod uction type PO (Table 4 9), which is approximately the counterpart of ES, showed significant differences between group P and the other three groups (p<.0001). This is an interesting result, for two principal reasons. First, it validates the creation of a s eparate group for those with previous instruction in Portuguese, in that these participants differed significantly from the other groups in their production (at least of items). Second, it suggests that instruction in Portuguese does make a difference when it comes to acquisition of pronunciation rules. Group P significantly outperformed the other three groups (the percentage for P on this production type was almost double the percentage of the next highest group), demonstrating better production of the Portuguese correspondence rule for intervocalic than was found with the other groups. (The fact that there is a statistical difference seen here for group P which was not observed in the previous chapter, concerning error resolution, is due to the fac t that here the data are broken down further than in the previous chapter, into different possible production types, allowing a more detailed perspective on what different participant groups are producing in different contexts.) Differences on the last pr oduction type, SP (Table 4 10), the approximate counterpart of EP, showed those groups consisting primarily of native English speakers with low or no Spanish proficiency (E1 and P) differing from those groups with proficient Spanish participants (E2 and S) These results would seem to indicate that E1, P and S are transferring significantly from their L1 (English for E1 and P, and Spanish for S), while E2 seems to be transferring significantly from the L2 (Spanish). This would indicate that where proficienc y in the typologically more similar language (here, Spanish) is higher, transfer from that language (be it from the L1 or L2)

PAGE 107

107 to the L3 is higher, suggesting that in the acquisition of related languages, typological similarity or relatedness is more important than language status (L1/L2). Consideration will now be given to comparisons between different types of production within each of the four participant groups. Those comparisons of interest (marked with + in the tables) are EP SP and ESPO. For group E1 (Table 4 11), there were significant differences for the two key comparisons EP SP and ESPO (for both, p<.0001). ES production was significantly greater than PO production. While this finding is not particularly unexpected, given that ES is indicative of transfer from English and/or Spanish (the L1 and/or the L2), it is noteworthy that SP production was significantly greater than EP production, which indicates negative interference from Spanish (this groups L2). Group E2 (Table 4 12) also had significa ntly greater SP production than EP production (p<.0001), suggestive of more negative transfer from Spanish than positive transfer from English (and/or acquisition of the Portuguese rule). Additionally, this group showed significantly greater production of ES than PO (p<.0001), with negative transfer from English and/or Spanish occurring more often than target Portuguese production. For group P (Table 413), there was no significant difference between EP and SP (p=1.0000), indicating that positive transfer from English (and/or acquisition of the Portuguese rule) and negative transfer from Spanish were roughly equivalent for the relevant words. However, as for groups E1 and E2, there was a significant difference between ES and PO (p<.0001). Although group P s howed a great deal more of production type PO than did the other groups, the fact remains that negative transfer from English and/or Spanish was still occurring abundantly, even for the group with previous instruction in Portuguese.

PAGE 108

108 Group S (Table 4 14) sh owed significant differences on both of the comparisons under consideration (EP SP and ES PO). In both cases, S showed more negative transfer from Spanish (SP and ES) than positive transfer from English and/or target Portuguese pronunciation (EP and PO). G iven that this is the group of native Spanish speakers, it is not unexpected that the bulk of their production should show transfer from Spanish (or possibly English, for production types ES and EP). It is encouraging, though, that they achieve some target Portuguese pronunciation (evidenced clearly by production type PO). 4.2.2 Production for Grapheme With these findings for the grapheme in mind, production for the grapheme will be examined in this section Results are presented in terms of raw production scores, percentages of items for each production type (out of all items on a test), and statistically significant diffe rences (output from the model), which are marked with an asterisk in the tables. For this grapheme, there was a reduced numbe r of possible production types, since English and Portuguese orthographic phonological rules always coincide to produce [z], and Spanish always differs, with its rule of [s] (or [ ] in Peninsular Spanish).45There were, therefore, four possible production types for this grapheme: EP, NO, OTH, and SP. All four of these types could occur on all of the 19 words, as participants produced a sound expected according to the English an d/or Portuguese phonological systems (EP), according to the phonological system of Spanish (SP), according to none of the three systems (OTH), or no sound at all for the target grapheme (NO). All the sounds produced were coded according to these four produ ction types, and the codes summed for each test. These sums (raw 45 The voicel ess interdental fricative [ the first test; all were allocated the production code SP.

PAGE 109

109 scores) were then converted to percentages, with these adding up to 100% on each test (Table 4 15), and grouped according to production type for graphic representation (Figure 43). Numerica lly, increases in correct production of as [z] (EP) were seen throughout the semester, as were decreases in incorrect production of as [s] (SP), while the marginal occurrences of NO and OTH remained roughly the same across all three tests (1 2% for each type ). Negative transfer from Spanish (SP) accounted for 37% of the total production on the first test, 23% on the second test, and 16% on the third test. Target Portuguese (EP), on the other hand, accounted for 60% on Test 1, 73% on Test 2, and 80% on Test 3. (NO and OTH accounted for the rest of the production, with both types adding up to approximately 3% on Tests 1 and 3, and 4% on Test 2.) The ANOVA results (Table 4 16) indicated that production type was a significant factor, and the interaction between production type and test was significant, as was the interaction between production type and participant group. The interaction between group and test was not found to be significant, but the three way interaction between group, test and productio n type was significant for . Tukey post hoc t tests revealed a number of significant differences on Test 1 (Table 4 17), Test 2 (Table 418) and Test 3 (Table 419). Those comparisons of interest are between production types EP and SP (marked with + in the tables). Despite what appears to be a sizeable numerical difference between EP and SP on the first test, no statistically significant difference was found between these production types. However, they did differ from each other on Tests 2 and 3 (p<.0001), with EP occurring significantly more than SP. These results suggest that there was no statistically significant difference between transfer from English and transfer from Spanish before instruction in Portuguese (i.e., at Test 1),

PAGE 110

110 but once instruction took place, negative transfer from Spanish decreased while acquisition of Portuguese GPCRs increased (possibly facilitated by positive transfer from English) Between tests, there were some significant differences found for EP: between Tests 1 and 2 (p=0.0006), and between Tests 1 and 3 (p<.0001). However, there was no significant difference found between Tests 2 and 3 (p=0.1343), suggesting that greater gains were made earlier in the semester. The pattern was similar for SP, with significant differences b etween Tests 1 and 2 (p=0.0009), and between Tests 1 and 3 (p<.0001), but no significant difference between Tests 2 and 3 (p=0.1242), indicating that negative transfer decreased more markedly at the beginning of the semester. Although instruction was not a factor that was specifically taken into account in this study, it does seem to have had an effect. 4.2.2.1 Production for grapheme by group and type Considering the interaction between participant group and production type, the comparisons of interest are between groups for the same production type (e.g., groups E1 and E2 for type EP), and comparisons between different production types for the same group (e.g., EP and SP for group E1). The production types for each group were summed for all three tests (raw scores) and then calculated as percentages of the total production, such that each groups production adds up to 100% (Table 420). These results are also shown graphically for the four participant groups (Figure 44). Tukey post hoc t tests were c arried out to determine where the significant differences could be found. These results are given first according to production type for all groups. No significant differences between groups were found for production types NO or OTH; the production types t o have significant inter group differences were EP and SP. For production type EP (Table 4 21), there were a number of significant differences to be considered. In fact, the only inter group difference which was found not to have statistical significance w as the one

PAGE 111

111 between E1 and P. Recalling that most of the participants in group P (with previous instruction in Portuguese) would otherwise have been included in group E1, it is perhaps not unexpected that their production should be similar when considered a cross all three tests. It is also perhaps not unexpected that these two groups should behave differently from the groups of speakers proficient in Spanish (E2 and S). It is noteworthy that the two groups of Spanish speakers, however, should differ signific antly from each other. It seems that the different L1s of these two groups may have played a role in their Portuguese production, despite the similarity they shared in their proficiency in Spanish. For production type SP (Table 4 22) the only inter group difference that was not statistically significant was that found between groups E1 and P. Again, it seems reasonable that these two groups, comprised primarily of native English speakers with low or no proficiency in Spanish, should have less negative tran sfer from Spanish than the other groups, comprised of proficient Spanish speakers. It is also worth highlighting that a significant difference was found between the two groups of proficient Spanish speakers (E2 and S), where the native Spanish speakers sho wed more negative transfer from Spanish than did the non native Spanish speakers. Consideration will now be given to comparisons between different types of production within each of the four participant groups. The noteworthy comparisons (EP SP) are again marked with + in the tables. The EP SP comparison showed statistically significant differences for all of the participant groups. However, the differences were not the same for all groups. That is, for groups E1 (Table 4 23), E2 (Table 424) and P (Table 425), production of EP was significantly greater than production of SP, while for group S (Table 4 26), the reverse was true production type SP occurred significantly more often than production type EP. Thus, it seems that the L1 was an important factor when it came to production: for those groups where the L1

PAGE 112

112 was primarily English (E1, E2 and P), target Portuguese exceeded negative transfer from Spanish for (perhaps facilitated by the possibility of positive transfer from English), while the group of participants whose L1 was Spanish transferred principally from Spanish for (from where there would be no possibility of positive transfer to L3 Portuguese) These results may be somewhat oversimplified, however, when test is not taken into account. T hus, it is necessary to examine the three way interaction of group, production type and test. 4.2.2.2 Production for grapheme by group, type and test For each participant group, raw scores were summed for each production type and test, and percentages were calculated, with each groups production for each test adding up to 100% (Table 4 27)46For production type EP (Table 4 28), there were a number of significant differences on the first test. Groups E1 and P differed significantly from groups E2 and S, but not from each other, nor did E2 and S differ significantly from each other. That is, the two groups of primarily native English speakers (E1 and P) were seen to differ from the two groups of proficient Spanish speakers (E2 and S), but the groups within the pairs did not differ significantly from each other. By Test 2, the picture was slightly different: E1 and P still differed significantly from S, but they no longer differed significantly from E2; and E2 continued not to differ significantly from S. For ease of reading, tests are indicated in parentheses. As would be hoped for language learners, there were increases in target production of [z] (EP) and decrea ses in incorrect production of [s] (SP) across all groups, between all tests. As is to be expected, groups differed in the proportions of their production, and post hoc Tukey tests revealed those differences which were statistically significant for EP and SP. (Test times are shown in parentheses.) 46 Production for NO and OTH is not shown, for the sake of simplicity, since there were no s ignificance differences between groups or tests for these production types.

PAGE 113

113 This would seem to indicate that the production of E2 was somewhere in the middle in terms of accuracy: it was not quite as accurate as E1 and P, but it was also not quite as inaccurate as S. This progression continued to Test 3: E1 and P remained significantly different from S and there continued to be no significant difference between E1, P and E2, but at this test E2 differed significantly from S. These results would suggest that native English speakers proficient in Spanish (E2) behave more like native Spanish speakers (S) initially than like native Eng lish speakers with no or low proficiency in Spanish (E1). However, by the end of the semester, the E2 participants largely overcame the negative transfer from Spanish, at least to the same extent as the native English speakers in E2 (and P). The native Spa nish speakers in S, however, did not seem as capable of overcoming the negative transfer from Spanish as the native English speakers proficient in Spanish in E2, suggesting that the L1 may be more involved in determining acquisition of the Portuguese GPCR [z] than proficiency in a typo logically similar language. For production type SP (Table 4 29), there was roughly the same significance pattern as for EP, albeit with the raw scores and percentages decreasing for SP where they increased for EP. The only divergence from the previous pattern was that for SP there was no significant difference between groups E2 and S on Test 3 (although it approached significance at p=0.0752). The reason for this non significant difference is unclear, given that there was a significant difference found between E2 and S on Test 3 for production type EP, but it may be related to the relatively small number of participants in group E2. Between tests, o nly four significant differences occurred. For group E2, there were significa nt differences between Tests 1 and 3 for production type EP (p<.0001) and for type SP SP (p=0.0008). For group S, there were significant differences between Tests 1 and 3 for production type EP (p=0.0132) and for type SP (p=0.0055). In the case of both gro ups, EP

PAGE 114

114 production increased and SP production decreased. These results show that, for , groups E1 and P started out with a relatively high level of accuracy, whether due to previous Portuguese instruction (for P) or due to the orthographic phonological rule [z] that is shared by English and Portuguese allowing for the possibility of positive transfer from English in the case of these items Due to this relatively high level of accuracy to begin with, it is not entirely unexpected that there should be no great gains made by these groups for . Also, it is perhaps to be expected that the other two groups, E2 and S, should display lower levels of accuracy initially, given their proficiency in Spanish and the typological similarity between Spanish and Portuguese. It is encouraging, then, to see both of the latter groups making significant gains in accuracy by the end of the semester. 4.2.3 Relationship between Spanish Production and Portuguese Production In this section, the results of the Spanish read ing task given at Test 1 are considered in relation to Portuguese production, in order to examine the hypothesis that accuracy of production in Spanish is inversely related to accuracy of production in Portuguese. The results ( Appendix J ) are presented in ascending order, according to level of accuracy on the Spanish reading task. A comparison between production on the Spanish reading task and an average of the Portuguese production on the three tests is shown graphically in Figure 4 5. (Production on the S panish reading task was first compared to each of the three Portuguese tests separately, but statistical tests did not reveal any significant differences between the three comparisons, so it was decided to calculate an average of the participants producti on on the Portuguese tests and use that average as a basis for comparison with the production on the Spanish task.)

PAGE 115

115 A regression model was used to analyze these results, in order to determine the relationship between participants production on the Spanis h reading task and production on the three Portuguese reading tasks. The results of the regression model are shown in Figure 4 6. The slope relating Spanish to Portuguese is negative; that is, there is a negative relationship between accuracy in Spanish an d accuracy in Portuguese. In other words, the greater the accuracy in Spanish was the lower the accuracy in Portuguese was This pattern appears to suggest that accuracy in Spanish pronunciation has a negative bearing on accuracy in Portuguese pronunciati on, an observation that is corroborated by othe r results seen in this chapter. 4.3 Other Observations In this section, production is considered at the level of individual participants and words. Production was summed per production type for each participan t, and then percentages of possible production calculated by grapheme for each participant, such that production of grapheme adds up to 100% and production of grapheme adds up to 100% for each participant. These percentage results were then ordered according to the four participant groups (Appendix K ). The items of most interest to consider are those where participants produced something other than what was expected, given the pronunciation rules for and in English, Spanish and Portuguese. For grapheme , production type OTH could logically occur on any of the 29 words in each test, yet there was a relatively low proportion of OTH for most participants, where production ranged from 0% to 3% across all three tests. Two participants had greater OTH production: participant S 4, with 5%, and participant E114, with 8%. Participant S 4 produced the voiceless palatal fricative [ ] for two words in the first test audiovisual (audiovisual) and televiso (television), and for the same word in Tests 2 and 3 presidente (president). Participant E1 14 produced the voiceless palatal fricative [ ] for six words on the

PAGE 116

116 fi rst test museu (museum), viso (vision), feserel (a nonce word), quasano (nonce), casaco (coat), preguioso (lazy) and [sj] for one word on the first test deciso (decision). It would seem that for the cognate words where English would have the voiced palatal fricative [ ], the participants were producing something approximating this sound with [ ] and [sj]. In the other cases, it is not clear what would have led the participants to produce the sound they did. In fact, the voiceless palatal fricative [ ] was the most comm on sound categorized as OTH for all of the participants, and generally occurred on cognate words where English would have the voiced palatal fricative [ ], though not always. Although this production type accounted for a very small proportion of the overal l production, it is interesting that it is not altogether absent, supporting claims in the literature that in acquisition of a language after the L1 (i.e., L2 and beyond), elements often emerge that are not consistent with the grammar of the L1 n or with th e L2 grammar (Broselow, Chen & Wang, 1998) For grapheme , production type OTH could logically occur on any of the 19 words. Proportions of this production type were relatively high for , compared with . Two participants E1 18 and E124 pr oduced something other than the expected phonemes for 18% of their total production, which equates to roughly 10 words across the three tests. Another 18 participants also registered OTH production, with percentages ranging from 2% to 12% of their tota l production. These other sounds tended to be the voiced palatal fricative [ ] or the affricate [ts], neither of which would be expected, given the pronunciation rules for in English, Spanish and Portuguese. However, again, it is interesting to observe that participants did not produce just any sound in these OTH cases; rather their production was generally some sort of approximation to the target in the case where [ ] was produced instead of [z], the approximation is in place of articulation (palatal instead of alveolar) rather than voicing, as was

PAGE 117

117 often observed for graphe me . These observations would indicate universal tendencies toward unmarked features. Consideration will now be given to the production of particular words. The 48 words are listed (Appendix L) along with their word type (C for cognate, N for nonce, and NC for non cognate) and the sound expected according to English pronunciation rules (the Portuguese rules would always produce [z]; and the Spanish rules would always produce [s], or possibly [ ] for ). Production was summed for each word, according to production type, and percentages calculated, with production for each word totaling 100%. For the five cognate words where English would have [z], target Portuguese pronunciation (possibly due to positive English transfer) was recorded approximately half of the time for four of the words; the fifth word, presidente (president), was produced correctly on roughly two thirds of instances. It is possible that this word appeared more like the E nglish form of the cognate than the other four words in this category, which may have led to its correct production more often. For the six cognate words where English would have [s] as well as Spanish, there was an overwhelming production of [s] (clos e to 100%) on all but one word casos (cases) where incorrect [s] was produced on 73% of instances. It is possible that this word was more common in the classroom, which led to its higher correct production in comparison with the other words in this c ategory. In the case of the six cognate words where English would have [ it is interesting to find relatively low proportions of EN ([ ]) on all the words apart from audiovisual (audiovisual), where participants produced [ ] on 35% of occasions. Again, it is possible that this word was produced with English [ ] more often than the other words in this category due to its exact orthographic equivalence in English. It is also noteworthy that participants produced something other than what would be expected by English,

PAGE 118

118 Portuguese and Spanish rules on 8% of occasions for this w ord. This OTH production tended to be [ ] which approximates the English [ ]. Across the words in this category, production of SP ([s]) tended to be quite a bit higher than that of PO ([z]), except in the case of audiovisual (audiovisual) and viso (vision) where production was fairly evenl y split between SP and PO. For the nonce and non cognate words, participants generally produced inaccurate ES ([s]) roughly 85% of the time, although the non cognate word casaco (coat) yielded only 70% inaccurate production. This may be due to this w ord being more common in the classroom than the other non cognate words. It is interesting that the results for the non cognate words and for the nonce words appear very similar.47Results for the words were mixed. The percentages of accuracy ranged fr om 55% on the non cognate rapazes (young men) to 86% on the cognate horizonte (horizon). The target in idealizao (idealization) was skipped on 15% of occasions, by far the greatest percentage of NO production for any word. Additionally, pastiza (nonce) and rapazes (young men) had relatively high OTH production, with 8% and 6%, respectively. The other sounds produced for these words tended to be [ ] and [ts], which are not attributable to Spanish or English. 4.4 Summary With the original research questions on interference in mind, the results presented in this chapter support Majors (2001) OPM, which predicts interference from the L1 and/or the L2 and this interference reducing over time as the L3 increases. Evidence was found for interference mainly from the L1 for groups E1 and S, while group E2 had interference from the L2. All groups tended to have increases in target Portuguese production thro ughout the semester, 47 This will be discussed in greater detail in Chapter 5.

PAGE 119

119 indicative of a reduction of the role of the L1/L2 and an increase in the L3 in the learners IL. Additionally, there were occurrences of some items which were not consistent with the orthographic phonological correspondence systems of the L1, L2 or L3 These other sounds (production type OTH) tended to be approximations of the sounds expected by the L1, L2 or L3 correspondence rule systems. For example, where English would have the [s] [ words like visual visual), the voiceless counterpart [ Rather than just being rogue sounds in the production, t hese approximations suggest evidence for the universal elements (U) predicted by the OPM, with less marked ( e.g., voiceless) sounds appearing in th e place of more marked ( e.g., voiced) sounds. Hammarberg (2001) proposed that for an L2 to be the basis of interference, proficiency in it would have to be relatively high. The results in this chapter support that claim to a certain extent. E2, the group o f L1 English speakers highly proficient in Spanish, tended to show more interference from the L2 (at least initially) than E1, the group of L1 English speakers with no or low proficiency in Spanish. Group S, the L1 Spanish speakers, showed little evidence of clear transfer from the L2 however; their transfer tended to be from the L1 even though their L2 was, presumably, every bit as strong as the L2 of group E2. Thus, it seems that the L1 was the principal basis of transfer for groups E1, P and S. For gro up E2, it appears that the more typologically similar L2 was the basis of transfer initially (before instruction in Portuguese pronunciation), but transfer from this language was reduced later on, in favor of positive transfer from the L1 and/or target L3 production. Evidence was also found to support Muller and Mullers (1968) prediction that the more difficult GPCR to acquire would be the one which differed from the L1 and L2 (i.e., the [z] rule, in this case). In general, participants produced i tems correctly more often than

PAGE 120

120 items, perhaps due in part to the possibility of positive transfer from English in the case of , but not (except in the case of those cognate items where English would have [z]). A negative relationship was al so found between participants production in the Spanish reading task and in the Portuguese reading tasks: the greater the accuracy in the Spanish reading task was the lower the accuracy in Portuguese tended to be. This inverse relationship supports obser vations and studies in the literature reviewed above, which found persistent difficulties for Spanish speakers learning Portuguese pronunciation. Finally, the section on variation showed some of the other sounds which participants produced, and displayed s ome of the disparity found in the production of different words. All of the words showed evidence of negative transfer: a cognate item, horizonte (horizon), suffered the least negative transfer (on only 14% of its production), while a cognate ite m where both E nglish and Spanish would have [ s ] generosidade (generosity), suffered the most negative transfer (on 99% of its production). It is interesting that these two ends of the spectrum, in terms of transfer (negative and possibly positive, too), should be found on cognate words. The discussion of the effect of word type on production, and the question regarding participants ability to generalize the pronunciation rules from seen to unseen word s is taken up in more detail in the next chapter.

PAGE 121

121 T able 4 1. Number and percentage of items by production type and test Type Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 % EN 21 1 34 2 24 2 EP 113 7 148 10 146 10 ES 860 57 780 52 741 49 NO 7 0 6 0 6 0 OTH 16 1 10 1 11 1 PO 144 10 251 17 305 20 SP 347 23 279 19 275 18 Table 4 2. ANOVA for all production types for Effect Num DF F Value P Value test 2 0.00 1.0000 group 3 0.00 1.0000 production type 6 697.13 <.0001 test* production type 12 9.12 <.0001 group production type 18 13 .25 <.0001 group *test 6 0.00 1.0000 group *test* production type 36 0.74 0.8681 Table 4 3. Post hoc t test results for all production types for on Test 1 Type EN EP ES NO OTH PO SP EN 0.0187 <.0001 1.0000 1.0000 <.0001 <.0001 EP *< .0001 0 .0033 0.0147 0.8903 + <.0001 ES <.0001 <.0001 + <.0001 <.0001 NO 1.0000 < .0001 <.0001 OTH < .0001 <.0001 PO <.0001 SP Table 4 4. Post hoc t test results for all production types for on Test 2 Type EN EP ES NO OTH PO SP EN 0.0003 <.0001 1.0000 1.0000 <.0001 <.0001 EP < .0001 <.0001 <.0001 <.0001 + <.0001 ES <.0001 <.0001 + <.0001 <.0001 NO 1.0000 <.0001 <.0001 OTH <.0001 <.0001 PO 0.9473 SP

PAGE 122

122 Table 4 5. Post hoc t test results for all production types for on Test 3 Type EN EP ES NO OTH PO SP EN <.0001 <.0001 1.0000 1.0000 <.0001 <.0001 EP <.0001 <.0001 <.0001 <.0001 + <.0001 ES <.0001 <.0001 + <.0001 <.0001 NO 1.0000 <.0001 <. 0001 OTH <.0001 <.0001 PO 1.0000 SP Table 4 6. Number and percentage of items, by production type and participant group EN # EN % EP # EP % ES # ES % NO # NO % OTH # OTH % PO # PO % SP # SP % E1 53 2 242 10 1267 54 15 1 26 1 333 14 413 18 E2 0 0 31 7 228 52 1 0 3 1 52 12 120 28 P 23 3 94 12 361 46 2 0 4 1 207 26 92 12 S 3 0 40 4 525 55 1 0 4 0 108 11 276 29 Table 4 7. Post hoc t test results for production type EP for < s > items, by participant group Group E1 E2 P S E1 0.9992 1.0000 0.0137 E2 0.9517 1.0000 P 0.0104 S Table 4 8. Post hoc t test results for production type ES for < s > items, by participant group Group E1 E2 P S E1 1.0000 0.0005 1.0000 E2 0.5945 1.0000 P 0.0013 S Table 4 9. Post hoc t test results for production type PO for < s > items, by participant group Group E1 E2 P S E1 1.0000 <.0001 0.9813 E2 <.0001 1.0000 P <.0001 S Table 4 10. Post hoc t test results for production type SP for < s > items, by pa rticipant group Group E1 E2 P S E1 0.0004 0.0713 <.0001 E2 <.0001 1.0000 P <.0001 S

PAGE 123

123 Table 4 11. Post hoc t test results for all production types for items for participant group E1 Type EN EP ES NO OTH PO SP EN <.0001 <.0001 0. 9999 1.0000 <.0001 <.0001 EP <.0001 <.0001 <.0001 0.1370 + <.0001 ES <.0001 <.0001 + <.0001 <.0001 NO 1.0000 <.0001 <.0001 OTH <.0001 <.0001 PO 0.3669 SP Table 4 12. Post hoc t test results for all productio n types for items for participant group E 2 Type EN EP ES NO OTH PO SP EN 0.6029 <.0001 1.0000 1.0000 0.0024 + <.0001 EP <.0001 0.6724 0.7977 0.9913 <.0001 ES <.0001 <.0001 + <.0001 <.0001 NO 1.0000 0.0034 <.0001 OTH 0.0 071 <.0001 PO <.0001 SP Table 4 13. Post hoc t test results for all production types for items for participant group P Type EN EP ES NO OTH PO SP EN 0.0016 <.0001 0.9999 1.0000 <.0001 0.0029 EP <.0001 <.0001 <.0001 <.0 001 + 1.0000 ES <.0001 <.0001 + <.0001 <.0001 NO 1.0000 <.0001 <.0001 OTH <.0001 <.0001 PO <.0001 SP Table 4 14. Post hoc t test results for all production types for items for participant group S Type EN EP ES NO OTH PO SP EN 0.9289 <.0001 1.0000 1.0000 <.0001 <.0001 EP <.0001 0.8812 0.9469 0.0206 + <.0001 ES <.0001 <.0001 + <.0001 <.0001 NO 1.0000 <.0001 <.0001 OTH <.0001 <.0001 PO <.0001 SP

PAGE 124

124 Table 4 15. Numbe r and percentage of items, by production type and test Type Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 % EP 588 60 719 73 793 80 NO 13 1 17 2 8 1 OTH 17 2 21 2 24 2 SP 370 37 231 23 163 16 Table 4 16. ANOVA for all production types for Effect Num DF F Value P value test 2 0.00 1.0000 group 3 0.00 1.0000 production type 3 405.24 <.0001 test* production type 6 18.33 <.0001 group production type 9 46.20 <.0001 group *test 6 0.00 1.0000 group *test* production type 18 1.83 0.0196 Table 4 17. Post hoc t test results for all production types for on Test 1 Type EP NO OTH SP EP <.0001 <.0001 + 0.9280 NO 1.0000 <.0001 OTH <.0001 SP Table 4 18. Post hoc t test results for all production types for on Test 2 Type EP NO OTH SP EP <.0001 <.0001 + *< .0001 NO 1.0000 *< .0001 OTH *< .0001 SP Table 4 19. Post hoc t test results for all production types for on Test 3 Type EP NO OTH SP EP <.0001 <.0001 + <.0001 NO 1.0000 0.0003 OTH 0.0014 SP Table 4 20. Number and percentage of items, by production type and participant group EP # EP % NO # NO % OTH # OTH % SP # SP % E1 1261 82 25 2 40 3 213 14 E2 176 62 4 1 3 1 102 36 P 426 83 3 1 10 2 74 14 S 237 38 6 1 9 1 375 60

PAGE 125

125 Table 4 21. Post hoc t test results for production type EP for items, by participant group Type E1 E2 P S E1 0.0009 1.0000 <.0001 E2 0.0042 0.0002 P <.0001 S Table 4 22. Post hoc t test results for production type SP for it ems, by participant group Type E1 E2 P S E1 0.0002 1.0000 <.0001 E2 0.0039 0.0002 P <.0001 S Table 4 23. Post hoc t test results for all production types for items for participant group E1 Type EP NO OTH SP EP <.0001 <.0001 + <.0001 NO 1.0000 0.0002 OTH 0.0010 SP Table 4 24. Post hoc t test results for all production types for items for participant group E2 Type EP NO OTH SP EP <.0001 <.0001 + 0.0011 NO 1.0000 <.0001 OTH <.0001 SP Tab le 4 25. Post hoc t test results for all production types for items for participant group P Type EP NO OTH SP EP <.0001 <.0001 + <.0001 NO 1.0000 0.1070 OTH 0.2339 SP Table 4 26. Post hoc t test results for all production types for items for participant group S Type EP NO OTH SP EP <.0001 <.0001 + <.0001 NO 1.0000 <.0001 OTH <.0001 SP

PAGE 126

126 Table 4 27. Number and percentage of items, by production type, participant group and test EP # EP % SP # SP % E1 380 (T 1) 432 (T 2) 449 (T 3) 74 (T 1) 84 (T 2) 88 (T 3) 115 (T 1) 52 (T 2) 46 (T 3) 22 (T 1) 10 (T 2) 9 (T 3) E2 31 (T 1) 62 (T 2) 83 (T 3) 33 (T 1) 65 (T 2) 87 (T 3) 59 (T 1) 32 (T 2) 11 (T 3) 62 (T 1) 34 (T 2) 12 (T 3) P 131 (T 1) 142 (T 2) 153 (T 3) 77 ( T 1) 83 (T 2) 89 (T 3) 37 (T 1) 25 (T 2) 12 (T 3) 22 (T 1) 15 (T 2) 7 (T 3) S 46 (T 1) 83 (T 2) 108 (T 3) 22 (T 1) 40 (T 2) 52 (T 3) 159 (T 1) 122 (T 2) 94 (T 3) 76 (T 1) 58 (T 2) 45 (T 3) Table 4 28. Post hoc t test results for production type EP for items, by participant group and test Type E1 E2 P S E1 0.0002 (T 1) 0.9530 (T 2) 1.0000 (T 3) 1.0000 (T 1) 1.0000 (T 2) 1.0000 (T 3) <.0001 (T 1) *<.0001 (T 2) *<.0001 (T 3) E2 0.0011 (T 1) 0.9986 (T 2) 1.0000 (T 3) 1.0000 (T 1) 0.6390 (T 2) *0.0298 (T 3) P <.0001 (T 1) *<.0001 (T 2) *0.0002 (T 3) S Table 4 29. Post hoc t test results for production type SP for items, by participant group and test Type E1 E2 P S E1 0.0005 (T 1) 0.5856 (T 2) 1.0000 (T 3) 1.0000 (T 1) 1.0000 (T 2) 1.0000 (T 3) <.0001 (T 1) *<.0001 (T 2) <.0001 (T 3) E2 0.0061 (T 1) 0.9944 (T 2) 1.0000 (T 3) 1.0000 (T 1) 0.7211 (T 2) 0.0752 (T 3) P <.0001 (T 1) *<.0001 (T 2) *0.0002 (T 3) S

PAGE 127

127 Figure 41. Percentage of items, by test for each production type Figure 42. Percentage of items, by participant group for each production type

PAGE 128

128 Figure 43. Percentage of items, by test for each production type Figure 44. Percentage of items, by participant group for each prod uction type

PAGE 129

129 Figure 45. Percentage of correct items on Spanish reading task and average percentage of correct items on Portuguese reading tasks Participant numbers are shown on the horizontal axis, ordered according to accuracy of production on the Spa nish reading task. For the sake of readability, only some of the participant numbers are shown (one in four). portug1 = 64.092 -0.5035 spanish N 52 Rsq 0.4894 AdjRsq 0.4792 RMSE 15.797 0 10 20 30 40 50 60 70 80 spanish 0 10 20 30 40 50 60 70 80 90 100 Figure 46. Relationship between production on Spanish reading task and average production on Portuguese reading tasks

PAGE 130

130 CHAPTER 5 GENERALIZAB ILITY RESULTS AND DISCUSSI ON 5.1 Introduction This chapter addresses the third broad research question, regarding generalizability, by presenting and discussing the relevant results. Before considering these, it is useful to recall the research questions asked and hypotheses stated (in section 2.2.3) concerning generalizability. The term generalizability was used to refer to participants ability to apply the GPCRs that they ha d learned for s een words to unseen words. The principal question was what diffe rences in production might arise due to word type : cognates, non cognates and nonce words In light of Defior et al.s (2002) and Lords (2007) findings that real words were read faster and more accurately than nonce words it was postulated that the ortho graphic phonological correspondences in seen words (cognates and non cognates) w ould be produced more accurately than those in unseen words (nonce words), due to learners greater familiarity with and exposure to the former. The results will be presented a nd discussed first for all production, then production by grapheme. Next, production of the cognate items will be examined separately, in view of the three sounds that occur for the English equivalents of these items. Variation between groups and indiv iduals will then be discussed, with a general summary of results to follow. 5.2 Results and Discussion Results are presented in terms of raw production scores, proportions of correct items out of all possible items per word type, and statistically signific ant differences (output from the model) which are marked with an asterisk in the tables 5.2.1 Production for Graphemes < s> and Considering together the production of all the participants, irrespective of group and grapheme, the data showed correct pr oduction of [z] on 1437 (out of 3588) cognate items, 842

PAGE 131

131 (out of 1872) nonce items, and 928 (out of 2028) non cognate items. Expressed as percentages of possible correct items, these raw scores are equivalent to 40%, 45% and 46%, respectively (Table 5 1). When all production (both graphemes and all three tests) was considered, correct production appeared to be roughly the same across all three word types (Figure 51). The ANOVA ( refer back to Table 37) showed that the interaction between grapheme and word type was significant. Other interactions which were shown to be significant by the ANOVA did not involve word type, which is of principal concern in this chapter, so only the significant interaction with word type is discussed here. Other significant inte ractions have been discussed in previous chapters. Since the ANOVA results show ed the grapheme by word type interaction to be significant, it is necessary to consider the graphemes separately 5.2.2 Production by Grapheme Correct production was summed (raw scores) for each word type, by grapheme, and then calculated as a percentage of all possible production for that word type and grapheme (Table 5 2). The percentage of correct tokens is also shown graphically for each word type and grapheme (Figure 5 2). P ost hoc Tukey t tests were carried out to determine statistically significant differences between graphemes for each word type, and between word types for each grapheme. Results of the ttests revealed significant differences between and on all thr ee word types, with produced correctly significantly more often than (Table 5 3). This result suggest s that accuracy in production ha s less to do with word type and more to do with orthography an observation that app ears to be confirmed by the fac t that no significant differences were found between word types for items (Table 5 4). Since participants produced items with relatively the same accuracy across all word types, it would appear that there were no facilitatory or inhibitory effects for cognates as compared with words which were not cognates (i.e., non cognates and nonce words).

PAGE 132

132 Additionally, the lack of significant differences observed between seen words (cognates and non cognates) and unseen words (nonce words) does not perm it any conclusion to be drawn concerning whether production is lexically driven or rule driven In the case of tokens, however, there appeared to be a facilitatory effect for cognates compared with non cognates and nonce words (Table 5 5). Cognate items were accurately produced significantly more often than non cognate and nonce items. As with , though, no significant difference was observed between non cognate and nonce items, which again makes it impossible to draw any conclusion a bout learners ability to generalize from seen to unseen words. The ostensibly facilitatory effect found for cognates bears further investigation, considering that this group of tokens included words which would have different sounds in English for 5.2.3 Cognate Items The group of cognate items had an unusual characteristic: one orthographic phonological correspondence rule in Portuguese ( [z]), and one rule, albeit a different one, in Spanish ( [s]), but three rules in English: five words pronounced with [z], six words pronounced with [s], and six words pronounced with [ ] (Table 5 6). This group of items allows a more in depth exploration of orthographic phonological correspondences. Till now, th is dissertation has examined only differences between graphemes ( and ), and between word types (cognate, non cognate and nonce); now, the consideration is of items within a single word type with one grapheme. What makes this analysis interesting is that the three sets of English sounds for these cognates fall into different production categories with respect to the two other languages under consideration, namely Portuguese and Spanish. That is, those words where English would have [z] share a correspondence rule with Portuguese but not with Spanish; those words where English would have [s] share a correspondence rule wit h Spanish but not with

PAGE 133

133 Portuguese; and those words where English would have [ ] do not share a correspondence rule with either Portuguese or Spanish. Thus, it would seem plausible to infer that the English correspondence rules play a part where differences are observed between the three sets of cognate tokens. Correct production for these three sets of words (where the set name is the sound in English) was summed (raw scores) and then calculated as a percentage of possible items (Table 57). Result s are shown graphically for the three sets of cognates for each test (Figure 5 3). Numerically, scores tend to increase between tests, for the three sound groups. Tokens with [z] in English are seen to be produced correctly more often than tokens with [ ] in English, which in turn are observed to be produced accurately more regularly than tokens with [s] in English. The ANOVA (Table 5 8) shows that each of the individual factors group test and English sound is significant, as is the interaction be tween group and sound; thus it is necessary to examine the production of cognate words according to the four participant groups. Correct production for the four language groups was summed (raw scores) and then calculated as a percentage of possible pro duction (Table 5 9). Results are shown graphically by participant group for each of the three sets of cognate items (Figure 5 4). A number of statistically significant differences were found, both between and within participant groups. For cognate tokens where English would have [z] in the pronunciation (Table 5 10), groups E1 and P produced correctly as [z] (according to the target Portuguese rules) significantly more often than group S. There were no significant differences between L1 English speakers with low or no Spanish (E1 and P, mostly) and the L1 English speakers proficient in Spanish (E2), nor was there a significant difference between the L1 English speakers proficient in Spanish (E2) and the native Spanish speakers (S). In other word s, for this set of items, group

PAGE 134

134 E2 fell in the middle (between groups E1 and P at the higher end, and group S at the lower end) with production that proved not to be significantly different from any other group. When it came to the cognate tokens where English would have [s] (Table 5 11), no significant differences were found between groups. That is, L1 English speakers, irrespective of proficiency in Spanish, L1 Spanish speakers, and those with previous exposure to Portuguese all produced these tokens with relatively the same (low) degree of accuracy. For those cognate items where English would have [ ] (Table 5 12), there were only significant differences between groups P and E2, and between P and S. In both comparisons, group P participants produc ed the items with greater accuracy. For this set of words, group E1 fell in the middle (between group P at the higher end, and groups E2 and S at the lower end) with production that proved not to differ significantly from any other group. Considering each group separately, it can be seen that E1 production differed significantly for the three sets of cognate tokens (Table 5 13), as did production by group P (Table 5 14). For both groups, those tokens where English would have [z] were produced accurately most often, followed by those tokens where English would have [ ], followed by those tokens where English would have [s]. For group E2 (Table 515), those tokens where English would have [z] were produced accurately significantly more often than the other tokens, but no significant differences were found between the other two sets of tokens (where English would have [s] or [ ]). Finally, group S showed no significant differences between any of the three sets of cognate tokens (Table 5 16). The results for these three sets of cognate tokens are noteworthy. It bears repeating that all of these words are cognates containing the grapheme , but that is clearly not a sufficient description. Since the orthographic phonological correspondence rules for t hese words

PAGE 135

135 remain constant in Portuguese ( [z]) and in Spanish ( [s]), it seems fair to state that the differences which were observed between the three sets of words arose as a result of the different English correspondence rules for the three sets. As seen in previous chapters, here again it was observed that linguistic background played a part in production. For the native Spanish speakers (group S), who had a consistent correspondence rule in their L1, there were no statistical differences observe d across the three sets of cognate words. For the native English speakers (groups E1, E2 and most of P), who had different correspondence rules in their L1, there were statistical differences in production between the sets. For all of the L1 Englis h sp eakers, accurate production, according to the target Portuguese rule, was significantly greater when the L1 correspondence rule coincided with the target rule ( [z]). When the L1 and L2 rules coincided with each other ( [s]) but differed from the L3 Portuguese rule ( [z]), no differences were observed between groups, corroborating what has been discussed in previous chapters: when two similar rules in a learners previous languages coincide, there is greater inaccuracy in production according to a different rule in the new language (L3). When the three languages had different correspondence rules, it appeared that differences in production were due primarily to previous instruction in the new language (group P significantly outperformed groups E2 an d S), rather than due to the L1 (groups E1 and E2 did not differ significantly from each other ), or due to proficiency in Spanish (groups E2 and S did not differ significantly from each other ). Consideration will now be given to some of the variation found for particular words and word categories.

PAGE 136

136 5.3 Other Observations Here, production of individual words is discussed. The percentages of correct production by word (Appendix M ) relate to overall production, that is, by all participants on all tests.48Clearly, there was a great deal of variation between the high and l ow points of the ranges, yet the distance between these points was relatively similar across the different word types and graphemes. These tendencies are interesting to consider as they illustrate the variability that occurs, in terms of accurate productio n, within a single word category for a single grapheme. Why there is such variability is difficult to pinpoint, but possible reasons may include how often the participants were exposed to the words in class (for seen words), word length, closeness in It is interesting to observe tendencies within groups of words. For cognate words where English would have [z], accuracy ranged from 46% for visvel (visible) to 68% for presidente (president). For cognate words where English would have [s], accurac y ranged from 1% for generosidade (generosity) to 27% for casos (cases). For cognate words where English would have [ accuracy ranged from 13% for divises (divisions) to 48% for viso (vision). For nonce words, accuracy ranged from 11% for lomosa to 24% for isgio ; while for non cognate words, accuracy ranged from 6% for manhoso (whiney) to 29% for casaco (coat). For cognate words, accuracy ranged from 62% for idealizao (idealization) to 86% for horizonte (horizon). For non ce words, accuracy ranged from 53% for paimozes to 83% for trazentar Finally, for non cognate words, accuracy ranged from 55% for rapazes (young men) to 79% for azedo (sour). 48 Although certain differences in production were found between groups in the previous section, here production by all part icipants is considered together, for the sake of brevity and clarity, since the purpose of this section is merely to present some of the salient and interesting differences between words.

PAGE 137

137 form to known words, or a combination of such factors. While the effect of these things was not controlled for or looked at in this study, future studies addressing these issues may be beneficial It is noteworthy that the most accurately produced word was the cognate word horizonte (horizon), where English would have [z], and the least accurately produced word was the cognate word generosidade (generosity), where English and Spanish would have [s]. With the former being produced correctly on 86% of instances, and the latter being produced correctly on only 1% of instances, there is clearly still room for improvement on all items, which is to be expected for learners in the early stages of language acquisition. What seems almost incomprehensible is th at the least accurately produced word, generosidade (generosity) was produced correctly by only one participant on one test (that is, once out of a possible 156 times). This level of inaccuracy is difficult to explain, though it might be due to a combina tion of the aforementioned factors: it is a relatively long word (six syllables), its use in the classroom may have been relatively low, and its closeness to known words ( generous generosity in English, for example) is high. Additionally, it falls into th e category of words where the L1 and L2 share a correspondence rule which differs from the rule in the target L3, a case which is potentially subject to negative transfer from both of a learners previous languages. These observations about particular words and groups of words reinforce yet again the effect of the grapheme on accurate production, as well as the role of the learners previous languages in the acquisition of orthographic phonological correspondence rules in L3. 5.4 Summary In this chapter, w ord type (cognate, non cognate, or nonce) was considered as a factor which might affect production. Output from the statistical model showed a significant interaction between word type and grapheme, revealing that accurate production of grapheme items occurred significantly more than accurate production of grapheme items across all word

PAGE 138

138 types. Differences between word types were not found to be statistically significant for , and only showed statistically significant differences between cogn ates and other words (non cognates and nonce). Thus, evidence was not found to support the hypothesis that participants would do better on seen words than unseen words, based on the findings from the study by Defior et al. (2002) and Lord (2007) where real words were read faster and more accurately than nonce words A closer examination of the cognate words, which contained three sets of tokens, according to different grapheme phoneme correspondence rules in English, revealed an interaction between par ticipant groups and sets of cognate words with different sounds in English ([z], [s] and [ ]). It was found that accuracy of production of these sets of cognate words varied, depending largely on the L1 of the participants. Native Spanish speakers (group S), whose L1 correspondence rule would have been the same for all three sets of cognate tokens ( [s]), showed no significant differences in production across the three sets. L1 English speakers (groups E1, E2 and most of P) displayed greater accuracy on words where English would have [z], similar to the target Portuguese rule. When th e L1 and L2 correspondence rules coincided ( [s]), differing from the target L3 rule ( [z]), no significant differences were found between groups. All of these results appear to indicate that word type did not play as important a part in production i n this study as the factors discussed in previous chapters. Nevertheless, the closer examination of the cognate tokens with different English correspondence rules was revealing, as it shed light on the crucial role that the L1s system of correspondenc e rules played in L3 production. This analysis of the three sets of cognate items is useful because it may help to clarify the origin of transfer (English or Spanish) for non cognate and nonce items,

PAGE 139

139 where the rules for English and Spanish pronunci ation would coincide ( [z]). For the results discussed in previous chapters, it wasnt possible to tease apart whether incorrect production of [s] for grapheme items was suggestive of negative transfer from English or from Spanish. Now, however, it would appear that the groups of native English speakers (E1, E2 and most of P) were producing sounds according to the rules of their L1, given that there were significant differences between the sets of cognates for these groups; while the native Spani sh speakers (S) were producing sounds according to the rules of their L1, since there were no significant differences between the cognate sets for this group. Thus, it seems plausible to propose that the groups producing [s] on the non cognate and nonc e items would also be transferring largely from their L1. This suggestion bears further investigation, something which will be addressed further in the next chapters concluding remarks.

PAGE 140

140 Table 5 1. Number and percentage of correct items for all [z], by word type Word Type Raw Scores Possible Percentages Cognate 1437 3588 40 Nonce 842 1872 45 Non cognate 928 2028 46 Table 5 2. Number and percentage of correct items, by grapheme and word type Word Type by Grapheme Raw Scores Possible Percentages Cognate 777 2652 29 Nonce 166 936 18 Non cognate 164 936 18 Cognate 660 936 85 Nonce 676 936 72 Non cognate 764 1092 82 Table 5 3. Post hoc t test results comparing graphemes for each word type compared to Cognate <.0001 Nonce *<.0001 Non cognate <.0001 Table 5 4. Post hoc t test results comparing word type s for Cognate Nonce Non cognate Cognate 0.9618 0.9998 Nonce 0.8824 Non cognate Table 5 5. Post hoc t test results comparing word type s for Cognate Nonce Non cognate Cognate <.0001 <.0001 Nonce 1.0000 Non cognate

PAGE 141

141 Table 5 6. Cognate words with corresponding sound in English Cognate Word English apresentao (presentation) [z] esquisito (odd) [z] museu (museum) [z] presidente (president) [z] visvel (visible) [z] bsico (basic) [s] casos (cases) [s] curiosidade (curiosity) [s] filosofia (philosophy) [s] generosidade (generosity) [s] persuasivo (persuasive) [s] audiovisual (aud iovisual) [ ] deciso (decision) [ ] divises (division) [ ] reviso (revision) [ ] televiso (television) [ ] viso (vision) [ ] Table 5 7. Number and percentage of correct items, by cognate set Cognate Test 1 # Test 1 % Test 2 # Test 2 % Test 3 # Test 3 % Total # Total % English [z] 113 43 148 57 146 56 407 52 English [s] 11 4 29 9 37 12 77 8 English [ ] 76 24 101 32 116 37 293 31 Table 5 8. ANOVA for cognate items Effect Num DF F Value P Value group 3 4.60 0.0036 test 2 12.1 1 <.0001 sound 2 102.92 <.0001 group *test 6 0.65 0.6882 group *sound 6 12.53 <.0001 test*sound 4 0.59 0.6704 group *test*sound 12 0.58 0.8600 Table 5 9. Number and percentage of correct items for cognate sets, by participant group [z] # [z] % [s] # [s] % [ # [ E1 242 60 26 5 160 33 E2 31 41 14 16 10 11 P 94 70 18 11 84 52 S 40 24 19 10 39 20

PAGE 142

142 Table 5 10. Post hoc t test results for cognate set with [z] in English, by participant group English [z] E1 E2 P S E1 0.7097 0.9733 <.0001 E2 0.2561 0.8872 P <.0001 S Table 5 11. Post hoc t test results for cognate set with [s] in English, by participant group English [s] E1 E2 P S E1 0.9948 0.9998 1.0000 E2 0.9948 1.0000 1.0000 P 0.9998 1.0000 1.0000 S 1.0000 1.0 000 1.0000 Table 5 12. Post hoc t test results for cognate set with [ ] in English, by participant group English [ ] E1 E2 P S E1 0.4503 0.3014 0.7393 E2 0.0086 0.9996 P 0.0114 S Table 5 13. Post hoc t test results for cognate sets for participant group E1 Group E1 [z] [s] [ ] [z] <.0001 <.000 1 [s] <.0001 [ ] Table 5 14. Post hoc t test results for cognate sets for participant group P Group P [z] [s] [ ] [z] <.0001 0.0425 [s] <.0001 [ ] Table 5 15. Post hoc t test results for cognate sets for participant group E2 Group E2 [z] [s] [ ] [z] 0.0177 0.0017 [s] 1.0000 [ ] Table 5 16. Post hoc t test results for cognate sets for participant group S Group S [z] [s] [ ] [z] 0.1003 0.9986 [s] 0.6242 [ ]

PAGE 143

143 Figure 51. Percentage of correct ite ms by word type Figure 52. Percentage of correct items, by grapheme and word type

PAGE 144

144 Figure 53. Percentage of correct items, by test for each cognate set Figure 54. Percentage of correct items, by participant group for each cognate set

PAGE 145

145 CHAPTER 6 CONCLUSION 6.1 Introduction The purpose of the current study was to investigate the acquisition of orthographic phonological correspondence rules by learners of Portuguese as an L2 or L3, with L1 English and, where applicable, L2 Spanish or L1 S panish and L2 English. Acquisition was examined from the perspectives of error resolution, interference, and generalizability, in an effort to answer the studys overarching research questions regarding these three areas. In the next section, t hese questio ns and the initial hypotheses proposed (in section 2.2) are restated in summarized form, with the hypotheses confirmed or rejected, in light of the results found. Again, each of the questions is presented separately, for sake of clarity. 6.2 Acquisition of Orthographic Phonological Rules 6.2.1 Error Resolution R. Q. 1. To what extent do learners acquire the Portuguese grapheme phoneme correspondence rules (GPCRs) under consideration, throughout the course of the semester? H. 1. Significant increases in corr ect production were expected to occur throughout the semester of instruction, at least when all production data were considered together (both graphemes, all word types, all participant groups). This hypothesis was confirmed, since significant progress was indeed found between tests, although for this group of learners as a whole, there was still much room for improvement at the end of the semester, in terms of accuracy in production. R. Q. 2. What differences in acquisition exist between participant group s?

PAGE 146

146 H. 2. It was hypothesized that there would be differences in correct production between the language groups, with lower accuracy being found in the production of the more proficient Spanish speakers. This hypothesis was partially supported. Certainly t he native Spanish speakers tended to produce the target Portuguese sounds less accurately than the native English speakers with low or no proficiency in Spanish. However, the native English speakers with high proficiency in Spanish tended to have less accu racy in Portuguese pronunciation only initially, and by the end of the semester their accuracy was not significantly different from that of any other group R. Q. 3. What differences in acquisition are there between the two graphemes and intervocalic < s> corresponding to the phoneme /z/? H.3 It was anticipated that there would be significant differences between tests for both graphemes, and significant differences between the two graphemes. This hypothesis was partially confirmed While there were no significant differences found between tests for each grapheme (the ANOVA result for the test by grapheme interaction was not found to be significant), a significant difference was found between the two graphemes with produced significantly more accur ately than by all participant groups. 6.2.2 Interference R. Q. 1. What evidence is there of interference from the L1 and/or the L2? H.1 It was anticipated that there would be evidence of interference from both the L1 and the L2, with these reducing over time It was also hypothesized that there would be evidence in the data that participants were producing sounds which were neither consistent with the correspondence systems of the L1/L2 nor with th at of the L3 These hypothes e s were confirmed. As a whole, participants tended to show evidence of transfer from the L1 and L2 (to varying degrees), with the influence of these reducing over time

PAGE 147

147 as the role of the L3 increased in the learners IL, evidenced by increased accuracy in target production Addit ionally, participants produced sounds which were not consistent with the correspondence systems of the L1, L2 nor L3. Such production was almost always an approximation of a sound in one of the languages (e.g., a voiceless palatal fricative was produced wh ere the English cognate would have the voiced palatal fricative), rather than a rogue sound, suggesting a universal tendency toward some unmarked feature ( voicelessness, in the case of the example given ). R. Q. 2. What differences in interference are the re between participant groups? H. 2. It was hypothesized that the higher the proficiency level in the closely related language, Spanish the more evidence of transfer there would be from that language in the target language production. This hypothesis was supported to a certain extent. The proficient Spanish speakers (both native and non native) showed significantly greater interference from Spanish than the native English speakers with low or no proficiency in Spanish, at least initially. By the end of the semester, the native English speakers with high proficiency in Spanish were not found to differ significantly from any other group. It is interesting to note that the native English speakers with low proficiency in Spanish also showed evidence of transfer from Spanish their L2 although to a lesser extent than the participants proficient in Spanish. Additionally, it is worth mentioning that the native Spanish speakers, proficient in their L2, English, showed more evidence of transfer from the L1 (the more similar language to the target language) than the L2 (the less similar language). Thus, while it is evident that both language status (L1/L2) and language relatedness come into play in transfer, the results for the group of native Spanish speakers would s uggest that language relatedness trumps language status

PAGE 148

148 R. Q. 3. What differences in interference exist between the two graphemes and ? H. 3. It was expected that there would be differences between the two graphemes, in light of the GPCRs in Englis h, Portuguese and Spanish: it was anticipated that there would be greater accuracy in production of due to the possibility of positive transfer from English, and less accuracy in production of due to the possibility of only negative transfe r from S panish and generally negative transfer from English. This hypothesis was supported, since participants as a whole produced items correctly more often than items, perhaps in part due to the possibility of positive transfer from English in the case o f , but not (except in the case of those cognate items where English would have [z]). R. Q. 4. To what extent is production in Spanish related to acquisition (correct production) of Portuguese? H. 4. It was anticipated that production in Spanis h would be inversely related to acquisition of the Portuguese GPCRs in question. That is, it was expected that the greater the accuracy in Spanish production, the lower the accuracy in Portuguese production would tend to be. Again, this hypothesis was conf irmed, with a negative relationship being found between participants production in the Spanish reading task and in the Portuguese reading tasks: students with higher accuracy in Spanish tended to have lower accuracy in Portuguese 6.2.3 Generalizability R. Q. 1 What differences in production arise due to word type (cognates, non cognates and nonce words)? H. 1. It was predicted that the orthographic phonological correspondences in seen words (cognates and non cognates) would be produced more accurately t han those in unseen words

PAGE 149

149 (nonce words), due to learners greater familiarity with and exposure to the former. That is, production would generally be lexically driven rather than rule driven. This hypothesis was mostly rejected, since there were no signifi cant differences between word types for grapheme , and significant differences for grapheme were only found between cognates and non cognates, and between cognates and nonce words with cognates produced more accurately than the other words The res ults for grapheme would seem to suggest that all production was rule driven, while the results for grapheme would appear to indicate that there was a facilitatory effect for cognates, but this positive effect was found to be limited to the set of w ords where English would have the same correspondence rule as the target language, Portuguese ( [z]), and was only found for native English speakers, suggesting that for grapheme items production is lexically driven, and more specifically, driven b y the lexicon of the L1. In summary, a general trend was found toward increased accuracy in production and decreased interference throughout the course of the semester of data collection. Some differences in production were observed between participant gro ups, graphemes, and word types. Proficient Spanish speakers, native and non native, showed greater interference from Spanish than did non proficient Spanish speakers, at least initially. This finding would suggest two things: 1) proficiency in a language h as to be relatively high in order for it to be a significant source of transfer; and 2) structural relatedness (i.e., between Spanish and Portuguese here) appears to play a greater role in interference than language status (i.e., L1/L2), at least in the be ginning. Significantly greater accuracy was found on grapheme items than on grapheme items, demonstrating that acquisition of pronunciation is closely related to orthography. Differences between word types were only found on items, where cognat es were produced more

PAGE 150

150 accurately than other words. A closer examination of these cognate items revealed that groups with L1 English were significantly more accurate on those items where their L1 had the same correspondence rule as the target L3 rule (< s> [z]), while no such difference was found for L1 Spanish speakers. The findings of the present study corroborate the complexity of a multilinguals linguistic system(s), as discussed previously in the literature. Both the L1 and the L2 play a role in the acquisition of an L3, to varying extents, depending on such factors as proficiency, language status, and language typology. The specific contributions and implications of the current study are considered further in the next section. 6.3 Contributions and Implications This study makes a contribution in several areas, and has theoretical and applied implications. I t contributes to the field s of L2 acquisition, L3 acquisition, and phonology, by specifically addressing the area of orthographic phonological cor respondence rules and systems, something which has not received considerable attention to date. This study also contributes to the existing literature on the acquisition of Portuguese, a less commonly taught and researched language 6.3.1 Theoretical Implications T he results from the present study lend support to previous work done in several areas. Evidence was found to corroborate Muller and Muller s (1968) prediction that the more difficult correspondence rule to acquire in an L3 is that which has the sa me grapheme representing different phonemes in the L1/L2 and the L3. The findings here al so support Majors (2001) OPM, showing an initial IL consisting of the L1 and L2, with these decreasing over time as the L3 increases and with language universals (U) also playing a part in the IL

PAGE 151

151 Some of the conflicting results found in the literature for instance, with regard to the effects of bilingualism on L3 acquisition, transfer from the L2 to the L3, the role of language status in transfer, the effects of a typologically similar language, etc. are also reflected in this studys results. On the one hand, bilingualism at first seems to be disadvantageous for L1 English speakers proficient in Spanish, since they appear to transfer negatively from Spanish initially; but later production is at least as accurate as that of native English speakers not proficient in Spanish, suggesting that bilingualism may have positive effects (or maybe no effect) later on. On the other hand, bilingualism does not appear to be of much help to native Spanish speakers fluent in English, since their production seems to generally reflect negative transfer from their L1, even when positive transfer from the L2 would be possible. As to whether proficiency in the L2 has to be low or high for it to be the basis of transfer in the L3, the present study shows greater influence from the L2 when proficiency is high, but only in the case of a more typologically similar language, since L1 Spanish speakers with high proficiency in L2 English tend ed not to transfer as much from their L2 as from their L1. Thus, the evidence found in this study indicates that language relatedness (similar vs. dissimilar) is more important than language status (L1 vs. L2) but only when proficiency in the related lang uage is high Given these findings, this study proposes that examination of orthographic phonological correspondence rules, while uncommon, may perhaps be a clearer way to to pinpoint more specifically the source of interference in an L3. As for the role o f different word types in the acquisition of grapheme phoneme correspondence rules, this study did not generally support previous work which found facilitation or inhibition with cognates Additionally it is hard to determine whether production here was le xically driven or rule driven, given that there was generally no difference between production

PAGE 152

152 for seen words and for unseen words. The implication of these inconclusive results here is that additional research is required in this area, something which wil l be discussed in the section on future directions. The findings of the current study not only support previous observations in the literature with regard to the persistent problem that Spanish speakers have in devoicing Portuguese intervocalic , but also show that English speakers have the same problem. The fact that English has the same GPCR for intervocalic as Spanish (in most cases) is not discussed in the literature, but is important to remember, in light of the fact that many Portuguese learner s are also speakers of English, albeit not always as a native language. This point is discussed further in the next section, which considers some of the applied implications of the present study. 6.3.2 Applied Implications This study further corroborates the complexity of orthographic phonological correspondence systems, making salient the fact that acquisition of pronunciation in a new language involves more than simply learning new sounds, stress or intonation patterns; orthography also has a significant part to play in th e acquisition of pronunciation. In order to assist learners in acquiring new or different orthographic phonological correspondence rules, it may be helpful to have them participate in self analysis projects, which would allow them to rec ord, examine and monitor their own progress in acquisition (Lord, 2005) It is also worth pointing out that, o ften, it is the contrasts between languages that are highlighted to learners, when maybe in this case explicit instruction regarding the similarities between a known language, English, and the target language, Portuguese, would be more advantageous to learners. For example, by illustrating to learners that English shares a rule for in tervocalic with Portuguese, and by providing specific example s of this, students may be

PAGE 153

153 able to make helpful analogies in the L3, enabling them to extend that specific English rule for intervocalic to all instances of intervocalic in Portuguese. The fact that the learners in this study did not appear to gene ralize the GPCRs to nonce words suggests that they are using lexical knowledge or stored exemplars to arrive at pronunciation, indicating that they may not have learned the rules for the realization of Portuguese grapheme phoneme correspondences. Again, this highlights the need for explicit instruction regarding these rules and specific activities to practise their implementation, in order to assist in the acquisition of pronunciation. While the findings of the present study have been inconclusive in certai n respects, it has raised interesting questions regarding the acquisition of L3 orthographic phonological correspondence rules, and the results firmly indicate how complex these rules and systems are which must prompt further studies in this field, a poin t which will b e taken up in the next section. 6.4 Limitations and Future Directions Although the contributions are noteworthy, as with any study there are also several limitations to be addressed here. T he method of participant selection resulted in a number of potential complications. Participants were selected from students enrolled in introductory Portuguese classes at the time of the study. While this form of participant selection was convenient, it made factors such as age, gender, and length of instru ction in or exposure to previous languages impossible to control for something which future studies might do well to take into account. Furthermore, it was difficult to find a good way to classify native English speakers according to their proficiency in Spanish. Grouping them according to the class in which they were enrolled (beginner or accelerated) seemed the most objective way to divide the participants, but some participants may have been assigned to one group when it might have been better to assign them to another group. The grouping criteria may have obfuscated some of

PAGE 154

154 the results, and future studies might achieve different results if different criteria for grouping were used say grouping according to some kind of pronunciation or proficiency test Additionally, having students participate in the study as they would in any normal class activity, without penalty or reward as far as their grade was concerned, may have resulted in some participants rushing through the activities, without giving due ca re and attention to the task. Perhaps offering some form of incentive (grade or monetary) would address this limitation, though it may prove difficult to quantify the effect of an incentive (or lack thereof). The materials and tasks developed are not with out their limitations. In order to obtain data for the graphemes targeted by the present study, production was elicited rather than spontaneous. While elicited production may have its limitations (being contrived, it may not reflect what learners would do in natural speech), it would not have been possible for students in introductory classes of Portuguese to produce spontaneous data prior to instruction in the language, as was discussed in an earlier chapter. Thus, the materials and tasks were specifically developed for the present study in order to elicit production by all participants on the same words containing the particular GPCRs under investigation. In creating the materials, words found in the initial chapters of the textbook were used to ensure all participants had equal opportunities for exposure to them, with the additional inclusion of some words not found in the textbook in order to balance word categories as much as possible. Due to the limited number of words in the textbook containing the tar get grapheme, it was not possible to control for word length or frequency, nor for similarity or dissimilarity to words which participants might have known in their other language(s). Such factors might have affected the results (recall how the target grap heme in the longer word idealizao idealization, was often not produced), and it would be good if future studies could limit some of these differences.

PAGE 155

155 In terms of the reading task itself, having participants move through the words (slides) at their o wn pace, rather than controlling the speed at which words were presented, may have resulted in some participants skipping over words inadvertently or on purpose in an effort to finish the activity more quickly. On the other hand, allowing participants to g o at their own pace may have achieved two things: 1) it may have minimized the tedium for those participants who were fast readers and who would otherwise have spent a good deal of time waiting for words to appear automatically on the screen; and 2) it may have minimized anxiety for those participants who were not fast readers and who would otherwise not have had sufficient time to read words carefully. It would be interesting to have post hoc feedback from participants as to whether they perceived this asp ect of the study design to be positive or negative, something which another study might consider incorporating in order to inform future research As well as the method of data collection not being as controlled as possible, the equipment used was certain ly not the most precise available, but realistically and logistically speaking, it was more accessible, easy to use, and inexpensive than experimental psycholinguistic equipment. Future work might consider using more precise equipment to record and analyze the data, rather than relying on rater judgments. Additionally, it would be interesting to consider response times along with accuracy in production. Response times might enable a difference to be seen in different word types, something which was not poss ible in the current studys examination of accuracy alone. Another factor which was not controlled for in the present study was the effect of instruction. This clearly had an impact, although the extent to which instruction played a role is not known, nor what type of instruction was used or found more helpful (i.e., classroom vs. supplementary materials). Future studies can try to isolate this factor and see whether and how it

PAGE 156

156 helps, by incorporating specific treatments (instruction), and by having a cont rol group (with no explicit instruction). As mentioned in the previous section, examining the acquisition of orthographic phonological correspondence rules has real potential for pinpointing more precisely the source(s) of interference in an L3. Thus, it w ould be interesting to pursue this line of investigation with other correspondence rules and participants. It would be worth looking at graphemes or phonemes that would be new to participants with different linguistic backgrounds, such as the <> [s] corre spondence rule in Portuguese, which does not exist in English or Spanish. It might also be worthwhile to compare elicited (reading) and spontaneous (spoken) data from more advanced learners to establish the effect of written stimuli in later stages of acqu isition. Longitudinal studies with students majoring or minoring in Portuguese might also be of interest, to determine how acquisition progresses over a longer period ; and studies looking at the effect of study abroad programs could also be considered. In sum, there are numerous ways to explore the potential of this particular area of L3 phonological acquisition in an effort to determine its viability as an indicator of acquisition of pronunciation and of interference from other languages. 6.5 Closing The c urrent study examined the role of the grapheme in the acquisition of pronunciation, in light of the fact that different languages have different GPCRs. This aspect of L2 and L3 acquisition has not received a great deal of attention in the literature to date, yet has been shown here to have a significant impact on accurate production in the target language. Further studies in this area will provide more empirical evidence with regard to common pronunciation mistakes and may provide teachers with useful means of correction, based on learners linguistic differences.

PAGE 157

157 In addition, it is proposed here that real potential exists for this type of examination in order to determine more precisely the nature of interference from an L3 learners previous languages. Th rough f uture studies examining the role of the L1 and L2 in the acquisition of orthographic phonological correspondence rules in the L3 it may be possible to gain new insights into how bilingual and multilingual learners organize not only the GPCR systems of their different languages but perhaps also their mental lexicon(s) which would be of great interest to researchers, and potentially very helpful to instructors and learners alike

PAGE 158

158 APPENDIX A INFORMED CONSENT FOR M Protocol Title: Developing Pronunciat ion in Portuguese as a Foreign Language Please read this consent document carefully before you decide to participate in this study. Purpose of the research study: The purpose of this study is to investigate development in pronunciation in a foreign langu age class. What you will be asked to do in the study: You will be asked to read lists of words in English, Spanish and Portuguese, while being audiorecorded. Time required: No additional time outside of class. Risks and Benefits: After the study is con cluded, you may benefit by receiving feedback concerning your development in pronunciation in Portuguese. There are no risks involved. Compensation: None. Confidentiality: Your identity will be kept confidential to the extent provided by law. Your inform ation will be assigned a code number. The list connecting your name to this number will be kept in a locked office. When the study is completed and the data have been analyzed, the list will be destroyed. Your name will not be used in any report. Voluntar y participation: Your participation in this study is completely voluntary. There is no penalty for not participating and participation will not affect your course grade in any way. Right to withdraw from the study: You have the right to withdraw from the study at any time without consequence. Whom to contact if you have questions about the study: Sharon Barkley, Graduate Student, Linguistics Program, (Supervisor: Dr. Gillian Lord, Romance Language and Literatures Department, sbarkley@ufl.edu glord@rll.uf l.edu ) Whom to contact about your rights as a research participant in the study: UFIRB Office, Box 112250, University of Florida, Gainesville, FL 326112250; Phone (352) 3920433 Agreement: I have read the procedure described above. I voluntarily agree t o participate in the procedure and I have received a copy of this description. Participant: __________________________________ Date: _________________ Principal Investigator: __________________________ Date: _________________

PAGE 159

159 APPENDIX B LANGUAGE BACKGROU ND QUESTIONNAIRE Name: ________________________ Class: POR ______ Time: ________ The information you provide here will help to inform us about your language experience, enabling us to place you in the right class. Please answer all questions as fully as possible. For the sections on ratin g your language skills, please refer to the descriptions at the end of this questionnaire. Sex: __________ Age: __________ Have you ever been diagnosed with a reading disorder (e.g., dyslexia) or speech deficit (e.g., stuttering)? If so, please explain. ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ Language 1: English 1. Are you a native speaker of English? ______________________________________________________________________________ ______________________________________________________________________________ 2. If not, where did you learn English (at school, somewhere else)? In what grade did you start taking classes in English? How long have you studied English? ______________________________________________________________________________ ______________________________________________________________________________ 3. Where do you currently s peak this language (at home, at school, at work)? ______________________________________________________________________________ ______________________________________________________________________________ 4. With whom do you speak this language (parents, grandparents, siblings, spouse, children, coworkers, friends)? ______________________________________________________________________________ ______________________________________________________________________________ Language 2: Spanish 5. Are you a native speaker of Spanish? ______________________________________________________________________________ ______________________________________________________________________________

PAGE 160

160 6. Have you had formal i nstruction in Spanish? In what grade did you start taking classes? How long have you studied Spanish? Have you taken any college classes in Spanish? If so, what level was the last class you took (1000, 2000, 3000 or 4000)? ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ 7. Have you lived in a count ry or community where Spanish is spoken? If so, explain. ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ 8. Where do you currently speak Spanish (at home, at school, at work)? ______________________________________________________________________________ ______________________________________________________________________________ 9. With whom do you speak Spanish (parents, grandparents, siblings, spouse, children, coworkers, friends)? ______________________________________________________________________________ ______________________________________________________________________________ 10. Please rate your communication skills in Spanish, referring to the descriptions at the end of this questionnaire. Check only ONE number, corresponding to the description which you feel best matches your abilities. Listening 1 2 3 4 Speaking 1 2 3 4 Reading 1 2 3 4 Writing 1 2 3 4 Language 3: Portuguese 11. Are you a native speaker of Portuguese? ______________________________________________________________________________ ______________________________________________________________________________ 12. Have you had formal instruction in Portuguese? In what grade did you start taking classes? How long have you studied Portuguese? Have you taken any college classes in Portuguese? If so, what level was the last class you took (1000, 2000, 3000 or 4000)? ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________

PAGE 161

161 13. Have you lived in a country or community where Portuguese is spoken? If so, explain. ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ 14. Where do you currently speak Portuguese (at home, at school, at wo rk)? ______________________________________________________________________________ ______________________________________________________________________________ 15. With whom do you speak Portuguese (parents, grandparents, siblings, spouse, children, coworkers, friends)? ______________________________________________________________________________ ______________________________________________________________________________ 16. Please rate your communication skills in Portuguese, referring to the descriptions at the end of this questionnaire. Check only ONE n umber, corresponding to the description which you feel best matches your abilities. (The descriptions are at the end of this document.) Listening 1 2 3 4 Speaking 1 2 3 4 Reading 1 2 3 4 Writing 1 2 3 4 Language 4 (please specify ): ____________________ 17. Are you a native speaker of this language? ______________________________________________________________________________ ______________________________________________________________________________ 18. Have you had formal instruction in this language? In what grade did you start taking classes? How long have you studied this language? Have you taken any college classes in this language? If so, what leve l was the last class you took (1000, 2000, 3000 or 4000)? ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ 19. Have you lived in a country or community where this language is spoken? If so, explain. ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ 20. Where do you currently speak this language (at home, at school, at work)? ______________________________________________________________________________ ______________________________________________________________________________

PAGE 162

162 21. With whom do you speak this language (parents, grandparents, siblings, spouse, children, coworkers, friends)? _____________________________________________________________________________ _____________________________________________________________________________ 22. Please rate your communication skills in this language, referring to the descriptions at the end of this questionnaire. Check only ONE number, corresponding to the description which you feel best matches your abilities. Listening 1 2 3 4 Speaking 1 2 3 4 Reading 1 2 3 4 Writing 1 2 3 4 Language 5 (please specify): ____________________ 23. Are you a native speaker of this language? ______________________________________________________________________________ ______________________________________________________________________________ 24. Have you had formal instruction in this language? In what grade did you start taking classes? How long have you studied this language? Have you taken any college classes in this language? If so, what level was the last class you took (1000, 2000, 3000 or 4000)? ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ 25. Have yo u lived in a country or community where this language is spoken? If so, explain. ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ 26. Where do you currently speak this language (at home, at school, at work)? ______________________________________________________________________________ ______________________________________________________________________________ 27. With whom do you speak this language (parents, grandparents, siblings, spouse, children co workers, friends)? ______________________________________________________________________________ ______________________________________________________________________________

PAGE 163

163 28. Please rate your communication skills in this language, referring to the descriptions at the end of this questionnaire. Check only ONE number, corresponding to the description which you feel best matches your abilities. Listeni ng 1 2 3 4 Speaking 1 2 3 4 Reading 1 2 3 4 Writing 1 2 3 4

PAGE 164

164 Use the following descriptions to rate your language skills for listening, speaking, reading and writing. For each skill, you should choose ONLY the description which you feel best describes your abilities. 1. I understand only a few isolated words and phrases. Listening 2. I can recognize a number of phrases and can understand simple questions about myself, my activities, my likes and dislikes, as well as other peoples conversations about such th ings. I understand people when they speak slowly and use simple past, present and future tenses 3. I can recognize a number of grammatical structures, including conditional and subjunctive, and I understand vocabulary on a variety of topics, such as the weat her, work, studies, films, books and so on. I can understand and follow most directions, I can understand native speakers in a restaurant or store situation. 4. I have an extensive understanding of grammar and vocabulary, and I can understand in depth convers ations on most topics. I can understand a native speaker of any age or social background, and I can understand discussions on topics such as sports, politics and religion. 1. I know a few words and phrases but I cannot have a conversation. Speaking 2. I know a number of phrases and can have a simple conversation. I can ask people questions about themselves, their likes and dislikes, their usual activities, and I can talk about the same things for myself. I can use simple past, present and future tenses. 3. I know quite a bit of vocabulary and grammar, and I can carry on a conversation on a variety of topics. I can ask for directions and give directions, I can order a meal in a restaurant, I can give my opinion and have a simple discussion on common topics, such as the weather, work, studies, films and so on. I can use a variety of tenses, including conditional and subjunctive. 4. I have extensive knowledge of grammar and vocabulary, and I can carry on in depth conversations on most topics. I can talk with a native spe aker of any age or social background, and I can debate topics such as sports, politics and religion. I can use most tenses with little concern. 1. I can recognize a few words and phrases but cannot read or understand simple texts without looking up mo st words in a dictionary. Reading 2. I can get the gist of a simple text, but I need a dictionary to understand all of the words in the text. 3. I can understand a variety of texts with occasional use of a dictionary. 4. I can understand most texts with little use of a dic tionary. 1. I can write a few words and phrases but cannot put them into a grammatically correct sentence. Writing 2. I can write sentences containing simple grammatical structures, on a limited number of topics (descriptions of people, places, activities, and s o on). 3. I can write sentences containing complex grammatical structures (including subordinate and relative clauses), in a variety of registers (e mails, informal letters, compositions). I can write business letters and academic papers, using a variety of v ocabulary and grammatical structures.

PAGE 165

165 APPENDIX C SUMMARY OF PARTICIPANT QUESTIONNAIRES Table C 1. Summary of Participant Questionnaires No Class_ Per Sex Age Read Dis Eng Sp n Spn Origin Spn L/S/ R/W P rt Instr Pr t Origin Prt L/S/ R/W Oth Lang Oth Origin Oth L/S/ R/W 1 1130_3 F 18 N NNS NS DR; Miami; cl in elem & mid sch 4/4/2/2 Yes Sr yr HS 5 ms; teacher sick; cl cancel 2/1/1/1 French 2 yrs HS 2/1/1/1 2 3010 F 22 N NNS NS Colombia 4/4/4/4 No 3 1130_3 F 18 N NS NS Up to IB in HS; relatives 3/4/4/3 No 4 1130_3 F 22 N NNS NS Fr Venezuela 4/4/4/4 No 5 3010 F 19 N NS NNS Grades 2 5 3/1/3/2 Yes POR 1130; parents fr Portugal 4/2/3/2 French HS 3/1/2/2 6 1130_4 F 28 N NS NNS 2 yrs HS 2/2/2/2 No 7 1130_3 M 21 N NS NNS 2 yrs HS 1/1/1/1 No 8 1130_3 F 30 N NS None Yes Priv lessons; 2 ms Curitiba Urb Reg 1.5/1/1/ 1 French 2 yrs HS 2/1/1/1 9 1130_4 M 18 N NS None No 10 3010 F 24 N NS NNS SPN 3300 3/3/3/3 No 11 1130_4 M 18 N NS NNS 3 yrs HS 3/3/3/3 Yes Lev 1 Rosetta Sto 1/2/ 1/1 12 3010 F 21 N NS NNS 4000 level in Mexico 4/4/3/3 No French 2000 level 2/1/2/1

PAGE 166

166 Table C 1. Continued No Class_ Per Sex Age Read Dis Eng Sp n Spn Origin Spn L/S/ R/W P rt Instr Pr t Origin Prt L/S/ R/W Oth Lang Oth Origin Oth L/S/ R/W 13 3010 M 2 0 N NS NNS 3301 3/3/3/4 No 14 1130_4 F 20 N NS NNS 3 yrs HS 2/2/3/2 No Haitian NS 4/4/4/3 15 1130_3 M 23 N NS NNS 1 coll course 2/1/2/1 No Brazilian fiance 16 1130_4 F 22 N NNS NS K 10; El Salv till 15 4/4/4/4 No 17 1130_4 F 19 N NS NNS Chile, 1 4 y old 3/2/1/1 No French 2 yrs HS 2/2/1/1 18 1130_4 M 18 N NS None No French 4 yrs HS 3/3/3/3 19 1130_3 M 18 N NNS NS Venezuela till 5; IB Spanish 4/4/4/4 No 20 1130_4 M 18 N NS NNS 2 yrs HS 1/1/2/1 No 21 1130_3 F 19 N NS NNS 3 yrs HS 2/2/2/2 No 22 1130_3 M 19 N NS NNS 3 yrs HS 2/2/3/2 No 23 3010 F 23 N NS NNS Coll major; 7+ yrs study 4/4/4/3 No 24 1130_4 F 21 N NS NNS 4 yrs HS 2/2/2/2 No Italian 3 coll sem; Italy 3/3/3/3 25 1130_4 M 21 N NS NNS 10 days in Peru, 10 in Ecuador 2/2/2/1 No 26 3010 F 19 N NS NNS 2240 3/2/4/3 No

PAGE 167

167 Table C 1. Continued No Class_ Per Sex Age Read Dis Eng Sp n Spn Origin Spn L/S/ R/W P rt Instr Pr t Origin Prt L/S/ R/W Oth Lang Oth Origin Oth L/S/ R/W 27 1130_3 F 17 N NS NNS All through school 3/3/3/3 No 28 1130_3 F 17 N NS NNS K IB 3/3/3/3 No 29 1130_3 M 24 N NS None No Born in Brazil; moved to US at 4 2/2/1/1 30 1130_4 F 18 N NS NNS 2 cl HS 1/1/1/1 No 31 1130_4 M 18 N NS NNS 3 yrs HS 1/1/1/1 No 32 1130_3 F 19 N NS NNS 2 yrs HS 1/1/1/1 No 33 3010 M 20 N NS NNS 9 12 HS 2/1/2/1 Yes POR 1130; 1/2 family fluent in Port 3/2/3/3 French 2 yrs HS 2/1/2/2 34 3010 F 30 N NS None Yes Private tutor 3 yrs; 8 ms in Brazil 2/2/2/2 35 1130_3 F 18 N NS NNS 6 12 HS 3/3/3/3 No 36 1130_4 F 18 N NS NNS 7 10 HS 1/2/2/1 No French 2 yrs HS 2/2/2/2 37 3010 F 20 N NNS NS K 12, Colombia 4/4/4/4 No French 2241 3/3/3/2 38 1130_3 F 19 N NS NNS 3 yrs HS 2/2/3/2 No 39 3010 M 21 N NNS NS 4314; Me xico for 2 yrs 4/4/4/4 No Visited Brazil a few times 2/1/2/1 French Lvls 1 4 HS 3/2/3/2 40 1130_4 M 17 N NS NNS 9 10 HS 2/1/2/1 No

PAGE 168

168 Table C 1. Continued No Class_ Per Sex Age Read Dis Eng Sp n Spn Origin Spn L/S/ R/W P rt Instr Pr t Origin Prt L/S/ R /W Oth Lang Oth Origin Oth L/S/ R/W 41 1130_4 M 20 N NS NNS 7 9 HS 1/1/1/1 No Born in Br; moved to US at 4; parents only Eng 1/1/1/1 42 3010 F 18 N NS NNS 6 yrs 4/3/4/3 No Speaks at home with mom 3/3/1/1 43 3010 F 22 N NNS NS Graduated HS in Colom bia 4/4/4/4 No 44 1130_4 F 18 N NS None No 45 1130_3 M 18 N NS NS Since 5th grade 4/3/4/3 No 46 1130_4 M 21 N NS NNS 8 12 HS 2/2/2/2 No 47 1130_4 F 18 N NS None No French 7 10 HS 3/2/2/2 48 1130_4 M 19 N NNS NS Bolivia for 14 yrs 4/4/4/4 No 49 1130_3 M 22 N NS None No 50 3010 F 21 N NNS NS Studied for 6 yrs; Colombia 4/4/4/4 No 51 1130_3 F 19 N NS NNS 10 11 HS 3/2/3/2 No G'parents used to live in Br; never taught Port

PAGE 169

169 Table C 1. Continued No Class_ Pe r Sex Age Read Dis Eng Sp n Spn Origin Spn L/S/ R/W P rt Instr Pr t Origin Prt L/S/ R/W Oth Lang Oth Origin Oth L/S/ R/W 52 1130_4 M 22 N NS NNS 9 10 HS 2/2/1/1 No

PAGE 170

170 APPENDIX D WORDS USED IN READIN G TASKS Table D 1. Words used in reading tasks L anguage Number Word Gloss Category Symbol 1 Sound 1 Symbol 2 Sound 2 Portuguese 1 agenda agenda C g N/A N/A Portuguese 2 apiterrar N/A rr h N/A N/A Portuguese 3 apresentao presentation C s z N/A N/A Portuguese 4 arranjar* to arrange NC rr h j P ortuguese 5 arrogante arrogant C rr h N/A N/A Portuguese 6 atencioso attentive NC s z N/A N/A Portuguese 7 audiovisual audiovisual C s z N/A N/A Portuguese 8 azedo* sour NC z z N/A N/A Portuguese 9 barreida N/A N rr h N/A N/A Portuguese 10 bsico basi c C s z N/A N/A Portuguese 11 bazero N/A N z z N/A N/A Portuguese 12 bzios shells NC z z N/A N/A Portuguese 13 canja chicken soup NC j N/A N/A Portuguese 14 casaco coat NC s z N/A N/A Portuguese 15 casos cases C s z N/A N/A Portuguese 16 chato boring F N/A N/A N/A N/A Portuguese 17 chegar to arrive F N/A N/A N/A N/A Portuguese 18 chuberro N/A N rr h N/A N/A Portuguese 19 ciberrar N/A N rr h N/A N/A Portuguese 20 colonizao colonization C z z N/A N/A Portuguese 21 compromisso commitment F N/A N/A N/A N/A Portuguese 22 conjugar to conjugate C j N/A N/A Portuguese 23 corresponder to correspond C rr h N/A N/A Portuguese 24 cor rimo* banister NC rr h N/A N/A Portuguese 25 criana child F N/A N/A N/A N/A

PAGE 171

171 Table D 1. Continued Language Number Word Gloss Category Symbol 1 Sound 1 Symbol 2 Sound 2 Portuguese 26 curiosidade curiosity C s z N/A N/A Portuguese 27 curricular (sg*) curricular C rr h N/A N/A Portuguese 28 danar to dance F N/A N/A N/A N/A Portuguese 29 deciso decision C s z N/A N/A Portuguese 30 desenho drawing NC s z N/A N/A Portuguese 31 digital digital C g N/A N/A Portuguese 32 disudo N/A N s z N/A N/A Portuguese 33 divises divisions C s z N/A N/A Portuguese 34 dzia* dozen NC z z N/A N/A Portuguese 35 endereo address F N/A N/A N/A N/A Portuguese 36 esquisito weird C s z N/A N/A Portuguese 37 fecha r to close F N/A N/A N/A N/A Portuguese 38 feijoada bean stew NC j N/A N/A Portuguese 39 ferrete N/A N rr h N/A N/A Portuguese 40 ferrugem* rust NC rr h g Portuguese 41 feserel N/A N s z N/A N/A Portuguese 42 filosofia philosophy C s z N/A N/A Po rtuguese 43 garrafa* bottle NC rr h N/A N/A Portuguese 44 gemotal N/A N g N/A N/A Portuguese 45 general general (mil) C g N/A N/A Portuguese 46 generosidade generosity C g s z Portuguese 47 geral (in) general NC g N/A N/A Portuguese 48 gigante giant C g N/A N/A Portuguese 49 givido N/A N g N/A N/A Portuguese 50 gizes (pl*) chalks NC g z z Portuguese 51 guilajo N/A N j N/A N/A Portuguese 52 horizonte horizon C z z N/A N/A

PAGE 172

172 Table D 1. Continued Language Number Word Gloss Category Sy mbol 1 Sound 1 Symbol 2 Sound 2 Portuguese 53 idealizao idealization C z z N/A N/A Portuguese 54 igreja church NC j N/A N/A Portuguese 55 isgio N/A N s z g Portuguese 56 jao N/A N j N/A N/A Portuguese 57 jade* jade C j N/A N/A Portuguese 58 janela window NC j N/A N/A Portuguese 59 jantar dinner NC j N/A N/A Portuguese 60 jeletida N/A N j N/A N/A Portuguese 61 jilo N/A N j N/A N/A Portuguese 62 joaninha ladybug NC j N/A N/A Portuguese 63 jofiro N/A N j N/A N/A Portuguese 64 jornal newspaper C j N/A N/A Portuguese 65 junho June C j N/A N/A Portuguese 66 justificao justification C j N/A N/A Portuguese 67 laginho N/A N g N/A N/A Portuguese 68 lazer leisure NC z z N/A N/A Portuguese 69 localizado localized C z z N/A N/A Portuguese 70 lomosa N/A N s z N/A N/A Portuguese 71 manhoso* smart; whiny NC s z N/A N/A Portuguese 72 maresa N/A N s z N/A N/A Portuguese 73 moa young lady F N/A N/A N/A N/A Portuguese 74 mongendo N/A N g N/A N/A Portuguese 75 museu mus eum C s z N/A N/A Portuguese 76 narrar to narrate C rr h N/A N/A Portuguese 77 nerneja N/A N j N/A N/A Portuguese 78 nirruto N/A N rr h N/A N/A Portuguese 79 objeto object C j N/A N/A Portuguese 80 organizao organization C z z N/A N/A Portugues e 81 origem origin C g N/A N/A

PAGE 173

173 Table D 1. Continued Language Number Word Gloss Category Symbol 1 Sound 1 Symbol 2 Sound 2 Portuguese 82 paimozes N/A N z z N/A N/A Portuguese 83 pastiza N/A N z z N/A N/A Portuguese 84 persuasivo persuasive C s z N/A N/A Portuguese 85 pessoa person F N/A N/A N/A N/A Portuguese 86 pingente* pendant NC g N/A N/A Portuguese 87 pirralho* child (slang) NC rr h N/A N/A Portuguese 88 poloneses Polish (plural) NC s z N/A N/A Portuguese 89 preguioso lazy NC s z N/A N/A Portuguese 90 prenegem N/A N g N/A N/A Portuguese 91 presidente president C s z N/A N/A Portuguese 92 pressa hurry F N/A N/A N/A N/A Portuguese 93 prozida N/A N z z N/A N/A Portuguese 94 quasano N/A N s z N/A N/A Portuguese 95 rabicho N/A N r h N/ A N/A Portuguese 96 racial racial C r h N/A N/A Portuguese 97 ragaria N/A N r h N/A N/A Portuguese 98 rapazes boys NC r h z z Portuguese 99 reagir to react NC g r h Portuguese 100 real real C r h N/A N/A Portuguese 101 realizado realized C r h z z Portuguese 102 reciro N/A N r h N/A N/A Portuguese 103 refeio meal NC r h N/A N/A Portuguese 104 regio region C r h g Portuguese 105 regic N/A N r h g Portuguese 106 regional regional C r h g Portuguese 107 relatrio report NC r h N/A N/A Portuguese 108 renalda N/A N r h N/A N/A Portuguese 109 reviso revision C r h s z

PAGE 174

174 Table D 1. Continued Language Number Word Gloss Category Symbol 1 Sound 1 Symbol 2 Sound 2 Portuguese 110 rigor* rigor C r h N/A N/A Portuguese 111 rinagem N/A N r h g Portuguese 112 riscar* to risk NC r h N/A N/A Portuguese 113 rodumar N/A N r h N/A N/A Portuguese 114 romntico romantic C r h N/A N/A Portuguese 115 roxo purple NC r h N/A N/A Portuguese 116 rua street NC r h N/A N/A Portuguese 117 rulica N/A N r h N/A N/A Portuguese 118 rural rural C r h N/A N/A Portuguese 119 sergipano* from Sergipe NC g N/A N/A Portuguese 120 sorridente* smiling NC rr h N/A N/A Portuguese 121 televiso television C s z N/A N/A Portuguese 122 territrio territory C rr h N/ A N/A Portuguese 123 terrvel terrible C rr h N/A N/A Portuguese 124 transio transition C s z N/A N/A Portuguese 125 trazentar N/A N z z N/A N/A Portuguese 126 vagem* green bean NC j N/A N/A Portuguese 127 viso vision C s z N/A N/A Portuguese 128 visvel visible C s z N/A N/A Portuguese 129 zalito N/A N z z N/A N/A Portuguese 130 zangado angry NC z z N/A N/A English 1 ahead N/A N/A h h N/A N/A English 2 behalf N/A N/A h h N/A N/A English 3 bizarre N/A N/A z z N/A N/A English 4 cohort N/A N/ A h h N/A N/A English 5 decision N/A N/A s N/A N/A English 6 decisive N/A N/A s s N/A N/A English 7 engineer N/A N/A g N/A N/A English 8 fusion N/A N/A s N/A N/A

PAGE 175

175 Table D 1. Continued Language Number Word Gloss Category Symbol 1 Sound 1 Symbol 2 Sound 2 English 9 garage N/A N/A g(e) N/A N/A English 10 gender N/A N/A g N/A N/A English 11 generosity N/A N/A g s s English 12 gymnasium N/A N/A g s z English 13 happy N/A N/A h h N/A N/A English 14 heavy N/A N/A h h N/A N/A English 15 history N/A N/A h h N/A N/A English 16 hung ry N/A N/A h h N/A N/A English 17 jacket N/A N/A j N/A N/A English 18 jelly N/A N/A j N/A N/A English 19 journey N/A N/A j N/A N/A English 20 judge N/A N/A j dg English 21 leisure N/A N/A s N/A N/A English 22 philosophy N/A N/A s s N/A N/ A English 23 physics N/A N/A s z N/A N/A English 24 pleasure N/A N/A s N/A N/A English 25 present N/A N/A s z N/A N/A English 26 realization N/A N/A r z z English 27 reclusive N/A N/A r s s English 28 regenerate N/A N/A r g English 29 regi on N/A N/A r g English 30 reheat N/A N/A r h h English 31 result N/A N/A r s z English 32 television N/A N/A s N/A N/A English 33 treasure N/A N/A s N/A N/A English 34 visual N/A N/A s N/A N/A English 35 zebra N/A N/A z z N/A N/A Engli sh 36 zipper N/A N/A z z N/A N/A Spanish 1 antropologa anthropology N/A g h N/A N/A

PAGE 176

176 Table D 1. Continued Language Number Word Gloss Category Symbol 1 Sound 1 Symbol 2 Sound 2 Spanish 2 asno donkey N/A s s/z N/A N/A Spanish 3 caja box N/A j h N/A N/A Spanish 4 calle street N/A ll j/ N/A N/A Spanish 5 caza hunt N/A z s N/A N/A Spanish 6 desde since N/A s s/z N/A N/A Spanish 7 girasol sunflower N/A g h s s Spanish 8 jardn garden N/A j h N/A N/A Spanish 9 jefe boss N/A j h N/A N/A Spanish 10 llamar to call N/A ll j/ N/A N /A Spanish 11 lleno full N/A ll j/ N/A N/A Spanish 12 mismo same N/A s s/z N/A N/A Spanish 13 refrigerador refrigerator N/A r r g h Spanish 14 rezar to pray N/A r r z s Spanish 15 rojo red N/A r r j h Spanish 16 rollo roll N/A r r ll j/ Spanis h 17 zapato shoe N/A z s N/A N/A Spanish 18 zarpar to set sail N/A z s N/A N/A Spanish 19 zocato overripe N/A z s N/A N/A Spanish 20 zumbar to hum N/A z s N/A N/A

PAGE 177

177 APPENDIX E INSTRUCTION AND SAMP LE SLIDES FOR READIN G TASK Figure E 1. Instruction s lides and sample word slides from PowerPoint reading task

PAGE 178

178 APPENDIX F TRANSCRIPTION G UIDE FOR R ATERS Table F 1. Transcription guide for raters Language Grapheme Symbol for Sound Portuguese s apo/cal a s ca s a/ z ebra z a ch o tch au j acar d iabo r ato/ca rr o h English s ad s de s ign/ z ebra z sh ow ch ur ch vi s ual j acket h ospital h r eact Spanish z apato/ca s a s mi s mo s/z ca ll e j/ / ch ico j efe r ollo/pe rr o r Recall that the orig inal data set was much larger and included more sounds, but only the sounds [s] and [z] are analysed in this dissertation.

PAGE 179

179 APPENDIX G RATER TRANSCRIPTION SHEET Figure G 1. Rater Transcription Sheet Page 1 Words in capital letters (PESSOA and FECHAR) in dicate the start point of the two different PowerPoint versions created for the task

PAGE 180

180 Figure G 2. Rater Transcription Sheet Page 2 English and Spanish words were read only on the first test; for tests 2 and 3, only the Portuguese words appeared on the s econd page of the transcription sheet.

PAGE 181

181 APPENDIX H SAMPLE OF PARTICIPAN T PRODUCTION DATA AND ASSIGNED VALUES FO R < Z > AND < S > ITEMS Table H 1. Sample of participant production data and assigned values for and items Word Type C1 C1 C1 C1 C1 C1 C1 C1 C1 Grapheme s s s s s s s s s Value SYMBOL TYPE VALUE SYMBOL TYPE VALUE SYMBOL TYPE VALUE No apresentao apresentao apresentao esquisito esquisito esquisito museu museu museu 1 s SP 0 s SP 0 s SP 0 2 s SP 0 s SP 0 s SP 0 3 s SP 0 s SP 0 s SP 0 4 s SP 0 s SP 0 s SP 0 5 z EP 1 z EP 1 s SP 0 6 z EP 1 z EP 1 s SP 0 7 z EP 1 z EP 1 z EP 1 8 z EP 1 z EP 1 z EP 1 9 z EP 1 z EP 1 s SP 0 10 z EP 1 s SP 0 s SP 0 11 s SP 0 s SP 0 z EP 1 12 s SP 0 s SP 0 s SP 0 13 s SP 0 s SP 0 s SP 0 14 z EP 1 s SP 0 OTH 0 15 s SP 0 z EP 1 z EP 1 16 s SP 0 s SP 0 s SP 0 17 s SP 0 s SP 0 s SP 0 18 s SP 0 z EP 1 s SP 0 19 s SP 0 z EP 1 z EP 1 20 z EP 1 z EP 1 z EP 1 21 z EP 1 z EP 1 z EP 1

PAGE 182

182 APPENDIX I PERCENTAGES OF CORRE CT PRODUCTION BY PAR TICIPANT AND GROUP F OR EACH GRAPHEME AND TEST Table I 1. Percentages of correct production by participant and group for each grapheme and test Part Group Test 1 % Test 2 % Test 3 % Test 1 % Test 2 % Test 3 % Test 1 All % Test 2 All % Test 3 All % All % All % All % 6 E1 24 34 38 42 74 84 31 50 56 67 32 46 7 E1 21 21 24 100 100 100 52 52 54 100 22 53 9 E1 14 14 17 84 79 79 42 40 42 81 15 41 14 E1 7 17 34 32 95 89 17 48 56 72 20 40 17 E1 3 17 24 53 79 84 23 42 48 72 15 38 18 E1 7 38 45 37 8 4 89 19 56 63 70 30 46 20 E1 48 59 66 95 89 100 67 71 79 95 57 72 21 E1 14 28 24 100 100 100 48 56 54 100 22 53 22 E1 17 34 21 68 79 95 38 52 50 81 24 47 24 E1 10 7 3 74 21 47 35 13 21 47 7 23 25 E1 38 28 41 100 100 100 63 56 65 100 36 61 27 E1 0 0 1 0 42 74 68 17 29 33 61 3 26 28 E1 0 3 17 68 79 100 27 33 50 82 7 37 30 E1 3 45 24 0 84 74 2 60 44 53 24 35 31 E1 7 14 17 79 68 74 35 35 40 74 13 37 32 E1 17 31 24 89 89 95 46 54 52 91 24 51 35 E1 24 10 14 100 84 89 54 40 44 91 16 46 36 E1 10 45 45 47 89 89 25 63 63 75 33 50 38 E1 14 24 21 95 95 89 46 52 48 93 20 49 40 E1 21 55 38 74 79 84 42 65 56 79 38 54 41 E1 3 14 17 95 95 74 40 46 40 88 11 42 44 E1 14 41 34 89 89 95 44 60 58 91 30 54 46 E1 34 31 34 89 95 100 56 56 60 95 33 58 47 E1 41 48 52 89 79 100 60 60 71 89 47 64

PAGE 183

183 Table I 1. Continued Part Group Test 1 % Test 2 % Test 3 % Test 1 % Test 2 % Test 3 % Test 1 All % Test 2 All % Test 3 All % All % All % All % 49 E1 31 52 38 100 100 100 58 71 63 100 40 64 51 E1 28 34 17 84 84 74 50 54 40 81 26 48 52 E1 17 17 10 74 89 89 40 46 42 84 15 42 E1 Average 17 28 28 74 84 88 40 50 51 82 24 47 10 E2 3 14 38 32 47 84 15 27 56 54 18 33 12 E2 0 3 24 21 74 79 8 31 46 58 9 28 13 E2 3 31 55 16 58 79 8 42 65 51 30 38 23 E2 0 14 24 5 53 95 2 29 52 51 13 28 26 E2 28 28 21 89 95 100 52 54 52 95 25 53 E2 Average 7 18 32 33 65 87 17 37 54 62 19 36 1 P 0 17 28 21 37 68 8 25 44 42 15 26 5 P 31 24 38 89 100 89 54 54 58 93 31 56 8 P 28 24 21 95 89 84 54 50 46 89 24 50 11 P 2 4 24 31 84 68 74 48 42 48 75 26 46 15 P 31 48 59 95 100 100 56 69 75 98 46 67 29 P 45 59 69 100 100 100 67 75 81 100 57 74 33 P 41 52 55 74 79 95 54 63 71 82 49 63 34 P 52 55 66 95 100 100 69 73 79 98 57 74 42 P 21 45 52 37 74 95 27 56 69 68 39 51 P Average 30 39 46 77 83 89 49 56 63 83 38 56 2 S 3 3 14 5 26 68 4 13 35 33 7 17 3 S 0 0 7 0 11 32 0 4 17 14 2 7 4 S 3 0 0 16 26 16 8 10 6 19 1 8 16 S 0 0 0 5 5 11 2 2 4 7 0 3 19 S 31 48 34 58 89 84 42 65 54 77 38 53 37 S 7 21 79 5 58 89 6 35 83 51 36 42 39 S 55 41 41 95 95 95 71 63 63 95 46 65 43 S 0 3 7 0 5 37 0 4 19 14 3 8

PAGE 184

184 Table I 1. Continued Part Group Test 1 % Test 2 % Test 3 % Test 1 % Test 2 % Test 3 % Test 1 All % Test 2 All % Test 3 All % All % All % All % 45 S 0 7 3 37 37 53 15 19 23 42 3 19 48 S 10 28 10 21 21 11 15 25 10 18 16 17 50 S 0 24 28 0 63 74 0 40 46 46 17 28 S Average 10 16 20 22 40 52 15 25 33 38 15 24

PAGE 185

185 APPENDIX J PRODUCTION ON SPANIS H READING TASK AND A VERAGE PRODUCTION ON PORTUGUESE READ ING TASKS Table J 1. Production on Spanish reading task and average production on Portuguese reading tasks Part Group Accuracy on Spanish Items % Average Accuracy on Portuguese Items % 44 E1 5 54 41 E1 9 42 8 P 14 50 5 P 18 56 15 P 18 67 32 E1 18 51 25 E1 23 61 34 P 23 74 47 E1 23 64 33 P 27 63 7 E1 32 53 17 E1 32 38 29 P 36 74 51 E1 36 48 52 E1 36 42 9 E1 41 41 21 E1 41 53 31 E1 41 37 18 E1 45 46 46 E1 45 58 20 E1 50 72 35 E1 50 46 38 E1 50 49 49 E1 50 64 14 E1 59 40 22 E1 59 47 26 E2 59 53 40 E1 59 54 42 P 64 51 4 S 68 8 6 E1 68 46 24 E1 68 23 36 E1 68 50

PAGE 186

186 Table J 1. Continued Part Group Accuracy on Spanish Items % Average Accuracy on Portuguese Items % 45 S 68 19 27 E1 82 26 11 P 86 46 28 E1 86 37 48 S 86 17 12 E2 91 28 30 E1 91 35 50 S 95 28 1 P 100 26 2 S 100 17 3 S 100 7 10 E2 100 33 13 E2 100 38 16 S 100 3 19 S 100 53 23 E2 100 28 37 S 100 42 39 S 100 65 43 S 100 8 The results are ordered according to accuracy on the Spanish reading task.

PAGE 187

187 APPENDIX K PERCENTAGES OF PRODU CTION BY PARTICIPANT FOR EACH PRODUCTION TYPE AND GRAPHEME Table K 1. Percentages of Production by Participant for Each Production Type and Grapheme Part Group EN EP ES NO OTH PO SP EP N O O TH SP 6 E1 0 10 51 1 2 22 14 67 4 0 30 7 E1 3 16 59 0 0 6 16 100 0 0 0 9 E1 1 10 61 0 0 5 23 81 2 12 5 14 E1 2 5 52 0 8 15 18 72 2 2 25 17 E1 1 7 54 1 2 8 26 72 4 4 21 18 E1 1 10 53 0 1 20 15 70 0 18 12 20 E1 3 17 36 0 0 40 3 95 0 2 4 21 E1 1 14 6 0 0 0 8 17 100 0 0 0 22 E1 1 15 55 0 0 9 20 81 5 0 14 24 E1 0 7 62 0 1 0 30 47 2 18 33 25 E1 3 16 54 0 0 20 7 100 0 0 0 27 E1 0 0 59 0 0 3 38 61 2 0 37 28 E1 0 0 54 2 0 7 37 82 4 0 14 30 E1 1 6 54 0 0 18 21 53 0 0 47 31 E1 7 8 60 0 1 5 20 74 4 7 16 32 E1 3 11 57 0 0 13 15 91 0 0 9 35 E1 1 9 56 0 1 7 25 91 5 0 4 36 E1 3 9 47 0 0 24 16 75 4 2 19 38 E1 1 11 57 0 1 8 21 93 0 2 5 40 E1 2 16 51 0 0 22 9 79 0 0 21 41 E1 10 6 57 6 2 6 13 88 2 2 9 44 E1 1 10 47 5 3 20 14 91 7 0 2 46 E1 3 10 54 1 2 23 6 95 0 0 5 47 E1 2 17 44 0 0 30 7 89 0 2 9 49 E1 0 15 53 0 0 25 7 100 0 0 0 51 E1 3 14 53 0 3 13 14 81 0 2 18 52 E1 2 7 57 1 0 8 24 84 0 0 16 10 E2 0 8 52 0 3 10 26 54 0 0 46 12 E2 0 1 54 0 0 8 37 58 0 0 42 13 E2 0 9 43 0 0 21 28 51 4 0 46 23 E2 0 5 55 0 0 8 32 51 4 5 40 26 E2 0 13 59 1 0 13 15 95 0 0 5 1 P 0 2 55 0 2 13 28 42 0 4 54 5 P 3 11 49 1 0 20 15 93 2 0 5 8 P 10 16 61 0 0 8 5 89 0 5 5

PAGE 188

188 Table K 1. Continued Part Group EN EP ES NO OTH PO SP EP N O O TH < z> SP 11 P 3 7 49 0 0 20 21 75 4 7 14 15 P 7 16 40 1 0 30 6 98 0 0 2 29 P 0 14 36 0 0 44 7 100 0 0 0 33 P 0 15 47 0 1 34 2 82 0 2 16 34 P 0 14 32 0 0 44 10 98 0 0 2 42 P 2 13 45 0 1 26 13 68 0 0 32 2 S 0 0 56 0 0 7 37 33 0 2 65 3 S 0 1 61 0 0 1 37 14 2 0 84 4 S 1 0 62 1 5 1 30 19 0 12 68 16 S 0 0 62 0 0 0 38 7 0 0 93 19 S 0 9 47 0 0 29 15 77 0 0 23 37 S 0 11 45 0 0 24 20 51 7 0 42 39 S 1 13 39 0 0 33 14 95 0 0 5 43 S 0 0 59 0 0 3 38 14 0 0 86 45 S 1 2 61 0 0 1 34 42 2 2 54 48 S 0 6 56 0 0 10 28 18 0 0 82 50 S 0 3 55 0 0 14 28 46 0 0 54

PAGE 189

189 APPENDIX L PERCENTAGES OF PRODU CTION BY PRODUCTION TYPE FOR EACH WORD Table L 1. Percentages of Production by Production Type for Each Word Word Type English sound EN % EP % ES % N O % O TH % PO % SP % a presentao C [z] 0 53 0 0 0 0 47 esquisito C [z] 0 47 0 3 1 0 49 museu C [z] 0 47 0 0 1 0 53 presidente C [z] 0 68 0 0 1 0 31 visvel C [z] 0 46 0 2 0 0 53 bsico C [s] 0 0 92 0 1 8 0 casos C [s] 0 0 73 0 0 27 0 curiosidade C [s] 0 0 93 1 0 6 0 fi losofia C [s] 0 0 97 1 0 2 0 generosidade C [s] 0 0 99 0 0 1 0 persuasivo C [s] 0 0 92 3 0 6 0 audiovisual C [ 35 0 0 0 8 28 30 deciso C [ 3 0 0 1 4 27 66 divises C [ 1 0 0 0 1 13 85 reviso C [ 2 0 0 0 0 35 63 televiso C [ 5 0 0 0 1 37 57 viso C [ 5 0 0 0 1 48 46 disudo N 0 0 87 0 0 13 0 feserel N 0 0 80 0 1 19 0 isgio N 0 0 76 0 0 24 0 lomosa N 0 0 89 0 0 11 0 maresa N 0 0 83 0 0 17 0 quasano N 0 0 77 0 1 22 0 atencioso NC 0 0 87 1 1 10 0 casaco NC 0 0 70 1 1 29 0 desenho NC 0 0 72 0 1 27 0 manhoso NC 0 0 94 0 0 6 0 poloneses NC 0 0 81 0 0 19 0 preguioso NC 0 0 84 1 2 13 0 idealizao C 0 62 0 15 0 0 23 localizado C 0 67 0 0 1 0 33 organizao C 0 71 0 4 1 0 24 realizado C 0 71 0 0 2 0 27 coloniza o C 0 67 0 4 1 0 28 horizonte C 0 86 0 0 0 0 14 bazero N 0 71 0 0 0 0 29

PAGE 190

190 Table L 1. Continued Word Type English sound EN % EP % ES % N O % O TH % PO % SP % paimozes N 0 53 0 0 3 0 44 prozida N 0 77 0 0 2 0 21 trazentar N 0 83 0 0 1 0 15 pasti za N 0 70 0 1 8 0 21 zalito N 0 79 0 0 1 0 20 azedo NC 0 79 0 0 1 0 20 bzios NC 0 66 0 0 4 0 29 dzia NC 0 73 0 1 4 0 22 gizes NC 0 65 0 0 2 0 33 lazer NC 0 77 0 0 1 0 22 rapazes NC 0 55 0 0 6 0 39 zangado NC 0 74 0 0 1 0 25

PAGE 191

191 APPENDIX M PERCENTAGES OF PRODU CTION BY WORD Table M 1. Percentages of Production by Word Word Grapheme Word Type English Cognate Sound Accuracy (Raw Scores) Accuracy (Percentage) apresentao C [z] 83 53 esquisito C [z] 74 47 museu C [z] 73 47 presidente C [z] 106 68 visvel C [z] 71 46 bsico C [s] 12 8 casos C [s] 42 27 curiosidade C [s] 10 6 filosofia C [s] 3 2 generosidade C [s] 1 1 persuasivo C [s] 9 6 audiovisual C [ 43 28 deciso C [ 42 27 divises C [ 20 13 reviso C [ 55 35 televiso C [ 58 37 viso C [ 75 48 disudo N 20 13 feserel N 30 19 isgio N 38 24 lomosa N 17 11 maresa N 26 17 quasano N 35 22 atencioso N C 16 10 casaco N C 45 29 desenho N C 42 27 manhoso N C 10 6 poloneses N C 30 19 preguioso N C 21 13 idealizao C 97 62 localizado C 104 67 organizao C 110 71 realizado C 111 71 colonizao C 104 67 horiz onte C 134 86

PAGE 192

192 Table M 1. Continued Word Grapheme Word Type English Cognate Sound Accuracy (Raw Scores) Accuracy (Percentage) bazero N 110 71 paimozes N 83 53 prozida N 120 77 trazentar N 130 83 pastiza N 109 70 z alito N 124 79 azedo NC 123 79 bzios NC 103 66 dzia NC 114 73 gizes NC 102 65 lazer NC 120 77 rapazes NC 86 55 zangado NC 116 74

PAGE 193

193 LIST OF REFERENCES Altmann, G. T. M. (1997). The ascent of Babel: An explora tion of language, mind and understanding. Oxford, UK: Oxford University Press. Aronin, L., & Hufeisen, B. (eds). (2009). The exploration of multilingualism: Development of research on L3, multilingualism and multiple language acquisition. Philadelphia, PA: John Benjamins. Azevedo, M. M. (1978). Identifying Spanish interference in the speech of learners of Portuguese. The Modern Language Journal, 62, 1823. Baker, C. (2006). Foundations of bilingual education and bilingualism. Clevedon, UK: Multilingual Matters. Baker, W., & Trofimovich, P. (2005). Interaction of native and second language vowel system(s) in early and late bilinguals. Language and Speech, 48, 127. Blanche, P. (2004). Using dictation to teach pronunciation. Linguagem & Ensino, 7, 175191. B loomfield, L. (1933). Language. New York, NY: Holt, Rinehart and Winston. Boxer, D., & Cohen, A. (Eds.) (2004). Studying speaking to inform second language learning. Clevedon, UK: Multilingual Matters. Broselow, E., Chen, S., & Wang, C. (1998). The emerge nce of the unmarked in second language phonology. Studies in Second Language Acquisition, 20, 261280. Brown, R. (1973). A first language: The early stages. Cambridge, MA: Harvard University Press. Carvalho, A. M. (2002). Portugus para falantes de espanho l: Perspectivas de um campo de pesquisa. Hispania, 85, 597608. Cenoz, J. (2000). Research on multilingual acquisition. In J. Cenoz & U. Jessner (Eds.), English in Europe: The acquisition of a third language (pp. 3953). Clevedon, UK: Multilingual Matters Cenoz, J. (2001). The effect of linguistic distance, L2 status and age on cross linguistic influence in third language acquisition. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross linguistic influence in third language acquisition: Psycholinguistic perspectives (pp. 820). Clevedon, UK: Multilingual Matters. Cenoz, J. (2003). The additive effect of bilingualism on third language acquisition: A review. The International Journal of Bilingualism, 7, 71 89. Cenoz, J., & Genesee, F. (eds.) (1998). Beyond bilingualism: Multilingualism and multilingual education. Clevedon, UK: Multilingual Matters.

PAGE 194

194 Cenoz, J., & Hoffmann, C. (2003). Acquiring a third language: What role does bilingualism play? The International Journal of Bilingualism, 7, 1 6. Cenoz, J., Huf eisen, B., & Jessner, U. (eds). (2001). Cross linguistic influence in third language acquisition: Psycholinguistic perspectives. Clevedon, UK: Multilingual Matters. Chandler, R. E. (1958). Spanish Portuguese transfer. The Modern Language Journal, 42, 329336. Chomsky, N. (1965). Aspects of the Theory of Syntax Cambridge, MA: MIT Press. Chomsky, N. (1959). Review of verbal behavior. Language, 35, 2658. Cohen, S. P., Tucker, R., & Lambert, W. E. (1967). The comparative skills of monolinguals and bilinguals in perceiving phoneme sequences. Language and Speech, 10, 159168. Corder, S. (1967). A role for the mother tongue. In S. Gass & L. Selinker (Eds.), Language transfer in language learning (pp. 85 97). Rowley, MA: Newbury House. Costa, A., Caramazza, A., & Sebastin Galles, N. (2000). The cognate facilitation effect: Implications for models of lexical access. Journal of Experimental Psychology: Learning, Memory and Cognition, 26, 12831296. Costa, A., Santesteban, M, & Cao, A. (2005). On the facilitatory effects of cognate words in bilingual speech production. Brain and Language, 94, 94103. Cummins, J. (1979). Linguistic interdependence and the educational development of bilingual children. Review of Educational Research, 49, 221251. Davine, M., Tucker, R., & Lambert, W. E. (1971). The perception of phoneme sequences by monolingual and bilingual elementary school children. Canadian Journal of Behavioral Science, 3, 7276. De Angelis, G. (2005). Multilingualism and non native lexical tr ansfer: An identification problem. International Journal of Multilingualism, 2, 125. De Angelis, G. (2007). Third or additional language acquisition. Clevedon, UK: Multilingual Matters. Defior, S., Martos, F., & Cary, L. (2002). Differences in reading acq uisition development in two shallow orthographies: Portuguese and Spanish. Applied Psycholinguistics, 23, 135148. Dewaele, J. (1998). Lexical inventions: French interlanguage as L2 versus L3. Applied Linguistics, 19, 471490.

PAGE 195

195 Dewaele, J. (2001). Activatio n or inhibition? The interaction of L1, L2 and L3 on the language mode continuum. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross linguistic influence in third language acquisition: Psycholinguistic perspectives (pp. 6989). Clevedon, UK: Multilingual Matters. Ecke, P. (2001). Lexical retrieval in a third language: Evidence from errors and tipof the tongue states. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross linguistic influence in third language acquisition: Psycholinguistic perspectives (pp. 90114). Clevedon, UK: Multilingual Matters. Ecke, P., & Hall, C. J. (2000). Lexicalische Fehler in Deutsch als Drittsprache: Translexikalischer Einfluss auf 3 Ebenen der mentalen Reprsentation. Deutsch als Fremdsprache, 1, 3137. Ellis, R. (2008). The s tudy of s econd l anguage a cquisition (2nd ed.). Oxford, UK: Oxford University Press. Ellison, F. P., & Andrews, N. (1969). Portuguese in the high schools. In D. D. Walsh (Ed.), A h andbook for t eachers of Spanish and Portuguese (pp. 250264) Lexington, MA: D. C. Heath. Enomoto, K. (1994). L2 perceptual acquisition: The effect of multilingual linguistic experience on the perception of a less novel contrast. Edinburgh Working Papers in Applied Linguistics, 5, 1529. Errasti, M. P. S. (2003). Acquiring writin g skills in a third language: The positive effects of bilingualism. The International Journal of Bilingualism, 7, 27 42. Faerch, C., & Kasper, G. (1986). Cognitive dimensions of language transfer. In M. Sharwood Smith & E. Kellerman (Eds.), Crosslinguistic influence in second language acquisition (pp. 4965). Oxford, UK: Pergamon Press. Faraco, C. E., & Moura, F. M. (1990). Gramtica. So Paulo, Brazil: Editora tica. Flege, J. E. (1986). The production and perception of foreign language speech sounds. In H Wintz (Ed.), Human Communication and its Disorders, Vol. II (224401). Norwood, NJ: Ablex Publishing. Flege, J. E. (1987). The production of new and similar phones in a foreign language: evidence for the effect of equivalence classification. Journal of Phonetics, 15, 4765. Fromkin, V., Rodman, R., & Hyams, N. (2007). An introduction to language. Boston, MA: Thomson Wadsworth. Garrison, D. L. (1979). Teaching the relatedness of Spanish and Portuguese. The Modern Language Journal, 63, 812.

PAGE 196

196 Gass, S. (1 983). Language transfer and universal grammatical relations. In S. Gass & L. Selinker (Eds.), Language transfer in language learning (pp. 69 82). Rowley, MA: Newbury House. Gass, S. (1984). A review of interlanguage syntax: Language transfer and language universals. Language Learning, 34, 115131. Gass, S., & Selinker, L. (2001). Second language acquisition: An introductory course. Mahwah, NJ: Lawrence Erlbaum Associates. Gibson, M., Hufeisen, B., & Libben, G. (2001). Learners of German as an L3 and their production of German prepositional verbs. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross linguistic influence in third language acquisition: Psycholinguistic perspectives (pp. 138148). Clevedon, UK: Multilingual Matters. Gleitman, L. R. (1985). Ortho graphic resources affect reading acquisition If they are used. Remedial and Special Education, 6, 24 36. Goodwin, J. (2001). Teaching pronunciation. In M. Celce Murcia (Ed.), Teaching English as a s econd or f oreign l anguage (3rd ed.) (pp. 117133) Bosto n, MA: Heinle & Heinle. Goto, H. (1971). Auditory perception by normal Japanese adults of the sounds l and r. Neuropsychologia, 9, 317323. Grosjean, F. (2001). The bilinguals language modes. In J. Nicol (Ed.), One mind, two languages: Bilingual langu age processing (pp. 122). Oxford, UK: Blackwell. Hadley, A. O. (1986). Teaching language in context: Proficiency oriented instruction. Boston, MA: Heinle & Heinle. Hammarberg, B. (2001). Roles of L1 and L2 in L3 production and acquisition. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross linguistic influence in third language acquisition: Psycholinguistic perspectives (pp. 2141). Clevedon, UK: Multilingual Matters. Hansen Edwards, J. G. H. & Zampini, M. L. (eds.). (2008). Phonology and second language acquisition. Philadelphia, PA: John Benjamins. Hermans, D., Bongaerts, T., De Bot, K., & Schreuder, R. (1998). Producing words in a foreign language: can speakers prevent interference from their first language? Bilingualism: Language and Cognition, 1, 213230. Herwig, A. (2001). Plurilingual lexical organization: evidence from lexical processing in L1L2 L3 L4 translation. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross linguistic influence in third language acquisition: Psycholinguistic perspectives (pp. 115137). Clevedon, UK: Multilingual Matters. Hockett, C. (1950). Learning pronunciation. The Modern Language Journal, 34, 261269. Holton, J. S. (1954). Portuguese for Spanish speakers. Hispania, 37, 446452.

PAGE 197

197 Hunt, K. (1965). Grammatical structures written at three grade levels NCTE Research Report, 3. Champaign, I L: NCTE. Jarvis, S., & Odlin, T. (2000). Morphological type, spatial reference, and language transfer. Studies in Second Language Acquisition, 22, 535556. Jebsen, H., & Biel, S. H (Eds.). (1986). Lets go: The budget guide to Spain, Portugal and Morocco. New York, NY: St. Martin. Jensen, J. B. (1989). On the mutual intelligibility of Spanish and Portuguese. Hispania, 72, 848852. Jordan, I. (1991). Portuguese for Spanish speakers: A case for contrastive analysis. Hispania, 74, 788792. Kasper, G., & Rose, K. R. (2002). Pragmatic development in a second language. Oxford, UK: Blackwell. Keys, K. G. (2000). Discourse level phonology in the language curriculum: A review of current thinking in teaching pronunciation in EFL courses. Linguagem & Ensino, 3, 89105. Khuri, A.I. (2009). Linear m odel m ethodology. London, UK: Chapman & Hall/CRC. Klobucka, A., Jout Pastr, C. M. C., Moreira, M. L. de B., Sobral, P. I., & Hutchinson, A. P. (2007). Ponto de e ncontro: Portuguese as a w orld l anguage. Upper Saddle River, NJ : Pearson Prentice Hall. Koda, K. (1999). Development of L2 intraword orthographic sensitivity and decoding skills. The Modern Language Journal, 83, 5164. Kohnert, K. (2004). Cognitive and cognate based treatments for bilingual aphasia: A case study. Brai n and Language, 91, 294302. Krashen, S. (1982). Principles and practice in s econd l anguage a cquisition. London, UK: Pergamon. Krashen, S. (1983). Newmarks Ignorance Hypothesis and current SLA theory. In S. Gass & L. Selinker (Eds.), Language transfer i n language learning (pp. 135153). Rowley, MA: Newbury House. Kroll, J. F., Dijkstra, A., Janssen, N., & Schriefers H. ( 2000). Selecting the language in which to speak: Experiments on lexical access in bilingual production. Paper presented at the 41st ann ual meeting of the psychonomic society, New Orleans, LA. Lazaraton, A. (2001). Teaching Oral Skills. In M. Celce Murcia (Ed.), Teaching English as a Second or Foreign Language (3rd ed.) (pp. 103115) Boston, MA: Heinle & Heinle.

PAGE 198

198 Lord, G. (2005). (How) Can we teach foreign language pronunciation? On the effects of a Spanish phonetics course. Hispania, 88, 557567. Lord, G. (2007). The role of the lexicon in learning second language stress patterns. Applied Language Learning, 17, 114. Major, R. C. (2001). F oreign accent: The ontogeny and phylogeny of second language phonology. Mahwah, NJ: Lawrence Erlbaum Associates. Margolis, M. (2009). Brazils Oil Rush. Newsweek, Oct. 2. Muller, D. H., & Muller, T. V. (1968). The problem of interference in beginning Portuguese. The Modern Language Journal, 52, 201205. Murphy, S. (2003). Second language transfer during third language acquisition. Columbia University Working Papers in TESOL and Applied Linguistics, 3, 121. Odlin, T. (1989). Language transfer. Cambridge, UK : Cambridge University Press. Olshtain, E. (2001). Functional tasks for mastering the mechanics of writing and going just beyond. In M. Celce Murcia (Ed.), Teaching English as a second or foreign language (pp. 207217). Boston, MA: Heinle & Heinle. Ornstei n, J. (1942). Problems in the teaching of Portuguese. The Modern Language Journal, 26, 512516. Orr, G. (1987). Aspects of the second language acquisition of Chichewa noun class morphology Unpublished doctoral dissertation, University of California, Los A ngeles. Peal, E., & Lambert, W. E. (1962). The relationship of bilingualism to intelligence. Psychological Monographs, 76, 1 23. Poulisse, N., & Bongaerts, T. (1994). First language use in second language production. Applied Linguistics, 15, 3657. Resnick S. (1945). Pitfalls in Spanish Portuguese homonyms and cognates. Hispania, 28, 233235. Roberts, P., & Deslauriers, L. (1999). Picture naming of cognate and non cognate nouns in bilingual aphasia. Journal of Communication Disorders, 32, 123. Schwegler, A., & Kempff, J. (2007). Fontica y fonologa espaolas (3rd ed). Hoboken, NJ: John Wiley & Sons. Selinker, L. (1972). Interlanguage. International Review of Applied Linguistics, 10, 209231. Selinker, L., & Baumgartner Cohen, B. (1995). Multiple language acquisition: Damn it, why cant I keep these two languages apart?. Language, C ulture and C urriculum, 8, Specia I ssue: Multilingualism and L anguage L earning, 115123.

PAGE 199

199 Shanon, B. (1991). Faulty language selection in polyglots. Language and Cognitive Proce sses, 6, 339350. Singleton, D. (1999). Exploring the second language mental lexicon. Cambridge, UK: Cambridge University Press. Skinner, B. F. (1957). Verbal behavior. New York, NY: Appleton Century Crofts. Tarquinio, L. T. L. (1977). The interference of Spanish in beginning Portuguese classes. Hispania, 60, 8287. Timberlake, C. (1989). Portugal still wary of closest neighbor: Report from Portugal. The Miami Herald, May 31, 2A. Vitevitch, M. S. (2002). The influence of phonological similarity neighborhoods on speech production. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 735747. Vitevitch, M. S. (2003). The influence of sublexical and lexical representations on the processing of spoken words in English. Clinical Linguistics an d Phonetics, 17, 487499. Wade Woolley, L. (1999). First language influences on second language word reading: All roads lead to Rome. Language Learning, 49, 447471. Werker, J., & Tees, R. (1984). Phonemic and phonetic factors in the adult cross language s peech perception. Journal of the Acoustical Society of America, 75, 18661878. White, L. (2003). Second language acquisition and universal grammar. Cambridge, UK: Cambridge University Press. Yamada, R., & Tohkura, Y. (1992). Perception of American English /r/ and /l/ by native speakers of Japanese. In E. Tohkura et al. (Eds.), Speech perception, production and linguistic structure (pp. 155174). Tokyo, Japan: Ohmsha.

PAGE 200

200 BIOGRAPHICAL SKETCH Sharon Barkley earned a JointHonours Bachelor of Arts in Linguistics and Portuguese from the University of Leeds, Leeds, U.K., in 1993. From the University of Florida, she earned a Certificate in Teaching English as a Second Language in 2006, and her PhD in Linguistics in 2010. Sharons research interests include second an d third language acquisition, phonology (in particular, pronunciation), and Portuguese pedagogy.