Citation
The effect of syntax on readability for Spanish-speaking adult students of English as a second language

Material Information

Title:
The effect of syntax on readability for Spanish-speaking adult students of English as a second language
Creator:
Blau, Eileen Kay, 1946-
Publication Date:
Copyright Date:
1980
Language:
English
Physical Description:
v, 108 leaves : ; 28 cm.

Subjects

Subjects / Keywords:
College students ( jstor )
English as a second language ( jstor )
Readability ( jstor )
Reading comprehension ( jstor )
Sentence structure ( jstor )
Sentences ( jstor )
Students ( jstor )
Syntactics ( jstor )
Syntax ( jstor )
Verbs ( jstor )
Curriculum and Instruction thesis Ph. D
Dissertations, Academic -- Curriculum and Instruction -- UF
English language -- Study and teaching -- Spanish students ( lcsh )
English language -- Syntax ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis--University of Florida.
Bibliography:
Bibliography: leaves 101-107.
Additional Physical Form:
Also available on World Wide Web
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Eileen Kay Blau.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
023404604 ( AlephBibNum )
06952866 ( OCLC )
AAL4664 ( NOTIS )

Downloads

This item has the following downloads:


Full Text
















THE EFFECT OF SINTAX ON READABILITY FOR SPAIISH-SPEAKING
ADULT STUDENTS OF ENGLISH AS A SECOND LANGUAGE








BY



EILEEN KAY BL64U


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



UNIVERSITY OF FLORIDA


19380










ACKNrO'.*JLE DGrMlE rrS


My sincere appreciation is expressed to Dr. Ruthellen Crews

for her guidance, support, encouragement, and cooperation throughout

the course of this dissertation. I wish also to express my gratitude

to the other members of my committee: Dr. James Algina for his

extreme helpfulness during all stages of this investigation, Dr.

Jayne C. Harder, especially for her help in the development of the

instrument used in the study and her editorial suggestions, and

Dr. Clemens L. Hallman for his impetus in the realization of the

Bilingual Education Program.

I wish to extend my gratitude to the administration and staff

of the English Department at the Mayaguez Campus of the University of

Puerto Rico who so generously cooperated in the collection of the data

for this study. I would also like to thank Dr. Patricia Byrd of the

English Language Institute at the University of Florida and Dr. Ellen

West of the Alachua County Adult Education Program for their coopera-

tion in the pilot studies.

I thank those friends who were so generous with their hospitality

and encouragement during the necessary visits to Gainesville this past

year. And I thank my husband, Robert W. Smith, for his constant sup-

port, encourageni nt and tolerance.

Finally I would like to express my gratefulness for the financial

support of the Department of Health, Education and Wclfare which made

my doctoral studies possible with a Title VII Fellowship.






TABLE OF CONTENTS
PAGE

ACKNOWL E DGMENTS ii

ABSTRACT iv

CHAPTER I INlTRODUCTIOI 1
Statement of Problem 1
Delimitations and Limitations 1
Justification 2
Assumptions 7
Definition of Terms 7
Null Hypotheses 9
Procedures 10
Organization of the Research Report 12

CHAPTER II REVIEW OF RELATED LITERATURE 13
Syntactic Development and Reading Ability 13
Syntax and the Human Mind 17
Conflicting Notions of Syntactic Complexity 23
Syntactic Complexity and Readability 34
Summary 40
Notes 42

CHAPTER III METHODOLOGY 43
Puerto Rico Study 43
Pilot Studies 50
Subjective Judgments of Relative Difficulty 52

CHAPTER IV PRESENTATION AND ANALYSIS OF THE DATA 54
Puerto Rico Study 54
Pilot Studies 57
Subjective Judgments of Relative Difficulty 61
Discussion 65

CHAPTER V SUMMARY AND CONCLUSIONS 72
Overview 72
Conclusions and Recommendations 76

APPENDIX A INSTRUMENT 80

APPENDIX B COVARIATE AND SDCT SCORES 97

REFERENCES 101

BIOGRAPHICAL SKETCH 108











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

THE EFFECT OF SYNTAX ON READABILITY FOR SPANISH-SPEAKING
ADULT STUDENTS OF ENGLISH AS A SECOND LANGUAGE

By

Eileen Kay Blau

August, 1980

Chairperson: Dr. Puthellen Crews
Major Department: Curriculum and Instruction


The purpose of this study was to determine the effect of manip-

ulating the degree of sentence combining on reading comprehension for

mature students of English as a second language (ESL). In this respect

the concept of the effect of syntax on readability differed from the

usual sentence length criterion.

A series of reading passages was developed in three versions.

Version 1 consisted of short, primarily simple sentences; version 2 con-

tained complex sentences with as many clues to underlying relationships

as possible left intact, and version 3 also contained complex sentences

but without those surface clues to underlying relationships. Vocabulary

and content were held constant across the three versions.

After several pilot studies, the Syntactically Different Compre-

hension Test (SDCT) was administered to 85 randomly selected under-

graduates enrolled in English 001 at the Mayaquez Campus of the University

of Puerto Rico. They were randomly assigned to one of three groups.








Each group read one of the versions of the SDCT and responded to the

same multiple choice comprehension questions.

The data were analyzed by analysis of covariance using the

English as a Second Language Achievement Test (ESLAT) as a covariate.

Although the differences in comprehension were not significant,

version 2 yielded higher scores than version 1 or version 3 in the

Puerto Rico study as well as in the pilot studies.

A secondary study was carried out to obtain the subjective

judgments of ESL students, ESL teachers, and pre-service teachers with

regard to the relative difficulty of the three syntactic versions.

Students tended to consider version 1 the most difficult while ESL

teachers and pre-service teachers tended to judge version 1 as the

easiest for students.

It was concluded that lower readability level material, as

measured by common readability formulas, does not facilitate the

comprehension of mature Spanish-speaking ESL students, Although none

of the three versions was significantly better than the others, all

results of this study suggest that version 2 may in fact be more

readable than the others. The short, primarily simple sentences

characteristic of low readability level material may actually be an

obstacle to comprehension. Consequently, such material is not

recommended for this population.

It was recommended that the study be replicated with native

speakers of languages other than Spanish as well as with younger ESL

students. A series of finer-grained studies was recommended. The

need to investigate factors other than syntax that might affect read-

ability was recognized.











CHAPTER I


INTRODUCTION

Because educational achievement, to a great degree, is dependent

on reading, there is great concern among educators about the measure-

rent of readability of materials. There are several readability for-

mulas that can be used to find the appropriate grade level of written

material. These formulas consider two characteristics of printed

material: word difficulty and sentence length. The effect of syntax

on readability is thus treated superficially and inadequately.


Statement of the Problem

The problem under investigation in this study was to determine

whether readability would be affected by manipulating the degree and

type of sentence combining or chunkingg" in reading material prepared

for mature students of English as a second language (ESL) who are

literate in Spanish, their native language, and who are at an inter-

rmediate level in their study of English. The study departed from the

usual sentence length criterion of the most commonly used readability

formulas in its consideration of the effect of syntax on readability.


Delinmiiations and Limitations

This study was confined to 85 undergraduate students enrolled

in English 001 at the Mayaquez Campus of the University of Puerto Rico.

All were native speakers of Spanish. They were at an intermediate

1









level in their study of English having scored between 400 and 599 on

the English as a Second Language Achievement Test (ESLAT), a grammar

and reading comprehension test administered by the College Board of

Puerto Rico to high school seniors.

In this study an effort was made to isolate only some of the

syntactic factors that might influence readability. The purpose of

the study was not to negate the effects of vocabulary and serrmantics

on readability nor to deny the possibility of an interaction between

syntax and semantics, or any other factors that might influence

readability.

The study relied on a researcher-developed instrument con-

sisting of 13 short passages, each followed by at least one multiple

choice comprehension question. The questions were limited to literal

.comprehension, which was considered a necessary prerequisite to other

types of comprehension.

The extent to which the findings of this study can be general-

ized is limited to a population similar to that from which the sample

was drawn.


Justification

This study attempted to fill two voids. The first is the lack

of attention paid to syntax in measures of readability. The second is

the paucity of research in readability that deals with the population

under study. Very little research has been reported on the effect of

syntax on readability for native speakers of English and virtually none

for the large population of students learning English as a second or








foreign language both in the United States and in the non-English-

speaking world.

Indeed, syntax does not receive much attention in the measure-

rent of readability beyond the crude sentence length criterion. Yet

one might expect the grammatical organization of sentences to play an

important part in determining how readable a passage is.

A reader may have all the necessary cognitive concepts
for interpreting the separate words in sentences and
paragraphs, but if sentence patterns are foreign to him
. or restrictive clauses unclear, meaning will not
be apparent. He must be able to understand the underlying
meaning of "deep structure" of what Chomsky . calls
the "surface" of print: how words are placed in varying
ways to produce meaning through their order in sentences,
or sy c.ti,. (Ransom, 1978, p. 282)

The reader must be able to "unpack" the surface structure (Larkin,

1977).

Surely a great deal of meaning is revealed by the arrangement

of words alone. A popular example of what has been termed "gram-

matical meaning" is taken from Alice in Wonderland.

'Twas brillig and the slithy toves
Did gyre and gimble in the wabe;
All mimsy were the borogroves,
And the more raths outgrabe.
(Dodgson, 1960, p. 85)

Moreover, with regard to reading, the phrase is the critical linguistic

unit worked on by mature readers. Gibson and Levin (1975) report that

people who read word by word are helped if words are grouped into

phrases (pp. 832-834). They also note that eye-voice spans end at

phrase boundaries to an extent well beyond the chance level.

Despite its potential importance, though, syntax does not

receive the systematic attention it deserves in the development and









selection of reading material either for native speakers (Blossom,

1969; Kachuck, 1975) or for ESL students (Gaies, 1979; Zukowsky/Faust,

1978). This is due in part to current methods of determining read-

ability. The most commonly used readability formulas consider only

two variables: words and sentence length. Although some readability

formulas do look at syntax more carefully than those that merely con-

sider sentence length (Botel, Dawkins, and Granowsky, 1973; Klare,

1974-75), these are more difficult to apply and are therefore not

commonly used. One must remember, too, that a readability formula is

merely intended to be an index of difficulty; it does not indicate how

to write readably. Yet such a formula determines the mold into which

most publishers force their writers. The doze procedure is yet

another device that is useful in determining whether students can

handle reading material (Anderson, 1971; Haskell, 1978; Owens, 1971).

It also is merely an indication of whether or not existing material

is too difficult. It does not in any way prescribe how to write

readably.

Despite the inadequacies of readability formulas in general,

there is, nevertheless, a tendency to apply the same criteria to

reading material for ESL students as those used in assessing materials

for native English-speaking students. Thonis (1970, p. 172) sug-

gests material with lower readability ratings for such students. This

is a quantitative rather than a qualitative suggestion, based on the

assumption that readability formulas are in fact accurate indices

applicable across populations.








Even Zukowsky/Faust (1978), whose concern is specifically

with the ESL/bilingual education situation, is bound to readability

formulas, albeit in a somewhat broader sense. Aware of the "hy-

bridized set of criteria" (p. 2) used in the ESL/bilingual education

setting (native speaker criteria plus an intuitive sense on the part

of more experienced teachers of what will be difficult for one's

students), she judges readability perception by teachers in terms of

agreement with traditional readability formulas. She does realize,

though, that an ESL readability formula would require a different basic

word list and a more precise means of assessing sentence structure.

In view of the lack of an ESL readability formula, are

specially prepared ESL readers more successful in meeting the needs

of limited English speaking (LES) and foreign students? ESL writers

may be doing an excellent job, but if they do, it is by chance--

without any research-based foundation. The progression from simple

to complex syntax in ESL readers is neither universal nor systematic

(Gaies, 1979).

What is syntactically simple or syntactically complex anyway?

Those who adhere to the sentence length criterion of readability

formulas believe that short, simple sentences are simplest, in other

words, easiest to comprehend. But this may actually l)e misleading.

There are others who believe that nrammatical complexity may in fact

be an aid to comprehension rather than a hindrance (Betts, 1977;

Dawkins, 1975; Pearson, 1974-75). This may be particularly true for cog-

nitively mature ESL students who are literate in their own lanruane.

They may be "unsophisticated in English but not l inguistically









unsophisticated" (Shook, 1977, p. 2). Shook also believes that our

capacity to deal with unintegrated information, characteristic of low

readability level material, is extremely limited. Our tendency to

integrate it is supported by the research of Bransford and Franks

(1971). Shook continues by saying that "simple beads-on-a-string

sentences . are like nothing the student will ever encounter in

the real world" (p. 6). Such "series of isolated propositions [are]

stacked together like bricks without mortar" (p. 11). The devices

used to join simple propositions into more complex sentences, i.e..

the devices of chunking, may actually clarify relationships and meaning

even though they create longer sentences and thus higher readability

levels. Again, there is a lack of research to support this notion

about sentence complexity.

Schlesinger (1968), in his monograph on sentence structure

and the reading process, advises that future research concentrate on

systematic exploration of variables by e>perimental manipulation

rather than dwell on analysis of te>ts in the effort to devise bigger

and better formulas (p. 149). Within the limitations described above,

this researcher tried to follow Schlesinger's advice and Shook's

hypothesis in an effort to begin filling the two voids mentioned above.

It was hoped that the results of this study would ultimately help bi-

lingual education and ESL programs better meet the special needs of

their students without depending so heavily on the intuition of

educators.








Assumptions

The assumptions basic to this study were as follows:

1. The subjects would be able to read the passages.

2. The passages were appropriate for purposes of measuring

reading comprehension.

3. Comprehension could be adequately measured for purposes

of this study.

4. Any differences in scores among the three groups would

result only from the differences in sentence structure among the

three versions of reading passages.


Definition of Terms

The following terms are defined in relation to their' use in

this study.

Chunking. Chunking refers to the combining of single under-

lying sentences to form grammatically complex sentences. Examples

are as follows:

(a) Birds migrate. They arrive in the North in early
spring.
(b) Birds that migrate arrive in the North in early spring.

(a) Some animals have a "weather sense." The weather
governs their movements and their very existence.
Man has never been able to equal this.
(b) Animals whose movements and very existence are governed
by the weather have a "weather sense" never equalled
by man.

The (b) sentences are considered more compressed or chunked than the

(a) sentences in which each underlying sentence is expressed in a

single simple sentence on the surface.










Complex sentence. A complex sentence is a sentence with

more than one underlying subject and one underlying verb.

Example: Voltaire, who was one of the greatest
writers of the eighteenth century, said
that the two essentials of style are
precision and color.

Intermediate level. ESL students at the intermediate level

scored between 400 and 599 on the ESLAT, a grammar and reading compre-

hension test of English as a second language that is administered by

the College Board of Puerto Rico.

Native language. Native language refers to the language

spoken in the students' childhood homes, i.e., the language their

parents used and thus the first language the students acquired. It

would be the language with which they feel most comfortable. A

synonym of this term is mother tongue. People whose native language

is Spanish, for example, are referred to as native speakers of Spanish.

For purposes of this study, subjects from homes in which two

or more languages were acquired simultaneously during childhood were

eliminated.

Readability. Readability is a function of a combination of

attributes of written language that determine its comprehensibility.

Reading comprehension. In the context of this study reading

comprehension refers to the meaning obtained from a written text as

measured by performance on multiple choice questions based on the

literal meaning of the passages.

Semantics. In the context of this study semantics refers to


meaning.








Simple sentence. A simple sentence is generally a sentence

with only one underlying subject and only one underlying verb.

Several exceptions are made for the purposes of this study. Pre-

nominal adjectives are not considered separate underlying sentences.

Three types of complex sentences occasionally appear in version 1

passages.

1. a. The puppets seem to become real people.

b. The plant is equipped to retain moisture.

2. Host valleys are wider than the streams.

3. Suppose the manufacturer and the market are a long
distance apart.

Syntax. Syntax refers to the grammatical organization of

sentences.


Null Hypotheses

The following null hypotheses were investigated in the

principal study:

1. There is no significant difference between comprehension
of version 1 and comprehension of version 2 passages.

2. There is no significant difference between comprehension
of version 3 and comprehension of version 2 passages.

In addition to the principal study, a subjective judgment

exercise was conducted to test the following null hypotheses:

3. Judgments by ESL students of relative difficulty of
passages is independent of syntactic version.

'1. Judgments by ESL teachers of relative difficulty of
passages is independent of syntactic version.

5. Judgments by pre-service teachers of relative difficulty
of passages is independent of syntactic version.









Procedures

Instrumentation

Development of Syntactically Different Comprehension Test (SDCT).

The 18 passages in the SDCT were adapted from passages from the

Science Research Associates' Reading for Understanding General

reading kit (Thurstone, 1969).

While holding content and vocabulary constant, three versions

of each passage were constructed. The versions differ from one another

onlyin syntax, or more precisely, in degree of chunking. Version 1

passages are unchunked and contain only simple sentences except for

those noted above under the definition of simple sentence.

Version 2 passages contain complex sentences and thus have

fewer and longer sentences than their version 1 counterparts.

Version 3 passages also contain complex sentences, but they

reflect a higher degree of chunking than their version 2 counterparts.

At least one multiple choice question follows each short

passage. There are a total of 24 literal comprehension questions

which are identical for the three versions of the test.

The SDCT was administered to 12 randomly selected English

Language Institute (ELI) students at the University of Florida. All

were native speakers of Spanish. All had scored between 35 and 85 on

the Institute's Reading and Vocabulary Test. The pilot study admin-

C(istration of the SDCT yielded a Kuder-Richardson 20 coefficient of .82.

Another randomly selected group of 15 ELI students took the SDCT with

the comprehension questions in Spanish. That administration yielded

a Kuder-Richardson 20 coefficient of .75. Finally, a 'group of 17 adult








education students who were native speakers of English took the SDCT.

That administration yielded a Kuder-Richardson coefficient of .81.

The purpose of the pilot studies was to determine if this study was

feasible and if the instrument was useful. Six of the 24 items and

two of the passages were slightly altered before the administration

of the instrument in Puerto Rico (see the end of Appendix A).

ESLAT. The ESLAT, which was used as a covariate to measure

proficiency in English, is a grammar and reading comprehension test

of English as a second language. It is administered by the College

Board to high school seniors in Puerto Rico.


Subjects

A sample of 85 students was randomly selected from a population

of undergraduate (mostly freshmen) ESL students enrolled in the

University of Puerto Rico's English 001 course. They were native

speakers of Spanish, literate in their brother tongue, and raised in

monolingual homes. One third were randomly assigned to each of three

groups. One group read the passages in version 1; one group read

version 2 passages, and the third group read version 3 passages.

All were given the same comprehension questions.


Design and Data Analysis

The design of the principal study included three treatment

groups. The ESLAT test was used as a covariate, and analysis of co-

variance was used to analyze the data. The model was

Y = b + bC + b2V1 + b3V2 + h4V3










where Y was the SDCT score, C the score on the covariate, and V

the version of the reading passages.


Secondary Study--Subjective Judgments

A secondary study was carried out which sought to reveal the

nature of subjective judgments of the relative difficulty of the

three syntactic versions. Seventy-nine students from the same popu-

lation as the principal study were asked to choose which of the three

versions they thought was the easiest and which they thought was the

most difficult. Three sets of passages were chosen for this exercise.

Seven of their teachers made the same judgment from the point of view

of what they thought would be easiest and most difficult for their

students. And 14 pre-service teachers at the University of Florida

made the sare judgment from the point of view of relative difficulty

for students of limited English proficiency.


Organization of the Research Report

Chapter II is a Review of Related Literature. Chapter III

describes the methodology used in the study while the data are pre-

sented, analyzed, and discussed in Chapter IV. Chapter V includes a

summary, conclusions, and recommendations for further research.









CHAPTER II


REVIEW OF RELATED LITERATURE

Chapter II reviews literature and research in four areas

that are related to the topic under investigation: (a) the relation

of syntactic development to reading ability, (b) syntax and the human

mind, (c) conflicting notions of syntactic complexity, and (d) syn-

tactic complexity and readability. Though most of this research has

been done with native speakers of English as subjects, it can, never-

theless, shed light on the area of reading for students learning

English as a second language (ESL) and thus serve as a basis for

further research focusing specifically on ESL students.


Syntactic Development and Reading Ability

A number of studies have been done that positively relate

one's grasp of syntax, both productive and receptive. to reading

ability, hence demonstrating the relationship between this component

of language and reading.

Using fourth grade students, Ruddell (1965a, 1965b) found

support for the hypothesis that reading comprehension is a function

of the similarity of the language structure of the written passages

and the language structure used orally hy children (1965a1. p. 40-1).

One version of each of three passanes was written with patterns of

high frequency in the oral language of children, and one version was

written with low frequency patterns. Comprehension, measured by the

cloze technique in which every fifth word was deleted, was superior

13









for passages containing structures frequently used in the oral

language of fourth graders.

Tatham (1969-70) measured the comprehension of second and fourth

graders by having them choose the one of three similar pictures that

best represented the sentence content. Results indicated better

comprehension of sentences that contained oral language patterns fre-

quently used by the students. Tatham concluded that "control over

vocabulary is not the only logical and desirable control when compre-

hension of language structures is essential" (p. 424). Her findings

supported the fact that ability to decode and knowledge of vocabulary

alone are not sufficient for successful reading; familiarity with and

control of the language patterns are also important.

In a study with over 240 fourth graders in the semi-rural

upper midwest, Bormuth, Carr, Manning, and Pearson (1970) used four

question types to test comprehension of various sentence structures,

anaphoric devices, and intersentential relationships. They discovered

that large proportions of their subjects were unable to demonstrate

comprehension of many of the basic syntactic structures by which

information is signaled.

Control of syntactic structures is a developmental matter.

Carol Chomsky (1969, 1972) demonstrated that children between the

ages of five and ten were at various states of development and did

not yet have complete mastery of their language. Mavrogenes (1977)

replicated Chomsky's stud), with working class 14- and 15-year-olds

and found that 17 our of 20 of her subjects had not mastered all five

stages of development with which Chomsky dealt.








Chomsky (1972) found a positive relationship between

children's stages of linguistic development as determined by her

testing of their knowledge of five different structures and the

following reading measures: scores on the Huck Inventory of

Children's Literary Background, the number of books at the top

level of syntactic complexity on the Master Book List with which

a child is familiar, and scores based on interviews with both

parents and children regarding such matters as time spent reading

and being read to and frequency of trips to the library.

Van Metre (1972) replicated Chomsky's (1969) study with a

different population. She used two groups of bilingual third graders

in the Tucson, Arizona, schools. One group included students who ob-

tained high scores on a state mandated reading test; the other group

obtained low scores on the same standardized test. Van Metre compared

the two groups on mastery of the structures tested by Chomsky. She

also used two monolingual comparison groups, one group each of high

and low scorers on the same reading test. Major differences in

mastery of the syntactic structures were between high and low groups

rather than between bilingual and monolingual groups. Van Metre also

found that the order of acquisition of the structures was less con-

sistent among those scoring low on the reading test. Her findings not

only lent support to the notion that there are indeed close ties be-

tween syntactic development and reading ability; they extended this

relationship to a bilingual population.

A study by Ribovich (1976) was designed in part to determine

the nature of the relationship between comprehension of syntactic










structures in oral language, as measured by a researcher-developed

instrument, and reading comprehension of first grade native English-

speaking children of low, mid and high socioeconomic status. Reading

comprehension was measured by the reading.su.test of the Metropolitan

Achievement Test. Ribovich looked at this relationship both with

and without controlling for intelligence. She found that compre-

hension of selected oral syntactic structures did relate significantly

to reading comprehension. Twenty percent of the variance in reading

comprehension was explained by the variance in comprehension of

selected oral syntactic structures. She also found, however, that

the relationship between intelligence, as measured by the California

Short-Form Test of Mental Maturity, and reading comprehension was

even stronger.

Simons' (1970) research with 87 fifth graders of above-average

intelligence aimed to relate skil.lat recovering deep structure _to

readina_comprehension. The first variable was measured by the

researcher-developed Deep Structure. Recovery...Test (DSRT), a multiple

choice instrument in which subjects chose the sentence with the same

meaning as the given sentence. Reading comprehension was measured

both by a cloze test and by the reading subtest of the Metropolitan

Achievement Test (MAT). The correlation between the DSRT scores and

cloze test scores was .73? while the correlation between the DSRT and

MAT scores was .476. Simons concluded that the MAT measured many

other things in addition to reading comprehension. He therefore con-

sidered the correlation between the DSRT and the cloze scores more
g............................ -:.. ................
mean ingful.








Nadler (1968) conducted a study with university foreign

students beginning an intensive English course. Reading material

for the experimental group included only previously studied syntactic

structures while reading material for the control group was not syn-

tactically controlled. Overall, the experimental group made greater

gains than the control group from vocabulary pre-tests to re-tests.

A similar study with reading comprehension instead of acquisition of

lexical items as dependent variable would also be of interest.

The above studies support the belief that knowledge of syntax

is positively related to reading ability, which suagests, in turn,

that the syntax of reading material might well affect thereadability

of that material for readers at different levels of syntactic develop-

ren t.


Syntax and the Human Mind

Some psycholinguistic research has been carried out in an

attempt to understand how the human mind processes different types

of sentences, which could indicate which syntactic characteristics

are harder to process. Unless otherwise stated, the subjects in all

of the studies reviewed in this section were college or university

students who were native speakers of English.

A number of studies (Bever, 1970; Epstein, 1961 1967; Fodor

and Bcvcr, 1965; Miller and Isard, 196-1) indicated that human per-

ception and learning of sentences is in terms of syntactic constit-

uents rather than individual words, attesting to the psychological

reality of the phrase structure of surface forms.









Other studies have attempted to relate human mental processes

to transformational-generative grammar and were based on the assumption

that comprehension of a sentence involves recovering its underlying

structure by means of a device that corresponds to a transformational-

generative grammar--despite the fact that transformational grammarians

make no claims that their models of language reflect human processing

devices.

Studies such as Miller (1962) and Miller and McKean (1964)

suggested that transformations such as negative, passive, interroga-

tive and combinations of these take time for human beings to perform.

Performance time was measured by matching a passive sentence, for

example, with its negative-passive counterpart.

A study by Mlehler (1963) suggested that kernel, or simple,

active, declarative sentences, were more easily recalled than trans-

formed sentences. Savin and Perchonock (1965) claimed that kernel

sentences took up less space in memory than transformed sentences.

Space in memory was measured by the number of words from a list

following the sentence that were recalled: fewer words remembered

from the list indicated more space taken up in memory by the sentence

itself. This group of studies suggested that people store a kernel

plus a separate notation as to which transformations go with it. They

implied that the greater the number of transformations the greater

the complexity.

Gough (1965) examined the effect of transformations on speed

of understanding by having subjects verify the truth or falsity of a

sentence based on a picture presented at the end of the sentence.








Actives were verified faster than passives which, in turn, were

verified faster than negative actives. The latter were verified

fas.ter_than negative passives. There was. however, an interaction

between truth value and the negative-affirmative dimension reflecting

the operation of a semantic factor. True sentences were verified

faster than false ones.

In a follow-up study. Gough (1966) attempted to isolate

verification time by allowing a three second delay between the end

of the sentence and the presentation of the picture. Mean verification

times were smaller with the delay, but not significantly smaller.

Differences among verification times for the different transforma-

tions paralleled Gough's earlier findings. Consequently. Gough

raised several questions. Perhaps these results cast doubt on the

hypothesis that a transformed sentence is reduced to underlying

structure at the moment it is read or heard. It may be de-transformed

when put into long term storage, but this may not occur at the time

of initial decoding. Without the opportunity to verify the sentence

right away, the subject may hold it in immediate memory postponing

the de-transformation process until presentation of the picture which

is the evidence on which to base the verification. Gouqh thus con-

cluded

. that a transformational description of sentence
complexity has not at all been discredited, for veri-
fication tire is a function of a number of trans-
fornations even after delay of evidence. But a
hypothesis of a transformational p.- :.-.-. in decoding
is less defensible. . (p. 191)








Sentence length did not seem to be a factor in processing

time according to Gough's findings. Active sentences, although

longer passivess and negative passives were shortened by deleting

agent phrases), were still verified faster.

These researchers assumed that human language processing

should parallel a particular linguistic model. But should it? There

are problems both with their oversimplified assumption, as Gough's

findings indicated, and with the nature of the experiments themselves.

Greene (1972) noted that a basic problem with experiments

of the type described above is a lack of realism or naturalness.

This problem is recognized by some of the experimenters themselves

(Miller and McKean, 1964). Even if the speed required for certain

transformations or de-transformations can be measured and contrasted.

the measurement in such research occurs under contrived conditions

and is no indication that such processes are what actually take place

.when people normally encode and decode sentences.

When subjects are performing the more natural 1anggaqge
function of extracting meaning from sentences, any exact
one-to-one correspondence between transformational
complexity and performance disappears. (Greene, 1972, p. 114)

Other factors undoubtedly operate at the same time. Perceptual

mechanisms and conceptual structure interact with language processing

(Bever. 1970). to say nothing of the interaction betv.een syntax and

senmntics. In fact this interaction has been a topic of considerable

linguistic research. Greene (1970) found that subjects handled a

syntactic form more easily when it fulfilled its natural semantic

function. Slobin (1966) demonstrated the role of reversibility in








processing passive sentences. Franks and Bransford (1974) have

demonstrated the role of extra-sentential context in dealing with

passive sentences. Glucksberg, Trabasso, and Wald (1973) mentioned

selectional restrictions on verbs and whether or not both of two

nouns appearing in a sentence could act as agents in processing

active and passive sentences.

In an attempt to put syntax in proper perspective, Greene

(1972, p. 191) stated

. the only purpose of syntactic rules is to
express semantic relationships, the meaningof sen-
tences being determined by the syntactic relations
holding between individual words.

Such a statement attests to the importance of syntax at the

same time that it recognizes its inseparability from semantics or

content. Syntax provides the structural information necessary for

arriving at a semantic reading of a sentence (p. 139), but it is only

one tool used by the hfiran mind in processing language, a fact that

was overlooked in the earlier research reviewed above.

In addition to neglected pragmatic considerations, a limitation

of the earlier experiments, aside from their unnaturalness, was that

they dealt only with a limited class of grammatical operations in the

context of isolated sentences. Among the limited sentence types there

are enormous discrepancies in frequency of use which may distort the

findings of these studies.

Goldman-Eisler and Cohen (1970) analyzed speech samples of

approximately equal length (about 100 clauses each) by speakers of

various levels of intelligence and mental health. They found that










simple active affirmative declarative, or kernel, sentences comprised

80-90% of the speech_samples. Negatives were.a distant second at

4-10%; passives accounted for .7-10% while passive neg .t..ives.....e.re

practically non-occurring. The use of different sentence types not

based on the same universe of behavior may account for differences

observed in experimental settings.

Another problem is that the results of the experiments could

be affected by the task employed to measure comprehension or relative

-ease or-difficulty in linguistic processing. A variety of tasks has

been used as dependent variables: recall, recognition, transformation/

de-transformation, verification. Furthermore, the tasks typical of

psycholinguistic experiments may not be so appropriately related to

the reading process as one might expect.

Shore (1976) found no significant differences between the

ability of good and poor readers in performance on the Linguistic

Processing Test (LPT). Her sample consisted of 40 Caucasian adult

education students who were reading at least on the seventh grade

level. The LPT is a recall test in which seven sentence types are

followed by strings of digits. Although there were significant dif-

ferences among sentence types for this sample, performance on the

LPT did not relate to subjects' reading ability.

Even recognizing the problems involved in analyzing syntactic

processing--simultaneous operations, dependence on type of task,

appropriateness of task--the basic premise that additional trans-

formaton..s. create additional comple.jity is still questionable. The
............-. .............. .-- ............. -.... ...








research reviewed in this section that does support that premise

reveals an overemphasis on singulary transformations.


Conflicting Notions of Syntactic Complexity

LThis section examines several notions of what constitutes

syntactic complexity- Many of the studies reviewed in the previous

section were inspired by what is sometimes called the Derivational

Theory of Complexity (DTC), which implies that the greater the

number of transformational rules employed in the derivation of a

sentence, the greater the complexity of that sentence. Thus kernel

sentences are the simplest. The language user, according to the

theory, follows a path that corresponds to the derivation of a

sentence in the manner of transformational-generative grammar. But,

as was noted in the previous section, it is not at all certain that

this is what occurs. Fodor and Garrett(1967) note

S. it is far from obvious that one ouqht to expect
a correspondence between the complexity of the ;7..:.: 'ea
operation of converting a base structure into a surface Lsructure
and the complexity of the pe,r'p:. ,p operation of con-
verting a surface structure into a base structure. (p. 289)

They go on to suggest

. that the complexity of a sentence is a function
not (or not only) of the transformational distance from
its base structure to its surface structure but also of
the degree to which the arrangement of elenients in the
surface structure provides clues to the relations of
elements in the deep structure. (p. 290)

Fodor and Garrett do agree that transformations often mask underlying

structure by eliminating surface clues. But at the same time they pro-

vide examples of counterintuitive predictions arising from the DTC.









According to the DTC the (a) sentences below should be more complex

than the (b) sentences.

(a) The small cat is on the grass mat.
(b) The cat that is small is on the mat that is made of grass.

(a) It amazed Bill that John left the party angrily.
(b) That John left the party angrily amazed Bill.

(a) The operator looked the address up.
(b) The operator looked up the address.

(a) Fred runs faster than the girl.
(b) Fred runs faster than the girl runs.

In lire with the hypothesis that the presence or absence of

surface structure clues to deep structure is the key to simplicity or

complexity with regard to comprehension, Fodor and Garrett chose

relative _pcooouns as an example of a surface structure clue to deep

structure relationships that lent itself to examination. Two groups

of fIT students listened to a set of nine sentences with two embeddings

apiece. For group one the relative pronouns were deleted; for group

two they were retained.

Example: The oen (which) the author (whom) the
editor liked used was new.

Subjects were asked to restate each sentence in their own words.

Scoring was based on the figure obtained when the number of correct

subject-object relations reported was divided by response delay.

Group two did significantly better than group one. Four variations

of the experiment were carried out in order to consider alternative

explanations, but results were always the same. In one of these

experiments adjectives were added before each of the first two noun

phrases of sentences in which relative pronouns were retained in order








to increase the derivational history. Despite the longer derivational

history, group two still outperformed group one. Surface clues.,_such

as retained relative pronouns, apparently outweighed the DTC.

A study done by Charrow and Charrow (1979) and reported in

the Linquistic Reporter ("You must follow the law as I state it,"

1978) revealed that one of the characteristics rendering instructions

to juries difficult to comprehend was deletion of relative pronouns

followed by a form of the verb be. This supported the findings of

Fodor and Garrett.

Hakes (1972) found supporting evidence for the negative

effect on sentence comprehension of reducing t'at-clause complement

constructions by deleting the optional t+,at, a clue to underlying rela-

tions. He believed there would be more local ambiguity in sentences

with deleted complementizers. Hakes used 40 freshmen native speakers

of English as subjects and sentences such as, The Lonld-rfa'ous

phylsicist f.'orot (thaLi) hi"'. old pr'o.fes.o)' ha.d be,'i tlihe frnt to sup:cst

the crucizl sxperienent to him. Two tasks were used to assess compre-

hension of sentences with and without the optional complementizer.

The first was a phoneme monitoring task in which subjects were

instructed to listen for a target phoneme and press a button upon

hearing it. The second measure of comprehension was a paraphrase task.

Phoneme rnnitorinq was indeed faster when the complermentizer was pres-

ent. The difference between the presence or absence of the comple-

mentizer was not significant for tie paraphrase task. In a second

experiment, however, in which supposedly more difficult sentences were

used, subjects performed significantly better on both tasks for









sentences in which the complementizer was present. Hakes concluded

that the paraphrasing task may be sensitive only when the overall

level of difficulty of the sentences ishigh. He considered his

results evidence that presence of the complementizer does facilitate

comprehension.

Fodor, Garrett, and Bever (1968) examined the effect of verb

type as another clue employed by language users to determine under-

lying structure. If the main verb of a sentence could appear in a

variety of deep structure configurations, the sentence was more

complex, harder to process, than one in which the main verb was less

ambiguous. Pairs of sentences were constructed that differed only

in verb type. One sentence had a verb that was transitive only:

Example: The letter the secretary the manager
employed -,i7e.d was late.

The other sentence in the pair had a verb that could take a variety

of complement types:

Example: The letter the secretary the manager
employed axrected was late.

The verb expect could be followed by a ti-at-clause or an infinitive,

i.e.,a sentential complement, as well as a simple direct object. The

verb :i.l, on the other hand, could be followed only by a simple direct

object. Subjects (MIT students) heard and were asked to paraphrase

bo th_type.s_of sentences. Exact words were not permitted in order to

avoid rote repetition. Scores were based on the number of correct

subject-verb-object triples reported in paraphrases. Subjects did

significantly better on sentences with simple transitive verbs whether
.. .. ....................... .... ...................... ........................ .. ............................








the sentences were read or heard. Similar results were reported for

versions in which subjects had to unscramble the sentences.

Holmes and Forster (1972) conducted a study with 40 Australian

university students who had been speaking English for at least ten

years. Written recall of four different types of sentences was

examined. There was no significant difference between recaLl _ofone-

clause and two-clause sentences, contradicting the notion that complex

sentences are more difficult than simple sentences. Holmes and

Forster also found that verb type in one-clause sentences affected

recall. Those that could take a variety of complement types were

more difficult to recall than simple transitive verbs confirming the

findings of Fodor, Garrett and Bever (1968).

Rohrman (1970) observed a significant difference in recall

between intransitive forms such as c.. i.on3 (lions growl) and

transitive forms such as 'i pig:-, hol Ze (PRO [dig holes]). The

intransitive forms were easier to recall. Rohrman hypothesized that

they constituted smaller chunks which were thus easier to store

despite the fact that the surface forms of both types appear the same.

Verb type itself may indeed operate as some sort of lexical clue to

underlying structure.

Carol Chomsky's (1969) concept of complexity is that the

exceptional structures are more difficult. This, too, is usually a

question of arriving at the appropriate deep structure when surface

structures look alike. This is true in the case of t.I,' j;l ..

Y.Rs! to r. r where, although ,!ol1 is in subject position, someone

else sees it. In a structure such as ;;,:.* ': rn.- ', :.r, however,









the surface subject of the sentence is the eager one. Recovery of

the correct deep structure is again the issue in as;: as a question

versus a.--r as a request. As: as a question, as in J.;, ak .d

ZBill 1~:at to A-, violates the Principle of Minimal Distance (PtD):

Joll: is the subject of .d rather than the closest noun phrase. Aok'

as a request, on the other hand, does not violate the PMD and is

therefore easier. In the sentence *To.:' ,z: a.zke BL: to ;,el th

drimvle.:!, Bill is the one who will do the shoveling. The verb

pc',rise- is difficult because its use also violates the PMD. In

Jc.;:, p 'Ci..ej BilZ to sov-l. thle Ji',e'' u, John will do the shoveling.

Because surface structure clues do not follow the usual pattern,

violations of the PMD and structures such as eaLz. t.o see are indeed

exceptional and therefore more difficult.

Another way of conceiving of sentence complexity that has

received some attention is based on Yngve (1960). According to his

model, sentence depth is the factor that creates difficulty. Depth

is defined as the amount of temporary memory needed to produce a given

sentence. Regressive structures--those that branch to the left, as in

the sentence, is .' ti:e tha t tc'S ti at tlat 1: ct thzt t7h doa

Lri-e.i ;:?.'P-d atL require more temporary storage space as they get

longer. Progressive structures, those that branch to the right, on

the other hand, do not require more than a minimum of temporary

storage space as they get longer as in the sentence, TFa:. i. tbz ,o,

t'i't t vo1iPd t4.1 cat t"i"t 17t i Z :if t]jt ate t1"2 at,.lt tih t au;

in tklc hom11'r t!at Jak b:iZt. Because of limitations on the immediate

memory of human beings, Yngve hypothesized that languages tend to limit







regressive constructions and include alternative structures of

lesser depth that maintain expressive power. Passivization can per-

form this function. Schlesinger (1968) gives the example of .Jo':,?

wlrom Jzoue v,,om Paul p'treers detested oL's ,' .'3a3 (p. 83) which can be

expressed as -:aky is loved bi J.ohni, -,ho is detested by. Jlc.Ju, .,'IL is

pre f;er-' by Paul (p. 85).

Wearing (1970) found that on a multiple choice recognition

test 120 undergraduate male subjects recognized low depth sentences

more easily than sentences of greater depth. Whether the sentences

were active or passive made no difference.

Martin and Roberts (1966) supported the depth hypothesis as

opposed to the DTC. They used sentence recall as the dependent vari-

able. They claimed, in addition, that the findings of Mehler (1963)

could be explained in terms of the depth hypothesis rather than the

DTC. They called into question Miller and McKean (1964) on the same

grounds.

Wright (1969) conducted two studies of the depth hypothesis

in which she attempted to compare the relative effects of depth and

transformational complexity as predicted by the DTC on the amount of

memory space taken up. The first study supported the depth hypothesis,

and the second lent somn support to the DTC.

Other findings have also indicated inconsistent predictive

success of the depth hypothesis. Prenominal qualifiers constituting

regressive structure should, according to the depth hypothesis, be

difficult. Matthews' (1968) subjects found them more of an obstacle









than transformational "complexity." Qualifiers were also a source

of difficulty for ESL students in sentence repetition tasks

(Stieglitz, 1973). Yet Schwartz, Sparkman, and Deese's (1970) sub-

jects, graduate students in psychology and native speakers of

English, judged sentences such as ':The esetric po ,ed te chormpi,

roc': t2;:'c.'Il ~ iaZ '": cw' 'e~' ru, t o.er it cord to be highly compre-

hensible regardless of length. Sever (1970, p. 339) also doubted

that this type of left-branching constitutes perceptual complexity.

He supposed that information provided early in a sentence, i.e.,

prenominal qualifiers, should make prediction of the latter part of

the sentence easier. Semantic aspects may also play a role in the

comprehensibility of these structures.

For the subjects in Schwartz, Sparkman and Deese's study, who

found left-branching easy to comprehend, the comprehensibility of

center-embedded and right-branching sentences declined rapidly as

clauses were added. Hamilton and Deese (1971), however, found that

for their 30 subjects the important issue was contiguity of subject

and verb rather than the number of clauses alone.

As noted in the previous section, experimental results are,

at least to some extent, dependent on the task required of the sub-

jects. In addition, there are numerous ways of viewing complexity

that yield findings which fail to provide clear-cut patterns. Wang

(1970) attempted to ascertain the role of syntactic complexity in

determining judgments of comprehensibility of a large number of sen-

tence types. Eignt potentially relevant measures of surface structure








complexity were computed for each sentence type. (This itself is

indicative of the variety of ways to "measure" complexity.)

1. ratio of nodes in a phrase marker to terminal nodes

2. total number of nodes (Trii)

3. total number of terminal nodes (TNITN) or sentence length

4. depth

5. mean depth2

6. embedding transformations (T-emb) or number of S nodes

7. conjoining transformations (T-conj)

8. self-embedding (S-emb) or recursiveness

The sentences were presented auditorily to undergraduate native

speakers of English. The two measures that revealed the strongest

negative correlation with comprehensibility ratings were mean depth

and self-embedding. When these were partial led out, the negative

correlation between sentence length and comprehensibility was reduced

implying that it is not sentence length per se that makes compre-

hension difficult.

Another factor that might affect ease of sentence comprehension

is the order of main and subordinate clauses. Bever (1970, p. 294)

referred to a study by Bever and Weksel that indicated stylistic

preference for sentences in which the subordinate clause (marked by

a conjunction) followed the main clause. Bever reported that Clark

and Clark observed that complex sentences with the subordinate clause

first were relatively hard to iiemorize. Holmes (1973) observed that

university students had significantly nore difficulty recalling









sentences with proposed adverbial clauses. He also found that, for

his sample, subject complements were more difficult than object

complements. Unexpectedly, center-embedded relative clauses were

easier than right-branching relative clauses, but this was with only

one relative clause. Th,:t-clauses as object complements were easier

than -i object complements, but tk..:t-clauses in subject position

were more difficult than -::- subject complements. Apparently dif-

ferent explanations are necessary for different types of complex

sentences.

The variety of approaches taken in the studies described above

indicate that narrow adherence to the DTC would be a mistake. There

are many ways to view the question of syntactic complexity/simplicity.

One outlook that has not received much attention is, in a way,

the opposite side of the coin from the DTC. Miller (1956) suggested

that the limitations on our capacity to process information are related

to what we can hold in short term memory (STM). STM capacity falls

in the range of seven plus or minus two items. In order to stay within

this range we tend to chunk material or compress it. Exactly what

constitutes a chunk in a given situation for a given person is unclear.

Miller gave the example of a telegraph operator packing an increasing

number of bits into each chunk he can hold in his mind as he gains

experience. By using sentence combining transformations rather than

expressing each proposition as a single simple sentence, language

users create convenient chunks that enhance their memory capacity.

Bransford and Franks (1971) found that their subjects, college

students, did in fact tend to integrate separate ideas and actually








believed that the complex sentences they thus created were those

that they had heard. Chunking, as Miller has suggested, may be a

powerful device for increasing memory capacity, and, as Bransford

and Franks have demonstrated, it may also be a natural one for humans

to employ.

Do people prefer to chunk or to string out each idea in its

own simple sentence? As far as the simplicity/complexity issue is

concerned, Betts (1977) states that short sentences may interfere with

rather than facilitate concept development. Dawkins (1975) notes

that short, choppy sentences are not necessarily easier to read.

Schlesinger (1963) cites Bar-Hillel's suggestion that some things

can not be expressed in sentences with a low degree of syntactic

complexity without a loss in other communicatively important respects.

Thus there is a price to be paid for simplification of sentence

structure. Schlesinger suggests that complex sentences render more

salient the organization of content. He also found that readers'

speed and comprehension were not inhibited by nested sentences. But

Miller and Isard (1964) did find that oral recall decreased as degree

of nesting increased. Fuggins (1977) sees syntax as a tool for ad-

justing complexity and compactness of a message in the process of

helping to transfer meaning from producer to receive. Hle agrees

with Fodor and Garrett that recoverability is the key. and surface

structure clues such as relative pronouns are indeed helpful. He

believes that subjects should prefer the long thin nr ssage only if

the more compact form strains their processing capacities. But long

messages may in turn tax memory capacity. Huggins therefore opts for








a middle road. Some balance that satisfies both processing and

memory capacities probably defines the optimum level of syntactic

complexity/simplicity for reading material.

This line of reasoning matches that of Shook (1977). He

sees the issue not so much as a matter of simplicity versus complexity,

but rather as the way in which a sentence is complex.

(a) Judy loved her mother. Her mother lived in New Jersey.
(b) Judy loved her mother, who lived in New Jersey.

In (b) the relative clause contributes to rather than hinders pro-

cessing. Syntax that is too simple_may inhibit and slow down the

reader-a-smuch as syntax that is too complex (Shook, 1977, p. 19).

ESL students therefore need to be taught to take advantage of

the help offered by sentence combining.

Pupils need guide words to help them follow the sequence
of ideas and to let them know when transitional thoughts
or summary statements are being offered. Words such as
7.-, t 0, .Y.- :7 I
?C2c2r.i2gq to th7ze conildti-ois, and countless other such
expressions which carry the author's reasoning can help
the reader see the direction of an argument or a view-
point. . Materials which follow an orderly progression,
signaled by words and phrases to guide the reader's
thinking, are more likely to be understood. (Thonis,
1970, p. 172)


Syntactic Complexity and Readability

The research reviewed in this section involves the manipu-

lation of the syntax of reading passages to see the effect this has

on some measure of reading comprehension.

The question how syntactic structure affects the ease
of reading is obviously of the greatest importance
to teachers, writers, editors, in short--to anyone
interested in written communication. (Schlesinger,
1968, p. 15)








Schmidt (1977) studied the effect of increased complexity

on reading comprehension using two versions of a biographical

selection about Helen Keller. The complex version was created with-

out altering the vocabulary, by merely proposing prepositional and

participial phrases, using gerunds instead of noun structures, and

using appositives and passives.

Example: She also wrote many books and articles
and included an autobiography of her
early years. (simple version)

Her writing, many books and articles,
included an autobiography of early years.
(complex version)

Cloze tests of one of the versions were given to 84 university students.

Those who had the less complex version did significantly better.

Coleman (1964) also used university students as subjects to

compare comprehension on what he considered simplified and unsimplified

versions of the same passages. Without altering content morphemes

and while holding sentence length nearly constant, Coleman simplified

by changing passives to actives, nominalizations to their active verb

derivatives (his explnation of theL de:i;gn i It,? cxplai,,d thl d. sin:),

and adjectivalizations to adjective or adverb forms (IeL has ~ social

po,' -' hi i.n ociaoli1 po!,rfJiul,). More multiple choice questions were

correctly answered on the simplified versions. In a similar experiment

by the same investigator, comprehension was measured by written recall.

The simplified versions yielded better recall. In a third experinwmnt,

20 sentences of equal length--ten witli nominalizations and ten with

active verb equivalents--were shown to subjects for four seconds each.

Recall by writing was significantly higher for the sentences with








active verbs. Although more multiple choice questions were cor-

rectly answered for the active verb versions, the difference did not

reach significance. Simplified sentences were also memorized faster.

In a finer-grained study Coleman (1965) aimed to discover

more precisely which sentence patterns within the broader categories

he had already looked at were more easily comprehended. In the first

of a series of experiments he looked at ten types_of nominalizations.

For six of the ten types the de-transformed active verb versions

provedeas.ier:to learn.

Example: His discussion of the reason for the
decision will be appreciated.
(nominalization)

If he discusses the reason for the
decision, it will be appreciated.
(active verb version)

Coleman concluded that two explicit clauses could be more easily

processed than one more highly chunked clause_(p. 334). In the

second experiment written recall was superior for active sentences as

opposed to passive sentences. In the third experiment Coleman found

no significant difference between adjectivalizations and de-trans-

formed adjective versions.

Example: The complexity of milk fat may be compre-
hended when the formula is known.
(adjectivalization)

We may comprehend how complex milk fat is
when we know the formula.
(de-transformed adjective version)

In the last experiment of the series Coleman compared cloze scores

on two versions of sentences--one he considered more embedded than

the other.








Example: The man who can sell it is Bill. (embedded)
Bill is the man who can sell it. (non-embedded)

Overall there was a significant difference in favor of the second

type of sentence. With regard to content words, there was no sig-

nificant difference, but with regard to function words, significantly

more were correctly supplied in the less embedded sentences. Coleman

concluded that, other things equal, writers should choose the more

easily comprehended versions of sentences.

Coleman (1971) reported, in addition, several correlational

findings that indicated an association between difficult-to-read

prose and such characteristics as a low proportion of verbs and

numerous adjectives and prepositions. The low proportion of verbs

could be a result of sentence nominalizations which he had found to

be more difficult than their active verb counterparts.

Using de-transformations to simplify passages (changing

passives to actives, nominalizations to noun + verb, expressing rela-

tive clauses in terms of their underlying propositions, and replacing

deleted items). Evans (1972) conducted a study with 24 high school

seniors reading on seventh, eighth and ninth grade level. Those who

read the simplified versions did significantly better on multiple

choice questions and on the last three of five cloze tests.

Fagan (1971) conducted a study with 4.10 fourth, fifth and sixth

graders. He identified 43 types of transformations found in randomly

selected passages from three 4th grade basal readers. He grouped these

into five categories: embedding, conjoining, deletion, simple (singu-

lary), and position shift. Based on different versions of adapted









passages he found that deletions and embeddings in particular

rendered information more difficult to process for the population

in question.

Pearson (1974-1975), in contrast to other studies, supported

the chunk model. Middle class suburban third and fourth graders

showed a preference for the more cohesive form of syntactic variants

of sentences with causal and adjectival relations.

Examples: Because John was lazy, he slept all day.
John was lazy. He slept all day.

The tall man thanked the young woman.
The man thanked the woman. He was tall.
She was young.

In addition, recall of causal relations seemed to be aided by cues

such as bcau. i.F and o. Results with regard to reading comprehension,

however, were of little value as all forms were correctly understood.

Perhaps semantic content was so simple that it masked possible dif-

ferences of form. Nevertheless, overall results supported the chunk

model.

Marshall and Glock (1978-1979) tested comprehension by both

free and probed recall of 16 versions of two 115 word texts. Each

version of a text maintained the same semantic network. Different

combinations of four different manipulations gave rise to the 16 ver-

sions of each of the passages. First, the dependency system was

either present or not. If present, '."-:':; relationships .wereex-

.plicitJy expressed. If not, one encountered a.sentence beginning

with s:<;:os n and. a separate sentence for the result clause. Second,

the relative system was either present or not. If present, superlative








forms of adjectives were used. If not, the common form of the ad-

jective was used thus providing less information. Third, the main

idea appeared either at the beginning of the text or at the end.

Finally, an evaluative statement about the topic was placed either

at the beginning or at the end of a sentence.

Significant differences were unexpectedly found between com-

munity college and Ivy League college students in the sample. For

community college students, manipulation of the relative system made

the greatest difference. The difference was also significant for the

Ivy League students. Manipulation of the dependency system also had

a strong effect for the community college students, but not for the

Ivy League students. The authors concluded that truly fluent readers

communicated with the author through discourse by responding to its

semantic structure. Less fluent readers, on the other hand, could

not infer existence of structures in text base unless they were

explicitly referenced in the surface structure. Less fluent readers

responded to surface structure, and their recall was negatively

affected by lack of explicitness.

Smith (1971) studied a broader spectrum of age groups than the

above studies and used a different approach to manipulating the text.

He used the output of a study by Hunt (1965) to construct four versions

of the same text. The fourth, eighth, twelfth grade and "skilled

adult" versions were each typical of the production of the respective

age groups. Sentence length was equalized over versions principally

by coordinating T-units (main clause plus any subordinate clause or










non-clausal structure attached to or embedded within it) in the

fourth grade version. Comprehension of 120 fourth through twelfth

graders was measured by means of cloze tests in which every fifth

word was deleted beginning with the second, third, fourth, fifth or

si

writing best while eleventh graders read the fourth grade version

with less facility than they read the other three versions. Highly

reedundan t -wr-Li ngseemed to_be helpful only to young ge read rs.

Moreover, redundancy was apparently not the onl. factor used in the

predictions called for by the cloze procedure. Smith concluded that

perhaps people read best what is syntactically closest to what they

produ,:e. provided vocabulary and content are notan obstacle. The

productive level may be the best determinant of the nost appropriate

receptive level which supports the findings of Ruddell (1965a, 1965b)

and Tatham (1970). It might be that these findings relate to

Shook's (1977) notion and that construction of the most appropriate

reading material for' ESL students who are linguistically mature in

a non-English language should not totally neglect the linguistic

maturity achieved in the native language.


Summa ry

Four areas of research and literature dealing with the rela-

tionship between syntax arid reading have been r-ev'iewed. The first

indicated a positive association between syntactic development and

reading ability. The second group of studies investigated how the

human mind handles different types of sentences. Results were







inconclusive and did not follow a single, undeviating line of rea-

soning. The third area attempted to get at an aspect of the

inconsistency noted by considering different conceptions of complexity.

These conflicts were not resolved by the experimental studies re-

viewed in the fourth section. Although some characteristics that

cause more difficulty than others have been isolated, research is

still spotty.

Perhaps one of the most glaring gaps in the research to date

is the paucity of studies along these lines that is done with ESL

students. It seems reasonable that this study should aim to extend

to ESL students, in a limited way, what has been done with native

speakers in the area of syntax and reading. Furthermore, research

of this kind which targets ESL students should underlie the pro-

duction of reading textbooks aimed at that population.







42


Notes


Singulary transformations: Hunt (1970) refers to trans-
formations that change single sentences to other single sentences
as singular transformations. These include question, imperative,
passive, and negative transformations. In contrast, sentence
combining transformations yield complex sentences.

2Mean depth (MD) is depth, as defined by Yngve (1960),
divided by the number of words in the sentence.











CHAPTER III


METHODOLOGY

The purpose of this study was to consider the effect of syntax

on readability and to determine what degree of sentence combining or

chunking is the most readable for mature students learning English as

a second language (ESL). The study departs from and questions the

usual sentence length criterion of syntactic complexity of the most

commonly used readability formulas in its consideration of the effect

of syntax on readability.

This chapter will provide information on the subjects, the

design, the materials and the procedures used in conducting the main

study and the pilot studies. In addition, subjective judgments of

relative difficulty of the three types of sentence structure char-

acteristic of each of the three treatments were obtained from students,

ESL teachers, and pre-service teachers taking a course in the teaching

of reading. A description of this secondary study is also included.


Puerto Rico Study

Subjects

The 85 subjects in this study were undergraduate students at

the Mayaguez Campus of the University of Puerto Rico. All were native

speakers of Spanish and literate in their, mot'ier tonaue. All were

enrolled in English 001 during the fall semester of 1979 and had









scored between 400 and 599 on the English as a Second Language

Achievement Test (ESLAT) taken during their senior year in high

school. English 001 is the most basic English course currently

offered at the university, but none of the students could be classi-

fled as beginners in their study of English. All had studied English

throughout their school years, whatever the quality of the instruc-

tion may have been. They have also been surrounded by Enqlish on an

everyday basis through commercial media. For example, labels and

instructions on the many products from the United States are usually

in English. Any students who had lived on the mainland of the United

States or who were from bilingual homes, however, were not included in

the study.


Design

Each subject was randomly assigned to one of the three

treatment groups in a one-way design with three levels. The three

levels are referred to as versions 1, 2, and 3. The reading passages

of each version are characterized by a different type of sentence

structure or syntax which will be described in detail in the following

section.


Ma trials

Syntactically Different Comprehension Test. The 13 passages

of the Syntactically Different Comprehension Test (SDCT) are based on

passages from the Science Research Associates' Reading for ULnder-

standing General reading kit (Thurstone, 1969).









While holding content and vocabulary constant, three ver-

sions of each of the 18 passages were constructed. The versions

differ from one another only in syntax, or, more precisely, in

degree of chunking. Sentence structure is the only characteristic

that is manipulated (see Appendix A).

The readability levels of the passages that are referred to

below are based on the Maginnis (1969) adaptation of the Fry read-

ability scale for use with passages shorter than those required when

using the Fry readability scale. An extended graph (Madsen and

Bowen, 1978) was used to determine readability levels in the upper

range.

Version 1 passages vary in length from 20 to 91 words. Where

it is obtainable, the readability level, as presented in Table 1,

varies from first to fourth grade. The passages are unchunked and

contain only simple sentences except those occasional deviations as

described under the definition of simple sentence (Chapter I, p. 9).

Conditional sentences are broken into two sentences. What is most

commonly expressed as an i.-clause is expressed in the form of a

sentence beginning with c-rrpon.?. The result clause constitutes a

separate sentence.

Version 2 passages vary in length from 19 to 84 words. The

passages contain complex sentences and thus have fewer and lonQer

sentences than their version 1 counterparts. Readability level, where

obtainable, varies from fifth to sixteenth grade level. In forming

complex sentences for version 2, relative pronouns are not deleted

even though they may be opLional. Subjects and finite verbs are









Table 1

Readability Levels of SDCT Passages* Based on Maginnis'
Adaptation of Fry and Extended Graph


Passaqe

1

2

3

4

5

6

7



9

10

11

12

13

14

15

16

17

18


Version 1

1st grade

high 2nd grade

3rd-4th grade

N.D.**

2nd grade

N.D.

3rd grade

IN.D.

1. D.

2nd grade

N.D.

low 2nd grade

4th grade

2nd-3rd grade

N.D.

2nd grade

3rd grade

N.D.


Version 2

7th grade

7th grade

7th grade

9th grade

8th grade

12th-13th grade

9th grade

12th grade

13th grade

8th grade

16th grade

low 7th grade

9th grade

5th grade

N.D.

6th grade

8th grade

15th grade


* See Appendix A for SDCT.

** N.D. = No Data
Fry makes no claims re material with these combinations of
sentences per 100 words and syllables per 100 words.


Version 3

7th grade

10th grade

7th grade

12th-13th grade

8th-9th grade

17+

7th-8th grade

llth-12th grade

10th grade

8th grade

16th grade

8th grade

9th grade

low 6th grade

N.D.

8th grade

3th grade

N.D.








retained in subordinate clauses wherever possible. Where a_fini.te

verb is not possible, infinitives are usually used rather than the

"nounier" (Ross, 1973) -.na forms or derived nouns.

Example: . it was hard .-'Or !:.r, to rsai~, i
that they had really escaped.

Conditional sentences are explicitly expressed with the subordinate

clause introduced by if. Subordinating conjunctions such as slce.,

bec:2se, wkil.^, and :,,rias are used wherever possible. Indirect

objects follow direct objects and are preceded by a preposition.

Example: . they promise that they will give
your money back to uc:.

Version 3 passages vary in length from 17 to 72 words and in

readability level, where obtainable, from sixth grade to 17+. These

passages also contain complex sentences, but they reflect a higher

degree of chunking than their version 2 counterparts. In version 3

passages relative pronouns, and a form of the verb b'.? where that is

what follows, are deleted wherever they are optional. Such deletions

were found to render aural comprehension more difficult by Charrow

and Charrow (1979). Subordinate clauses often do not contain surface

subjects or finite verbs. Where non-finite verbs are used, -in. forms

or derived nouns are usually used rather than infinitives.

Example: Because of ahirin17 few social contacts, the
farnm r .

These services improve employee n~rale and
help the employer through a z.li.'ti'ci in
absenteeism.

The latter type of nominalization is the type that Coleman (1964. 1965)









found to have an adverse effect on reading comprehension and written

recall as opposed to active verb expressions with the same meaning.

Occasionally an adverbial prepositional phrase takes the

place of an adverbial clause.

Example: T .r7 t: ,,9 C?._, .. you
will find . (version 3)viersus

.... :'' r.2 M' & .;!, you
"will find . (version 2)--

Conditional sentences in version 3 are not always explicit. Instead

of .t'-clauses there may be a verb form preceding the subject.

Example: .Z. i. the puppeteer come out after the show
for a bow, people are often surprised

Finally, indirect objects precede direct objects and thus appear with-

out prepositions.

Example: . they promise to give ,o; your
money back.

At least one multiple choice question follows each short

passage. There are a total of 24 comprehension questions. The aim

of the questions is to test literal comprehension, which is considered

a necessary prerequisite to other types of comprehension. Because

the passages are short, they often test whether or not the reader

understood the main idea rather than more specific facts or details.

The questions cannot be answered merely on the basis of outside

knowledge; the test taker must read the passages in order to answer

the questions correctly. The stem of each test item is followed by

four choices of more or less equal length or two short and two longer

choices so that a single choice is never conspicuously different from

all the others. Wherever possible the test items are worded differently








from the passage itself to prevent recognition of the correct answer

without comprehension.

The test administration yielded a Kuder-Richardson 20 co-

efficient of .672.

The dittoed 13-page test booklet had one or two double-spaced

passages per page. Each passage was immediately followed by its

comprehension questionss.

Covariate. The ESLAT was used as a covariate to control for

varying degrees of proficiency in English among the subjects. The

ESLAT scores correlated .7 (Pearson's Product-Moment Correlation

Coefficient) with the scores on the dependent variable, the SDCT.

ESLAT is a 60 item grammar and reading comprehension test

administered by the College Board to high school seniors in Puerto

Rico. The test includes 40 grammar and 20 reading comprehension

questions. The test is used to place students in the appropriate

English course at the university. Test reliability from 1964 through

1974 is reported by the Puerto Rico College Board to be consistently

greater than .9. They do not, however, indicate what type of re-

liability coefficient was calculated.


Procedure

Four classes of English 001 participated in the study, and

all were tested on the saTe day. The students were given the full

class period to complete the SDCT. They marked their answers in the

test booklet. Both the researcher and the regular teacher were

present in three of the classes. The teacher alone administered the









test in the fourth class after having done so with the researcher

earlier that day. ESLAT scores were made available to the researcher

by the university administration. With the ESLAT scores as a co-

variate, the data were analyzed by analysis of covariance using the

Statistical Analysis System (SAS).


Pilot Studies

Before conducting the main study, several small pilot studies

were done to determine whether or not use of the SDCT was feasible

and whether the study was practical. Because the results will be

referred to in Chapter IV, the pilot studies will be briefly described

here.


Subjects

Two of the pilot study samples consisted of randomly selected

English Language Institute (ELI) students at the University of Florida.

All were native speakers of Spanish. All were undergraduates, at

least 18 years of age, who had graduated from high school in their

respective Latin American homelands. All had scored between 35 and

35 on a version of the Reading and Vocabulary section of an Institute-

developed test that is used for purposes of initial placement and meas-

urement of progress throughout the term. One group consisted of 12

students and the other of 15.

The third sample was a group of 17 native speakers of English

enrolled in a reading course offered by Alachua County Adult Education.

Their reading levels varied from third to tenth grade as measured by

the reading and vocabulary section of the Test of Adult Basic Education

(TABE).








Design

The design of these studies was the same as that of the

Puerto Rico study.


Ma terials

Several alterations were made in the SDCT before it was ad-

ministered in Puerto Rico. Passages 10 and 13 were slightly reworded

to comply better with the specifications for syntax previously des-

cribed, and several comprehension questions (2, 5b, 8, la, 12a, and

14) were altered (see the end of Appendix A). The pilot study samples

were given the earlier version. The format of the test booklet was

the same as in the main study. One major difference, however, was

that the group of 15 ELI students took the SDCT with the comprehension

questions in Spanish. The intention was to assure that the language

of the questions themselves would not affect students' performance.

The Kuder-Richardson reliability coefficients based on the

pilot study administrations were .75 for the group that had the ques-

tions in Spanish, .82 for the group that had the questions in English,

and .81 for the adult education students.

The ELI's Reading and Vocabulary Test was used as a covariate

for the two samples of ELI students to control for varying degrees of

proficiency in English among the subjects. The test scores correlated

.85 (Pearson's Product-Moment Correlation Coefficient) with scores on

the dependent variable for the group that had the comprehension ques-

tions in English and .61 for the qroup that had the comprehension

questions in Spanish. The test consists of a single three-page passage.










Vocabulary items are underlined in the passage and tested by means

of 20 multiple choice questions. The student must choose the best

synonym based on the use of the vocabulary item in the context of

the passage. Comprehension is measured by 20 true-false items. The

test is not timed.

The reading and vocabulary section of the Test of Adult

Basic Education (TABE), used as a covariate for the sample of adult

education students, correlated .8 with SDCT scores.


Procedure

The ELI students who participated in the pilot study attended

a special session conducted by the researcher instead of their reading

class for that day. The test was not timed and all finished easily

within the class hour. There was a separate session for each of the

two samples.

The adult education students took the SDCT during their class.

They were allowed as much time as necessary.

Because it could not be certain that the assumptions of a

parametric statistical procedure were met, a non-parametric analysis

of covariance was used (Quade, 1967) to analyze the data.


Subjective Judgments of Relative Difficulty


Subjects

The subjects who participated in this secondary study were

seventy-nine undergraduates at the Mayaguez Campus of the University








of Puerto Rico who were enrolled in English 002 during the spring

semester of 1980.

Seven teachers of English 001-002 also made subjective judg-

ments of the relative difficulty of examples of the three syntactic

versions from the point of view of what they thought would be easiest

and hardest for their students. Fourteen undergraduate students

taking a course in the teaching of reading at the University of Florida

also participated, making the judgment from the point of view of

limited English proficiency students.


Materials

Three sets of passages of the SDCT (2, 8, 16) were each on

a separate dittoed page. The order in which the three versions ap-

peared was varied as was the order of the three sets of passages.

There was a line to the left of each version of each passage for

subjects to indicate the easiest of the three and the most difficult

of the three.


Procedure

Students were asked, as an in-class exercise, to choose the

paragraph they thought was the easiest in each set of three as well

as the one they thought was the most difficult. The version left

blank was ranked as "in-between." Teachers were asked to make the

same- judgment from the point of view of their students. Pre-service

teachers did the exercise in class.

The data were analyzed by the Friedman two-way analysis of

variance by ranks (Siegel, 1956).












CHAPTER IV


PRESENTATION AND ANALYSIS OF THE DATA

The basic problem addressed by the study was to determine

what degree of sentence combining is the most readable for mature

students learning English as a second language. This chapter des-

cribes and presents the statistical analysis of the data collected

for the Puerto Rico study, the pilot studies, and the subjective

judgments of relative difficulty of the three versions of the Syn-

tactically Different Comprehension Test (SDCT).


Puerto Rico Study

The sample consisted of 85 students enrolled in English 001

at the Mayaguez Campus of the University of Puerto Rico during the

fall semester of 1979. Thirty each read versions 1 and 2 of the SDCT

while 25 read version 3. The scores on the covariate, the English as

a Second Language Achievement Test (ESLAT), and the SDCT for each

student appear in Appendix B.

Figure 1 shows the least squares lines for the plots of ESLAT

scores on the x-axis versus SDCT scores on the y-axis for each of the

three syntactically different versions. The least squares line cor-

responding to version 2 is consistently higher than the line for

version 1 within the range of this sample. The least squares line

for version 3, however, is neither consistently lower than nor higher

than the other two lines.





version 1

version 2
version 3


'3

-a~.5o


390 410 430 450


470 490 510 530 550 570 590 610

ESLAT scores


Figure 1

Least Squares Lines for Each Version
Puerto Rico Study


'' -'''


r









In the statistical analysis of the data,ESLAT scores, which

correlated .7 (Pearson's Product-Moment Correlation Coefficient)

with the SDCT, were included in the model as a covariate. A test of

horrogeneity of regression was performed to determine if there was an

interaction between the covariate and the dependent variable. The

interaction term was not significant. Thus, it was acceptable to

proceed with analysis of covariance in order to test the null hypo-

theses that there is no difference in comprehension between versions 1

and 2, nor between versions 3 and 2.

Table 2 indicates that, overall, the effect of the version

of the SDCT read by the subjects does not reach significance (p < .2388).


Table 2

Analysis of Covariance Using SDCT as Dependent
Variable and ESLAT as Covariate

Source df SS M.S F Prob

Version 2 23.44 11.72 1.46 .2388

Covariate 1 539.18 539.18 67.06 .0001

Error 81 651.23 8.04


Although the overall test was non-significant, the compari-

sons between versions 1 and 2 and between versions 3 and 2 are pre-

sented in Table 3 as they were of special interest. Neither of the

first two null hypotheses stated in Chapter I can be rejected, but

the difference between versions 1 and 2 approaches significance

(p < .09).








Table 3

Pairwise Comparisons

T Prob

Version 1 vs. Version 2 -1.69 .09

Version 3 vs. Version 2 -0.98 .33


Although neither hypothesis can be rejected, the least

squares means, which estimate the average level of response for

subjects scoring at the mean of the covariate, indicate that compre-

hension of version 2 is-higher than comprehension of versions_1 and

3 as shown in Table 4.


Table 4

Least Squares Means

Version I1 LS Mean

1 30 13.43

2 30 14.67

3 25 13.92


Pilot Studies

he results for the 12 English Language Institute (ELI) stu-

dents who took the SDCT with the comprehension questions in Enqlish

Were similar to, though nore pronounced than, those of the Puerto

Rico study) Figure 2 shows that within the range of this sample the

least squares line of version 2 is consistently higher than the lines

corresponding to versions 1 and 3, but the difference between versions

1 and 2 is greater than the difference between versions 3 and 2.





version 1
version 2

version 3


- -V


6,

-
C;

*.; /j


35 45 55 65 75 85

Reading and Vocabulary Scores




Figure 2

Least Squares Lines for Each Version
ELI-Questions in English








Parametric tests indicated significant differences with an

overall F (2, 11) of 13.61 (p < .0027). Pairwise comparisons were

also significant: the t value for version 1 versus version 2 was

-5.04 (p < .001), and the t value for version 3 versus version 2 was

-3.24 (p < .0104). But because it was uncertain if violations of

assumptions of the parametric statistical procedure would affect the

results with a sample of 12 subjects, the data were analyzed non-

parametrically using Quade's Rank Analysis of Covariance (Quade, 1967).

A version of the ELI Reading and Vocabulary Test, which correlated

.85 with the SDCT, was used as the concomitant variable to control for

the varying degrees of proficiency in English among the subjects.

Scores on both the Reading and Vocabulary Test and the SDCT appear in

Appendix B. In the non-parametric test the calculated F (2, 9)

statistic was 2.379 which falls short of significance at the experi-

ment-wise alpha level of .05. Nevertheless, the least squares means

shown in Table 5 indicate that comprehension of version 2 is superior

to that of versions 1 and 3.


Table 5

Least Squares Means
ELI-Questions in English

Version N LS Mean

1 4 12.00

2 5 16.27

3 3 13.20









In the other administrations of the pilot study a similar

trend can be detected: version 2 cons.istently_yieldsthe highest

comprehension though the differences are not statistically sig-

nificant. Scores on the SDCT and the covariates appear in Appendix B

while Tables 6 and 7 present the least squares means for the ELI

students who had the comprehension questions in Spanish and for the

adult education students respectively.


Table 6

Least Squares Means
ELI-Questions in Spanish

Version IN LS Mean

1 5 14.6

2 5 17.2

3 5 13.5


Table 7

Least Squares leans
Adult Education

Version I\ LS IMean

1 7 12.5

2 6 14.2

3 4 12.1


The consistency of this trend as shown in Figures 1 and 2

and Tables 4, 5, 6 and 7 plus the nearly significant difference

between version 1 and 2 in the Puerto Rico study strongly suggest









that characteristics of version 2 sentence structure render version

2 more readable than the others, especially more than version 1.


Subjective Judgments of Relative Difficulty

Students' Judgments

Seventy-nine students enrolled in English 002 at the Mayaguez

Campus of the University of Puerto Rico during the spring semester

of 1980 were asked to rank the three versions of passages 2, 8 and 16

of the SDCT. They did so by indicating the version they considered

the easiest and the one they considered the most difficult leaving

blank the one tallied as "in between."

Each of the three sets of passages was separately analyzed
2
using the Friedman two-way analysis of variance by ranks. The ,.r

statistic for the three versions of passage 2 was 14.99.The null

hypothesis that judgments of difficulty are independent of syntactic

version can thus be rejected (p < .001). .Follow-up tests indicated

that version 2 was ranked significantly easier than version 1

(p < .001) and version 3 was ranked as easier than version 2 (p < .005).

Table 9 shows that students tended to rank version 1 as the most

difficult while versions 2 and 3 were almost equally ranked easiest

or in between.

For passage 16 version 2 was most often ranked as easiest

while version 1 was most often ranked as hardest. The *,r statistic

was 33.4413 indicating independence of judgment of difficulty and

syntactic version (p < .001). Follow-up tests indicated that










version 2 was ranked as significantly easier than both versions 1

(p < .005) and 3 (p < .005).


Table 9

Contingency Table-Passage 2

version 1 version 2


hardest


easiest


in between


hardest


easiest


in between


Table 10

Contingency Table-Passaae 16

,ion 1 version 2


The null hypothesis was also rejected for passage 8

(.,r = 11.87, p .01), but in this case version 3 was most often

ranked as the hardest. The follow-up test contrasting versions 2

and 3 was significant (p < .005). Versions 1 and 2 were almost

equally ranked easiest or in between. The follow-up contrasting





63


versions 1 and 2 was not significant. Perhaps subject matter or

conceptual load accounts for the different pattern of judgment on

this particular set of passages.


Table 11

Contingency Table-Passage 8

version 1 version 2 version 3
21 18 40
hardest 21.6% 22.3' 50.6%


easiest


in between


35 30 14
44.3% 38.o0; 17.71

23 31 25
29.1% 39.2"; 31.6;


Teachers' Judgments

Seven English 001-002 teachers at the University of Puerto

Rico performed the same subjective judgment exercise from the point

of view of what they thought would be easiest and hardest for their

students. Separate analyses of each passage using the Friedman

two-way analysis of variance by ranks were not significant. However,

Table 12, which represents all three passages together, illustrates

a pattern. Overall, teachers are more likely to consider version 1

easiest for their students, version 2 in between, and version 3

hardest.









Table 12

Contingency Table
Teachers' Judgments of Three Sets of Passages

v\rcinn 1 erscinn 2 r i 3


hardest


easiest


in between



Judgments of


Pre-Service Teachers


A class of 14 pre-service teachers enrolled in a basic course

in the teaching of reading at the University of Florida also judged

the three sets of passages. Each set was analyzed using the Friedman

two-way analysis of variance by ranks. The ,,r statistic (3.999) was

significant (p < .02) for passage 3 only. Follow-up tests indicate

that version 1 of passage 3 was judged to be significantly easier

than version 2. The follow-up contrasting versions 2 and 3 was not

significant.


Table 13

Contingency Table-Passage 8

version 1 version 2 version 3
2 7 5
hardest 14.3, 50.0', 35.7%

11 1 2
easiest 78.6' 7 .1, d 14


in between


V' 3j i 1 "i .^ I .l'kII "'c I )i3 ,J
4 5 12
19.0W.- 23.8" 57. 1

10 5 6
47.6'. 23. 23. 2.6'

7 11 3
33. 3 52.4' 14.3'








Although the null hypothesis of independence of difficulty judgment

and syntactic version could not be rejected for passages 2 and 16,

the overall pattern of the judgments of the pre-service teachers, as

shown in Table 14, was similar to that of the ESL teachers.


Table 14

Contingency Table
Pre-Service Teachers' Judgments of Three Sets of Passages

version 1 version 2 version 3


hardest


easiest


in between


Because the pre-service teachers had recently been exposed to read-

ability formulas in their course, it is not surprising that they

tended to judge version 1 as the easiest, version 2 as in between, and

version 3 as hardest or in between.


Discussion

The results of this study indicate that lower readability

levels, as measured by common readability formulas, do not facili-

tate the reading comprehension of mature Spanish speakers learning

English as a second language. In fact, it is suggested that the

short, primarily simple sentences generally considered to be typical

of the low readability levels of the version 1 passages actually are


13 13 16
31.0% 31.0% 38.1%

25 6 11
59.5% 14.3% 26.2%

4 23 15
9.5% 54.8% 35.7.'










an obstacle to comprehension. As Shook (1977) hypothesized, rela-

tionships and meaning, it seems, are lost with such sentence struc-

ture. Although the differences among the three versions of the

SDCT were not significant, the fact that on all administrations opf

the instrument version 2 yielded the highest comprehension scores
.... - -. .
does suggest that version,2 senP-_ t structure is more readable and

more comprehensible than-that of.version 1 and to some extent that

of version 3. To what degree this is so, however, remains uncertain.

It is strange that, as seen in Figure 1, at the lower end of

the range of proficiency in English, version 3 yielded slightly

higher scores than versions 1 and 2, though a test of equality of

slopes was not significant. Perhaps at that level of proficiency

students are so lost that sentence structure simply makes no dif-

ference, and these results merely reflect random variation. Also,

the studies reviewed in Chapter II that led the researcher to

believe certain characteristics of version 3 sentence structure

would prove difficult may have been misleading. The fact that these

characteristics did not affect reading comprehension in this study

may be due to the use of a different dependent variaE....jErmoLJ ose

used in the original studies. Some of those studies, for example,

measured listening comprehension while others measured recall or

memory zatian.._ The dependent variable in this study, in contrast, was

comprehension measured solely by response to multiple choice ques-

tions which were answered immediately after reading.
i-:_ :->:: -; -; ; -; -.: a .;. a








'Another consideration is that the SDCT may attempt to

encompass too much at once. A series of finer-grained studies

might render the greater readability of version 2 more apparent,

as well as clarify the puzzling results of version 3 with students

of lower proficiency. These studies would examine single struc-

tures such as nominalization versus noun + verb constructions

Coleman, 1964, 1965),'inclusion versus exclusion of relative pro-

neRis-+....be (Charrow and Charrow, 1979,, "You must follow the law as

I state it," 1978), inclusion versus exclusion of optional comple-

mentizers (Hakes, 1972), explicit versus implicit conditionals

(Marshall and Glock, 1978-79) and the like. Such studies might

prove more fruitful than examining something as gross as_"degree

of chunking." Lumping together the various leads gleaned from such

research as the studies mentioned above may have blurred any poten-

tial findings.

It is of interest that the subjective judgments of the

students support the suggestion of the main study. Students do

not tend to judge version 1 as the easiest to read, demonstrating
u cw ===== .................... ... ..... ....
tl inappropriateness_of the most commonly used readability formulas
for this population. In fact, they often judge version 1 to be the

most difficult. They tend to judge version 2 as the easiest while

judgments of version 3 are more evenly dispersed. In contrast, both

pre-service teachers and experienced ESL teachers often judge

version 1 to be the easiest, version 2 in between and version 3 the

hardest for their students. Thus well-meaning teachers or curriculum

planners may, based on their tendency to judge version 1 as the









easiest, select reading material with sentence structure similar

to that of version 1 passages for their mature ESL students. Even

for a basic course like Enolish 001-002, teachers may unwittingly

be doing students a disservice. In the first place, this study has

shown that such reading material does not result in higher compre-

hension. Nor is such material considered easiest by the students

themselves as demonstrated by their responses to the subjective

judgment exercise. And finally, as Shook (1977, p. 6) suggested,

such sentence structure is "like nothing they'll ever encounter in

the real world."

The notion that lower readability levels, as measured by

traditional formulas, are more suitable for non-native speakers is

cast in doubt, at least for adult learners who are literate in their

mother tongue, Spanish. In this case their own language is one

that is not so radically different in structure from English as the

languages of other foreign or ESL students may be. It is also a

language that, as a matter of style, prefers complex to simple

sentences. It is possible that for native speakers of non-western

languages, in contrast to Spanish speakers, conventionally measured

lower readability levels are an aid to comprehension. This remains

to be investigated.

If the merely suggestive results of this study are inter-

preted as a weak effect of syntax on readability, one can speculate

why this is so for native speakers of Spanish, and in particular for

Puerto Ricans. Many Puerto Ricans have learned to cope with reading

in English more successfully than they have learned to cope with







oral/aural skills. Information supplied by English 002 students

reveals that they have all studied English since elementary school,

usually beginning as early as the first grade. They generally

began to read in English a year or two later. Some had their

science and math textbooks in high school in English. All confront

textbooks in English in college although the majority of 001-002

students are freshmen without lengthy experience with college text-

books in English. Many claim to read comics and magazines in

English, and some state that they read the San Juan Star, the

island's English language newspaper. They constantly encounter

labels and advertisements in English because most products available

on the island are from the U.S. Directions and indications on these

products are often in English. Thus this poDulation may be ac-

customed to extracting, or trying to extract, meaning from English

with any type of syntax. The least likely t xye of syntax for them

to encounter in these real-life settings is the simple sentences of
...................... ................................ ............................................. ........... .............. .... ,
version 1, explaining the lower comprehension of those subjects who

read version 1.

In addition to exposure to English, much of the Spanish read

by Puerto Ricans is influenced by English. Aside from the inter-

ference due to language contact, much of what is printed in Spanish

in Puerto Rico is clearly translation from English (wire service news

columns, Cosmopolitan en Espaiol, etc.). American English syntax has

in this way infiltrated Puerto Rican Spanish. Furthermore, as men-

tioned above, as a question of taste and style, Spanish prefers

complex sentences. Also, near literal translation of several passages









of the SDCT revealed that version 1, in the judgment of a native
speaker of Spanish, is very "un-Spanish" while versions 2 and 3

were perceived as being much more in line with Spanish sentence
structure. This opinion mirrors the fact that version 1 resulted

in almost significantly lower comprehension while the difference
between versions 2 and 3 was inconclusive.

f indeed syntax is less important for this population than

one might expect, and if we discard the sentence length criterion

as a result of this study, what are we left with other than vocabu-

lary?( Are there other criteria that should be considered in

selecting reading material? Rhetorical devices such as organization,

example, restatement, coherence and the like may play an important
role in determining reada:Jlt..jyX Densjy.sof-ideas, or .the..n.umber

of propositions presented in a text and the structure of that text
may affect its readability.(.~Aai -aQ=L^.ZiggQnl;;~ J 9). These factors
were not manipulated in this study. .Kintsch and Vipond go in a

different direction from this study in searching for neglected ele-

ments that might influence readability. If syntax does not have a

great effect, perhaps their model points in a more fruitful direction.

Even if syntax has a relatively unimportant effect on read-

ability, this study indicates that Smith's (1971) findings with

English speakers may hold for Spanish speakers within the limited

age range of the sample. He found that older students find more
mature syntax to be more readable. In this study, it was found that

the higher readability level material with more mature syntax is
at least as readable as the material considered to be lower if not





71


more so. This may be evidence that cognitji...Ljam ur.i.ty_.j one's

native language does carry over as an advantage. in..deaJlng. with a

second language.









CHAPTER V


SUMMARY AND CONCLUSIONS

Overview

The purpose of this stud, was to determine whether read-

ability would be affected by manipulating the degree and type of

sentence combining or "chunkinq" in reading material prepared for

mature students of Enqlish as a second language (ESL) who are

literate in Spanish, their native language, and who are at an

intermediate level in their study of English.

Studies done with Young native speakers of English have

shown that there is a relationship between reading ability and

syntactic development (Chomsky, 1972; P.ibovich. 1976; Ruddell.

1965a, 1965b; Simons. 1970; Tatham. 1969-70). Van Metre (1970)

showed that a similar relationship exists for bilingual third

graders. However, virtually no research relating syntax and reading

ability has been done with mature ESL students.

Research done on how the human mind handles different types

of sentences has exclusively used native speakers of English as sub-

jects. One group of such studies (Gough. 1965, 1966; Mehler. 1963;

Miller. 1962; Savin and Perchonock, 1965) aimed to learn whether

transformed sentences are more of a burden on memory than kernel

sentences. These studies, of limited scope, have been criticized

on several grounds and are not conclusive.








Studies that attempted to define_sy.n.tactic co.mplexi,.ty .....a.re

also inconclusive. Several studies (Martin and Roberts, 1966;

Wright, 1969) aimed to resolve the conflict between Yngve's (1960)

depth hypothesis and the theory that transformations create com-

plexity. Others support the notion that integration or chunking

of ideas is both natural and more comprehensible (Bransford and

Franks, 1971; Pearson, 1974-75). Fodor and Garrett (1967) at-

tempted to directly isolate surface clues .who.se presence .or absence

facilitated or..hinder.ed._p.roQ.ggsi.ng. Again, the subjects in these

studies focusing on the question of what is syntactically complex

wereall native speakers of Englis.h. The field is therefore open

to choosing a population of ESL students and manipulating syntax

along the lines of questions raised by previous research with native

speakers of English.

In this study, 85 students enrolled in an ESL course at the

University of Puerto Rico were randomly divided into three groups.

Each group read 18 short passages of the Syntactically Different

Comprehension Test (SDCT) and responded to 24 multiple choice

comprehension questions on the passages. The SDCT was constructed

by the researcher by adapting passages from the Science Research

Associates' Reading for Understanding General reading kit (Thurstone,

1969). Three versions of each passage were constructed. The ver-

sions differed from one another principally in sentence structure;

each version represented a different degree of chunking or sentence

combining. Version 1 passages consisted of short, primarily simple










sentences; version 2 included complex sentences with as many clues

to underlying relationships as possible left intact, and version 3

also included complex sentences, but, because of the higher degree

of chunking, fewer clues to underlying relationships were explicitly

expressed. Vocabulary and content were held constant. There were

several pilot administrations after which a few minor alterations

in the instrument were made. The purpose was to determine if this

manipulation of syntax affected the readability of the passages as

reflected in the comprehension scores.

The hypotheses tested in the principal study were:

1. There is no significant difference between comprehen-
sion of version 1 and comprehension of version 2
passages.
Uvl "v2 = 0

2. There is no significant difference between compre-
hension of version 3 and comprehension of version 2
passages.
uv3 Uv2 = 0

The data were analyzed by analysis of covariance using the

Statistical Analysis System (SAS). Scores on the English as a

Second Language Achievement Test (ESLAT), a grammar and reading

comprehension test administered to high school seniors by the College

Board of Puerto Rico, were used as a covariate to control for

varying degrees of proficiency in English among the subjects. As

in all the pilot studies, version 2 yielded the highest comprehension

scores, but with an experinent-wise alpha level of .05 neither null

hypothesis could be rejected. Hypothesis 1 came close to being

rejected.








Because the results suggested that version 1 sentence struc-

ture is not an aid to comprehension, a supplement to the main study

was carried out. Three different groups of subjects participated

in the secondary study: a sample of 79 students, different from

those used in the main study, but from the same population, seven

ESL teachers who teach that population, and a class of 14 pre-service

teachers enrolled in a basic course in the teaching of reading at

the University of Florida. They were asked to subjectively judge

which of the three versions of several of the passages they thought

was the easiest and which they thought was the most difficult. The

one not ranked was considered "in between." Overall, the students

tended to rank version 1 as the most difficult (or in between) and

version 2 as the easiest (or in between) while judgments of version 3

were more evenly dispersed. For two of the three passages judged,

this pattern was significant using the Friedman two-way analysis of

variance by ranks to test the null hypothesis that syntactic version

and judgment of difficulty are independent. Although the same null

hypothesis was not rejected for University of Puerto Rico ESL

teachers' judgments of the three versions of the passages, the

teachers tended to judge version 1 as the easiest, version 2 as in

between and version 3 as the hardest; i.e., chunking, they often

thought, created difficulty for their students. For the pre-service

teachers who judged the passages from the point of view of students

whose native language was not English, the null hypothesis was rejected









for one of the passages. Overall, they usually judged version 1

as the easiest, version 2 as in between, and version 3 as the

hardest or in between.


Conclusions and Recommendations

_Based on the results of the study, it was concluded that

reading materials with the short, primarily simple sentences char-

acteristic of conventionally determined low readability levels do

not enhance reading comprehension for the population from which the

sample used in this study was drawn. This supports Shook's (1977)

hypothesis that relationships and meaning are sacrificed for these

lower readability levels, which, moreover, result in an unnatural,

unrealistic type of sentence structureD It was also concluded that

students do not necessarily judge this type of syntax as the easiest

to read.

The warning, then, is to teachers and curriculum planners

who may choose for their ESL students material considered easy to

read according to commonly used formulas as well as according to

their own subjective judgments. First, this study suggests that

such reading material does not result in higher comprehension.

Second, the students themselves do not consider it easiest. And

third, as Shook (1977, p. 6) suggested, such sentence structure is

"like nothing they'll ever encounter in the real world."

But there are limitations to this conclusion that must be

borne in mind. Two primary limitations are based on the native

language of the subjects and their age. Spanish is not so radically
""""" ------=.. .--








different in structure from English as some other foreign languages.

Furthermore, Puerto Rico has a unique relationship to American

English. Because of political and economic ties to the United

States, exposure to English is often broader and deeper than it is

in other parts of the Spanish-speaking world. Whether or not simple

sentences typical of what is commonly considered lower readability

levels are an aid to other populations of ESL students remains to

be investigated. It is noteworthy that the pilot studies suggest

the same pattern as the Puerto Rico study. The question of what

type of sentence structure is most readable for younger ESL students

also remains open.

One must consider, too, the possibility that the failure to

reject the two null hypotheses of the principal study is an indica-

tion that syntax does not have a great effect_..nrj.bja 1 lity, at

least for the population under study. What then, other than vocabu-

lary, does influence readability for ESL students? Other features

of a passage such as organization, coherence, restatement, density

of ideas, conceptual difficulty must be considered.

Before abandoning the relationship between syntax and read-

ability, however, the following recommendations for further research

are made. They may succeed in strengthening and expanding the

present study.

1. As already mentioned, the study should be replicated

with ESL students who are native speakers of other languages,

especially non-western languages. Radically different effects of









syntax on readability for different populations would render a

uniform approach to reading inappropriate.

2. Conducting the study with younger ESL students would

be similar to the work of Smith (1971) with non-native speakers

of English. If younger students do comprehend version 1 better than

the other versions, this would parallel Smith's findings, and the

results of this study would moreclearlyindicate positivetransfer

of cognitiv.e_Ldeyelopment and an advantage for adult learners. The

differences across age groups may not be parallel across native

language groups. Whether or not this is so would be of interest.

3. replication of some of the psycholinguistic studies re-

viewed in Chapter II with ESL students as subjects would reveal

whether or not non-native speakers respond in a fashion parallel to

the responses of native speakers of English. Do they find trans-

formed or kernel sentences easier toprocess (Gough, 1965, 1966;

Mehler, 1963; Miller, 1962; Miller and McKean, 1964; Savin and

Perchonock, 1965)? Are chunked ideas preferred (Franks and Bransford,

1971)? Wha.t_Atdegree._Q.f_.ne.stJ.ia.can be handled (Schlesinger, 1968)?

Is center-embedding easier to process than right-branching (Hamilton

and Deese, 1971; Miller and Isard, 1964)? Is there an effect_of

seQatence depth (Yngve, 1960)?

4. A series of finer-grained studies manipulating one item

at a time would provide more definite and specific findings than

those resulting from use of the SDCT. Such studies might demonstrate

that the inconclusiveness of this study is due to insensitivity of

the SDCT, possibly because it attempts to encompass too much.








Finer-grained studies might also help_ UwA.vel the mytrigus oc-

currence of higher scores on version 3 at the lower end of the

range .of pro.fi.cyiaSyin.. EgJ-i sh. These studies could include,

along the lines of Coleman (1964, 1965), examining whether a noun +

verb combination is more comprehensible than the corresponding

nominalization. Taking Hakes (1972) and Charrow and Charrow (1979)

as models, the effect of omission or inclusion of optional comple-

mentizers and relative pronouns could be examined. Following

Marshall and Glock (1978-79), one could examine whether or not

explicitly expressed conditions are more comprehensible than alter-

nate forms. Comprehensibility of two simple sentences versus a

.correspondinng complex sentence with a relative, adverbial, or noun

clause might be contrasted. Several dependent variables are

possible: recall, paraphrase, memorization, time, translation, as

well as comprehension questions.

Until further research is done permitting more specific

pedagogical implications, the principal recommendation of the

present study is that low level reading material, as measured by

conventional readability formulas with primarily simple sentences

is not recommended for college age ESL students in Puerto Rico.

This does not mean that care should not be given to the selection

of material. Further work on the effect of syntax needs to be done

and other factors affecting readability need to be considered.








APPENDIX A


INSTRUMENT

The Syntactically Different Comprehension Test*


Please read the paragraphs in this booklet. Try to answer
the question or questions that are after each paragraph based only
on the information that is in the paragraph.

Mark the letter of the correct answer with an X. There is
only one correct answer for each item.

If you cannot answer the question based on the information
that is in the paragrph, go on to the next one.


Practice Paragraph


version 1





version 2





version 3


The stems of the cactus olant are provided with a large
quantity of special tissue. This tissue stores water.
The plant is equipped to retain moisture, and this is
one of the ways. This feature makes it particularly
suitable to life in arid areas.

The stems of the cactus plant are provided with a large
quantity of social tissue which stores water. This is
one of the ways that this plant is equipped to retain
moisture, which is a feature that makes it particularly
suitable to life in arid areas.

The stems of the cactus plant are provided with a large
quantity of special tissue for water storage. This is
one of the ways this plant is equipped to retain moisture,
a feature making it particularly suitable to life in arid
areas.

A cactus plant does not need frequent rain because it


A. does not require water.
B. can save water for later use.
C. uses other liquids.
D. grows slowly.
A B C D
( ) ( ) ( ) ( )


*In the actual test booklet for each version the comprehension
questions) immediately follow each paragraph on the same page. The
paragraphs and the choices are double-spaced.











version 1



version 2



version 3


Steve couldn't swim. He walked along the side of the
creek. He looked for a shallow place. He would wade
across at that place.

Since Steve couldn't swim, he walked along the side of
the creek looking for a olace that was shallow enough for
him to wade across.

Since Steve couldn't swim, he walked along the side of the
creek looking for a place shallow enough to wade across.

Steve looked for a place where the water wasn't
deep so he could


swim to the other side.
walk to the other side.
throw rocks to the other side.
see the other side.


A B C D


version 1




version 2




version 3


Disease germs may be present in food. Cook food for a
long enough time. This will kill any disease nerms. Food
may not be clean. Cook it thoroughly. In this way you
can combat possible uncleanliness of food.

If you cook food for a long enough time, you will kill any
disease germs that may be present. Therefore, one way that
you can combat possible uncleanliness of food is by cooking
it thoroughly.

Cooking food for a long enough time will kill any disease
germs possibly present. Therefore, cooking it thoroughly
is one way of combatting possible uncleanliness of food.

To be certain that food is safe to eat


wash it.
cook it completely.
cook it immediately.
fry it.


A B C D
( ) ( ) ( ) ( )









version 1






version 2





version 3


They use olives to make olive oil. The olives ripen.
They pick them promptly. They send them to the oil mill
immediately. The olives spoil rapidly. They cannot ship
them far. In some places olives are grown extensively.
You will find a small oil mill every few miles in these
places.

Olives that are used in making-olive oil are picked as
soon as they ripen and are immediately sent to the oil
mill. They cannot be shipped far because they spoil
rapidly. Where olives are grown extensively, you will
find a small oil mill every few miles.

Olives used in making olive oil are picked as soon as
they ripen and are sent immediately to the oil mill.
Because of rapid spoiling, they cannot be shipped far.
In extensive olive growing regions you will find a small
oil mill every few miles.

According to this paragraph, olive oil mills are
close to where the olives are grown because


it saves shipping costs.
the growers also make the oil.
the olive oil is consumed locally.
picked olives will not last a long time.


A B C D
( ) ( ) ( ) ( )


version 1





version 2





version 3


Some animals have a "weather sense." The weather governs
their movements and their very existence. Man has never
been able to equal this. The American Indian was a fine
weather prophet. Even he relied almost entirely on the move-
ments of the creatures of the wild for his weather predictions.

Animals whose movements and whose very existence are
governed by the weather have a "weather sense" that man has
never been able to equal. Even the American Indian, who was
a fine weather prophet, relied almost entirely on the move-
ments of the creatures of the wild for his weather predictions.

Animals whose movements and very existence are governed by
the weather have a "weather sense" never equalled by man.
Even the American Indian, a fine weather prophet, relied
almost entirely on the movements of the creatures of the
wild for his weather predictions.








Those who are most sensitive to climatic changes are

A. American Indians.
B. people with rheumatism.
C. certain animals.
D. meteorologists.


A B C D
( ) ( ) ( ) ( )


version 1





version 2





version 3


We were concealed behind the bushes. We watched our
victim. Our victim approached the pit. We had dug the
pit. We had covered it with leaves and light branches.
Suppose the animal sensed something. It was about to be
captured. Surely it would not have walked along so casually.

We were concealed behind the bushes and watched our victim
approach the pit that we had dug and covered with leaves
and light branches. If the animal had sensed that it was
about to be captured, surely it would not have walked along
so casually.

Concealed behind the bushes, we watched our victim approach
the pit we had dug and covered with leaves and light
branches. Surely the animal would not have walked along so
casually had it sensed it was about to be captured.

The animal that was beinn hunted

A. saw the hunters.
B. was looking for food.
C. tried to hide.
D. was seen by the hunters.


A B C D
( ) ( ) ( ) ( )


The trap that the hunters set was

A. visible.
B. under the bushes.
C. camouflaged.
D. dug by their guide.


A B C D
( ) ( ) ( ) ( )







The hunters believed that the animal

A. had sensed danger.
B. was nervous because it was afraid.
C. was not aware that it was being hunted.
D. had walked away.

A B C D
( ) ( ) ( ) ( )


version 1





version 2


Educators and parents have been complaining. Children
spend too much time at something. They watch television
too much. They don't do their homework. Educators and
parents got support this week. A company is now producing
a television with doors. The doors can be locked.

Educators and parents, who have been complaining that
children spend too much time watching television instead
of doing their homework, got support this week from a
company that is now producing a television with doors that
can be locked.


version 3 Educators and parents, complaining children spend too much
time watching television instead of doing their homework,
got support this week from a company now producing a tele-
vision with lockable doors.

According to this paragraph, teachers and parents
believe that

A. children should spend more time studying.
B. children should participate in television
quiz shows.
C. television is an excellent educational device.
D. children's play time should be limited.

A B C D


The company that is referred to in this paragraph

A. promotes television watching.
B. wants to help parents and teachers.
C. is supported by the television networks.
D. produces children's books.

A B C D
()()()()










version 1


version 2


version 3


The jury returned the verdict after a very short interval.
They were very prompt. The judge thanked them for this.

Because the jury returned the verdict after a very short
interval, the judge thanked them for being so prompt.

The jury returned the verdict after so short an interval
the judge thanked them for their promptness.


The judge was grateful because the jury

A. went home early.
B. acted quickly.
C. made a good decision.
D. was chosen quickly.


A B C D
( ) ) ( ) ( )


version 1




version 2




version 3


Manufacturers have to get goods to market. Suppose the
manufacturer and the market are a long distance apart.
This can be a big expense. The manufacturer can minimize
this expense. He can establish himself near a market.

If the manufacturer and the market are a long distance
apart, then it can be a big expense for the manufacturer
to get goods to market. The manufacturer can minimize
this expense by establishing himself near a market.

Getting goods to market can he a big expense if the manu-
facturer and the market are a long distance apart.
Establishment of the manufacturer near a market can
minimize this expense.

According to this paragraph, manufacturers can
save money if

A. they do not have to advertise in out-of-town
newspapers.
B. products do not have to be shipped far.
C. natural resources are nearby.
D. goods are handled carefully during shipping.

A B C D
()()()()




86




version I Voltaire was one of the greatest writers of the eighteenth
century. He said, "The two essentials of style are pre-
cision and color." The remark is interesting. Voltaire's
style is exceptionally precise. But it is not outstanding
for color.

version 2 Voltaire, who was one of the greatest writers of the
eighteenth century, said that the two essentials of style
are precision and color. The remark is interesting because
Voltaire's style, although exceptionally precise, is not
outstanding for color.

version 3 Voltaire, one of the greatest writers of the eighteenth
century, said the two essentials of style are precision
and color. The remark is interesting in view of Voltaire's
style, which is precise but not outstanding for color.

Voltaire's writing

A. has all the stylistic characteristics that
he considers indispensable.
B. is not particularly colorful.
C. is difficult to understand.
D. is less exact than it should be.

A B C D



version 1 The two boys were pale and shaken. They hurried to the
safety of their home. There they told of their fright-
ening experience. They sat in the kitchen. They were warm
and comfortable. They had really escaped. They realized
it, but this was hard.

version 2 The two boys, who were pale and shaken, hurried to the
safety of their home, where they told of their frightening
experience. As they sat in the kitchen, where they were
warm and comfortable, it was hard for them to realize that
they had really escaped.

version 3 The two pale and shaken boys hurried to the sa-fety of
their home, where they told of their frightening experi-









The boys returned home

A. after having a good time.
B. after escaping some danger.
C. without understanding what happened.
D. without telling their story.

A B C D
()()()()


(11)


version 1






version 2





version 3


The farmer has few social contacts. Therefore, he tends
to be offended by minor insults. The resident of the
city has acquired immunity to these. He is accustomed to
the give-and-take of casual interpersonal relationships.
He fails even to notice discourtesy. The rural person
would take this very seriously.

Because he has few social contacts, the farmer tends to be
offended by minor insults against which the resident of
the city has acquired immunity. The resident of the city
is accustomed to the give-and-take of casual interpersonal
relationships and fails even to notice discourtesy that
the rural person would take very seriously.

Because of having few social contacts, the farmer tends
to be offended by minor insults against which the resident
of the city has acquired immunity. The resident of the
city, accustomed to the give-and-take of casual inter-
personal relationships, fails even to notice discourtesy
the rural person would take very seriously.

According to this paragraph, rural people are
often more socially sensitive because they

A. exoect people to be kind to strangers.
B. are raised that way.
C. are isolated from other people.
D. are from large families and are used to a
lot of people.

A B C D
()()()()







City people, according to this paragraDh,


are easily offended by gruffness.
have frequent personal contact.
are quiet.
are friendly.


A B C D
( ) ( ) ( ) ( )


(12)


version 1








version 2







version 3


Puppets are small doll-like figures. They can be moved
by a person's hand. The hand is inside the doll. They
can be moved by rods. The rods stick up through the body.
Or they can be moved by strings from above. The audience
does not see the puppeteer. People watch a puppet show.
The puppets seem to become real people. The audience no
longer thinks of them as dolls. Suppose the puppeteer
comes out after the show. He takes a bow. People are
often surprised. He looks like a giant.

Puppets are small doll-like figures that can be moved by
a person's hand which is inside the doll, by rods that
stick up through the body, or by strings from above. The
puppeteer is not seen by the audience. When people watch
a puppet show, the puppets seem to become real people. The
audience no longer thinks of them as dolls. If the
puppeteer comes out after the show to take a bow, people
are often surprised because he looks like a giant.

Puppets are small doll-like figures movable by a person's
hand inside the doll, rods sticking up through the body,
or strings from above. The puppeteer is unseen by the
audience. The puppets seem to become real people, and the
audience watching a puppet show no longer thinks of them as
dolls. Should the puppeteer come out after the show for a
bow, people are often surprised because he looks like a giant.

Puppets are animated by

A. a remote control device.
B. a special kind of battery.
C. one of three methods.
D. a magnetic system.

A B C D
()()()()








During a puppet show the audience

A. feels as though they are puppets themselves.
B. takes part in the show.
C. operates the puppets.
D. begins to think of the puppets as being alive.

A B C D
()()()()


(13)


version 1







version 2






version 3


Streams flow through valleys. Most valleys are wider
than the streams. We conclude something. Factors operate
to form the valley. There are usually factors other than
the wearing-down action of the stream. Suppose a valley
were formed only by the action of the stream. The valley
would be no wider than the stream. Its sides would be
vertical.

Since most valleys are wider than the streams that flow
through them, we can conclude that there are usually
factors other than the wearing-down action of the stream
that operate in forming the valley. If a valley were
formed only by the action of the stream, the valley would
be no wider than the stream and its sides would be vertical.

Since most valleys are wider than the streams flowing
through them, we can conclude there are usually factors
other than the wearing-down action of the stream operating
in the formation of the valley. Were the valley formed
only by the action of the stream, the valley would be no
wider than the stream and would have vertical sides.

The creation of a valley involves


several forces.
the force of a river alone.
the influence of weather.
forming steep sides.


A B C D
( ) ( ) ( ) ( )






(14)


version 1





version 2




version 3


Suppose you subscribe to this magazine. You can't lose.
The editors believe something. You will consider it
well worth the subscription cost. Suppose you do not
like it. They make a promise. They will give your money
back to you.

You can't lose if you subscribe to this magazine. The
editors believe that you will consider it well worth the
subscription cost. If you do not like it, they promise
that they will give your money back to you.

Should you subscribe to this magazine, you can't lose.
The editors believe you will consider it well worth the
subscription cost. Should you not like it, they promise
to give you your money back.

The magazine editors make a promise which
shows that they

A. are confident that most people will
like the magazine.
B. do not care if people are pleased with
the magazine.
C. do not-care about making money.
D. want to make money.

A B C D
()()()()


(15)


version 1




version 2



version 3


Factories provide medical and nursing services. These
help to keep employees healthy. These services improve
employee morale. They help the employer. They reduce
absenteeism.

Factories provide medical and nursing services that help
to keep employees healthy. These services improve employee
morale and help the employer by reducing absenteeism.

By providing medical and nursing services, factories
help keep employees healthy. These services improve
employee morale and help the employer through a reduction
in absenteeism.








According to this paragraph, giving benefits
to employees

A. causes prices to go up.
B. is worthwhile for the employers.
C. is compulsory.
D. is tax deductible.

A B C D
( ) ( ) ( ) ( )


(16)


version 1





version 2





version 3


Birds migrate. They arrive in the North in early spring.
The days are a certain length. The birds leave again in
the fall. The days in fall are the same length. This
fact has led some scientists to believe something. The
birds' urge to travel is the result of the amount of
sunshine.

Birds that migrate arrive in the North in early spring.
They leave again in the fall when the days are the same
length as when they arrived. This fact has led some
scientists to believe that the birds' urge to travel is
the result of the amount of sunshine.

Migrating birds arriving in the North in early spring
leave again in the fall when the days are the same length
as upon their spring arrival. This fact has led some
scientists to believe the birds' urge to travel to be the
result of the amount of sunshine.

According to the theory that is proposed in
this paragraph, the migration of birds


depends on variation in humidity.
depends on the number of hours of light
per day.
is related to mating habits.
is related to the search for food.

A B C D
C ) ( ) C ) C )







(17)


version 1





version 2


The baby elephant watched the two bull elephants. They
fought. It is difficult to be king bull of a herd. The
baby realized this. The king has to be strong and ob-
servant. He must use good judgment. Nevertheless, the
baby elephant resolved to be king some day.

After watching the two bull elephants fight, the baby
elephant realized that it is very difficult to be king
bull of a herd. The king has to be strong and observant
and must use good judgment. Nevertheless, the baby
elephant resolved to be king some day.


version 3 After watching the two bull elephants fight, the baby
elephant realized the difficulty of being king bull of
a herd. The king has to be strong, observant and must
use good judgment. Nevertheless, the baby elephant re-
solved to be king some day.

To be king of a herd of elephants


A. requires a good memory.
B. may be dangerous.
C. one's father must have been king before.
D. one must be the oldest male elephant
in the herd.

A B C D


Despite all odds, the young elephant demonstrated


determination.
bravery.
intelligence.
power.


A B C D
()()()()








(18)


version 1







version 2






version 3


Science can be regarded either from a static or from a
dynamic point of view. Consider it from a static point
of view. Science is a body of knowledge. It describes
the universe. It is regarded as explanatory. Some
people have a dynamic orientation. To them science im-
plies continuous activity. Today's state of knowledge
is a basis for further operations.

Science can be regarded either from a static or from a
dynamic point of view. If one considers it from a static
point of view, science is a body of knowledge that des-
cribes the universe and is regarded as explanatory, whereas
to anyone with a dynamic orientation science implies con-
tinuous activity that has today's state of knowledge as a
basis for further operations.

Science can be regarded from a static or a dynamic point
of view. Considered statically, science is a body of
knowledge descriptive of the universe and regarded as
explanatory, whereas to anyone with a dynamic orientation
science implies continuous activity with today's state of
knowledge as a basis for further operations.

This paragraph presents two ways to view science:

A. one view is physical; the other is
biological.
B. one views science as a store of knowledge
to apply to problems; the other views
science as a device for preventing problems.
C. one views science as a fixed set of
knowledge; the other views science as
building on that set of knowledge.
D. one view is historical; the other is
futuristic.

A B C D
( ) () ( ) ( )







SDCT Items Used in Pilot Studies that Differ from
Form Used in Puerto Rico Study


In the pilot studies versions 2 and 3 of passage 10 and
versions 1 and 2 of passage 13 appeared as follows:


(10)


version 2





version 3


version 1


version 2


The two boys, who were pale and shaken, hurried to the
safety of their home, where they told of their frigntening
experience. Even as they sat in the kitchen, where they
were warm and comfortable, it was hard for them to rea-
lize that they had really escaped.

The two pale and shaken boys hurried to the safety of
their home, where they told of their frightening experi-
ence. Even as they sat, warm and comfortable, in the
kitchen realizing they had really escaped was hard.


Streams flow through valleys. Most valleys are wider than
the streams. We conclude something. Factors operate in
the formation of the valley. There are usually factors
other than the wearing-down action of the stream. Suppose
a valley were formed only by the action of the stream.
The valley would be no wider than the stream. Its sides
would be vertical.

Since most valleys are wider than the streams that flow
through them, we can conclude that there are usually
factors other than the wearing-down action of the stream
that operate in the formation of the valley. If a valley
were formed only by the action of the stream, the valley
would be no wider than the stream and its sides would be
vertical.


Questions on passages 2, 5, 8, 11, 12, and 14 appeared as
follows in the pilot studies:


Question on
passage 2


According to this paragraph, food should be
cooked completely because


it may be cold.
it will be more attractive.
that will destroy bacteria.
that will make it taste better.


A B C D
( ) ( ) ( ) ( )


(13)









Second ques-
tion on
passage 5


The trap that the hunters set was

A. visible.
B. camouflaged.
C. square.
D. deep.


A B C D
( ) ( ) ( ) ( )


Question on
passage 8


If factories and markets are close,

A. products can be sold sooner.
B. there will be less chance of damage
during shipping.
C. it will not be necessary to advertise
in out-of-town newspapers.
D. manufacturers can save money.


A B C D
( ) ( ) ( ) ( )


First ques-
tion on
passage 11


Rural people are often more socially sensitive
because they are


kind to strangers.
raised that way.
isolated from other people.
from large families.


A B C D
( ) ( ) ( ) ( )


First ques-
tion on
passage 12


Puppets are

A. powered by a special kind of battery.
B. life-sized.
C. animated by one of several methods.
D. unattractive.


A B C D
( ) ( ) ( ) ( )




Full Text

PAGE 1

THE EFFECT OF SYNTAX ON READABILITY FOR SPANISH-SPEAKING ADULT STUDENTS OF ENGLISH AS A SECOND LANGUAGE BY EILEEN KAY BLAU A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1980

PAGE 2

ACKNOWLEDGMENTS My sincere appreciation is expressed to Dr. Ruthellen Crews for her guidance, support, encouragement, and cooperation throughout the course of this dissertation. I wish also to express my gratitude to the other members of my committee: Dr. James Algina for his extreme helpfulness during all stages of this investigation, Dr. Jayne C. Harder, especially for her help in the development of the instrument used in the study and her editorial suggestions, and Dr. Clemens L. Hallman for his impetus in the realization of the Bilingual Education Program. I wish to extend my gratitude to the administration and staff of the English Department at the Mayaguez Campus of the University of Puerto Rico who so generously cooperated in the collection of the data for this study. I would also like to thank Dr. Patricia Byrd of the English Language Institute at the University of Florida and Dr. Ellen West of the Alachua County Adult Education Program for their cooperation in the pilot studies. I thank those friends who were so generous with their hospitality and encouragement during the necessary visits to Gainesville this past year. And I thank my husband, Robert W. Smith, for his constant support, encouragement and tolerance. Finally I would like to express my gratefulness for the financial support of the Department of Health, Education and Welfare which made my doctoral studies possible with a Title VII Fpllowship. i i

PAGE 3

TABLE OF CONTENTS PAGE ACKNOWLEDGMENTS ii ABSTRACT iv CHAPTER I INTRODUCTION 1 Statement of Problem Delimitations and Limitations 1 Justification 2 Assumptions 7 Definition of Terms 7 Null Hypotheses 9 Procedures 10 Organization of the Research Report 12 CHAPTER II REVIEW OF RELATED LITERATURE 13 Syntactic Development and Reading Ability 13 Syntax and the Human Mind 17 Conflicting Notions of Syntactic Complexity 23 Syntactic Complexity and Readability 34 Summary 40 Notes " 42 CHAPTER III METHODOLOGY 43 Puerto Rico Study 43 Pilot Studies 50 Subjective Judgments of Relative Difficulty 52 CHAPTER IV PRESENTATION AND ANALYSIS OF THE DATA 54 Puerto Rico Study 54 Pilot Studies 57 Subjective Judgments of Relative Difficulty 61 Discussion 65 CHAPTER V SUMMARY AND CONCLUSIONS 72 Overview 72 Conclusions and Recommendations 76 APPENDIX A INSTRUMENT 80 APPENDIX B COVARIATE AND SDCT SCORES 97 REFERENCES 101 BIOGRAPHICAL SKETCH 108 m

PAGE 4

Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE EFFECT OF SYNTAX ON READABILITY FOR SPANISH-SPEAKING ADULT STUDENTS OF ENGLISH AS A SECOND LANGUAGE By Eileen Kay Blau August, 1980 Chairperson: Dr. Ruthellen Crews Major Department: Curriculum and Instruction The purpose of this study was to determine the effect of manipulating the degree of sentence combining on reading comprehension for mature students of English as a second lanquage (ESL). In this respect the concept of the effect of syntax on readability differed from the usual sentence length criterion. A series of reading passages was developed in three versions. Version 1 consisted of short, primarily simple sentences; version 2 contained complex sentences with as many clues to underlying relationships as possible left intact, and version 3 also contained complex sentences but without those surface clues to underlying relationships. Vocabulary and content were held constant across the three versions. After several pilot studies, the Syntactical ly Different Comprehension Test (SDCT) was administered to 85 randomly selected undergraduates enrolled in English 001 at the Mayaguez Campus of the University of Puerto Rico. They were randomly assigned to one of three groups. i v

PAGE 5

Each group read one of the versions of the SDCT and responded to the same multiple choice comprehension questions. The data were analyzed by analysis of covariance using the English as a Second Language Achievement Test (ESLAT) as a covariate. Although the differences in comprehension were not significant, version 2 yielded higher scores than version 1 or version 3 in the Puerto Rico study as well as in the pilot studies. A secondary study was carried out to obtain the subjective judgments of ESL students, ESL teachers, and pre-service teachers with regard to the relative difficulty of the three syntactic versions. Students tended to consider version 1 the most difficult while ESL teachers and pre-service teachers tended to judge version 1 as the easiest for students. It was concluded that lower readability level material, as measured by common readability formulas, does not facilitate the comprehension of mature Spanish-speaking ESL studentsT\ Although none of the three versions was significantly better than the others, all results of this study suggest that version 2 may in fact be more readable than the others. The short, primarily simple sentences characteristic of low readability level material may actually be an obstacle to comprehension. Consequently, such material is not recommended for this population. It was recommended that the study be replicated with native speakers of languages other than Spanish as well as with younger ESL students. A series of finer-grained studies was recommended. The need to investigate factors other than syntax that might affect readability was recognized. v

PAGE 6

CHAPTER I INTRODUCTION Because educational achievement, to a great degree, is dependent on reading, there is great concern among educators about the measurement of readability of materials. There are several readability formulas that can be used to find the appropriate grade level of written material. These formulas consider two characteristics of printed material: word difficulty and sentence length. The effect of syntax on readability is thus treated superficially and inadequately. Statement of the Problem The problem under investigation in this study was to determine whether readability would be affected by manipulating the degree and type of sentence combining or "chunking" in reading material prepared for mature students of English as a second language (ESL) who are literate in Spanish, their native language, and who are at an intermediate level in their study of English. The study departed from the usual sentence length criterion of the most commonly used readability formulas in its consideration of the effect of syntax on readability. Del [ml taj Ions and i Invj tatl on; This study was confined to 85 underqraduate students enrolled in English 001 at the Mayaguez Campus of the University of Puerto Rico. All were native speakers of Spanish. They were at an intermediate 1

PAGE 7

level in their study of English having scored between 400 and 599 on the English as a Second Language Achievement Test (ESLAT), a grammar and reading comprehension test administered by the College Board of Puerto Rico to high school seniors. In this study an effort was made to isolate only some of the syntactic factors that might influence readability. The purpose of the study was not to negate the effects of vocabulary and semantics on readability nor to deny the possibility of an interaction between syntax and semantics, or any other factors that might influence readabi 1 ity. The study relied on a researcher-developed instrument consisting of 18 short passages, each followed by at least one multiple choice comprehension question. The questions were limited to literal comprehension, which was considered a necessary prerequisite to other types of comprehension. The extent to which the findings of this study can be generalized is limited to a population similar to that from which the sample was drawn. Justification This study attempted to fill two voids. The first is the lack of attention paid to syntax in measures of readability. The second is the paucity of research in readability that deals with the population under study. Mery little research has been reported on the effect of syntax on readability for native speakers of English and virtually none for the large population of students learning English as a second or

PAGE 8

foreign language both in the United States and in the non-Englishspeaking world. Indeed, syntax does not receive much attention in the measurement of readability beyond the crude sentence length criterion. Yet one might expect the grammatical organization of sentences to play an important part in determining how readable a passage is. A reader may have all the necessary cognitive concepts for interpreting the separate words in sentences and paragraphs, but if sentence patterns are foreign to him ... or restrictive clauses unclear, meaning will not be apparent. He must be able to understand the underlying meaning of "deep structure" of what Chomsky . . . calls the "surface" of print: how words are placed in varying ways to produce meaning through their order in sentences, or syntax. (Ransom, 1978, p. 282) The reader must be able to "unpack" the surface structure (Larkin, 1977). Surely a great deal of meaning is revealed by the arrangement of words alone. A popular example of what has been termed "grammatical meaning" is taken from Alice in Wonderland. 'Twas brillig and the si i thy toves Did gyre and gimble in the wabe; All mimsy were the borogroves, And the mome raths outgrabe. (Dodgson, 1960, p. 85) Moreover, with regard to reading, the phrase is the critical linguistic unit, worked on by mature readers. Gibson and Levin (1975) report that Deople who read word by word are helped if words are grouped into phrases (pp. 832-834). They also note that eye-voice spans end at phrase boundaries to an extent well beyond the chance level. Despite its potential importance, thouqh, syntax does not receive the systematic attention it deserves in the development and

PAGE 9

selection of reading material either for native speakers (Blossom, 1969; Kachuck, 1975) or for ESL students (Gaies, 1979; Zukowsky/Faust, 1978). This is due in part to current methods of determining readability. The most commonly used readability formulas consider only two variables: words and sentence lenqth. Although some readability formulas do look at syntax more carefully than those that merely consider sentence length (Botel, Dawkins, and Granowsky, 1973; Klare, 1974-75), these are more difficult to apply and are therefore not commonly used. One must remember, too, that a readability formula is merely intended to be an index of difficulty; it does not indicate how to write readably. Yet such a formula determines the mold into which most publishers force their writers. The cloze procedure is yet another device that is useful in determining whether students can handle reading material (Anderson, 1971; Haskell, 1978; Owens, 1971). It also is merely an indication of whether or not existing material is too difficult. It does not in any way prescribe how to write readably. Despite the inadequacies of readability formulas in general, there is, nevertheless, a tendency to apply the same criteria to reading material for ESL students as those used in assessing materials for native English-speaking students. Thorn's (1970, p. 172) suggests material with lower readability ratings for such students. This is a quantitative rather than a qualitative suggestion, based on the assumption that readability formulas are in fact accurate indices applicable across populations.

PAGE 10

Even Zukowsky/ Faust (1978), whose concern is specifically with the ESL/bilingual education situation, is bound to readability formulas, albeit in a somewhat broader sense. Aware of the "hybridized set of criteria" (p. 2) used in the ESL/bilingual education setting (native speaker criteria plus an intuitive sense on the part of more experienced teachers of what will be difficult for one's students), she judges readability perception by teachers in terms of agreement with traditional readability formulas. She does realize, though, that an ESL readability formula would require a different basic word list and a more precise means of assessing sentence structure. In view of the lack of an ESL readability formula, are specially prepared ESL readers more successful in meeting the needs of limited English speaking (LES) and foreign students? ESL writers may be doing an excellent job, but if they do, it is by chance-without any research-based foundation. The progression from simple to complex syntax in ESL readers is neither universal nor systematic (Gaies, 1979). What is syntactically simple or syntactically complex anyway? Those who adhere to the sentence length criterion of readability formulas believe that short, simple sentences are simplest, in other words, easiest to comprehend. But this may actually bo misleading. There are others who believe that grammatical complexity may in fact be an aid to comprehension rather than a hindrance (Betts, 1977; Dawkins, 1975; Pearson, 1974-75). This may be particularly true for cognitively mature ESL students who are literate in their own lanquaqe. They may be "unsophisticated in English but not linguistically

PAGE 11

unsophisticated" (Shook, 1977, p. 2). Shook also believes that our capacity to deal with unintegrated information, characteristic of low readability level material, is extremely limited. Our tendency to integrate it is supported by the research of Bransford and Franks (1971). Shook continues by saying that "simple beads-on-a-string sentences . . . are like nothing the student will ever encounter in the real world" (p. 6). Such "series of isolated propositions [are] stacked together like bricks without mortar" (p. 11). The devices used to join simple propositions into more complex sentences, i.e., the_deyices of chunking, may actually clarify relationships and meaning even though they create longer sentences and thus higher readability levels. Again, there is a lack of research to support this notion about sentence complexity. Schlesinger (1968), in his monograph on sentence structure and the reading process, advises that future research concentrate on systematic exploration of variables by experimental manipulation rather than dwell on analysis of texts in the effort to devise bigger and better formulas (p. 149). Within the limitations described above, this researcher tried to follow Schlesinger' s advice and Shook's hypothesis in an effort to begin filling the two voids mentioned above. It was hoped that the results of this study would ultimately help bilingual education and ESL programs better meet the special needs of their students without depending so heavily on the intuition of educators.

PAGE 12

Assumptions The assumptions basic to this study were as follows: 1. The subjects would be able to read the passages. 2. The passages were appropriate for purposes of measuring reading comprehension. 3. Comprehension could be adequately measured for purposes of this study. 4. Any differences in scores among the three groups would result only from the differences in sentence structure among the three versions of reading passages. Definition of Terms The following terms are defined in relation to their use in this study. Chunking . Chunking refers to the combining of single underlying sentences to form grammatically complex sentences. Examples are as follows: (a) Birds migrate. They arrive in the North in early spring. (b) Birds that migrate arrive in the North in early spring. (a) Some animals have a "weather sense." The weather governs their movements and their \/ery existence. Man has never been able to equal this. (b) Animals whose movements and very existence are governed by the weather have a "weather sense" never equalled by man. The (b) sentences are considered more compressed or chunked than the (a) sentences in which each underlying sentence is expressed in
PAGE 13

8 Complex sentence . A complex sentence is a sentence with more than one underlying subject and one underlying verb. Example: Voltaire, who was one of the greatest writers of the eighteenth century, said that the two essentials of style are precision and color. Intermediate level . ESL students at the intermediate level scored between 400 and 599 on the ESLAT, a grammar and reading comprehension test of English as a second language that is administered by the College Board of Puerto Rico. Native language . Native language refers to the language spoken in the students' childhood homes, i.e., the language their parents used and thus the first language the students acquired. It would be the language with which they feel most comfortable. A synonym of this term is mother tongue . People whose native language is Spanish, for example, are referred to as native speakers of Spanish. For purposes of this study, subjects from homes in which two or more languages were acquired simultaneously during childhood were el iminated. Readabil ity . Readability is a function of a combination of attributes of written language that determine its comprehensibility. Reading comprehension . In the context of this study reading comprehension refers to the meaning obtained from a written text as measured by performance on multiple choice questions based on the literal meaning of the passages. Semantics . In the context of this study semantics refers to meaning.

PAGE 14

Simple sentence . A simple sentence is generally a sentence with only one underlying subject and only one underlying verb. Several exceptions are made for the purposes of this study. Prenominal adjectives are not considered separate underlying sentences. Three types of complex sentences occasionally appear in version 1 passages. 1. a. The puppets seem to become real people, b. The plant is equipped to retain moisture. 2. Most valleys are wider than the streams. 3. Suppose the manufacturer and the market are a long distance apart. Syntax . Syntax refers to the grammatical organization of sentences. Null Hypotheses The following null hypotheses were investigated in the principal study: 1. There is no significant difference between comprehension of version 1 and comprehension of version 2 passages. 2. There is no significant difference between comprehension of version 3 and comprehension of version 2 passages. In addition to the principal study, a subjective judgment exercise was conducted to test the following null hypotheses: 3. Judgments by ESL students of relative difficulty of passages is independent of syntactic version. 4. Judgments by ESL teachers of relative difficulty of passages is independent of syntactic version. 5. Judgments by pre-service teachers of relative difficulty of passages is independent of syntactic version.

PAGE 15

10 Procedures Instrumentation Development of Syntactically Different Comprehension Test (SDCT) The 18 passages in the SDCT were adapted from passages from the Science Research Associates' Reading for Understanding General reading kit (Thurstone, 1969). While holding content and vocabulary constant, three versions of each passage were constructed. The versions differ from one another only in syntax, or more precisely, in degree of chunking. Version 1 passages are unchunked and contain only simple sentences except for those noted above under the definition of simple sentence. Version 2 passages contain complex sentences and thus have fewer and longer sentences than their version 1 counterparts. Version 3 passages also contain complex sentences, but they reflect a higher degree of chunking than their version 2 counterparts. At least one multiple choice question follows each short passage. There are a total of 24 literal comprehension questions which are identical for the three versions of the test. The SDCT was administered to 12 randomly selected English Language Institute (ELI) students at the University of Florida. All were native speakers of Spanish. All had scored between 35 and 85 on the Institute's Reading and Vocabulary Test. The pilot study adminq istration of the SDCT yielded a Kuder-Richardson 20 coefficient of .82. Another randomly selected group of 15 ELI students took the SDCT with the comprehension questions in Spanish. That administration yielded a Kuder-Richardson 20 coefficient of .75. Finally, a 'group of 17 adult

PAGE 16

11 education students who were native speakers of English took the SDCT. That administration yielded a Kuder-Richardson coefficient of .81. The purpose of the pilot studies was to determine if this study was feasible and if the instrument was useful. Six of the 24 items and two of the passages were slightly altered before the administration of the instrument in Puerto Rico (see the end of Appendix A). ESLAT . The ESLAT, which was used as a covariate to measure proficiency in English, is a grammar and reading comprehension test of English as a second language. It is administered by the College Board to high school seniors in Puerto Rico. Subjects A sample of 85 students was randomly selected from a population of undergraduate (mostly freshmen) ESL students enrolled in the University of Puerto Rico's English 001 course. They were native speakers of Spanish, literate in their mother tongue, and raised in monolingual homes. One third were randomly assigned to each of three groups. One group read the passages in version 1; one group read version 2 passages, and the third group read version 3 passages. All were given the same comprehension questions. Design and Data Analysis The design of the principal study included three treatment groups. The ESLAT test was used as a covariate, and analysis of covariance was used to analyze the data. The model was Y b n + b.C + b 9 V, + b,V 9 + b.V, o 2 '.4 3

PAGE 17

12 where Y was the SDCT score, C the score on the covariate, and V the version of the reading passages. Secondary Study—Subjective Judgments A secondary study was carried out which sought to reveal the nature of subjective judgments of the relative difficulty of the three syntactic versions. Seventy-nine students from the same population as the principal study were asked to choose which of the three versions they thought was the easiest and which they thought was the most difficult. Three sets of passages were chosen for this exercise. Seven of their teachers made the same judgment from the point of view of what they thought would be easiest and most difficult for their students. And 14 pre-service teachers at the University of Florida made the same judgment from the point of view of relative difficulty for students of limited English proficiency. Organization of the Research Report Chapter II is a Review of Related Literature. Chapter III describes the methodology used in the study while the data are presented, analyzed, and discussed in Chapter IV. Chapter V includes a summary, conclusions, and recommendations for further research.

PAGE 18

CHAPTER II REVIEW OF RELATED LITERATURE Chapter II reviews literature and research in four areas that are related to the topic under investigation: (a) the relation of syntactic development to reading ability, (b) syntax and the human mind, (c) conflicting notions of syntactic complexity, and (d) syntactic complexity and readability. Though most of this research has been done with native speakers of English as subjects, it can, nevertheless, shed light on the area of reading for students learning English as a second language (ESL) and thus serve as a basis for further research focusing specifically on ESL students. Syntactic Development and Reading Ability A number of studies have been done that positively relate one's grasp of syntax, both productive and receptive, to reading ability, hence demonstrating the relationship between this component of language and reading. Using fourth grade students, Ruddell (1965a, 1965b) found support for the hypothesis that reading comprehension is a function of the similarity of the language structure of the written passages and the language structure used orally by children (1965a, p. -104 ) . One version of each of three passaqes was written with patterns of high frequency in the oral language of children, and one version was written with low frequency patterns. Comprehension, measured by the cloze technique in which every fifth word was deleted, was superior 13

PAGE 19

14 for passages containing structures frequently used in the oral language of fourth graders. Tatham (1969-70) measured the comprehension of second and fourth graders by having them choose the one of three similar pictures that best represented the sentence content. Results indicated better comprehension of sentences that contained oral language patterns frequently used by the students. Tatham concluded that "control over vocabulary is not the only logical and desirable control when comprehension of language structures is essential" (p. 424). Her findings supported the fact that ability to decode and knowledge of vocabulary alone are not sufficient for successful reading; familiarity with and control of the language patterns are also important. In a study with over 240 fourth graders in the semi-rural upper midwest, Bormuth, Carr, Manning, and Pearson (1970) used four question types to test comprehension of various sentence structures, anaphoric devices, and intersentential relationships. They discovered that large proportions of their subjects were unable to demonstrate comprehension of many of the basic syntactic structures by which information is signaled. Control of syntactic structures is a developmental matter. Carol Chomsky (1969, 1972) demonstrated that children between the ages of five and ten were at various states of development and did not yet have complete mastery of their language. Mavrogenes (1977) replicated Chomsky's study with working class 14and 15-year-olds and found that 17 our of 20 of her subjects had not mastered all five stages of development with which Chomsky dealt.

PAGE 20

15 Chomsky (1972) found a positive relationship between children's stages of linguistic development as determined by her testing of their knowledge of five different structures and the following reading measures: scores on the Huck Inventory of Children's Literary Background, the number of books at the top level of syntactic complexity on the Master Book List with which a child is familiar, and scores based on interviews with both parents and children regarding such matters as time spent reading and being read to and frequency of trips to the library. Van Metre (1972) replicated Chomsky's (1969) study with a different population. She used two groups of bilingual third graders in the Tucson, Arizona, schools. One group included students who obtained high scores on a state mandated reading test; the other group obtained low scores on the same standardized test. Van Metre compared the two groups on mastery of the structures tested by Chomsky. She also used two monolingual comparison groups, one group each of high and low scorers on the same reading test. Major differences in mastery of the syntactic structures were between high and low groups rather than between bilingual and monolingual groups. Van Metre also found that the order of acquisition of the structures was less consistent among those scoring low on the reading test. Her findings not only lent support to the notion that there are indeed close ties between syntactic development and reading ability; they extended this relationship to a bilingual population. A study by Ribovich (1976) was designed in part to determine the nature of the relationship between comprehension of syntactic

PAGE 21

16 structures in oral language, as measured by a researcher-developed instrument, and reading comprehension of first grade native Englishspeaking children of low, mid and high socioeconomic status. Reading comprehension was measured by the reading subtest of the Metropolitan Achievement Test. Ribovich looked at this relationship both with and without controlling for intelligence. She found that comprehension of selected oral syntactic structures did relate significantly to reading comprehension. Twenty percent of the variance in reading comprehension was explained by the variance in comprehension of selected oral syntactic structures. She also found, however, that the relationship between intelligence, as measured by the California Short-Form Test of Mental Maturity, and reading comprehension was even stronger. Simons 1 (1970) research with 87 fifth graders of above-average intelligence aimed to relate skill at recovering deep structure to reading comprehension. The first variable was measured by the researcher-developed Deep Structure Recovery Test (DSRT), a multiple choice instrument in which subjects chose the sentence with the same meaning as the given sentence. Reading comprehension was measured both by a cloze test and by the reading subtest of the Metropolitan Achievement Test (MAT). The correlation between the DSRT scores and cloze test scores was .732 while the correlation between the DSRT and MAT scores was .476. Simons concluded that the MAT measured many other things in addition to reading comprehension. He therefore considered the correlation between the DSRT and the cloze scores more meaningful .

PAGE 22

17 Nadler (1968) conducted a study with university foreign students beginning an intensive English course. Reading material for the experimental group included only previously studied syntactic structures while reading material for the control group was not syntactically controlled. Overall, the experimental group made greater gains than the control group from vocabulary pre-tests to re-tests. A similar study with reading comprehension instead of acquisition of lexical items as dependent variable would also be of interest. The above studies support the belief that knowledge of syntax is positively related to reading ability, which suqgests, in turn, that the syntax of reading material might well affect the readability of that material for readers at different levels of syntactic development. Syntax and the Human Mind Some psychol inguistic research has been carried out in an attempt to understand how the human mind processes different types of sentences, which could indicate which syntactic characteristics are harder to process. Unless otherwise stated, the subjects in all of the studies reviewed in this section were college or university students who were native speakers of English. A number of studies (Bever, 1970; Epstein, 1961, 1967; Fodor and Bever, 1965; Miller and Isard, 1964) indicated that human perception and learning of sentences is in terms of syntactic constituents rather than individual words, attesting to the psychological reality of the phrase structure of surface forms.

PAGE 23

18 Other studies have attempted to relate human mental processes to transformational-generative grammar and were based on the assumption that comprehension of a sentence involves recovering its underlying structure by means of a device that corresponds to a transformational generative grammar--despi te the fact that transformational grammarians make no claims that their models of language reflect human processing devices. Studies such as Miller (1962) and Miller and McKean (1964) suggested that transformations such as negative, passive, interrogative and combinations of these take time for human beings to perform. Performance time was measured by matching a passive sentence, for example, with its negative-passive counterpart. A study by Mehler (1963) suggested that kernel, or simple, active, declarative sentences, were more easily recalled than transformed sentences. Savin and Perchonock (1965) claimed that kernel sentences took up less space in memory than transformed sentences. Space in memory was measured by the number of words from a list following the sentence that were recalled: fewer words remembered from the list indicated more space taken up in memory by the sentence itself. This group of studies suggested that people store a kernel plus a separate notation as to which transformations go with it. They implied that the greater the number of transformations the greater the complexity. Gough (1965) examined the effect of transformations on speed of understanding by having subjects verify the truth or falsity of a sentence based on a picture presented at the end of the sentence.

PAGE 24

19 Actives were verified faster than passives which, in turn, were verified faster than negative actives. The latter were verified faster than negative passives. There was, however, an interaction between truth value and the negative-affirmative dimension reflecting the operation of a semantic factor. True sentences were verified faster than false ones. In a follow-up study, Gough (1966) attempted to isolate verification time by allowing a three second delay between the end of the sentence and the presentation of the picture. Mean verification times were smaller with the delay, but not significantly smaller. Differences among verification times for the different transformations paralleled Gough's earlier findings. Consequently, Gough raised several questions. Perhaps these results cast doubt on the hypothesis that a transformed sentence is reduced to underlying structure at the moment it is read or heard. It may be de-transformed when put into long term storage, but this may not occur at the time of initial decoding. Without the opportunity to verify the sentence right away, the subject may hold it in immediate memory postponing the de-transformation process until presentation of the picture which is the evidence on which to base the verification. Gouqh thus concluded . . . that a transformational description of sentence complexity has not at all been discredited, for verification time is a function of a number of transformations even afti r delay of evidence. But a hypothesis of a transformational • -..-.• in decoding is less defensible. ... (p. 494)

PAGE 25

20 Sentence length did not seem to be a factor in processing time according to Gough's findings. Active sentences, although longer (passives and negative passives were shortened by deleting agent phrases), were still verified faster. These researchers assumed that human language processing should parallel a particular linguistic model. But should it? There are problems both with their oversimplified assumption, as Gough's findings indicated, and with the nature of the experiments themselves. Greene (1972) noted that a basic problem with experiments of the type described above is a lack of realism or naturalness. This problem is recognized by some of the experimenters themselves (Miller and McKean, 1964). Even if the speed required for certain transformations or de-transformations can be measured and contrasted, the measurement in such research occurs under contrived conditions and is no indication that such processes are what actually take place when people normally encode and decode sentences. When subjects are performing the more natural language function of extracting meaning from sentences, any exact one-to-one correspondence between transformational complexity and performance disappears. (Greene, 1972, p. 114) Other factors undoubtedly operate at the same time. Perceptual mechanisms and conceptual structure interact with language processing (Bever, 1970), to say nothing of the interaction between syntax and semantics. In fact this interaction has been a topic of considerable linguistic research. Greene (1970) found that subjects handled a syntactic form more easily when it fulfilled its natural semantic function. Slobin (1966) demonstrated the role of reversibility in

PAGE 26

21 processing passive sentences. Franks and Bransford (1974) have demonstrated the role of extra-sentential context in dealing with passive sentences. Glucksberg, Trabasso, and Wald (1973) mentioned selectional restrictions on verbs and whether or not both of two nouns appearing in a sentence could act as agents in processing active and passive sentences. In an attempt to put syntax in proper perspective, Greene (1972, p. 191) stated . . . the only purpose of syntactic rules is to express semantic relationships, the meaning of sentences being determined by the syntactic relations holding between individual words. Such a statement attests to the importance of syntax at the same time that it recognizes its inseparability from semantics or content. Syntax provides the structural information necessary for arriving at a semantic reading of a sentence (p. 139), but it is only one tool used by the human mind in processing language, a fact that was overlooked in the earlier research reviewed above. In addition to neglected pragmatic considerations, a limitation of the earlier experiments, aside from their unnaturalness, was that they dealt only with a limited class of grammatical operations in the context of isolated sentences. Among the limited sentence types there are enormous discrepancies in frequency of use which may distort the findings of these studies. Goldman-Eisler and Cohen (1970) analyzed speech samples of approximately equal length (about 100 clauses each) by speakers of various levels of intelligence and mental health. They found that

PAGE 27

22 simple active affirmative declarative, or kernel, sentences comprised 80-90* of the speech samples. Negatives were a distant second at 4-10%; passives accounted for .7-10% while passive negatives were practically non-occurring. The use of different sentence types not based on the same universe of behavior may account for differences observed in experimental settings. Another problem is that the results of the experiments could be affected by the task employed to measure comprehension or relative ease or difficulty in linguistic processing. A variety of tasks has been used as dependent variables: recall, recognition, transformation/ de-transformation, verification. Furthermore, the tasks typical of psycholinguistic experiments may not be so appropriately related to the reading process as one might expect. Shore (1976) found no significant differences between the ability of good and poor readers in performance on the Linguistic Processing Test (LPT). Her sample consisted of 40 Caucasian adult education students who were reading at least on the seventh grade level. The LPT is a recall test in which seven sentence types are followed by strings of digits. Although there were significant differences among sentence types for this sample, performance on the LPT did not relate to subjects' reading ability. Even recognizing the problems involved in analyzing syntactic processing—simultaneous operations, dependence on type of task, appropriateness of task--the basic premise that additional transformations create additional complexity is still questionable. The

PAGE 28

23 research reviewed in this section that does support that premise reveals an overemphasis on singulary transformations. Conflicting Notions of Syntactic Complexity This section examines several notions of what constitutes syntactic complexity. Many of the studies reviewed in the previous section were inspired by what is sometimes called the Derivational Theory of Complexity (DTC), which implies that the greater the number of transformational rules employed in the derivation of a sentence, the greater the complexity of that sentence. Thus kernel sentences are the simplest. The language user, according to the theory, follows a path that corresponds to the derivation of a sentence in the manner of transformational -generative grammar. But, as was noted in the previous section, it is not at all certain that this is what occurs. Fodor and Garrett (1967) note . . . it is far from obvious that one ought to expect a correspondence between the complexity of the • ' operation of converting a base structure into a surface structure and the complexity of the perc operation of converting a surface structure into a base structure, (p. 289) They go on to suggest . . . that the complexity of a sentence is a function not (or not only) of the transformational distance from its base structure to its surface structure but also of the degree to which the arrangement of elements in the surface structure provides clues to the relations of elements in the deep structure, (p. 290) Fodor and Garrett do agree that transformations often mask underlying structure by eliminating surface clues. But at the same time they pro vide examples of counterintuitive predictions arising from the DTC.

PAGE 29

24 According to the DTC the (a) sentences below should be more complex than the (b) sentences. (a) The small cat is on the grass mat. (b) The cat that is small is on the mat that is made of grass. (a) It amazed Bill that John left the party angrily. (b) That John left the party angrily amazed Bill. (a) The operator looked the address up. (b) The operator looked up the address. (a) Fred runs faster than the girl. (b) Fred runs faster than the girl runs. In line with the hypothesis that the presence or absence of surface structure clues to deep structure is the key to simplicity or complexity with regard to comprehension, Fodor and Garrett chose relative pronouns as an example of a surface structure clue to deep structure relationships that lent itself to examination. Two groups of MIT students listened to a set of nine sentences with two embeddings apiece. For group one the relative pronouns were deleted; for group two they were retained. Example: The Den (which) the author (whom) the editor liked used was new. Subjects were asked to restate each sentence in their own words. Scoring was based on the figure obtained when the number of correct subject-object relations reported was divided by response delay. Group two did significantly better than group one. Four variations of the experiment were carried out in order to consider alternative explanations, but results were always the same. In one of these experiments adjectives were added before each of the first two noun phrases of sentences in which relative pronouns were retained in order

PAGE 30

25 to increase the derivational history. Despite the longer derivational history, group two still outperformed group one. Surface clues, such as retained relative pronouns, apparently outweighed the DTC. A study done by Charrow and Charrow (1979) and reported in the Linguistic Reporter ("You must follow the law as I state it," 1978) revealed that one of the characteristics rendering instructions to juries difficult to comprehend was deletion of relative pronouns followed by a form of the verb be. This supported the findings of Fodor and Garrett. Hakes (1972) found supporting evidence for the negative effect on sentence comprehension of reducing tnat-clause complement constructions by deleting the optional that, a clue to underlying relations. He believed there would be more local ambiguity in sentences with deleted complementizers. Hakes used 40 freshmen native speakers of English as subjects and sentences such as, The world-famous physicist forgot (that) his old professor had been the first to suggest the crucial experiment to him. Two tasks were used to assess comprehension of sentences with and without the optional complementizer. The first was a phoneme monitoring task in which subjects were instructed to listen for a target phoneme and press a button upon hearing it. The second measure of comprehension was a paraphrase task. Phoneme monitoring was indeed faster when the complementizer was present. The difference between the presence or absence of the complementizer was not significant for the paraphrase task. In a second experiment, however, in which supposedly more difficult sentences were used, subjects performed significantly better on both tasks for

PAGE 31

26 sentences in which the complementizer was present. Hakes concluded that the paraphrasing task may be sensitive only when the overall level of difficulty of the sentences is high. He considered his results evidence that presence of the complementizer does facilitate comprehension. Fodor, Garrett, and Bever (1968) examined the effect of verb type as another clue employed by language users to determine underlying structure. If the main verb of a sentence could appear in a variety of deep structure configurations, the sentence was more complex, harder to process, than one in which the main verb was less ambiguous. Pairs of sentences were constructed that differed only in verb type. One sentence had a verb that was transitive only: Example: The letter the secretary the manager employed mailed was late. The other sentence in the pair had a verb that could take a variety of complement types: Example: The letter the secretary the manager employed expected was late. The verb expect could be followed by a thatclause or an infinitive, i.e., a sentential complement, as well as a simple direct object. The verb mail , on the other hand, could be followed only by a simple direct object. Subjects (MIT students) heard and were asked to paraphrase both types of sentences. Exact words were not permitted in order to avoid rote repetition. Scores were based on the number of correct subject-verb-object triples reported in paraphrases. Subjects did significantly better on sentences with simple transitive verbs whether

PAGE 32

27 the sentences were read or heard. Similar results were reported for versions in which subjects had to unscramble the sentences. Holmes and Forster (1972) conducted a study with 40 Australian university students who had been speaking English for at least ten years. Written recall of four different types of sentences was examined. There was no significant difference between recall of oneclause and two-clause sentences, contradicting the notion that complex sentences are more difficult than simple sentences. Holmes and Forster also found that verb type in one-clause sentences affected recall. Those that could take a variety of complement types were more difficult to recall than simple transitive verbs confirming the findings of Fodor, Garrett and Bever (1968). Rohrman (1970) observed a significant difference in recall between intransitive forms such as growling lions (lions growl) and transitive forms such as digging holes (PRO [dig holes]). The intransitive forms were easier to recall. Rohrman hypothesized that they constituted smaller chunks which were thus easier to store despite the fact that the surface forms of both types appear the same. Verb type itself may indeed operate as some sort of lexical clue to underlying structure. Carol Chomsky's (1969) concept of complexity is that the exceptional structures are more difficult. This, too, is usually a question of arriving at the appropriate deep structure when surface structures look alike. This is true in the case of the toll where, although doll is in subject position, someone else sees it. In a structure such as , however,

PAGE 33

28 the surface subject of the sentence is the eager one. Recovery of the correct deep structure is again the issue in ask as a question versus ask as a request. Ask as a question, as in John asked Bill what to do, violates the Principle of Minimal Distance (PMD): John is the subject of do rather than the closest noun phrase. Ask as a request, on the other hand, does not violate the PMD and is therefore easier. In the sentence John asked Bill to shovel the driveway, Bill is the one who will do the shoveling. The verb promise is difficult because its use also violates the PMD. In John promised Bill to shovel the driveway, John will do the shoveling. Because surface structure clues do not follow the usual pattern, violations of the PMD and structures such as easy to see are indeed exceptional and therefore more difficult. Another way of conceiving of sentence complexity that has received some attention is based on Yngve (1960). According to his model, sentence depth is the factor that creates difficulty. Depth is defined as the amount of temporary memory needed to produce a given sentence. Regressive structures—those that branch to the left, as in the sentence, TJiis is the malt that the rat that the cat that the dog worried killed ate, require more temporary storage space as they get longer. Progressive structures, those that branch to the right, on the other hand, do not require more than a minimum of temporary storage space as they get longer as in the sentence, This is the dog that worried the cat that "kitted the rat that ate the malt that lay in the house that Jack built. Because of limitations on the immediate memory of human beings, Yngve hypothesized that languages tend to limit

PAGE 34

29 regressive constructions and include alternative structures of lesser depth that maintain expressive power. Passivization can perform this function. Schlesinger (1968) gives the example of John whom June whom Paul prefers detests loves Mary (p. 83) which can be expressed as Mary is loved by John, who is detested by June, who is preferred by Paul (p. 85). Wearing (1970) found that on a multiple choice recognition test 120 undergraduate male subjects recognized low depth sentences more easily than sentences of greater depth. Whether the sentences were active or passive made no difference. Martin and Roberts (1966) supported the depth hypothesis as opposed to the DTC. They used sentence recall as the dependent variable. They claimed, in addition, that the findings of Mehler (1963) could be explained in terms of the depth hypothesis rather than the DTC. They called into question Miller and McKean (1964) on the same grounds. Wright (1969) conducted two studies of the depth hypothesis in which she attempted to compare the relative effects of depth and transformational complexity as predicted by the DTC on the amount of memory space taken up. The first study supported the depth hypothesis, and the second lent some support to the DTC. Other findings have also indicated inconsistent predictive success of the depth hypothesis. Prenominal qualifiers constituting regressive structure should, according to the depth hypothesis, be difficult. Matthews' (1968) subjects found them more of an obstacle

PAGE 35

30 than transformational "complexity." Qualifiers were also a source of difficulty for ESL students in sentence repetition tasks (Stieglitz, 1973). Yet Schwartz, Sparkman, and Deese's (1970) subjects, graduate students in psychology and native speakers of English, judged sentences such as T)ie electric powered toe chomping rock throwing lawn mower ran over its own cord to be highly comprehensible regardless of length. Sever (1970, p. 339) also doubted that this type of left-branching constitutes perceptual complexity. He supposed that information provided early in a sentence, i.e., prenominal qualifiers, should make prediction of the latter part of the sentence easier. Semantic aspects may also play a role in the comprehensibil ity of these structures. For the subjects in Schwartz, Sparkman and Deese's study, who found left-branching easy to comprehend, the comprehensibil ity of center-embedded and right-branching sentences declined rapidly as clauses were added. Hamilton and Deese (1971), however, found that for their 30 subjects the important issue was contiguity of subject and verb rather than the number of clauses alone. As noted in the previous section, experimental results are, at least to some extent, dependent on the task required of the subjects. In addition, there are numerous ways of viewing complexity that yield findings which fail to provide clear-cut patterns. Wang (1970) attempted to ascertain the role of syntactic complexity in determining judgments of comprehensibil ity of a large number of sentence types. Eignt potentially relevant measures of surface structure

PAGE 36

31 complexity were computed for each sentence type. (This itself is indicative of the variety of ways to "measure" complexity.) 1. ratio of nodes in a phrase marker to terminal nodes 2. total number of nodes (TNN) 3. total number of terminal nodes (TNTN) or sentence length 4. depth 2 5. mean depth 6. embedding transformations (T-emb) or number of S nodes 7. conjoining transformations (T-conj) 8. self-embedding (S-emb) or recursiveness The sentences were presented auditorily to undergraduate native speakers of English. The two measures that revealed the strongest negative correlation with comprehensibil ity ratings were mean depth and self-embedding. When these were partial led out, the negative correlation between sentence length and comprehensibil ity was reduced implying that it is not sentence length per se that makes comprehension difficult. Another factor that might affect ease of sentence comprehension is the order of main and subordinate clauses. Bever (1970, p. 294) referred to a study by Bever and Weksel that indicated stylistic preference for sentences in which the subordinate clause (marked by a conjunction) followed the main clause. Bever reported that Clark and Clark observed that complex sentences with the subordinate clause first were relatively hard to memorize. Holmes (1973) observed that university students had significantly more difficulty recalling

PAGE 37

32 sentences with preposed adverbial clauses. He also found that, for his sample, subject complements were more difficult than object complements. Unexpectedly, center-embedded relative clauses were easier than right-branching relative clauses, but this was with only one relative clause. Tfezt-clauses as object complements were easier than -ing object complements, but ttot-clauses in subject position were more difficult than -ing subject complements. Apparently different explanations are necessary for different types of complex sentences. The variety of approaches taken in the studies described above indicate that narrow adherence to the DTC would be a mistake. There are many ways to view the question of syntactic complexity/simplicity. One outlook that has not received much attention is, in a way, the opposite side of the coin from the DTC. Miller (1956) suggested that the limitations on our capacity to process information are related to what we can hold in short term memory (STM). STM capacity falls in the range of seven plus or minus two items. In order to stay within this range we tend to chunk material or compress it. Exactly what constitutes a chunk in a given situation for a given person is unclear. Miller gave the example of a telegraph operator packing an increasing number of bits into each chunk he can hold in his mind as he gains experience. By using sentence combining transformations rather than expressing each proposition as a single simple sentence, language users create convenient chunks that enhance their memory capacity. Bransford and Franks (1971) found that their subjects, college students, did in fact tend to integrate separate ideas and actually

PAGE 38

33 believed that the complex sentences they thus created were those that they had heard. Chunking, as Miller has suggested, may be a powerful device for increasing memory capacity, and, as Bransford and Franks have demonstrated, it may also be a natural one for humans to employ. Do people prefer to chunk or to string out each idea in its own simple sentence? As far as the simplicity/complexity issue is concerned, Betts (1977) states that short sentences may interfere with rather than facilitate concept development. Dawkins (1975) notes that short, choppy sentences are not necessarily easier to read. Schlesinger (1968) cites Bar-Hi 1 lei's suggestion that some things can not be expressed in sentences with a low degree of syntactic complexity without a loss in other communicatively important respects. Thus there is a price to be paid for simplification of sentence structure. Schlesinger suggests that complex sentences render more salient the organization of content. He also found that readers' speed and comprehension were not inhibited by nested sentences. But Miller and Isard (1964) did find that oral recall decreased as degree of nesting increased. /Huggins (1977) sees syntax as a tool for adjusting complexity and compactness of a message in the process of helping to transfer meaning from producer to receiver. He agrees with Fodor and Garrett that recoverabil i ty is the key, and surface structure clues such as relative pronouns are indeed helpful. He believes that subjects should prefer the long thin message only if the more compact form strains their processing capacities. But long messages may in turn tax memory capacity. Huggins therefore opts for

PAGE 39

34 a middle road. Some balance that satisfies both processing and memory capacities probably defines the optimum level of syntactic complexity/simplicity for reading material. This line of reasoning matches that of Shook (1977). He sees the issue not so much as a matter of simplicity versus complexity, but rather as the way in which a sentence is complex. (a) Judy loved her mother. Her mother lived in New Jersey. (b) Judy loved her mother, who lived in New Jersey. In (b) the relative clause contributes to rather than hinders processing. Syntax that is too simple may inhibit and slow down the reader as much as syntax that is too complex (Shook, 1977, p. 19). ESL students therefore need to be taught to take advantage of the help offered by sentence combining. Pupils need guide words to help them follow the sequence of ideas and to let them know when transitional thoughts or summary statements are being offered. Words such as first, consequently, thus, therefore, as a result of, according to these conditions, and countless other such expressions which carry the author's reasoning can help the reader see the direction of an argument or a viewpoint. . . . Materials which follow an orderly progression, signaled by words and phrases to guide the reader's thinking, are more likely to be understood. (Thonis, 1970, p. 172) Syntactic Complexity and Readability The research reviewed in this section involves the manipulation of the syntax of reading passages to see the effect this has on some measure of reading comprehension. The question how syntactic structure affects the ease of reading is obviously of the greatest importance to teachers, writers, editors, in short--to anyone interested in written communication. (Schlesinger, 1968, p. 15)

PAGE 40

6 Schmidt (1977) studied the effect of increased complexity on reading comprehension using two versions of a biographical selection about Helen Keller. The complex version was created without altering the vocabulary, by merely preposing prepositional and participial phrases, using gerunds instead of noun structures, and using appositives and passives. Example: She also wrote many books and articles and included an autobiography of her early years, (simple version) Her writing, many books and articles, included an autobiography of early years, (complex version) Cloze tests of one of the versions were given to 84 university students Those who had the less complex version did significantly better. Coleman (1964) also used university students as subjects to compare comprehension on what he considered simplified and unsimplified versions of the same passages. Without altering content morphemes and while holding sentence length nearly constant, Coleman simplified by changing passives to actives, nominal izations to their active verb derivatives {his explanation of the design * he explained the design), and adjectival izations to adjective or adverb forms (he has aoci power * h .-ooially powerful). More multiple choice questions were correctly answered on the simplified versions. In a similar experiment by the same investigator, comprehension was measured by written recall. The simplified versions yielded better recall. In a third experiment, 20 sentences of equal length--ten with nominal izations and ten with active verb equivalents—were shown to subjects for four seconds each. Recall by writing was significantly higher for the sentences with

PAGE 41

36 active verbs. Although more multiple choice questions were correctly answered for the active verb versions, the difference did not reach significance. Simplified sentences were also memorized faster. In a finer-grained study Coleman (1965) aimed to discover more precisely which sentence patterns within the broader categories he had already looked at were more easily comprehended. In the first of a series of experiments he looked at ten types of nominal izations . For six of the ten types the de-transformed active verb versions proved easier to learn. Example: His discussion of the reason for the decision will be appreciated, (nominal ization) If he discusses the reason for the decision, it will be appreciated, (active verb version) Coleman concluded that two explicit clauses could be more easily processed than one more highly chunked clause (p. 334). In the second experiment written recall was superior for active sentences as opposed to passive sentences. In the third experiment Coleman found no significant difference between adjectival izations and de-transformed adjective versions. Example: The complexity of milk fat may be comprehended when the formula is known. (adjectivalization) We may comprehend how complex milk fat is when we know the formula. (de-transformed adjective version) In the last experiment of the series Coleman compared cloze scores on two versions of sentences—one he considered more embedded than the other.

PAGE 42

37 Example: The man who can sell it is Bill, (embedded) Bill is the man who can sell it. (non-embedded) Overall there was a significant difference in favor of the second type of sentence. With regard to content words, there was no significant difference, but with regard to function words, significantly more were correctly supplied in the less embedded sentences. Coleman concluded that, other things equal, writers should choose the more easily comprehended versions of sentences. Coleman (1971) reported, in addition, several correlational findings that indicated an association between difficult-to-read prose and such characteristics as a low proportion of verbs and numerous adjectives and prepositions. The low proportion of verbs could be a result of sentence nominalizations which he had found to be more difficult than their active verb counterparts. Using de-transformations to simplify passages (changing passives to actives, nominalizations to noun + verb, expressing relative clauses in terms of their underlying propositions, and replacing deleted items), Evans (1972) conducted a study with 24 high school seniors reading on seventh, eighth and ninth grade level. Those who read the simplified versions did significantly better on multiple choice questions and on the last three of five cloze tests. Fagan (1971) conducted a study with 440 fourth, fifth and sixth graders. He identified 43 types of transformations found In randomly selected passages from three 4th qrade basal readers. He grouped these into five categories: embedding, conjoining, deletion, simple (singulary), and position shift. Based on different versions of adapted

PAGE 43

38 passages he found that deletions and embeddings in particular rendered information more difficult to process for the population in question. Pearson (1974-1975), in contrast to other studies, supported the chunk model. Middle class suburban third and fourth graders showed a preference for the more cohesive form of syntactic variants of sentences with causal and adjectival relations. Examples: Because John was lazy, he slept all day. John was lazy. He slept all day. The tall man thanked the young woman. The man thanked the woman. He was tall. She was young. In addition, recall of causal relations seemed to be aided by cues such as because and so. Results with regard to reading comprehension, however, were of little value as all forms were correctly understood. Perhaps semantic content was so simple that it masked possible differences of form. Nevertheless, overall results supported the chunk model . Marshall and Glock (1978-1979) tested comprehension by both free and probed recall of 16 versions of two 115 word texts. Each version of a text maintained the same semantic network. Different combinations of four different manipulations gave rise to the 16 versions of each of the passages. First, the dependency system was either present or not. If present, if-tlien relationships were explicitly expressed. If not, one encountered a sentence beginning with swpp , Qsff and a separate sentence for the result clause. Second, the relative system was either present or not. If present, superlative

PAGE 44

39 forms of adjectives were used. If not, the common form of the adjective was used thus providing less information. Third, the main idea appeared either at the beginning of the text or at the end. Finally, an evaluative statement about the topic was placed either at the beginning or at the end of a sentence. Significant differences were unexpectedly found between community college and Ivy League college students in the sample. For community college students, manipulation of the relative system made the greatest difference. The difference was also significant for the Ivy League students. Manipulation of the dependency system also had a strong effect for the community college students, but not for the Ivy League students. The authors concluded that truly fluent readers communicated with the author through discourse by responding to its semantic structure. Less fluent readers, on the other hand, could not infer existence of structures in text base unless they were explicitly referenced in the surface structure. Less fluent readers responded to surface structure, and their recall was negatively affected by lack of expl icitness. Smith (1971) studied a broader spectrum of age groups than the above studies and used a different approach to manipulating the text. He used the output of a study by Hunt (1965) to construct four versions of the same text. The fourth, eighth, twelfth grade and "skilled adult" versions v/ere each typical of the production of the respective age groups. Sentence length was equalized over versions principally by coordinating T-units (main clause plus any subordinate clause or

PAGE 45

40 non-clausal structure attached to or embedded within it) in the fourth grade version. Comprehension of 120 fourth through twelfth graders was measured by means of cloze tests in which every fifth word was deleted beginning with the second, third, fourth, fifth or sixth word of the passage. Fourth graders read the fourth grade writing best while eleventh graders read the fourth grade version with less facility than they read the other three versions. Highly redundant writing seemed to be helpful only to younger readers. Moreover, redundancy was apparently not the only factor used in the predictions called for by the cloze procedure. Smith concluded that perhaps people read best what is syntactically closest to what they produce, provided vocabulary and content are not an obstacle. The productive level may be the best determinant of the most appropriate receptive level which supports the findings of Ruddell (1965a, 1965b) and Tatham (1970). It might be that these findings relate to Shook' s (1977) notion and that construction of the most appropriate reading material for ESL students who are linguistically mature in a non-English language should not totally neglect the linguistic maturity achieved in the native language. Summary Four areas of research and literature dealing with the relationship between syntax and reading have been reviewed. The first indicated a positive association between syntactic development and reading ability. The second group of studies investigated how the human mind handles different types of sentences. Results were

PAGE 46

41 inconclusive and did not follow a single, undeviating line of reasoning. The third area attempted to get at an aspect of the inconsistency noted by considering different conceptions of complexity. These conflicts were not resolved by the experimental studies reviewed in the fourth section. Although some characteristics that cause more difficulty than others have been isolated, research is still spotty. Perhaps one of the most glaring gaps in the research to date is the paucity of studies along these lines that is done with ESL students. It seems reasonable that this study should aim to extend to ESL students, in a limited way, what has been done with native speakers in the area of syntax and reading. Furthermore, research of this kind which targets ESL students should underlie the production of reading textbooks aimed at that population.

PAGE 47

42 Notes Singulary transformations: Hunt (1970) refers to transformations that change single sentences to other single sentences as singulary transformations. These include question, imperative, passive, and negative transformations. In contrast, sentence combining transformations yield complex sentences. 2 Mean depth (MD) is depth, as defined by Yngve (1960), divided by the number of words in the sentence.

PAGE 48

CHAPTER III METHODOLOGY The purpose of this study was to consider the effect of syntax on readability and to determine what degree of sentence combining or chunking is the most readable for mature students learning English as a second language (ESL). The study departs from and questions the usual sentence length criterion of syntactic complexity of the most commonly used readability formulas in its consideration of the effect of syntax on readability. This chapter will provide information on the subjects, the design, the materials and the procedures used in conductinq the main study and the pilot studies. In addition, subjective judqments of relative difficulty of the three types of sentence structure characteristic of each of the three treatments were obtained from students, ESL teachers, and pre-service teachers taking a course in the teaching of reading. A description of this secondary study is also included. Puerto Rico Study Subjects The 85 subjects in this study were undergraduate students at the Mayaguez Campus of the University of Puerto Rico. All were native speakers of Spanish and literate in their mother tonque. All were enrolled in Englir.h 001 during the fall semester of 1979 and had 43

PAGE 49

44 scored between 400 and 599 on the English as a Second Language Achievement Test (ESLAT) taken during their senior year in high school. English 001 is the most basic English course currently offered at the university, but none of the students could be classified as beginners in their study of English. All had studied English throughout their school years, whatever the quality of the instruction may have been. They have also been surrounded by English on an everyday basis through commercial media. For example, labels and instructions on the many products from the United States are usually in English. Any students who had lived on the mainland of the United States or who were from bilingual homes, however, were not included in the study. Design Each subject was randomly assigned to one of the three treatment groups in a one-way design with three levels. The three levels are referred to as versions 1, 2, and 3. The reading passages of each version are characterized by a different type of sentence structure or syntax which will be described in detail in the following section. Ma te r i a 1 s Syntactically Different Comprehension Test . The 18 passages of the Syntactically Different Comprehension Test (SDCT) are based on passages from the Science Research Associates' Reading for Under standing General reading kit (Thurstone, 1969).

PAGE 50

45 While holding content and vocabulary constant, three versions of each of the 18 passages were constructed. The versions differ from one another only in syntax, or, more precisely, in degree of chunking. Sentence structure is the only characteristic that is manipulated (see Appendix A). The readability levels of the passages that are referred to below are based on the Maginnis (1969) adaptation of the Fry readability scale for use with passages shorter than those required when using the Fry readability scale. An extended graph (Madsen and Bowen, 1978) was used to determine readability levels in the upper range. Version 1 passages vary in length from 20 to 91 words. Where it is obtainable, the readability level, as presented in Table 1, varies from first to fourth grade. The passages are unchunked and contain only simple sentences except those occasional deviations as described under the definition of simple sentence (Chapter I, p. 9). Conditional sentences are broken into two sentences. What is most commonly expressed as an if"-clause is expressed in the form of a sentence beginning with r,upr>or.o . The result clause constitutes a separate sentence. Version 2 passages vary in length from 19 to 84 words. The passages contain complex sentences and thus have fewer and longer sentences than their version 1 counterparts. Readability level, where obtainable, varies from fifth to sixteenth grade level. In forming complex sentences for version 2, relative pronouns are not deleted even though they may be optional. Subjects and finite verbs are

PAGE 51

46 Table 1 Readability Levels of SDCT Passages* Based on Maginnis' Adaptation of Fry and Extended Graph Passage Version 1 Version 2 Version 3 1 1st grade 7th grade 7th grade 2 high 2nd grade 7th grade 10th grade 3 3rd-4th grade 7th grade 7th grade 4 N.D.** 9th grade 12th-13th grade 5 2nd grade 8th grade 8th-9th grade 6 N.D. 12th-13th grade 17+ 7 3rd grade 9th grade 7th-8th grade 8 N.D. 12th grade llth-12th grade 9 N.D. 13th grade 10th grade 10 2nd grade 8th grade 8th grade 11 N.D. 16th grade 16th grade 12 low 2nd grade low 7th grade 8th grade 13 4th grade 9th grade 9th grade 14 2nd-3rd grade 5th grade low 6th grade 15 N.D. N.D. N.D. 16 2nd grade 6th grade 8th grade 17 3rd grade 8th grade 8th grade 18 N.D. 15th grade N.D. * See Appendix A for SDCT. ** N.D. = No Data Fry makes no claims re material with these combinations of sentences per 100 words and syllables per 100 words.

PAGE 52

47 retained in subordinate clauses wherever possible. Where a finite verb is not possible, infinitives are usually used rather than the "nounier" (Ross, 1973) -ing forms or derived nouns. Example: ... it was hard for them to realize that they had really escaped. Conditional sentences are explicitly expressed with the subordinate clause introduced by if. Subordinating conjunctions such as since, because, while, and whereas are used wherever possible. Indirect objects follow direct objects and are preceded by a preposition. Example: . . . they promise that they will give your money back to you. Version 3 passages vary in length from 17 to 72 words and in readability level, where obtainable, from sixth grade to 17+. These passages also contain complex sentences, but they reflect a higher degree of chunking than their version 2 counterparts. In version 3 passages relative pronouns, and a form of the verb be where that is what follows, are deleted wherever they are optional. Such deletions were found to render aural comprehension more difficult by Charrow and Charrow (1979). Subordinate clauses often do not contain surface subjects or finite verbs. Where non-finite verbs are used, -'>.; forms or derived nouns are usually used rather than infinitives. Example: Because of • few social contacts, the farmer . . . These services improve employee morale and help the employer through a in absenteeism. The latter type of nominal ization is the type that Coleman (1964, 1965)

PAGE 53

48 found to have an adverse effect on reading comprehension and written recall as opposed to active verb expressions with the same meaning. Occasionally an adverbial prepositional phrase takes the place of an adverbial clause. Example: In extensive olive growing regions you will find . . . (version 3) versus Where olives are grown extensively, you will find . . . (version 2) Conditional sentences in version 3 are not always explicit. Instead of ^/-clauses there may be a verb form preceding the subject. Example: Should the puppeteer come out after the show for a bow, people are often surprised . . . Finally, indirect objects precede direct objects and thus appear without prepositions. Example: . . . they promise to give you your money back. At least one multiple choice question follows each short passage. There are a total of 24 comprehension questions. The aim qf_the questions is to test literal comprehension, which is considered a necessary prerequisite to other types of comprehension. Because the passages are short, they often test whether or not the reader understood the main idea rather than more specific facts or details. The questions cannot be answered merely on the basis of outside knowledge; the test taker must read the passages in order to answer the questions correctly. The stem of each test item is followed by four choices of more or less equal length or two short and two longer choices so that a single choice is never conspicuously different from all the others. Wherever possible the test items are worded differently

PAGE 54

49 from the passage itself to prevent recognition of the correct answer without comprehension. The test administration yielded a Kuder-Richardson 20 coefficient of .672. The dittoed 13-page test booklet had one or two double-spaced passages per page. Each passage was immediately followed by its comprehension question(s). Covariate . The ESLAT was used as a covariate to control for varying degrees of proficiency in English among the subjects. The ESLAT scores correlated .7 (Pearson's Product-Moment Correlation Coefficient) with the scores on the dependent variable, the SDCT. ESLAT is a 60 item grammar and readinq comprehension test administered by the College Board to high school seniors in Puerto Rico. The test includes 40 grammar and 20 reading comprehension questions. The test is used to place students in the appropriate English course at the university. Test reliability from 1964 through 1974 is reported by the Puerto Rico College Board to be consistently greater than .9. They do not, however, indicate what type of reliability coefficient was calculated. Procedure Four classes of English 001 participated in the study, and all were tested on the same day. The students were given the full class period to complete the SDCT. They marked their answers in the test booklet. Both the researcher and the regular teacher were present in three of the classes. The teacher alone administered the

PAGE 55

50 test in the fourth class after having done so with the researcher earlier that day. ESLAT scores were made available to the researcher by the university administration. With the ESLAT scores as a covariate, the data were analyzed by analysis of covariance using the Statistical Analysis System (SAS). Pilot Studies Before conducting the main study, several small pilot studies were done to determine whether or not use of the SDCT was feasible and whether the study was practical. Because the results will be referred to in Chapter IV, the pilot studies will be briefly described here. Subjects Two of the pilot study samples consisted of randomly selected English Language Institute (ELI) students at the University of Florida. All were native speakers of Spanish. All were undergraduates, at least 18 years of age, who had graduated from high school in their respective Latin American homelands. All had scored between 35 and 85 on a version of the Reading and Vocabulary section of an Institutedeveloped test that is used for purposes of initial placement and measurement of progress throughout the term. One group consisted of 12 students and the other of 15. The third sample was a group of 17 native speakers of English enrolled in a reading course offered by Alachua County Adult Education. Their reading levels varied from third to tenth grade as measured by the reading and vocabulary section of the Test of Adult Basic Education (TABE).

PAGE 56

51 Design The design of these studies was the same as that of the Puerto Rico study. Materials Several alterations were made in the SDCT before it was administered in Puerto Rico. Passages 10 and 13 were slightly reworded to comply better with the specifications for syntax previously described, and several comprehension questions (2, 5b, 8, 11a, 12a, and 14) were altered (see the end of Appendix A). The pilot study samples were given the earlier version. The format of the test booklet was the same as in the main study. One major difference, however, was that the group of 15 ELI students took the SDCT with the comprehension questions in Spanish. The intention was to assure that the language of the questions themselves would not affect students' performance. The Kuder-Richardson reliability coefficients based on the pilot study administrations were .75 for the group that had the questions in Spanish, .82 for the group that had the questions in English, and .81 for the adult education students. The ELI's Reading and Vocabulary Test was used as a covariate for the two samples of ELI students to control for varying degrees of proficiency in English among the subjects. The test scores correlated .85 (Pearson's Product-Moment Correlation Coefficient) with scores on the dependent variable for the group that had the comprehension questions in English and .61 for the group that had the comprehension questions in Spanish. The test consists of a single three-page passage

PAGE 57

52 Vocabulary items are underlined in the passage and tested by means of 20 multiple choice questions. The student must choose the best synonym based on the use of the vocabulary item in the context of the passage. Comprehension is measured by 20 true-false items. The test is not timed. The reading and vocabulary section of the Test of Adult Basic Education (TABE), used as a covariate for the sample of adult education students, correlated .8 with SDCT scores. Procedure The ELI students who participated in the pilot study attended a special session conducted by the researcher instead of their reading class for that day. The test was not timed and all finished easily within the class hour. There was a separate session for each of the two samples. The adult education students took the SDCT during their class. They were allowed as much time as necessary. Because it could not be certain that the assumptions of a parametric statistical procedure were met, a non-parametric analysis of covariance was used (Quade, 1967) to analyze the data. Subjective Judgments of Relative Difficult y Subjects The subjects who participated in this secondary study were seventy-nine undergraduates at the Mayaguez Campus of the University

PAGE 58

53 of Puerto Rico who were enrolled in English 002 during the spring semester of 1 980. Seven teachers of English 001-002 also made subjective judgments of the relative difficulty of examples of the three syntactic versions from the point of view of what they thought would be easiest and hardest for their students. Fourteen undergraduate students taking a course in the teaching of reading at the University of Florida also participated, making the judgment from the point of view of limited English proficiency students. Material s Three sets of passages of the SDCT (2, 8, 16) were each on a separate dittoed page. The order in which the three versions appeared was varied as was the order of the three sets of passages. There was a line to the left of each version of each passage for subjects to indicate the easiest of the three and the most difficult of the three. Procedure Students were asked, as an in-class exercise, to choose the paragraph they thought was the easiest in each set of three as well as the one they thought was the most difficult. The version left blank was ranked as "in-between." Teachers were asked to make the same judgment from the point of view of their students. Pre-service teachers did the exercise in class. The data wore analyzed by the Friedman two-way analysis of variance by ranks (Siegel , 1956).

PAGE 59

CHAPTER IV PRESENTATION AND ANALYSIS OF THE DATA The basic problem addressed by the study was to determine what degree of sentence combining is the most readable for mature students learning English as a second language. This chapter describes and presents the statistical analysis of the data collected for the Puerto Rico study, the pilot studies, and the subjective judgments of relative difficulty of the three versions of the Syntactically Different Comprehension Test (SDCT). Puerto Rico Study The sample consisted of 85 students enrolled in English 001 at the Mayaguez Campus of the University of Puerto Rico during the fall semester of 1979. Thirty each read versions 1 and 2 of the SDCT while 25 read version 3. The scores on the covariate, the English as a Second Language Achievement Test (ESLAT), and the SDCT for each student appear in Appendix B. Figure 1 shows the least squares lines for the plots of ESLAT scores on the x-axis versus SDCT scores on the y-axis for each of the three syntactically different versions. The least squares line corresponding to version 2 is consistently higher than the line for version 1 within the range of this sample. The least squares line for version 3, however, is neither consistently lower than nor higher than the other two lines. 54

PAGE 60

55 version 1 version 2 version 3 in o O Q 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 n 2 1 ,e<^°* 390 410 430 450 470 490 510 530 550 570 590 610 ESLAT scores Fiqure 1 Least Squares Lines for Each Version Puerto Rico Study

PAGE 61

56 In the statistical analysis of the data, ESLAT scores, which correlated .7 (Pearson's Product-Moment Correlation Coefficient) with the SDCT, were included in the model as a covariate. A test of homogeneity of regression was performed to determine if there was an interaction between the covariate and the dependent variable. The interaction term was not significant. Thus, it was acceptable to proceed with analysis of covariance in order to test the null hypotheses that there is no difference in comprehension between versions 1 and 2, nor between versions 3 and 2. Table 2 indicates that, overall, the effect of the version of the SDCT read by the subjects does not reach significance (p < .2388) Table 2 Analysis of Covariance Using SDCT as Dependent Variable and ESLAT as Covariate Source df SS MS F Prob Version

PAGE 62

57 Table 3 Pairwise Comparisons T Prob Version 1 vs. Version 2 -1.69 .09 Version 3 vs. Version 2 -0.98 .33 Although neither hypothesis can be rejected, the least squares means, which estimate the average level of response for subjects scoring at the mean of the covariate, indicate that comprehension of version 2 is higher than comprehension of versions 1 and 3 as shown in Table 4. Table 4 Least Squares Means Version N LS Mean 1

PAGE 63

58 version 1 version 2 version 3 CD So (J c_> Q t/1 24

PAGE 64

59 Parametric tests indicated significant differences with an overall F (2, 11) of 13.61 (p < .0027). Pairwise comparisons were also significant: the t value for version 1 versus version 2 was -5.04 (p < .001), and the t value for version 3 versus version 2 was -3.24 (p < .0104). But because it was uncertain if violations of assumptions of the parametric statistical procedure would affect the results with a sample of 12 subjects, the data were analyzed nonparametrically using Quade's Rank Analysis of Covariance (Quade, 1967) A version of the ELI Reading and Vocabulary Test, which correlated .85 with the SDCT, was used as the concomitant variable to control for the varying degrees of proficiency in English among the subjects. Scores on both the Reading and Vocabulary Test and the SDCT appear in Appendix B. In the non-parametric test the calculated F (2, 9) statistic was 2.379 which falls short of significance at the experiment-wise alpha level of .05. Nevertheless, the least squares means shown in Table 5 indicate that comprehension of version 2 is superior to that of versions 1 and 3. Table 5 Least Squares Means ELI-Questions in English Version N LS Mean 1 4 12.00 2 5 16.27 3 3 13.20

PAGE 65

60 In the other administrations of the pilot study a similar trend can be detected: version 2 consistently yields the highest comprehension though the differences are not statistically significant. Scores on the SDCT and the covariates appear in Appendix B while Tables 6 and 7 present the least squares means for the ELI students who had the comprehension questions in Spanish and for the adult education students respectively. Table 6 Least Squares Means ELI-Questions in Spanish Ve rsion N LS Mean 1 5 14.6 2 5 17.2 3 5 13.5 Table 7

PAGE 66

61 that characteristics of version 2 sentence structure render version 2 more readable than the others, especially more than version 1. Subjective Judgments of Relative Difficulty Students' Judgments Seventy-nine students enrolled in English 002 at the Mayaguez Campus of the University of Puerto Rico during the spring semester of 1980 were asked to rank the three versions of passages 2, 8 and 16 of the SDCT. They did so by indicating the version they considered the easiest and the one they considered the most difficult leaving blank the one tallied as "in between." Each of the three sets of passages was separately analyzed using the Friedman two-way analysis of variance by ranks. The xr statistic for the three versions of passage 2 was 14.99. The null hypothesis that judgments of difficulty are independent of syntactic version can thus be rejected (p < .001). Follow-up tests indicated that version 2 was ranked significantly easier than version 1 (p < .001) and version 3 was ranked as easier than version 2 (p < .005) Table 9 shows that students tended to rank version 1 as the most difficult while versions 2 and 3 were almost equally ranked easiest or in between. For passage 16 version 2 was most often ranked as easiest while version 1 was most often ranked as hardest. The xr statistic was 33.443 indicating independence of judgment of difficulty and syntactic version (p < .001). Follow-up tests indicated that

PAGE 67

62 version 2 was ranked as significantly easier than both versions 1 (p < .005) and 3 (p < .005). Table 9 Contingency Table-Passage 2 version 1 version 2 version 3 42 hardest easiest in between 42

PAGE 68

63 versions 1 and 2 was not significant. Perhaps subject matter or conceptual load accounts for the different pattern of judgment on this particular set of passages. Table 11 Contingency Table-Passage 8 version 1 version 2 version 3 21

PAGE 69

64 Table 12 Contingency Table Teachers' Judgments of Three Sets of Passages hardest easiest in between version 1 version 2 version 3 4 19.0%

PAGE 70

65 Although the null hypothesis of independence of difficulty judgment and syntactic version could not be rejected for passages 2 and 16, the overall pattern of the judgments of the pre-service teachers, as shown in Table 14, was similar to that of the ESL teachers. Table 14 Contingency Table Pre-Service Teachers' Judgments of Three Sets of Passages hardest easiest in between version 1

PAGE 71

66 an obstacle to comprehension. As Shook (1977) hypothesized, relationships and meaning, it seems, are lost with such sentence structure. Although the differences among the three versions of the SDCT were not significant, the fact that on all administrations of the instrument version 2 yielded the highest comprehension scores does suggest that version 2 sentence structure is more readable and more comprehensible than that of version 1 and to some extent that of version 3. To what degree this is so, however, remains uncertain. It is strange that, as seen in Figure 1, at the lower end of the range of proficiency in English, version 3 yielded slightly higher scores than versions 1 and 2, though a test of equality of slopes was not significant. Perhaps at that level of proficiency students are so lost that sentence structure simply makes no difference, and these results merely reflect random variation. Also, the studies reviewed in Chapter II that led the researcher to believe certain characteristics of version 3 sentence structure would prove difficult may have been misleading. The fact that these characteristics did not affect reading comprehension in this study may_be due to the use of a different dependent variable from those used in the original studies. Some of those studies, for example, measured listening comprehension while others measured recall or memorization. The dependent variable in this study, in contrast, was comprehension measured solely by response to multiple choice questions which were answered immediately after reading.

PAGE 72

67 Another consideration is that the SDCT may attempt to encompass too much at once. A series of finer-grained studies might render the greater readability of version 2 more apparent, as well as clarify the puzzling results of version 3 with students of lower proficiency. These studies would examine single structures such as nominal ization versus noun + verb constructions (Coleman, 1964, 1965), inclusion versus exclusion of relative pronouns + be (Charrow and Charrow, 1979, "You must follow the law as I state it," 1978), inclusion versus exclusion of optional complementizers (Hakes, 1972), explicit versus implicit conditionals (Marshall and Glock, 1978-79) and the like. Such studies might prove more fruitful than examining something as gross as "degree of chunking." Lumping together the various leads gleaned from such research as the studies mentioned above may have blurred any potential findings. It is of interest that the subjective judgments of the students support the suggestion of the main study. Students do not tend to judge version 1 as the easiest to read, demonstrating the inappropr^ateness of the most commonly used readability formulas for this population. In fact, they often judge version 1 to be the most difficult. They tend to judge version 2 as the easiest while judgments of version 3 are more evenly dispersed. In contrast, both pre-service teachers and experienced ESL teachers often judge version 1 to be the easiest, version 2 in between and version 3 the hardest for their students. Thus well-meaning teachers or curriculum planners may, based on their tendency to judge version 1 as the

PAGE 73

68 easiest, select reading material with sentence structure similar to that of version 1 passages for their mature ESL students. Even for a basic course like English 001-002, teachers may unwittingly be doing students a disservice. In the first place, this study has shown that such reading material does not result in higher comprehension. Nor is such material considered easiest by the students themselves as demonstrated by their responses to the subjective judgment exercise. And finally, as Shook (1977, p. 6) suggested, such sentence structure is "like nothing they'll ever encounter in the real world." The notion that lower readability levels, as measured by traditional formulas, are more suitable for non-native speakers is cast in doubt, at least for adult learners who are literate in their mother tongue, Spanish. In this case their own language is one that is not so radically different in structure from English as the languages of other foreign or ESL students may be. It is also a language that, as a matter of style, prefers complex to simple sentences. It is possible that for native speakers of non-western languages, in contrast to Spanish speakers, conventionally measured lower readability levels are an aid to comprehension. This remains to be investigated. If the merely suggestive results of this study are interpreted as a weak effect of syntax on readability, one can speculate why this is so for native speakers of Spanish, and in particular for Puerto Ricans. Many Puerto Ricans have learned to cope with reading in English more successfully than they have learned to cope with i-

PAGE 74

69 oral /aural skills. Information supplied by English 002 students reveals that they have all studied English since elementary school, usually beginning as early as the first grade. They generally began to read in English a year or two later. Some had their science and math textbooks in high school in English. All confront textbooks in English in college although the majority of 001-002 students are freshmen without lengthy experience with college textbooks in English. Many claim to read comics and magazines in English, and some state that they read the San Juan Star , the island's English language newspaper. They constantly encounter labels and advertisements in English because most products available on the island are from the U.S. Directions and indications on these products are often in English. Thus this poDulation may be accustomed to extracting, or trying to extract, meaning from English with any type of syntax. The least likely type of syntax for them to encounter in these real -life settings is the simple sentences of version 1, explaining the lower comprehension of those subjects who read version 1. In addition to exposure to English, much of the Spanish read by Puerto Ricans is influenced by English. Aside from the interference due to language contact, much of what is printed in Spanish in Puerto Rico is clearly translation from English (wire service news columns, Cosmopolitan en Espanol , etc.). American English syntax has in this way infiltrated Puerto Rican Spanish. Furthermore, as mentioned above, as a question of taste and style, Spanish prefers complex sentences. Also, near literal translation of several passages

PAGE 75

70 of the SDCT revealed that version 1, in the judgment of a native speaker of Spanish, is very "un-Spanish" while versions 2 and 3 were perceived as being much more in line with Spanish sentence structure. This opinion mirrors the fact that version 1 resulted in almost significantly lower comprehension while the difference between versions 2 and 3 was inconclusive. If indeed syntax is less important for this population than one might expect, and if we discard the sentence length criterion as a result of this study, what are we left with other than vocabulary? Are there other criteria that should be considered in selecting reading material? Rhetorical devices such as organization, example, restatement, coherence and the like may play an important role in determining readability. Density of ideas, or the number of propositions presented in a text and the structure of that text may affect its readability (Kintsch and Vipond, 1979). These factors were not manipulated in this study. Kintsch and Vipond go in a different direction from this study in searching for neglected elements that might influence readability. If syntax does not have a great effect, perhaps their model points in a more fruitful direction Even if syntax has a relatively unimportant effect on readability, this study indicates that Smith's (1971) findings with English speakers may hold for Spanish speakers within the limited age range of the sample. He found that older students find more mature syntax to be more readable. In this study, it was found that the higher readability level material with more mature syntax is at least as readable as the material considered to be lower if not

PAGE 76

71 more so. This may be evidence that cognitive maturity in one's native language does carry over as an advantage in dealing with a second language.

PAGE 77

CHAPTER V SUMMARY AND CONCLUSIONS Overview The purpose of this study was to determine whether readability would be affected by manipulating the degree and type of sentence combining or "chunking" in reading material prepared for mature students of English as a second language (ESL) who are literate in Spanish, their native language, and who are at an intermediate level in their study of English. Studies done with young native speakers of English have shown that there is a relationship between reading ability and syntactic development (Chomsky, 1972; Ribovich, 1976; Ruddell, 1965a, 1965b; Simons, 1970; Tatham, 1969-70). Van Metre (1970) showed that a similar relationship exists for bilingual third graders. However, virtually no research relating syntax and reading ability has been done with mature ESL students. Research done on how the human mind handles different types of sentences has exclusively used native speakers of English as subjects. One group of such studies (Gough, 1965, 1966; Mehler, 1963; Miller, 1962; Savin and Perchonock, 1965) aimed to learn whether transformed sentences are more of a burden on memory than kernel sentences. These studies, of limited scope, have been criticized on several grounds and are not conclusive. 72

PAGE 78

73 Studies that attempted to define syntactic complexity are also inconclusive. Several studies (Martin and Roberts, 1966; Wright, 1969) aimed to resolve the conflict between Yngve's (1960) depth hypothesis and the theory that transformations create complexity. Others support the notion that integration or chunking of ideas is both natural and more comprehensible (Bransford and Franks, 1971; Pearson, 1974-75). Fodor and Garrett (1967) attempted to directly isolate surface clues whose presence or absence facilitated or hindered processing. Again, the subjects in these studies focusing on the question of what is syntactically complex were all native speakers of English. The field is therefore open to choosing a population of ESL students and manipulating syntax along the lines of questions raised by previous research with native speakers of English. In this study, 85 students enrolled in an ESL course at the University of Puerto Rico were randomly divided into three groups. Each group read 18 short passages of the Syntactically Different Comprehension Test (SDCT) and responded to 24 multiple choice comprehension questions on the passages. The SDCT was constructed by the researcher by adapting passages from the Science Research Associates' Reading for Understanding General reading kit (Thurstone, 1969). Three versions of each passage were constructed. The versions differed from one another principally in sentence structure; each version represented a different degree of chunking or sentence combining. Version 1 passages consisted of short, primarily simple

PAGE 79

74 sentences; version 2 included complex sentences with as many clues to underlying relationships as possible left intact, and version 3 also included complex sentences, but, because of the higher degree of chunking, fewer clues to underlying relationships were explicitly expressed. Vocabulary and content were held constant. There were several pilot administrations after which a few minor alterations in the instrument were made. The purpose was to determine if this manipulation of syntax affected the readability of the passages as reflected in the comprehension scores. The hypotheses tested in the principal study were: 1. There is no significant difference between comprehension of version 1 and comprehension of version 2 passages. vl v2 2. There is no significant difference between comprehension of version 3 and comprehension of version 2 passages. \i -, u n v3 v2 The data were analyzed by analysis of covariance using the Statistical Analysis System (SAS). Scores on the English as a Second Language Achievement Test (ESLAT) , a grammar and reading comprehension test administered to high school seniors by the College Board of Puerto Rico, were used as a covariate to control for varying degrees of proficiency in English among the subjects. As in all the pilot studies, version 2 yielded the highest comprehension scores, but with an experiment-wise alpha level of .05 neither null hypothesis could be rejected. Hypothesis 1 came close to being rejected.

PAGE 80

75 Because the results suggested that version 1 sentence structure is not an aid to comprehension, a supplement to the main study was carried out. Three different groups of subjects participated in the secondary study: a sample of 79 students, different from those used in the main study, but from the same population, seven ESL teachers who teach that population, and a class of 14 pre-service teachers enrolled in a basic course in the teaching of reading at the University of Florida. They were asked to subjectively judge which of the three versions of several of the passages they thought was the easiest and which they thought was the most difficult. The one not ranked was considered "in between." Overall, the students tended to rank version 1 as the most difficult (or in between) and version 2 as the easiest (or in between) while judgments of version 3 were more evenly dispersed. For two of the three passages judged, this pattern was significant using the Friedman two-way analysis of variance by ranks to test the null hypothesis that syntactic version and judgment of difficulty are independent. Although the same null hypothesis was not rejected for University of Puerto Rico ESL teachers' judgments of the three versions of the passages, the teachers tended to judge version 1 as the easiest, version 2 as in between and version 3 as the hardest; i.e., chunking, they often thought, created difficulty for their students. For the pre-service teachers who judged the passages from the point of view of students whose native language was not English, the null hypothesis was rejected

PAGE 81

76 for one of the passages. Overall, they usually judged version 1 as the easiest, version 2 as in between, and version 3 as the hardest or in between. Conclusions and Recommendations Based on the results of the study, it was concluded that reading materials with the short, primarily simple sentences characteristic of conventionally determined low readability levels do not enhance reading comprehension for the population from which the sample used in this study was drawn. This supports Shook' s (1977) hypothesis that relationships and meaning are sacrificed for these lower readability levels, which, moreover, result in an unnatural, unrealistic type of sentence structure.! It was also concluded that students do not necessarily judge this type of syntax as the easiest to read. The warning, then, is to teachers and curriculum planners who may choose for their ESL students material considered easy to read according to commonly used formulas as well as according to their own subjective judgments. First, this study suggests that such reading material does not result in higher comprehension. Second, the students themselves do not consider it easiest. And third, as Shook (1977, p. 6) suggested, such sentence structure is "like nothing they'll ever encounter in the real world." But there are limitations to this conclusion that must be borne in mind. Two primary limitations are based on the native language of the subjects and their age. Spanish is not so radically

PAGE 82

77 different in structure from English as some other foreign languages. Furthermore, Puerto Rico has a unique relationship to American English. Because of political and economic ties to the United States, exposure to English is often broader and deeper than it is in other parts of the Spanish-speaking world. Whether or not simple sentences typical of what is commonly considered lower readability levels are an aid to other populations of ESL students remains to be investigated. It is noteworthy that the pilot studies suggest the same pattern as the Puerto Rico study. The question of what type of sentence structure is most readable for younger ESL students also remains open. One must consider, too, the possibility that the failure to reject the two null hypotheses of the principal study is an indication that syntax does not have a great effect on readability, at least for the population under study. What then, other than vocabulary, does influence readability for ESL students? Other features of a passage such as organization, coherence, restatement, density of ideas, conceptual difficulty must be considered. Before abandoning the relationship between syntax and readability, however, the following recommendations for further research are made. They may succeed in strengthening and expanding the present study. 1. As already mentioned, the study should be replicated with ESL students who are native speakers of other languages, especially non-western languages. Radically different effects of

PAGE 83

78 syntax on readability for different populations would render a uniform approach to reading inappropriate. 2. Conducting the study with younger ESL students would be similar to the work of Smith (1971) with non-native speakers of English. If younger students do comprehend version 1 better than the other versions, this would parallel Smith's findings, and the results of this study would more clearly indicate positive transfer of cognitive development and an advantage for adult learners. The differences across age groups may not be parallel across native language groups. Whether or not this is so would be of interest. 3. Replication of some of the psycho! inguistic studies reviewed in Chapter II with ESL students as subjects would reveal whether or not non-native speakers respond in a fashion parallel to the responses of native speakers of English. D o they find transformed or kernel sentences easier to process (Gough, 1965, 1966; Mehler, 1963; Miller, 1962; Miller and McKean, 1964; Savin and Perchonock, 1965)? Are chunked ideas preferred (Franks and Bransford, 1971)? What degree of nesting can be handled (Schlesinger, 1968)? Is center-embedding easier to process than right-branching (Hamilton and Deese, 1971; Miller and Isard, 1964)? Is there an effect of sentence depth (Yngve, I960)? 4. A series of finer-grained studies manipulating one item at a time would provide more definite and specific findings than those resulting from use of the SDCT. Such studies might demonstrate that the inconclusiveness of this study is due to insensitivity of the SDCT, possibly because it attempts to encompass too much.

PAGE 84

79 Finer-grained studies might also help unravel the mysterious occurrence of higher scores on version 3 at the lower end of the range of proficiency in English. These studies could include, along the lines of Coleman (1964, 1965), examining whether a noun + verb combination is more comprehensible than the corresponding nominal ization. Taking Hakes (1972) and Charrow and Charrow (1979) as models, the effect of omission or inclusion of optional complementizers and relative pronouns could be examined. Following Marshall and Glock (1978-79), one could examine whether or not explicitly expressed conditions are more comprehensible than alternate forms. Comprehensibility of two simple sentences versus a corresponding complex sentence with a relative, adverbial, or noun clause might be contrasted. Several dependent variables are possible: recall, paraphrase, memorization, time, translation, as well as comprehension questions. Until further research is done permitting more specific pedagogical implications, the principal recommendation of the present study is that low level reading material , as measured by conventional readability formulas, with primarily simple sentences is not recommended for college age ESL students in Puerto Rico. This does not mean that care should not be given to the selection of material. Further work on the effect of syntax needs to be done and other factors affecting readability need to be considered.

PAGE 85

APPENDIX A INSTRUMENT The Syntactically Different Comprehension Test* Please read the paragraphs in this booklet. Try to answer the question or questions that are after each paragraph based only on the information that is in the paragraph. Mark the letter of the correct answer with an X. There is only one correct answer for each item. If you cannot answer the question based on the information that is in the paragrph, go on to the next one. Practice Paragraph version 1 The stems of the cactus plant are provided with a large quantity of special tissue. This tissue stores water. The plant is equipped to retain moisture, and this is one of the ways. This feature makes it particularly suitable to life in arid areas. version 2 The stems of the cactus plant are provided with a large quantity of special tissue which stores water. This is one of the ways that this plant is equipped to retain moisture, which is a feature that makes it particularly suitable to life in arid areas. version 3 The stems of the cactus plant are provided with a large quantity of special tissue for water storage. This is one of the ways this plant is equipped to retain moisture, a feature making it particularly suitable to life in arid areas. A cactus plant does not need frequent rain because it A. does not require water. B. can save water for later use. C. uses other liquids. D. grows slowly. A B C D ()()()() *In the actual test booklet for each version the comprehension question(s) immediately follow each paragraph on the same page. The paragraphs and the choices are double-spaced. 80

PAGE 86

81 (1) version 1 Steve couldn't swim. He walked along the side of the creek. He looked for a shallow place. He would wade across at that place. version 2 Since Steve couldn't swim, he walked along the side of the creek looking for a place that was shallow enough for him to wade across. version 3 Since Steve couldn't swim, he walked along the side of the creek looking for a place shallow enough to wade across. Steve looked for a place where the water wasn't deep so he could A. swim to the other side. B. walk to the other side. C. throw rocks to the other side. D. see the other side. A B C D ()()()() (2) version 1 version 2 version 3 Disease germs may be present in food. Cook food for a long enough time. This will kill any disease germs. Food may not be clean. Cook it thoroughly. In this way you can combat possible uncleanliness of food. If you cook food for a long enough time, you will kill any disease germs that may be present. Therefore, one way that you can combat possible uncleanliness of food is by cooking it thoroughly. Cooking food for a long enough time will kill any disease germs possibly present. Therefore, cooking it thoroughly is one way of combatting possible uncleanliness of food. To be certain that food is safe to eat A. wash it. B. cook it completely. C. cook it immediately. D. fry it. A B C D ()()()()

PAGE 87

82 (3) version 1 They use olives to make olive oil. The olives ripen. They pick them promptly. They send them to the oil mill immediately. The olives spoil rapidly. They cannot ship them far. In some places olives are grown extensively. You will find a small oil mill every few miles in these places. version 2 Olives that are used in making olive oil are picked as soon as they ripen and are immediately sent to the oil mill. They cannot be shipped far because they spoil rapidly. Where olives are grown extensively, you will find a small oil mill every few miles. version 3 Olives used in making olive oil are picked as soon as they ripen and are sent immediately to the oil mill. Because of rapid spoiling, they cannot be shipped far. In extensive olive growing regions you will find a small oil mill every few miles. According to this paragraph, olive oil mills are close to where the olives are grown because A. it saves shipping costs. B. the growers also make the oil. C. the olive oil is consumed locally. D. picked olives will not last a long time. A B C D ()()()() (4) version 1 Some animals have a "weather sense." The weather governs their movements and their very existence. Man has never been able to equal this. The American Indian was a fine weather prophet. Even he relied almost entirely on the movements of the creatures of the wild for his weather predictions, version 2 Animals whose movements and whose very existence are governed by the weather have a "weather sense" that man has never been able to equal. Even the American Indian, who was a fine weather prophet, relied almost entirely on the movements of the creatures of the wild for his weather predictions, version 3 Animals whose movements and very existence are governed by the weather have a "weather sense" never equalled by man. Even the American Indian, a fine weather prophet, relied almost entirely on the movements of the creatures of the wild for his weather predictions.

PAGE 88

83 Those who are most sensitive to climatic changes are A. American Indians. B. people with rheumatism. C. certain animals. D. meteorologists. A B C D ()()()() (5) version 1 We were concealed behind the bushes. We watched our victim. Our victim approached the pit. We had dug the pit. We had covered it with leaves and light branches. Suppose the animal sensed something. It was about to be captured. Surely it would not have walked along so casually. version 2 We were concealed behind the bushes and watched our victim approach the pit that we had dug and covered with leaves and light branches. If the animal had sensed that it was about to be captured, surely it would not have walked along so casually. version 3 Concealed behind the bushes, we watched our victim approach the pit we had dug and covered with leaves and light branches. Surely the animal would not have walked along so casually had it sensed it was about to be captured. The animal that was beinn hunted A. saw the hunters. 3. was looking for food. C. tried to hide. D. was seen by the hunters. A B C D ()()()() The trap that the hunters set was \, visible. B. under the bushes. C. camouflaged. D. dug by their guide. A B C D ()()()()

PAGE 89

84 The hunters believed that the animal A. had sensed danger. B. was nervous because it was afraid. C. was not aware that it was being hunted. D. had walked away. A B C D ()()()() (6) version 1 Educators and parents have been complaining. Children spend too much time at something. They watch television too much. They don't do their homework. Educators and parents got support this week. A company is now producing a television with doors. The doors can be locked. version 2 Educators and parents, who have been complaining that children spend too much time watching television instead of doing their homework, got support this week from a company that is now producing a television with doors that can be locked. version 3 Educators and parents, complaining children spend too much time watching television instead of doing their homework, got support this week from a company now producing a television with lockable doors. According to this paragraph, teachers and parents believe that A. children should spend more time studying. B. children should participate in television quiz shows. C. television is an excellent educational device. D. children's play time should be limited. A B C D ()()()() The company that is referred to in this paragraph A. promotes television watching. B. wants to help parents and teachers. C. is supported by the television networks. D. produces children's books. A B C D ()()()()

PAGE 90

85 (7) version 1 The jury returned the verdict after a very short interval. They were very prompt. The judge thanked them for this. version 2 Because the jury returned the verdict after a very short interval, the judge thanked them for being so prompt. version 3 The jury returned the verdict after so short an interval the judge thanked them for their promptness. The judge was grateful because the jury A. went home early. B. acted quickly. C. made a good decision. D. was chosen quickly. A B C D ()()()() (8) version 1 Manufacturers have to get goods to market. Suppose the manufacturer and the market are a long distance apart. This can be a big expense. The manufacturer can minimize this expense. He can establish himself near a market. version 2 If the manufacturer and the market are a long distance apart, then it can be a big expense for the manufacturer to get goods to market. The manufacturer can minimize this expense by establishing himself near a market. version 3 Getting goods to market can he a big expense if the manufacturer and the market are a long distance apart. Establishment of the manufacturer near a market can minimize this expense. According to this paragraph, manufacturers can save money if A. they do not have to advertise in out-of-town newspapers. B. products do not have to be shipped far. C. natural resources are nearby. D. goods are handled carefully during shipping. A B C D ()()()()

PAGE 91

86 (9) version 1 version 2 version 3 Voltaire was one of the greatest writers of the eighteenth century. He said, "The two essentials of style are precision and color." The remark is interesting. Voltaire's style is exceptionally precise. But it is not outstanding for color. Voltaire, who was one of the greatest writers of the eighteenth century, said that the two essentials of style are precision and color. The remark is interesting because Voltaire's style, although exceptionally precise, is not outstanding for color. Voltaire, one of the greatest writers of the eighteenth century, said the two essentials of style are precision and color. The remark is interesting in view of Voltaire's style, which is precise but not outstanding for color. Voltaire's writing A. has all the stylistic characteristics that he considers indispensable. B. is not particularly colorful. C. is difficult to understand. D. is less exact than it should be. A B C D ()()()() (10) version 1 The two boys were pale and shaken. They hurried to the safety of their home. There they told of their frightening experience. They sat in the kitchen. They were warm and comfortable. They had really escaped. They realized it, but this was hard. version 2 The two boys, who were pale and shaken, hurried to the safety of their home, where they told of their frightening experience. As they sat in the kitchen, where they were warm and comfortable, it was hard for them to realize that they had really escaped. version 3 The two pale and shaken boys hurried to the safety of their home, where they told of their frightening experience. Realizing they had really escaped was hard as they sat, warm and comfortable, in the kitchen.

PAGE 92

87 The boys returned home A. after having a good time. S. after escaping some danger. C. without understanding what happened. D. without telling their story. A B C D ()()()() (11) version 1 The farmer has few social contacts. Therefore, he tends to be offended by minor insults. The resident of the city has acquired immunity to these. He is accustomed to the give-and-take of casual interpersonal relationships. He fails even to notice discourtesy. The rural person would take this very seriously. version 2 Because he has few social contacts, the farmer tends to be offended by minor insults against which the resident of the city has acquired immunity. The resident of the city is accustomed to the give-and-take of casual interpersonal relationships and fails even to notice discourtesy that the rural person would take very seriously. version 3 Because of havinq few social contacts, the farmer tends to be offended by minor insults against which the resident of the city has acquired immunity. The resident of the city, accustomed to the give-and-take of casual interpersonal relationships, fails even to notice discourtesy the rural person would take very seriously. According to this paragraph, rural people are often more socially sensitive because they A. exoect people to be kind to strangers. B. are raised that way. C. are isolated from other people. D. are from large families and are used to a lot of people. A 3 C D ()()()()

PAGE 93

88 City people, according to this paragraph, A. are easily offended by gruffness. B. have frequent personal contact. C. are quiet. D. are friendly. A B C D ()()()() (12) version 1 Puppets are small doll-like figures. They can be moved by a person's hand. The hand is inside the doll. They can be moved by rods. The rods stick up through the body. Or they can be moved by strings from above. The audience does not see the puppeteer. People watch a puppet show. The puppets seem to become real people. The audience no longer thinks of them as dolls. Suppose the puppeteer comes out after the show. He takes a bow. People are often surprised. He looks like a giant. version 2 Puppets are small doll-like figures that can be moved by a person's hand which is inside the doll, by rods that stick up through the body, or by strings from above. The puppeteer is not seen by the audience. When people watch a puppet show, the puppets seem to become real people. The audience no longer thinks of them as dolls. If the puppeteer comes out after the show to take a bow, people are often surprised because he looks like a giant. version 3 Puppets are small doll-like figures movable by a person's hand inside the doll, rods sticking up through the body, or strings from above. The puppeteer is unseen by the audience. The puppets seem to become real people, and the audience watching a puppet show no longer thinks of them as dolls. Should the puppeteer come out after the show for a bow, people are often surprised because he looks like a giant, Puppets are animated by A. a remote control device. B. a special kind of battery. C. one of three methods. D. a magnetic system. A B C D ()()()()

PAGE 94

89 During a puppet show the audience A. feels as though they are puppets themselves. B. takes part in the show. C. operates the puppets. D. begins to think of the puppets as being alive. A B C D ()()()() (13) version 1 Streams flow through valleys. Most valleys are wider than the streams. We conclude something. Factors operate to form the valley. There are usually factors other than the wearing-down action of the stream. Suppose a valley were formed only by the action of the stream. The valley would be no wider than the stream. Its sides would be vertical . version 2 Since most valleys are wider than the streams that flow through them, we can conclude that there are usually factors other than the wearing-down action of the stream that operate in forming the valley. If a valley were formed only by the action of the stream, the valley would be no wider than the stream and its sides would be vertical. version 3 Since most valleys are wider than the streams flowing through them, we can conclude there are usually factors other than the wearing-down action of the stream operating in the formation of the valley. Were the valley formed only by the action of the stream, the valley would be no wider than the stream and would have vertical sides. The creation of a valley involves A. several forces. B. the force of a river alone. C. the influence of weather. D. forming steep sides. A B C D ()()()()

PAGE 95

90 (14) version 1 Suppose you subscribe to this magazine. You can't lose. The editors believe something. You will consider it well worth the subscription cost. Suppose you do not like it. They make a promise. They will give your money back to you. version 2 You can't lose if you subscribe to this magazine. The editors believe that you will consider it well worth the subscription cost. If you do hot like it, they promise that they will give your money back to you. version 3 Should you subscribe to this magazine, you can't lose. The editors believe you will consider it well worth the subscription cost. Should you not like it, they promise to give you your money back. The magazine editors make a promise which shows that they A. are confident that most people will like the magazine. B. do not care if people are pleased with the magazine. C. do not care about making money. D. want to make money. A B C D ()()()() 05) version 1 Factories provide medical and nursing services. These help to keep employees healthy. These services improve employee morale. They help the employer. They reduce absenteeism. version 2 Factories provide medical and nursing services that help to keep employees healthy. These services improve employee morale and help the employer by reducing absenteeism. version 3 By providing medical and nursing services, factories help keep employees healthy. These services improve employee morale and help the employer through a reduction in absenteeism.

PAGE 96

91 According to this paragraph, giving benefits to employees A. causes prices to go up. B. is worthwhile for the employers. C. is compulsory. D. is tax deductible. A B C D ()()()() (16) version 1 Birds migrate. They arrive in the North in early spring. The days are a certain length. The birds leave again in the fall. The days in fall are the same length. This fact has led some scientists to believe something. The birds' urge to travel is the result of the amount of sunshine. version 2 Birds that migrate arrive in the North in early spring. They leave again in the fall when the days are the same length as when they arrived. This fact has led some scientists to believe that the birds' urge to travel is the result of the amount of sunshine. version 3 Migrating birds arriving in the North in early spring leave again in the fall when the days are the same length as upon their spring arrival. This fact has led some scientists to believe the birds' urge to travel to be the result of the amount of sunshine. According to the theory that is proposed in this paragraph, the migration of birds A. depends on variation in humidity. B. depends on the number of hours of light per day. C. is related to mating habits. D. is related to the search for food. A B C D ()()()()

PAGE 97

92 (17) version 1 The baby elephant watched the two bull elephants. They fought. It is difficult to be king bull of a herd. The baby realized this. The king has to be strong and observant. He must use good judgment. Nevertheless, the baby elephant resolved to be king some day. version 2 After watching the two bull elephants fight, the baby elephant realized that it is very difficult to be king bull of a herd. The king has to be strong and observant and must use good judgment. Nevertheless, the baby elephant resolved to be king some day. version 3 After watching the two bull elephants fight, the baby elephant realized the difficulty of being king bull of a herd. The king has to be strong, observant and must use good judgment. Nevertheless, the baby elephant resolved to be king some day. To be king of a herd of elephants A. requires a good memory. B. may be dangerous. C. one's father must have been king before. D. one must be the oldest male elephant in the herd. A B C D ()()()() Despite all odds, the young elephant demonstrated A. determination. B. bravery. C. intelligence. D. power. A B C D ()()()()

PAGE 98

93 (18) version 1 Science can be regarded either from a static or from a dynamic point of view. Consider it from a static point of view. Science is a body of knowledge. It describes the universe. It is regarded as explanatory. Some people have a dynamic orientation. To them science implies continuous activity. Today's state of knowledge is a basis for further operations. version 2 Science can be regarded either from a static or from a dynamic point of view. If one considers it from a static point of view, science is a body of knowledge that describes the universe and is regarded as explanatory, whereas to anyone with a dynamic orientation science implies continuous activity that has today's state of knowledge as a basis for further operations. version 3 Science can be regarded from a static or a dynamic point of view. Considered statically, science is a body of knowledge descriptive of the universe and regarded as explanatory, whereas to anyone with a dynamic orientation science implies continuous activity with today's state of knowledge as a basis for further operations. This paragraph presents two ways to view science: A. one view is physical; the other is biological . B. one views science as a store of knowledge to apply to problems; the other views science as a device for preventing problems. C. one views science as a fixed set of knowledge; the other views science as building on that set of knowledge. D. one view is historical; the other is futuristic. A B C D ()()()()

PAGE 99

94 SDCT Items Used in Pilot Studies that Differ from Form Used in Puerto Rico Study In the pilot studies versions 2 and 3 of passage 10 and versions 1 and 2 of passage 13 appeared as follows: (10) version 2 The two boys, who were pale and shaken, hurried to the safety of their home, where they told of their frigntening experience. Even as they sat in the kitchen, where they were warm and comfortable, it was hard for them to realize that they had really escaped. version 3 The two pale and shaken boys hurried to the safety of their home, where they told of their frightening experience. Even as they sat, warm and comfortable, in the kitchen realizing they had really escaped was hard. (13) version 1 Streams flow through valleys. Most valleys are wider than the streams. We conclude something. Factors operate in the formation of the valley. There are usually factors other than the wearing-down action of the stream. Suppose a valley were formed only by the action of the stream. The valley would be no wider than the stream. Its sides would be vertical . version 2 Since most valleys are wider than the streams that flow through them, we can conclude that there are usually factors other than the wearing-down action of the stream that operate in the formation of the valley. If a valley were formed only by the action of the stream, the valley would be no wider than the stream and its sides would be vertical. Questions on passages 2, 5, 8, 11, 12, and 14 appeared as follows in the pilot studies: Question on passage 2 According to this paragraph, food should be cooked completely because A. it may be cold. B. it will be more attractive. C. that will destroy bacteria. D. that will make it taste better. A B C D ()()()()

PAGE 100

95 Second quesThe trap that the hunters set was tion on passage 5 A. visible. B. camouflaged. C. square. D. deep. A B C D ()()()() Question on If factories and markets are close, passage 8 A. products can be sold sooner. B. there will be less chance of damage during shipping. C. it will not be necessary to advertise in out-of-town newspapers. D. manufacturers can save money. A B C D ()()()() First quesRural people are often more socially sensitive tion on because they are passage 11 A. kind to strangers. B. raised that way. C. isolated from other people, D. from large families. A B C D ()()()() First quesPuppets are tion on passage 12 A. powered by a special kind of battery. B. life-sized. C. animated by one of several methods. D. unattractive. A B C D ()()()()

PAGE 101

96 Question on The magazine editors make a promise which shows passage 14 that they A. are confident that most people will like the magazine. B. do not care if people are pleased with the magazine. C. will publish weekly. D. want to make money. A B C D ()()()()

PAGE 102

APPENDIX 5 COVARIATE AND SDCT SCORES* Puerto Rico Study

PAGE 103

98 Version ESLAT

PAGE 104

99 ELI p-

PAGE 105

TOO Adult Education Students

PAGE 106

REFERENCES Anderson, Jonathan. Selecting a suitable reader: procedures for teachers to assess language differences. Regional English Language Center Journal , 1971, II (2), 35-42. Betts, Emmett Albert. Readability: linguistic factors . Paper presented at the annual meeting of the International Reading Association, 1977. (ERIC Document Reproduction Service No. ED 141744). Bever, Thomas G. The cognitive basis for linguistic structures. In Hayes, John R., Ed. Cognition and the development of language . New York: John Wiley & Sons, Inc., 1970. Blossom, Grace A. Teaching the bilingual child to read. Changing Education , Winter, 1969, 3, 28-30 I 7Bormuth, J., Carr, J., Manning, J., & Pearson, D. Children's comprehension of betweenand within-sentence syntactic structures. Journal of Educational Psychology , 1970, 61_, 349-357. ./Botel , M., Dawkins, J., & Granowsky, A. A syntactic complexity formula. In Mac Gin i tie, W. H., Ed. Assessment problems in reading . Newark, Delaware: International Reading Association, 1973. Bransford, John D., & Franks, Jeffrey J. The abstraction of linguistic ideas. Cognitive Psychology , 1971, 2, 331-350. Charrow, Veda R. , & Charrow, Robert P. Characteristics of the language of jury instructions . Paper presented at Georgetown Roundtable on Language and Linguistics, 1979. Chomsky, Carol. The acquisition of syntax in children from 5 to 10 . Cambridge, Mass.: MIT Press, 1969. ^Chomsky, Carol. Stages in language development and reading exposure. Harvard Educational Review , 1972, 42 (1), 1-33. Coleman, E. B. The comprehensibil ity of several grammatical transformations. Journal of Applied Psychology , 1964, 48_, 186-190. 101 '

PAGE 107

102 Coleman, E. B. Learning of prose in four grammatical transformations. Journal of Applied Psychology , 1965, 49, 332-341. Coleman, E. B. Developing a technology of written instruction: some determiners of the complexity of prose. In Rothkopf, E., & Johnson, P., Eds. Symposium on verbal learning research and the technology of written instruction . New York: Teachers College Press, 1971 . ^-Dawkins, John. Syntax and readability . Newark, Delaware: International Reading Association, 1975. Dodgson, Charles L. The humorous verse of Lewis Carroll . New York: Dover Publications Inc., 1960. Epstein, William. The influence of syntactical structure on learning. The American Journal of Psychology , 1961, 74_, 80-85. ^ Epstein, William. Some conditions of the influence of syntactical structure on learning: grammatical transformations, learning instructions and "chunking." Journal of Verbal Learning and Verbal Behavior , 1967, 6_, 415-419. Evans, Ronald V. The effects of transformational simplification on the reading comprehension of selected high school students. Journal of Reading Behavior , 1972, 5_, 273-281. ^Fagan, William T. Transformations and comprehension. The Reading Teacher , 1971, 25_, 169-172. Fodor, J. A. , & Bever, T. G. The psychological reality of linguistic segments. Journal of Verbal Learning and Verbal Behavior , 1965, 4, 414-420. Fodor, J. A., & Garrett, M. F. Some syntactic determinants of sentential complexity. Perception and P sychophysics , 1967, 2, 289-296. Fodor, J. A., Garrett, M. F. , S Bever, T. G. Some syntactic determinants of sentential complexity, II: verb structure. Perception and Psychophysics , 1968, 3_> 453-461. Franks, Jeffrey J., & Bransford, John D. Memory for syntactic form as a function of semantic context. Journal of Experimental Psychology , 1974, 103 (5), 1037-1039. ,/Gaies, Stephen J. Linguistic input in formal second language learning: the issues of syntactic gradation and readability in ESL materials. TESOL Quarterly , 1979, 13 (1), 41-50.

PAGE 108

103 ^/Gibson, Eleanor J. , & Levin, Harry. The psychology of reading. Cambridge, Mass.: MIT Press, 1975. Glucksburg, Sam, Trabassa, Tom, & Wald, Jerry. Linguistic structures and mental operations. Cognitive Psychology , 1973, 5_, 338-370. Goldman-Eisler, Frieda, & Cohen, Michele. Is N, P, and PN difficulty a valid criterion of transformational operations? Journal of Verbal Learning and Verbal Behavior , 1970, 9, 161-166. Gough, Philip B. Grammatical transformations and the speed of understanding. Journal of Verbal Learning and Verbal Behavior , 1965, 4, 107-111. Gough, Philip B. The verification of sentences: the effects of delay of evidence and sentence length. Journal of Verbal Learning and Verbal Behavior , 1966, 5_, 492-496. ^Greene, Judith. Syntactic form and semantic function. Quarterly Journal of Experimental Psychology , 1970, 22, 14-27. Greene, Judith. Psychol inguistics , Chomsky and psychology . New York: Penguin Books, 1972. Hakes, David T. Effects of reducing complement constructions on sentence comprehension. Journal of Verbal Learning and Verbal Behavior, 1972, 11, 278-286. ? Hamilton, Helen, & Deese, James. Comprehensibil ity and subject-verb relations in complex sentences. Journal of Verbal Learning and Verbal Behavior , 1971, 10., 163-170. Haskell, John. Read this! But how do I know if they can? In Ilyin, Donna, & Tragardh, Thomas, Eds. Classroom practices in adult ESL . Washington, D.C.: Teachers of Engl ish to Speakers of Other Languages, 1978. Holmes, V. M. Order of main and subordinate clauses in sentence perception. Journal of Verbal Learning and Verbal Behavior , 1973, 12, 285-293. Holmes, V. M., & Forster, K. I. Perceptual complexity and underlying sentence structure. Journal of Verbal Learning and Verbal Behavior , 1972, 1J_, 148-156. ?Huggins, A. W. F. Syntactic aspects of reading comprehension . Technical report No. 33, Bolt, Beranek, & Newman, Inc., Cambridge, Mass., 1977. (ERIC Document Reproduction Service No. ED 142972).

PAGE 109

104 Hunt, Kellogg. Grammatical structures written at three grade level's . Champaign, Illinois: National Council of Teachers of English, 1965. Hunt, Kellogg. How little sentences grow into big ones. In Lester, Mark, Ed. Readings in applied transformational grammar . New York: Holt, Rinehart and Winston, Inc., 1970. ' Kachuck, Beatrice. Syntax patterns in elementary school readers . Brooklyn College, 1975. (ERIC Document Reproduction Service No. ED 138971). fKintsch, Walter, & Vipond, Douglas. Reading comprehension and readability in educational practice and psychological theory. In Nilsson, Lars-Goran, Ed. Perspectives on memory research: essays in honor of Uppsala University's 500th anniversary . Hillsdale, New Jersey: Lawrence Erlbaum Associates, Publishers, 1979. fKlare, G. R. Assessing readability. Reading Research Quarterly , *~ 1974-75, 10, 62-102. £ Larkin, Don. Grammar: an important component of reading. In Shuy, Roger, Ed. Linguistic theory: what can it say about reading? Newark, Delaware: International Reading Association, 1977. fMadsen, Harold S., & Bowen, J. Donald. Adaptation in language teaching . Rowley, Mass.: Newbury; House, 1978. f'Maginnis, George. The readability graph and informal reading inventories. The Reading Teacher , 1969, 22 (6), 516-519. larshall, Nancy, & Glock, Marvin D. Comprehension of connected discourse: a study into the relationships between the structure of text and information recalled. Reading Research Quarterly , 1978-79, 14, 10-56. ?Martin, E., & Roberts, K. H. Grammatical factors in sentence retention. Journal of Verbal Learning and Verbal Behavior , 1966, 5, 211-218. 7 Matthews, W. A. Transformational complexity and short term recall. Language and Speech , 1968, 1T_, 120-128. Mavrogenes, Nancy A. The language development of the disabled secondary reader . Paper presented at the annual meeting of the International Reading Association, 1977. (ERIC Document Reproduction Service No. ED 141763).

PAGE 110

105 Mehler, Jacques. Some effects of grammatical transformations on the recall of English sentences. Journal of Verbal Learning and Verbal Behavior , 1963, 2_, 346-351. Miller, George A. The magical number 7 ± 2. Psychological Review, 1956, 63, 81-97. Miller, George A. Some psychological studies of grammar. American Psychologist , 1962, 17, 748-762. Miller, George A., & Isard, Stephen. Free recall of self-embedded English sentences. Information and Control , 1964, 7_, 292-303. Miller, George A., & McKean, K. E. A chronometric study of some relations between sentences. Quarterly Journal of Experimental Psychology , 1964, 16_, 297-308. Nadler, Harvey. A study to determine how the use of English reading materials containing only previously learned syntactic structures compares with the use of unstructured English reading materials in enhancing the beginning foreign student's learning of new lexical items (Doctoral dissertation, New York University, 1968). Dissertation Abstracts, 28, 4618A (68-06163). Owens, R. J. Selecting a suitable reader. Regional English Language Center Journal , 1971, H_ (2), 29-34. Pearson, P. David. The effects of grammatical complexity on children's comprehension, recall, and conception of certain semantic relations Reading Research Quarterly , 1974-75, 10., 155-192. Quade, D. Rank analysis of covariance. Journal of the American Statistical Association , 1967, 62_, 1187-1200. Ransom, Grayce A. Preparing to teach reading . Boston: Little, Brown, and Co., 1978. Ribovich, Jerilyn K. Comprehension of syntactic structures in oral language and its relationship to reading comprehension in first grade children . West Virginia University, 1976. (ERIC Document Reproduction Service No. ED 144103). , Rohrman, Nicholas L. More on the recall of nominal izations . Journal of Verbal Lea rnin g an d Verbal Behavio r, 1970, 9, 534-536. Ross, John R. Nouniness. In Fujimura, Osamu, Ed. Three dimensions of 1 inguistic theory . Tokyo: TEC Company, Ltd. , 1973.

PAGE 111

106 Ruddell , Robert B. The effect of oral and written patterns of language structure on reading comprehension. The Reading Teacher , 1965, 18, 270-275. (a) Ruddell, Robert B. The effect of the similarity of oral and written patterns of language structure on reading comprehension. Elementary English , 1965, 42, 403-410. (b) Savin, H. B., & Perchonock, E. Grammatical structure and the immediate recall of English sentences. Journal of Verbal Learning and Verbal Behavior , 1965, 4, 348-353. f* Schlesinger, I. M. Sentence structure and the reading process . The Hague: Mouton, 1968. Schmidt, Eunice L. What makes reading difficult: the complexity of structures . Paper presented at the annual meeting of The National Reading Conference, 1977. (ERIC Document Reproduction Service No. ED 149297). Schwartz, D. , Sparkman, J. P., & Deese, J. The process of understanding and judgments of comprehensibil ity. Journal of Verbal Learning and Verbal Behavior , 1970, 9, 87-93. (Shook, Ron. Discourse structure in reading. TESL Reporter , 1977, L 10 (2, 3, 4). Shore, Anne Carter. The relationship of an adult education student's 2 ability to process linguistic transformations and his or her ability to perform on a test of reading comprehension (Doctoral dissertation, University of Florida, 1976). Dissertation Abstracts International , 38, 678-A (77-17,060). Siegel , Sidney. Nonparametric statistics for the behavioral sciences . New York: McGraw Hill Book Co., 1956. imons, Herbert D. Linguistic skills and reading comprehension . Office of Education Grant, August 1970. (ERIC Document Reproduction Service No. ED 047927). Slobin, Dan I. Grammatical transformations and sentence comprehension in childhood and adulthood. Journal of Verbal Learning and Verbal Behavior , 1966, 5, 219-227. \ Smith, William L. The effect of transformed syntactic structures on reading. In Braun, Carl, Ed. Language, reading, and the communication process . Newark, Delaware: International Reading Association, 1971 .

PAGE 112

107 Stieglitz, Francine B. Teaching a second language: sentence length and syntax . Urbana, Illinois: National Council of Teachers of English, 1973. Tatham, Susan M. Reading comprehension of materials with selected oral language patterns. Reading Research Quarterly , 1969-1970, 5, 402-426. Ponis, Eleanor. Teaching reading to non-English speakers . New York: Collier MacMi 11 an International, Inc., 1970. Thurstone, The! ma Gwinn. Reading for Understanding General reading kit (revised edition). Chicago: Science Research Associates, 1969. Van Metre, Patricia D. Syntactic characteristics of selected bilingual children (Doctoral dissertation, University of Arizona, 1972). Dissertation Abstracts International , 1973, 33, 5160A (73-07806T /Wang, M. D. The role of syntactic complexity as a determiner of comprehensibil ity. Journal of Verbal Learning and Verbal Behavior , 1970, 9, 398-404. Wearing, A. J. The storage of complex sentences. Journal of Verbal Learning and Verbal Behavior , 1970, 9, 21-29. Wright, Patricia. Two studies of the depth hypothesis. British Journal of Psychology , 1969, 60_, 63-69. Yngve, V. A model and an hypothesis for language structure. Proceedings American Philosophical Society , 1960, 104 , 444-466. "You must follow the law as I state it:" do jurors really understand this routine instruction? The Linguistic Reporter , September 1978, 2J_ (1), 2. Zukowski/Faust, Jean Ann. Readability criteria used in materials selection for English as a second language (Doctoral dissertation, University of Arizona, 1978). Dissertation Abstracts International 39, 6744-A (7909486).

PAGE 113

BIOGRAPHICAL SKETCH Eileen Kay Blau was born in New York City on April 22, 1946. She graduated from Scarsdale High School in 1963 and received her B.A. in French from the University of California in 1967. She received an M.A. in I be roAmerican Studies from the University of Wisconsin in 1971. Between 1967 and 1975 she taught in the Dominican Republic as a Peace Corps Volunteer and in Honduras and Uruguay as an English teaching fellow. In 1975 she entered graduate school at the University of Florida and in 1977 received an M.A. in linguistics with a specialization in TESL. She was a graduate teaching assistant at the English Language Institute while working on her master's degree. Upon completion of that degree she began work on a doctorate in bilingual education with the support of a Title VII fellowship from the Department of Health, Education and Welfare. Since January 1980 she has been on the faculty of the English Department at the Mayaguez Campus of the University of Puerto Rico. 108