<%BANNER%>

The Development of the Obligation System in Mexican Spanish

Permanent Link: http://ufdc.ufl.edu/UFE0044551/00001

Material Information

Title: The Development of the Obligation System in Mexican Spanish a Variationist Sociolinguistic Perspective
Physical Description: 1 online resource (212 p.)
Language: english
Creator: Honea, Katherine
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: deontic -- grammaticalization -- modality -- obligation -- spanish
Spanish and Portuguese Studies -- Dissertations, Academic -- UF
Genre: Romance Languages thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Research that examines diachronic change and modality positthat modal verbs follow certain universal paths of development (e.g.  Cornillie, 2007; Bybee & Fleischman, 1995;Bybee, Perkins & Pagliuca, 1994). The present study examines thedevelopment of Spanish modality in Mexico through the use of multivariateanalyses, relative frequencies, and the comparative method as a means touncover where in the grammar changes are taking place. In addition, this studyexposes the subtle semantic differences of four obligation markers – haber de, tener de, tener que, hay que-commonly used in Mexico between the 16th and 21st centuries. Results from this study aim to explain the use of onemarker over another by operationalizing various factors found to contribute tochanges in modal systems across languages. The factors examined include: typeof verb, animacy/grammatical person, type of sentence, tense-aspect-mood,temporal markers, polarity, objects (direct, indirect), preceding clitic,lexical verb type, sex and age. Working within the usage-based framework, 5691tokens were extracted, analyzed and compared across centuries using relativefrequencies and the statistical program GoldVarb (Cedergren and Sankoff, 1974). Results indicated that there is an overall shift inpreference from haber de to tener que, suggesting that the system isundergoing longitudinal renewal. This is supported by the dramatic increase infrequency of tener que between the 19thand 20th centuries and the sharp decrease of haber de during this same period. A statistical examination of thisshift indicates that change is occurring within grammatical person/animacy. Researchon modality (e.g. Bybee and Fleischman, 1995) suggests that, as obligationmarkers develop, they will occur more often in contexts with third-personsubjects so it is not surprising that animacy/grammatical person may be an areain the grammar where change takes place. In addition to the outcome from the 19thand 20th centuries, the results also expose the various factors thatcondition the use of each of these obligation markers over time. Throughevidence found in processes such as semantic weakening, decategorialization, theresults corroborate postulations of a universal path of change inmodality.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Katherine Honea.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Aaron, Jessica.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044551:00001

Permanent Link: http://ufdc.ufl.edu/UFE0044551/00001

Material Information

Title: The Development of the Obligation System in Mexican Spanish a Variationist Sociolinguistic Perspective
Physical Description: 1 online resource (212 p.)
Language: english
Creator: Honea, Katherine
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: deontic -- grammaticalization -- modality -- obligation -- spanish
Spanish and Portuguese Studies -- Dissertations, Academic -- UF
Genre: Romance Languages thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Research that examines diachronic change and modality positthat modal verbs follow certain universal paths of development (e.g.  Cornillie, 2007; Bybee & Fleischman, 1995;Bybee, Perkins & Pagliuca, 1994). The present study examines thedevelopment of Spanish modality in Mexico through the use of multivariateanalyses, relative frequencies, and the comparative method as a means touncover where in the grammar changes are taking place. In addition, this studyexposes the subtle semantic differences of four obligation markers – haber de, tener de, tener que, hay que-commonly used in Mexico between the 16th and 21st centuries. Results from this study aim to explain the use of onemarker over another by operationalizing various factors found to contribute tochanges in modal systems across languages. The factors examined include: typeof verb, animacy/grammatical person, type of sentence, tense-aspect-mood,temporal markers, polarity, objects (direct, indirect), preceding clitic,lexical verb type, sex and age. Working within the usage-based framework, 5691tokens were extracted, analyzed and compared across centuries using relativefrequencies and the statistical program GoldVarb (Cedergren and Sankoff, 1974). Results indicated that there is an overall shift inpreference from haber de to tener que, suggesting that the system isundergoing longitudinal renewal. This is supported by the dramatic increase infrequency of tener que between the 19thand 20th centuries and the sharp decrease of haber de during this same period. A statistical examination of thisshift indicates that change is occurring within grammatical person/animacy. Researchon modality (e.g. Bybee and Fleischman, 1995) suggests that, as obligationmarkers develop, they will occur more often in contexts with third-personsubjects so it is not surprising that animacy/grammatical person may be an areain the grammar where change takes place. In addition to the outcome from the 19thand 20th centuries, the results also expose the various factors thatcondition the use of each of these obligation markers over time. Throughevidence found in processes such as semantic weakening, decategorialization, theresults corroborate postulations of a universal path of change inmodality.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Katherine Honea.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Aaron, Jessica.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044551:00001


This item has the following downloads:


Full Text

PAGE 1

1 THE DEVELOPMENT OF T HE OBLIG ATION SYSTEM IN MEXI CAN SPANISH: A VARIATIONIST SOCIOLI NGUISTIC PERSPECTIVE By KATHERINE HONEA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMEN T OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Katherine Honea

PAGE 3

3 To my sister, the memory of your laugh, your smile, or those nights of us just laying on the bed talking has helped get me through the most stressful of times, I love you

PAGE 4

4 ACKNOWLEDGMENTS I want to express my gratitude to a number of people who have supported me throughout my education and through this dissertation process. First, I want to thank the faculty for all their support I am eternally indebted to Jessi Aaron, who has spent countless hours offering invaluable advice, both as an advisor and as a friend. I remember so clearly that first class I took and the overwhelming yet satisfying feeling I had as she opened my eyes to functionalism. That class was a turning point for me and I am truly grateful for that. The opportunity to work with such a remarkable scholar and to learn so much about functionalism and language change has been an expe rience that I will never forget I am also grateful to Ana de Prada Perez and Geraldine Nichols for the many lessons they have taught me about writing and organization. It is because of them that my writing has improved so much over the last four years. Many thanks also go to Gillian Lord, wh o has been supportive since the day I arrived in Gainesville. She has believed in me and given me the opportunity to teach content courses and has always encouraged me incorporate new ideas into my lessons She has been a tremendous support throughout my c areer and I feel lucky to have her as a friend and colleague. I am equally appreciative to David Pharies and Diana Boxer for being the voice s of reason throughout this process W hether it was through occasional visits to their office s or through email s t he y ha ve always managed to keep me grounded or make me laugh. I am grateful to Jason Rothman for being so supportive of my development as a professor. Whether it was his advice in my practice job talks or conference presentations or his supervision and tra ining of my teaching, he has helped build my confidence and has tru ly prepared me to teach content based courses. I would

PAGE 5

5 also like to thank Wind Cowles, who allowed me to look at this research project from a new perspective with her comments and suggestio ns. In general, t he Spanish and Portuguese Studies department ha s provided me with endless support, ranging from Bobbye and Tania making sure all the necessary paperwork was done, to mock interviews and job talks with half of the faculty present I am rem inded on a daily basis of how lucky I am to have be en part of such a supportive department. Of course, I cannot begin to say how grateful I am to my friends and family. The countless way s they have shown their support through letters, phone calls or just listening to me have been a guiding light throughout my life Knowing that I have support from friends and family like Diego, Laurie, Briane, Valerie, Donna, Don, Raciel, Diana, Kathleen, Sabrina, Brian, Jake, Shawna, Eric, Bill Sandy Pilla, Ariana and a ll my friends in Guanajuato (and also those who I have not mentioned) has been a constant comfort and driving force in my life I have always known that I am surrounded by some exceptional people but this process has made that fact even more apparent to m e. I am particularly grateful to Delano Lamy, Maria Fionda and Kimberly Louie who have been pillars of support for me. I would not you in my life. You three have listened to and sympathized with every painful (and joy ful) step in this process. I love you all so very much and am fortunate to count you three as part of my extended family A special thank you is in order for my parents, who have full heartedly supported me in every adventure I have ever embarked on in lif e, despite their own personal misgivings. They have some how always managed to know when I needed a little break,

PAGE 6

6 a quick laugh, a homemade card from my nieces or just a hug. They define true support and I am forever grateful to the two of them. I would al so like to thank my little Gainesville family my husband Chad and my cats Muki, Chucha, and Marley, who continue to love me unconditionally despite months and months of stress. I am truly blessed to be married to such an accommodating, dedicated and comp assionate man. The unfaltering support he gives me through the countless meals he cook s the constant words of encouragement, and the weekly serenades continue to keep me sane. Also, I do not know how I would have made it through each day, locked up writin g in my house without the presence of my three cute and devoted kitties Ultimately, t his research project would not have been possible without the careful guidance of my committee members, the continuous support of my department, and the love and confide nce from my friends and family. I have so many people to thank for helping me get through this program.

PAGE 7

7 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF T ABLES ................................ ................................ ................................ .......... 10 LIST OF FIGURES ................................ ................................ ................................ ........ 12 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 1.1 Defining Modality in a Dynamic System ................................ ............................ 16 1.1.1 Classification Systems and Terminology ................................ ................. 16 1.1.2 Diachronic Change in Modality ................................ ................................ 23 1.1.3 Modality in Spanish ................................ ................................ ................. 26 1.2 Usage Based Models ................................ ................................ ........................ 29 1.2.1 Grammaticalization Theory ................................ ................................ ...... 31 1.2.2 Mechanisms in Grammaticalization ................................ ......................... 36 1.2.3 Subjectification ................................ ................................ ........................ 40 1.2.4 Subjectification and Grammaticalization. ................................ ................. 44 1.3 The Obligation System in Spanish ................................ ................................ .... 46 1.3.1 The (Non)grammaticalization of These Constructions ............................. 47 1.3.2 The (Non)subjectification of These Constructions ................................ ... 51 1.4 A Variationist Perspective ................................ ................................ ................. 54 1.4.1 The Variationist Method ................................ ................................ ........... 54 1.4.2 Com parative Method ................................ ................................ ............... 56 1.5 The Present Study ................................ ................................ ............................ 57 2 METHODOLOGY ................................ ................................ ................................ ... 59 2.1 Corpora ................................ ................................ ................................ ............. 59 2.1.1 Written Corpora ................................ ................................ ....................... 59 2.1.2 Oral Corpus ................................ ................................ ............................. 61 2.1.2. 1 The speech community ................................ ................................ .. 62 2.1.2.2 Participant selection ................................ ................................ ....... 63 2.2 Data Collection ................................ ................................ ................................ 65 2.3 The Variable Context and Form Function Asymmetry ................................ ...... 66 2.4 Circumscribing the Variable Context ................................ ................................ 67 2.5 Variabl e Rule Analysis ................................ ................................ ...................... 68 2.6 Comparative Sociolinguistic Method ................................ ................................ 71 2.7 Variables and Hypotheses ................................ ................................ ................ 74 2.7.1 Verb Type ................................ ................................ ................................ 75 2.7.2 Subject and Animacy ................................ ................................ ............... 76

PAGE 8

8 2.7.3 Sentence Type ................................ ................................ ........................ 78 2.7.4 Polarity ................................ ................................ ................................ .... 80 2.7.5 Objects ................................ ................................ ................................ .... 81 2.7.6 Preceding Clitic ................................ ................................ ....................... 83 2.7.7 Temporal Adverbs ................................ ................................ ................... 84 2.7.8 Tense, Aspect, Mood ................................ ................................ ............... 86 2.7.9 Lexical Verb Type ................................ ................................ .................... 87 2.7.10 Extralinguistic Variables ................................ ................................ ........ 89 2.7.10.1 Age. ................................ ................................ .............................. 90 2.7.10.2 Sex ................................ ................................ ............................... 91 2.8 Summary ................................ ................................ ................................ .......... 92 3 COMPETITION, VARIATION AND LANGUAGE CHANGE ................................ .... 93 3.1 Gramm aticalization of Changing Forms ................................ ............................ 93 3.2 Frequency and Language Change ................................ ................................ ... 96 3.2.1 Verb Type ................................ ................................ ................................ 98 3.2.1.1 Tener que ................................ ................................ ....................... 98 3.2.1.2 Haber de ................................ ................................ ...................... 100 3.2.1.3 Haber que ................................ ................................ .................... 101 3.2.1.4 Tener de ................................ ................................ ....................... 102 3.2.2 Grammatical Person and Animacy ................................ ........................ 103 3.2.2.1 Haber de ................................ ................................ ...................... 105 3.2.2.2 Tener de ................................ ................................ ....................... 106 3.2.2.3 Tener que ................................ ................................ ..................... 107 3.2.2.4 Haber que ................................ ................................ .................... 108 3.2.3 Sentence Type ................................ ................................ ...................... 109 3.2.3.1 Haber de ................................ ................................ ...................... 109 3.2.3.2 Tener de ................................ ................................ ....................... 110 3.2.3.3 Tener que ................................ ................................ ..................... 111 3.2.3.4 Haber que ................................ ................................ .................... 111 3.2.4 Polarity ................................ ................................ ................................ .. 112 3.2.5 Preceding Clitic ................................ ................................ ..................... 113 3.2.6 Direct Objects ................................ ................................ ........................ 114 3.2.6.1 Haber de ................................ ................................ ...................... 115 3.2.6.2 Tener de ................................ ................................ ....................... 116 3.2.6.3 Tener que ................................ ................................ ..................... 116 3.2.6.4 Haber que ................................ ................................ .................... 117 3.2.7 Indirect Objects ................................ ................................ ..................... 118 3.2.8 Tense Aspect Mood ................................ ................................ .............. 120 3.2.8.1 Haber d e ................................ ................................ ...................... 121 3.2.8.2 Tener de ................................ ................................ ....................... 122 3.2.8.3 Tener que ................................ ................................ ..................... 12 3 3.2.8.4 Haber q ue ................................ ................................ .................... 124 3.2.9 Temporal Adverbs ................................ ................................ ................. 124 3.3 Multivariate Analyses ................................ ................................ ...................... 126 3.3.1 Haber de vs. Tener que : 19 th Century ................................ .................. 128

PAGE 9

9 3.3.2 Haber de vs. T ener que : 20 th Century ................................ ................... 130 3.3.3 Haber de vs. Haber que : 20 t h Century ................................ ................... 134 3.3.4 Tener que vs. Haber que: 20 th Century ................................ ................ 135 3.4 Comparative Analysis ................................ ................................ ..................... 137 3.5 DLNE and Oral Corpus ................................ ................................ ................... 139 3.6 Summary ................................ ................................ ................................ ........ 144 4 DATA ANALYSIS ................................ ................................ ................................ .. 145 4.1 Identifying the Path of Change ................................ ................................ ........ 147 4.1.1 Grammaticalization ................................ ................................ ................ 148 4.1.2 Subjectification ................................ ................................ ...................... 153 4.2 Data Analysis ................................ ................................ ................................ .. 155 4.2.1 Tener de ................................ ................................ ................................ 156 4.2.2 Tener que ................................ ................................ .............................. 161 4.2.3 Haber de ................................ ................................ ................................ 165 4.2.4 Haber que ................................ ................................ .............................. 168 4.3 General Discussion ................................ ................................ ......................... 172 5 FUTURE DIRECTIONS AND CONCLUSIONS ................................ .................... 181 5.1 Future Directions ................................ ................................ ............................. 181 5.2 General Summary ................................ ................................ ........................... 184 APPENDIX A LIST OF TEXTS USED IN WRITTEN CORPORA ................................ ................ 189 B SAMPLE QUESTIONS FOR SOCIOLINGUISTIC INTERVIEW ........................... 195 C COMPLETE LIST OF FACTOR GROUPS AND FACTORS ................................ 198 LIST OF REFERENCES ................................ ................................ ............................. 200 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 212

PAGE 10

10 LIST OF TABLES Table page 1 1 printed with permission) ................................ ................................ ................................ ......... 53 2 1 Frequency count of each variant per century in the written corpora ................... 60 2 2 Frequency count of each variant per century in the oral an d DLNE corpora ...... 61 2 3 Exclusions of tener que/de (T) and haber de/que diachronically ........................ 66 3 1 Absolute and relative frequency of e ach variant per century in the written corpora normalized per 5000 words ................................ ................................ ... 97 3 2 Frequency of tener que with the factor group verb type ................................ ..... 99 3 3 Frequency of haber de with the factor group verb type ................................ .... 101 3 4 Relative frequency of haber que with the factor group verb type ...................... 102 3 5 Frequency of tener de with the factor group Verb type ................................ ..... 103 3 6 Frequency of haber de with the factor group grammatical person .................... 105 3 7 Frequency of tener de with the factor group grammatical person ..................... 106 3 8 Frequency of tener que with the factor group grammatical person ................... 108 3 9 Frequency haber que with the factor group grammatical person ...................... 109 3 10 Frequency haber de with the factor group type of sentence ............................. 110 3 11 Frequency tener de with the factor group type of sentence .............................. 110 3 12 Frequency tener que with the factor group type of sentence ............................ 111 3 13 Frequency haber que with the factor group type of sentence ........................... 111 3 14 Frequency of all constructions with the factor group polarity ............................ 112 3 15 Frequency of all constructions with the factor group preceding clitic ................ 113 3 16 Frequency of haber de with the factor group di rect objects .............................. 115 3 17 Frequency of tener de with the factor group direct objects ............................... 116 3 18 Frequency of tener que with the factor group direct objects ............................. 117

PAGE 11

11 3 19 Frequency of haber que with the factor group direct objects ............................ 118 3 20 Frequency of all factors occ urring with the factor group indirect objects .......... 119 3 21 Frequency of haber de occurring with the factor group tense aspect mood ..... 122 3 22 Frequency of tener que occurring with the factor group tense aspect mood .... 123 3 23 The occurrence of all obligation markers with temporal adverbial Markers across centuries ................................ ................................ ............................... 125 3 24 A nalysis from the 19th century: Linguistic factors conditioning haber de occurrence in comparison with tener que in CORDE. ................................ ...... 130 3 25 Data from the 20th Century: Linguistic factors conditioning haber de occurrence in comparison with tener que in CORDE. ................................ ...... 133 3 26 Data from the 20 th century: Linguistic factors conditioning ha ber de occurrence in comparison with haber que in CORDE ................................ ..... 135 3 27 Data from the 20 th century: Linguistic factors conditioning tener que occurrence in comparison with haber que in CORDE. ................................ ..... 136 3 28 Variation of haber de and tener que across two centuries ................................ 137 3 29 Analysis of constrain hierarchies of haber de vs. tener que across the 19th and 20th centuries. ................................ ................................ ........................... 138 3 30 Data from the 21 st Century: Frequency of occurrence of each obligation marker within factor group verb type ................................ ................................ 140 3 31 Data from the 21 st Century: Frequency of occurrence of each obligation marker according to factor group ................................ ................................ ...... 142 3 32 Data from the 21 st Century: Frequency of occ urrence of each obligation marker within factor group tense aspect mood (TAM) ................................ ...... 143

PAGE 12

12 LIST OF FIGURES Figure page 4 1 Frequency of variants across cent uries ................................ ............................ 146 4 2 Comparison of tener de as an possession marker and as an obligation marker across centuries ................................ ................................ ................... 161

PAGE 13

13 Abstract of Dissertation Presented to the Gradu ate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE DEVELOPMENT OF THE OBLIGATION SYSTEM IN MEXICAN SPANISH : A VARIATIONIST SOCIOLINGUISTIC PERSPECTIVE By Katherine Honea August 2012 Chair: Jessi Aaron Major: Romance Languages Research that examines diachronic change and modality posit that modal verbs follow certain universal paths of development ( e.g. Cornillie, 2007; Bybee & Fleischman, 1995; Bybee, Perkins & Pagliuca, 1994). The present study examines the development of Spanish modality in Mexico through the use of multivariate analyses, relative frequencies and the comparative method as a means to uncover where in the grammar changes are taking place. In a ddition, this study exposes the subtle semantic differences of four obligation markers haber de tener de, tener que h ay que commonly used in Mexico between the 1 6 th and 2 1 st centuries. Results from this study aim to explain the use of one marker over another by operationalizing various factors found to contribute to changes in modal systems across languages. The factors examined include: type of verb, animacy/grammatical person, type of sentence, tense aspect mood, temporal markers, polarity objects (direct, indirect), preceding clitic lexical verb type, sex and age Working within the usage based framework, 5691 tokens were extracted, analyzed and compared across centuries using relative frequencies and the statistical program GoldVarb (Cedergren an d Sankoff, 1974).

PAGE 14

14 Results indicated that there is an overall shift in preference from haber de to tener que suggesting that the system is undergoing longitudinal renewal. This is supported by the dramatic increase in frequency of tener que between the 19 th and 20 th centuries and the sharp decrease of haber de during this same period A statistical examination of this shift indicate s that change is occurring within grammatical person/animacy. Research on modality (e.g. Bybee and Fleischman, 1995) suggests that as obligation markers develop they will occur more often in contexts with third person subjects so it is not surprising that animacy/grammatical person may be an area in the grammar where change takes place. In addition to the outcome from the 19 th and 20 th centuries the results also expose the various factors that condition the use of each of these obligation markers over time. Through evidence found in processes such as semantic weakening, decategorialization, the r esults corroborate postulations of a universal path of change in modality.

PAGE 15

15 CHAPTER 1 INTRODUCTION Modality is an essential element of human experience and thus a fundamental part of how we think and express ourselves. It comprises possible and necessary truths and allows us to form t houghts and/or statements about whether something is conceptually, logically, physica lly or metaphysically possible or necessary (Melia, 2003). These truths are realized in language through numerous channels. For example, in Spanish modality is expressed i n a variety of ways, from mood subjunctive, indicative (e.g. Solano Araya, 1984; Silva Corvaln, 1985), to modal adverbs probablemente ciertamente (e.g. Hengeveld, 1988), to lexical verbs obligar necesitar (e. g. Palmer, 2001), or modal periphrases (e.g. Yllera, 1980). It is the expression of modality through the use of modal periphrases that is the focus of the present study, particularly tener de (1.1) hay que (1.2) tener que (1.3) and haber de (1.4) all of which express strong obligation. 1.1 Fabio Yo tengo de ir con vos. (Pedro Caldern de la Barca, 1629 : CORDE ) Fabio I have to go with you. 1 1.2 No hay que mirarme, que vos no merecsteis descalzarme. (Moreto, Agustn, 1658: CORDE) 1.3 Siendo esa vuestra opinin, ya no tengo que os decir. (Caldern de la Barca,Pedro, 1649) 1.4 Seor, con qu mandamiento te he de matar? (Quirs, Francisco Bernardo de, 1656: CORDE) Sir, with what order do I have to kill you? 1 All translations are mine unless otherwise noted.

PAGE 16

16 The purpose of this study is to examine these strong obligation markers diachronically. This study will contribute to the limited literature on di achronic variation and change within modality in Spanish. 1. 1 Defining Modality in a Dynamic System The concept of modality has long been a topic in the field of logic. Outside of linguistics, it has traditionally been associated with possibility and nece ssity (Coates, 1983; Kiefer, 1987; Narrog, 2005a, 2005b). Unsurprisingly, this association has led to a tendency to classify linguistic modality in a similar way, which has been problematic at best, as we will see below in section 1.2.1 Subsequently, a la rge body of research on modality has focused primarily on two areas: 1) defining and classifying modality at both the macro and the micro level; and 2) diachronic changes that occur in modal systems. In the following sections, these areas will be examined. 1. 1 .1 Classification Systems and Terminology Defining modality is a particularly challenging task, given that it is a universal feature of human languages and therefore categorizing it should necessarily encompass modality in all languages (Coates, 1983; Bybee, Perkins, & Pagliuca, 1994; Narrog, 2005a). Although there is general agreement that modality is a semantic category (Fleischman, 1982; Bybee & Pagliuca, 1985; Bybee & Fleischman, 1995; Palmer, 2001), when it comes to a formal definition, researcher s have disagreed as often as they have agreed. Palmer (2001 : 1 ) for example, states that modality refers specifically to the proposition that describes an event while Kiefer (1987) views it as related to possibility and necessity, basing his conceptualiza tion on modal logic. On the other hand, Bybee et al. (1994 : 176 ) suggest it may be impossible to define modality in such a way that includes both a notional domain of modality and the part of it that is

PAGE 17

17 expressed grammatically They therefore take a diff erent tactic and avoid a definition altogether, instead opting to let the results of a cross linguistic diachronic study on modality speak for themselves. Coates (1983 : 10 ) also voices her concerns about achieving a uni forma with precise categories and have failed to acknowledge the problem of indeterminacy in natural language However, Coates (1983: 235) goes on to broadly categorize modali ty non factive Other linguists have echo ed Coates definition (e.g. Lyons, 1977; Narrog, 2005a, 2005b) and its proclivity to encompass the many types of modality. Narrog (2005a, 2005b), who uses the term FACTUALITY : Modality is a linguistic category referring to the factual status of a state of affairs. The expression of a state of affairs is modalized if it is marked for being undetermined with respect to its factual status, i.e. is neither positively nor negatively factual. (p. 679) Taking this definition into account, and everything factuality can include in a language, it is not surprising that linguists have deemed it necessary to establish a typology of the different categories of modality that the various levels of factuality include. It is widely accepted that modality is comprised of mood and modality systems; however, there is no consensus on how to separate these into more precise categories. Heine, Claudi and Hnnemeyer (1991) reiterate the amb iguity of modality and the general difficulty of defining forms by discrete categories. They claim that the use of discrete categories does not give a complete picture of the form because, in reality, the meaning of a form is part of a continuum, often ref erred to in the grammaticalization literature as a cline and can vary greatly: categories such as constituent types or morpheme classes, a more

PAGE 18

18 appropriate approach would be that wh ich highlights the continuum nature of linguistic structures (1991: 3). This approach, however, has not been widely adopted in the literature. Instead, a wide array of classifications has been postulated to account for its semantic (and grammatical) multi plicity. Generally, these typologies fall into one of two categories: bipartite systems, such as those of Lyons (1977), Givon ( 1990 [ 1984 ] ), Bybee and Fleischman (19 9 5), and Palmer (2001); and those systems that consist of more than two categories, like th ose proposed by Bybee et al. (1994), Skotarek, (1996) and Narrog (2005a, 2005b). Of those that have a binary classification, the most traditional typology is DEONTIC / EPISTEMIC ( e.g. Solano Araya, 1982; Brinton, 1991 ), where deontic relates to obligation or permission, emanating from a n external source (Palmer, 2001: 10) and of (Coates, 1983: 18). Some authors, however, have found this typology to be too restrictive. Palmer (2001), for example, also divides modality into two systems, EVENT MODALITY and PROPOSITIONAL MODALITY but differentiates his classification by asserting that his taxonomy is more general and thus covers a broader range of modality. H e cites DYNAMIC MODALITY as an example of event modality that would otherwise be excluded from the traditional deontic/epistemic typology. Coates (1983) also addresses this issue by using the term ROOT MODALITY in place of deontic modality in order to cover a larger subset of modals (essentially adding ability to the group) their study on modals but use the term AGENT ORIENTED MODALITY They attempt to

PAGE 19

19 broaden t he scope of classification by positing a binary system of agent oriented modality and epistemic modality. They define agent oriented modality as modalities that predicate conditions of either an internal or external nature on a willful agent, which essen tially includes ability, obligation, desire and intention (p. 63). Again, like Palmer (2001) and Coates (1983), the aim was to include a broader range of modals than the prototypical deontic/epistemic typology. 2 All of these linguists address a valid issue by modifying the traditional deontic/epistemic classification while maintaining the bipartite system. In addition to binary classifications there have also been typologies including more than two categories. The purpose of these typologies has been to no t only account for the various types of modality not included in the aforementioned binary classifications, but also to account for typical changes found in modal systems, which will be discussed in further detail in section 1.2.2. Narrog (2005a) examines modality from a dimensional perspective by postulating two modal axes. On one end of the first axis are modals that are more objective, or EVENT ORIENTED (e.g. grammatical third person would be more event oriented) and on the other end are those that are more subjective or SPEAKER ORIENTED ( e.g. grammatical first person would be more speaker oriented ) Sh e calls the other axis the VOLITIVITY dimension which on the volitive end s he equates to deontic modality and the other end, non volitive, she correlates this end to epistemic evidential and dynamic modality, In essence, o n one end of this dimensio action that has to 2 the subject is not always the agent but can also be an experiencer or ev en not expressed (2005a: 682). Nevertheless, for the present paper this term will be adopted since it is part of the terminology in the theory of change proposed by Bybee et al. (1994) that will be used throughout this paper.

PAGE 20

20 be performed volitionally by the addressee or some other p erson understood in the (2005a: 683). On require such an action. While her argument for defining changes in modality with volition is convincing and her attempt to include SUBJECTIFICATION (i.e. speaker oriented) as a principle dimension is valuable, her model only provides a general path of development for all modals. In the classification put forth by Bybee et al. (1994), on the other hand, they identify the paths of each of the different types of modals and suggest specific paths (e.g. obligation>intention>fu ture>purpose/to want or order) for each of the different types of modals thereby offering a more specific account of development for modality. Bybee et al. (1994) identify four categories within modality systems: AGENT ORIENTED EPISTEMIC SPEAKER ORIENTED and SUBORDINATING MOODS Similar to Bybee and Pagliuca (1984), they define agent oriented modality as an element of internal or external will that is placed on the agent. They include in this classification obligation, necessity, desire, and ability (1994 : 178). Speaker oriented modality includes directives and utterances where permission is granted to the listener. They emphasize that speaker oriented modality does not include conditions (external or internal) on the agent, but instead allows the speaker to enact these conditions (p. 179). Their designation of epistemic modality concurs with other definitions by labeling it as the commitment of the speaker to the truth of the proposition (p. 179). The last modality type, subordinating moods, comprises the same constructions that are used in agent (e.g. obligation) and speaker oriented modality (e.g. directives) but are markers of the verb in specific types of subordinate clauses such as in the example given by Bybee et al. (1994: 180) in 1. 5 This typology is complete in that it incorporates an extensive

PAGE 21

21 range of modality (like illocutionary force, lexical verbs etc.) and it also considers mood in its typology which is often regarded as separate. Consequently, it is the typology adopted for the present disse rtation This does not negate, however, that this typology (as well as the others) fails to address the semantic nuances of these modals on a more micro level. 1. 5 I suggested that he should call you immediately. Determining the semantics of these modal markers has been as polemic as categorizing the different types of modals. Much of the research conducted on the semantic side of modality has determined the meanings of modals based on the psychological motives of the speaker (De Maeseneer, 1998; Myhill, 1995; Sirbu Dumitrescu, 1988). For example, Myhill (1995) claims that the changes that have occurred in the English modal system are due to a n overall shift in societal norms pointing out that, in the last century, the United States has undergone a collec tive shift from a focus on power and speaker control to a society that is focused on more like cooperation, advice etc. He claims that this shift is reflected in the development of the American English obligation system. Although the shift in society is observable, and thus is a possible explanation for changes found in the obligation system, it is difficult to attribute these changes to the societal shift without having a way to operationalize it. Myhill (1996) attempts to do this by explaining that the interest of the speaker is what determines which obligation marker will be used He categorizes modals with such names as benefits relationsh or meeting societal expectations A shift based on the moti ves of the speaker is difficult to prove, however.

PAGE 22

22 In fact, t he use of psychological motives as a way to determine semantic categories has been questioned in recent research. This is highlighted in a study on Romance futures by Poplack and Malvar (2007 : 1 62 ), who state that psychological motives have little empirical basis and that, in order to determine what conditions the use of one variant or another, empirical based analyses are required: very few of the motivations ascribed to variant choice in the ( prescriptive or descriptive) literature are now, or ever have been, relevant to actual usage. Grammarians have been silent on the role of the operative contextual factors, focusing instead on semantic, psychological and other motivations which have no basi s in empirical fact. Other modality studies have focused less on psychological motives, and many of these have determined the semantics of these modals through the use of frequency distributions. Cornillie (2007), for example, uses relative frequency as a measure in his study of modality in Spanish. He looks at the frequency with which tener que occurs in certain surrounding contexts revealing specified patterns of use. However, he looks primarily at the patterns of use of the epistemic use of tener que 3 (2007) also look at obligation in their study of Toronto English, where they determine the use of the obligation markers by examining the effect that various linguistic and extralinguistic factors have on these constructions. Studies s uch as these provide an objective means of determining the semantics of these modals and are preferable because they allow for easier replication. As mentioned above, Heine et al. (1991) suggest that one of the possible reasons for the challenge in uncove ring the specific functions of these modals is because 3 Cornillie reached numerous c onclusions that are beyond the scope of this study since his research was related to modal grounding in relation to evidential and epistemic modality.

PAGE 23

23 sometimes the uses can be ambiguous. Take the following sentence taken from Traugott and Dasher (2002: 2) as an example. 1. 6 They must be married. They describe that this sentence, when taken out of co ntext, can have two separate meanings: one expressing obligation and one with an epistemic sense. In the obligation sense it would mean something like It is whereas in the epistemic sense there is a high sense of certainty essentially expressing that the speaker is certain that the couple is married. As Traugott and Dasher (2002) point out, modals such as this example are polysem ic and occur in ambiguous contexts and this ambiguity must be taken into consideration. Ultimat ely, all of these studies reveal that many modals are polysemic (e.g. Coates, 1983) and appear to develop and acquire new meaning s as well as lose old meanings over time. One question that is raised by this perspective, then, is precisely how do these moda l markers acquire multiple meanings? 1. 1 .2 Diachronic Change in Modality Change has long been recognized as part of language and modality is not spared in this area. In fact, as previously mentioned, many of the typological studies (e.g. Lyons, 1977; Coat es, 1983; Bybee et al., 1994; Narrog, 2005a, 2005b) have identified patterns of change in modal systems. With the identification of these patterns, some linguists (c.f. Lyons, 1977, Bybee & Pagliuca, 1985, Traugott 1989) have taken it a step further and id entified universal tendencies in modality. That is they postulate that certain types of modality appear to develop in parallel ways across languages. Most notable is the model for universal changes in modality put forth by Bybee et al. (1994), in which the y outline several important universal tendencies through their data.

PAGE 24

24 One such proclivity is the frequency with which they find agent oriented modality (obligation, necessity, desire and ability) to be periphrastic (e.g. have to, tener que ) while the other three types (speaker oriented, epistemic, and subordinating moods) were more likely to be inflectional (e.g. synthetic future tense in Spanish: hablar ) 4 Taking into consideration that many linguists (e.g. Heine et al., 1991; Bybee et al. 1994) suggest that modality is better viewed as a continuum, or what is often referred as a cline [t]his correlation suggest[s] a diachronic scenario whereby as agent oriented modalities GRAMMATIC AL IZE ( i.e. change) they develop into the other types and gradually (Bybee et al. 1994 : 181 and references therein). This assumption suggest s then, that agent oriented modality in Spanish should be comprised of periphrastics which we see with the strong obligation markers tener que tener de, haber de, and hay que. Westney (1995) and Coates (1983) go a step further in suggesting that periphrastics that operate as auxiliary verbs are more developed along this cline than those that do not. Westney (1995 : 31 ) claims that these moda l auxiliaries are more likely to occur in epistemic contexts and states that the periphrastic items [i.e. semi auxiliaries] tend to manifest epistemic senses only to a relatively restricted degree As a result of the differences in development among sem i auxiliaries and auxiliaries, criteria for determining the status of a modal verb as either auxiliary or (1979) (Westney, 1995, and references therein). If a verb d oes not meet these 4 Bybee et al. (1994: 180) find that, of the 76 languages examined, 75% of the obligation markers were p eriphrastic and only 25% were affixes. This was a similar trend found in the various agent oriented modals for which they coded

PAGE 25

25 criteria, it is not considered to be a modal auxiliary but rather a periphrastic or lexical verb (Westney, 1995: 15). Considering that these criteria are specific to English, it is questionable whether the criteria Palmer (1979) and othe rs use are valid in Spanish. Additionally, Bybee et al. (1994) do not identify the development of periphrastics into modal auxiliaries as a universal tendency in modality although they do use the term oriented m odality thus implying that modal verbs can and do become auxiliaries. Another tendency Bybee et al. (1994) document is a unidirectional path of change from agent oriented modality to epistemic modality. 5 This path of change, a process referred to as gramm aticalization and, in certain stages, SUBJECTIFICATION will be discussed in greater detail in section 1.3.1 Given that epistemic modals tend to be inflectional, it can be deduced that the continued progress of these agent oriented modals towards epistemi city is accompanied by phonological reduction or fusion that ultimately results in inflection. The results of Bybee et al. (1994) reveal that these patterns occur when lexical verbs covering a small range of specific meanings begin to generalize semantical ly and accordingly move into new contexts. Similarly, they find that once epistemicity is part of the semantics of a modal marker, further generalization could occur whereby these forms could develop into subordinating moods 6 The tendencies observed by By bee et al. (1994) highlight the universality in modality and the formation of modals and mood systems. Given the relative youth of the field, however, 5 Narrog tries to challenge this idea by positing his two dimensional model. He claims that there is indeed unidirectionality bu t the path is from objective to subjective, not agent oriented to epistemic (2005a). 6 There is another possible path for agent oriented modality proposed by Bybee et al. whereby these markers acquire a speaker oriented notion, then generalize to express s ubordinating mood (1995; 241). Owing to the fact that the obligation markers in this paper do not appear to take this path, it is not the central focus.

PAGE 26

26 the empirical evidence supporting or refuting claims of cross linguistic diachronic universality is contr oversial and has prompted much research and heated debate (e.g. Nordlinger & Traugott, 1997; Axelrod, 1999; Kim, 2009). In order to corroborate or refute these claims of universality more empirical evidence is needed. This study is a step in that direction 1. 1 .3 Modality in Spanish I n Spanish, as in many other languages, modality is comprised of both mood and a modal system and can be expressed through the use of inflection ( e.g. ra ), periphrastics ( e.g. tener que, deber de ), adverbs ( e.g. probablemente, posiblemente ) or even lexical verbs ( e.g. mandar, sugerir ). Despite the association between mood and modality, the focus here will primarily center on modal markers given that these strong obligation constructions have yet to become inflectional. Spanish has not escaped the disagreement described above in terms of defining and classifying modality but research documenting modality in Spanish is minimal. Besides research focused on developing a typology for modality across languages including Spanish the limited research available on modality in Spanish has centered on the same two areas previously mentioned: the semantics of these individual modal markers and the diachronic changes these modals have undergone Molina Plaza (2005) for example looks at mod ality in Spanish and identifies the various uses of the modals tener que, deber, haber de and hay que She bases her semantic classes on the classification system adopted by Coates (1983). Using relative frequencies to determine which uses are most common among these modals she does not find any examples of epistemic uses of tener que, haber de or hay que in her data. This is surprising given the universal tendency mentioned above for these agent

PAGE 27

2 7 oriented modals to move towards and eventually acquire an epi stemic sense. Sirbu Dumitrescu (1988) also looks at the semantics of agent oriented modality in Spanish but often bases the uses of these forms on the psychological motives of the speaker which, as already mentioned, cannot produce empirically reliable res ults. In examining the diachronic path of Spanish obligation markers 7 research has largely focused on only one or two aspects of these tendencies. For example Cornillie (2007) focuses on the change from deontic to epistemic modality in two modal markers. H e takes a diachronic look at tener que and compares the deontic and epistemic uses of this obligation marker in various contexts. Other studies have focused on the status of these modal markers in Spanish, focusing on the aforementioned claim that modal a uxiliaries are more susceptible to change than periphrastics (Coates, 1983; Westney, 1995). That being said, the Spanish auxiliary system is practically non existent if we were to adopt the criteria put forth by linguists such as Palmer (1990) and Coates ( 1983) 8 But a s Cornillie (2007: 224) states unlike English modal auxiliaries, the Spanish modals have tense and mood inflections and hence, do not belong to a well defined morphosyntactic category of modal There have been attempts, albeit f ew, to identify criteria for modal auxiliaries, periphrastics and lexical verbs in Spanish. For example, Fernndez (1999: 22) reviews a number of criteria that have been used to identify periphrases in general and claims the best way to identify a periphra stic construction is the combination of an impersonal 7 The specific paths for each of these obligation markers will be discussed more in detail in section 1 .4. 8 Criteria for determining modal auxiliaries in English have been dubbed NICE properties which define the modal auxiliaries: negation, inversion, code, and emphatic affirmation. For a more in depth discussion on English modal auxiliaries see Coates (19 83), Palmer (1987), Westney (1995), or Krug (2000).

PAGE 28

28 form of the purported periphrastic alongside a phrase containing the infinitive haber (to) be, have 7 a,b ). 1. 7 a. {There may/ must/ tends b. Fernndez (1999) is also aware that there are exceptions to this criterion, most notably with the construction hay que Nevertheless, he considers this to be t he most reliable way to recognize a periphrastic in Spanish 9 Unfortunately, Fernndez uses the terms auxiliary and periphrastic interchangeably so, although he gives a clear criterion for differentiating between per iphrastic verbs and lexical verbs, he does not try to identify criteria for recognizing an auxiliary. Yllera also identifies criteria for determining the status of a modal verb and lists several conditions but, like Fernndez, she does not appear to diffe rentiate between the modal periphrastics and auxiliaries (1980). These studies lead to the conclusion given by Olbertz (1998; 33) haber ser tener que as periphrastics. would imply that these constructions only express epistemicity to a restricted degree. Again, this is an empirical question that requires a quantitative evaluation of the modal system in Spanish. Regardless of the answer, we now can at least identify the modal system as consisting of modal periphrastics. 9 Fernndez claims that this test does not work well with haber que for systematic reasons. Nevertheless a Google search of hay que haber turned out more than 32 million occurrences of this phrase.

PAGE 29

29 1. 2 Usage Based Models One way to account for the difficulty in finding clear definitio ns for grammatical categories such as modality (Company 2004) is through a usage based framework. Instead of grammatical categories, usage based models are comprised of a continuum with the prototypical uses found in the middle of this continuum and other, less common uses found towards the ends of this continuum ( for more on this see Company 2004; Heine et al. 1991). Importantly, usage based models provide or a way to map language so as to account for both language function and gen eral cognition under one model (Langacker, 2000). Langacker (2000: ix) who coined the term, also describes this model as being maximali st, non showing that, contrary to generative theory, it is not necessary for language structur e to be economic, nor does the grammar need to be based on general rules (or categories). Rather, grammar is the cognitive organization of (Bybee, 2006 : 711 ). Thus, this model emphasizes the relationship between experience, cognitive representation and the construction of grammar. The focus on experience as something central to language and change is in contrast to other theories that have focused on minimalism and economy and the basic idea that expressions are generally derivable by rule, [and] to list them individually wo (Langacker, 2000: 2). Instead, a usage based model focuses on a bottom up orientation using experience as way to create connections and, ultimately, structures in the bra in. Each utterance is stored not as a part of a rule or a list, but rather as an individual USAGE EVENT A fundamental part of how grammar develops and thus changes in usage based models is largely determined by the frequency of these utterances. As Langac ker (2000 : 3) explains, each usage event

PAGE 30

30 leaves a trace th at facilitates [its] recurrence The continual recurrence of this usage event leads to ENTRENCHMENT as an autonomous unit. As these utt erances become entrenched they are stored in close proximity to similar units. As Langacker (2000:7) points out, [t]he occurrence of any pattern close to it in state sp and can therefore act as an indicator of significance. Once similarities can be drawn from these structures a schema emerges and is considered a standard from which novel but similar structures will be compared (Langacker, 2000). Bybee provides evidence to this notion as she shows the development of prefabs a nd the high frequency of related word patterns. 10 She posits that the high recurrence of these prefabs in the spoken and written corpora, as well as the new constructions derived from these prefabs, are evidence for CATEGORIZATION or cognitive storage of p articular exemplars of constructions (2006; 729) 11 The formation, entrenchment and resulting categorization of word patterns and constructions through usage and frequency are important tenets of language change in a usage based model. Grammaticalization a nd subjectification, processes associated with diachronic language change, are intricately bound to the usage event and thus related to the entrenchment and categorization of patterns and, ultimately the formation of grammar. 10 Acc prefab is a combination of at least two words favored by native speakers in preference to an alternative combination which would have been equivalent had there been no conventionalization 11 The review of usage based models will include pertinent components only, since some aspects are beyond the scope of this study. However, for a detailed outline of the various principles of usage based models see Biber (1999), Barlow (1999), Langacker (1999), and Barlow and Kemmer (2000).

PAGE 31

31 1. 2 .1 Grammaticalization Theo ry As previously mentioned, it has long been recognized that languages undergo change. Among the various types of linguistic change is grammaticalization. This type of change has been studied extensively in historical linguistics, which consequently produc ed a theory known as GRAMMATICALIZATION T HEORY Although the concept of grammaticalization dates back to 1912 with Meillet, who first coined the term, the theoretical framework is still relatively new. Company (2004) summarizes the various aspects and defi nitions the theory entails by identifying three general views on grammaticalization. The first group, which she calls the TRADITIONALISTS adopts a more historical approach. This approach can be summarized with a commonly used definition put forth by Kury lowicz (1976 :69 [1965]) who states that [g]rammaticalization consists in the increase of range of a morpheme advancing from a lexical to a grammatical or from a less grammatical to a more grammatical status, e.g. from a derivati ve formant to an inflectio n one This definition is in line with Meillet (1912) and will be the definition adopted here. Linguists who subscribe to this view primarily see grammaticalization as a gradual process. As Heine et al. (1991) point out; however, this definition can also signify lexicalization. In an effort to confine the meaning of grammaticalization, other definitions have been posited. Hyman (1984: 73, 83), for example, regards this process to be more narrowly 6 [1965]) definition. Hyman (1984: 73) claims that grammaticalization is the harnes sing of pragmatics by a grammar Company (2004) identifies this second group as looking at this process as PRAGMATICALIZATION Linguists in this group view grammaticalization as a change that occurs based on the contexts in

PAGE 32

32 which the forms are used. That is, the more these forms are used in certain contexts the more possible it is for them to acquire new meaning (c.f. Traugott, 1982, 1989, 1999, 2003; Traugott & Dasher, 2002). Company (2004) different iates this group from like Bybee et al. (1994), Heine et al. (1991), Bybee & Pagliuca (1985) based on where the change occurs. She states that traditionalists see this change as a gradual process that occurs to the form due to phonologica l erosion, loss of autonomy etc. instead of basing change on the context (Company, 2004). The third position sees grammaticalization as a synonym of grammar (Heine et al., 1991). Linguists supporting this position claim that structure did not exist prior t o grammaticalization. They also see grammaticalization as similar to any other creative cognitive process of human beings (e.g. Company, 2004). This has also been referred to by Hopper (1987) as emergent grammar. R egardless of the position one takes, all agree that grammaticalization is a process. This process can be viewed from two different perspectives: diachronic or synchronic. Traditionally, grammaticalization has been studied from a diachronic perspective, given that it is a slow, gradual process. Ad opting a diachronic approach allows for the identification of the source of the forms under examination and also illustrates the various steps these forms take as they become more grammatical (Hopper & Traugott, 2003 [1993]). From a diachronic perspective, identifying changes in progress is more difficult since the data available do not include social factors, which are often crucial to language change. Company (2004 : 24 ) states that, in order to identify these changes diachronically, it is necessary to loo k at the forms within a single text to see if alternation exists. In recent years linguists have also adopted a synchronic

PAGE 33

33 approach, focusing primarily on the variation within the system to determine whether a change is in progress. Synchronic grammatical ization is often identified when two or more forms are used for what appears to be the same function. A synchronic approach allows for the incorporation of social factors given that the data is recent. 12 A synchronic approach also reveals the lack of discre te categories of these forms and highlights the fluid (Hopper & Traugott, 2003: 3). Additionally, Labov (1972) indicates that synchronic variation is often a reflection of diachronic change so detecting changes in a modern system is essentially a window into the past. I n fact, i t has been pointed out in the literature that forms often retain connotations that are related to their original meaning (Bybee et al. 1994; Poplack & Tagliamonte, 2001). However, as Giv n (19 90 [1984]: 234 often necessary in order to explain the reasons why and how these patterns arose. Given the importa nce placed on both a diachronic and synchronic account of language, both approaches were adopted in this study. An important principle that has been widely accepted as a part of grammaticalization theory is the PRINCIPLE OF UNIDIRE CTIONALITY which crucial ly identifies grammaticalization as a process that is unidirectional. In recent years however, there have been numerous counterexamples to this principle (Newmeyer, 1998; Campbell, 2001), which have consequently called the theory of grammaticalization into question (Fischer, Norde & Perridon, 2004). 12 A more detailed look at the benefits of diachronic study vs. synchronic stu dy will be offered in Chapter 2

PAGE 34

34 Haspelmath (2004 : 3 ) confronts this issue and admits that there are a limited number of counterexamples like the change from an affix to a clitic in Continental Scandinavian, but that, overall, grammaticalizat ion is unidirectional. In addition, he adds that the goal in historical linguistics is to identif and classifies unidirectionality in grammaticalization as one of the most important universals in language change given that it is so widespread (2004: 35). Haspelmath (2004) also discredits many of the criticisms of claims of unidirectionality citing several reasons as to why they were not counterexamples. Other studies have addressed the principle of unidirectionality and its validity (c.f. Taeymans, 2004, Hopper & Traugott, 2003 [1993]) where general results may vary slightly from Haspelmath (2004) but ultimately all arrive at the same conclusion: the general pattern of grammaticalization is unidirectional. 13 Given the differe nt interpretations as to what grammaticalization theory entails and the counterexamples to the principle of unidirectionality, there has been debate as to whether or not grammaticalization is a theory (Fischer, Norde, & Perridon, 2004). Haspelmath (2004) a rgues that, although grammaticalization theory encompasses many approaches, it can still be a theory. of related approache s and basic issues in (2004: 23). Company (2004: 1) also sees grammaticalization as a theory and argues that if one adopts the functionalist perspective of what grammatic al ization is a predictive line of investigation that accounts for then one can say it i s a theory. For the sake of the present paper, grammaticalization will be considered a 13 For more on the discussion of unidirectionality and these counterexamples see also Hopper and Traugott (1993) Tabor and Traugott (1998), Traugott (1995) and Narrog (2005).

PAGE 35

35 theoretical framework and furthermore, the definition put forth by Kurylowicz (1975:52 [1965]) will b e the definition adopted. Research within the grammaticalization framework has concluded that areas of grammar that refer to specific conceptual domains are more apt to undergo grammaticalization. Bybee (1998 : 258) asserts that only a small set of words i n a language are susceptible to grammaticalization and ascerta ins that they typically relate to notions about human experience across cultures: she offers as specific examples of words referring to spatial concepts, such as the body, movement in space etc. Traugott & Dasher (2002 : 3 ) similarly identify areas as primary hot spots for grammaticalization but add to the list by including aspect ( have, finish ) and modality ( want, will ) The changes that typically occur to the last group modality are further de scribed as a general semantic pragmatic shift. This occurs as they acquire a more epistemic and thus subjective meaning, a process known as SUBJECTIFICATION (Traugott, 1989; Bybee et al., 1994; Tagliamonte, 2004; Narrog, 2005; Tagliamonte & Smith, 2006; Ta gliamonte An important aspect in grammaticalization theory to consider is frequency. As mentioned above particular consideration is given to frequency in a usage based model as it is through these high frequency forms that we find languag e change like grammaticalization. These forms become more grammatical through processes such as decategorialization, phonetic reduction, reanalysis, semantic bleaching or emancipation from the source form. It is these mechanisms that, when linked together, comprise the cline of grammaticalization (Hopper & Traugott, 2003 [1993]). However, not all these processes are required in order for a construction to be considered as undergoing

PAGE 36

36 grammaticalization Considering that grammaticalization is a gradual proces s it is difficult to witness a form that has gone through the entire process 14 Nonetheless, there are certain examples, such as the Latin verb habere going to that have undergone many of these stages, which will be discu ssed in greater detail in section 1.3.1.1 1. 2 2 Mechanisms in G rammaticaliz ation If we look at the example going to we see that, in its original sense (1. 8 ), this construction had a sense of direction or movement (Hopper & Traugott, 1993; Bybee, 2003). At the same time it is also evident that direction/movement is not the only connotation of this construction Intention can also be inferred in this con text in that the listener assumes that the speaker intends to travel to London in the future to get marr ied. It is the ambiguity of this sentence that can lead to the grammaticalization of the construction. 1. 8 I am going to London to marry Bill. (Hopper & Traugott, 1993: 2001) The fact that intentio n can be inferred is not enough, however, because in orde r for inferences to play a significant role in grammaticalization, they must be frequently occurring, since only standard inferences can plausibly be assumed to have a lasting impact on the meaning of an expression or to function cross linguistically (Hopp er & Traugott, 2003; 82). We know in the case of going to inference did play a significant role, sinc e we now have 1. 9 We are going to get married in June (Bybee, 2003; 147) 1.10 These trees are going to lose their leaves (Bybee, 2003; 147) 14 Again, certain features such as affix binding, clause combining are not discussed as they are outside the bounds of the current paper.

PAGE 37

37 examples s uch as 1. 9 and 1. 10 Of importance is th e fact that in 1. 9 we can still extract a sense of movemen t or direction, whereas in 1.10 this sense has been completely lost. This loss, known as SEMANTIC BLEACHING or sometimes DESEMANTICATION is also the result of increased frequency and entails the general shift of a construction into new contexts. Semantic bleaching can sometimes indicate the complete loss of semantic content such as in the case haber in Spanish. During the 13 th and 14 th centuries, this form w as still being used to express possession as we see in 1.11 although it appeared more frequently as an auxiliary and as an expression of obligation. 1.11 O Antigono, tu as muchos amigos, & de aga tu as muchos grandes hombres (Fernndez de Heredia, Juan, 1379 1384: Corde) Oh Antigono, you have many friends, and behind you, you have many By the 15 th century, however, this meaning cannot be found in the data 15 Semantic bleaching does not always entail loss, however. Crucially, the ambiguity seen in going to in 1. 8 and the resulting POLYSEMY of this construction (movement into new contexts) leads to autonomization even though they are semantically and historically related. In this way, the two semantic interpretations can exist for long peri ods of time without falling into disuse, like going to. Although polysemy indicates that the construction has emerged in new contexts, this development does not occur in order to fill a (Hopper & Traugott, 2003). Thus the semantic shift co mmonly entails moving into the territory of another 15 Nevertheless there are still remnants of the possession meaning in its auxiliary f orm, particularly in sentences like: porque muchos temores adelantados han abierto las puertas a la ofensa /because many anticipated fears have opened the door to insult. ( CORDE Mrquez, Fray Juan, 1612 a 1625)

PAGE 38

38 form. This process commonly known as RENEWAL or what Tagliamonte (2007) refers to as longitudinal renewal, is an alternat e way of saying the same thing and is ubiquitous in language (Hopper & Traugott, 1 993). Take for example the Spanish construction ir a go (ing) going to (Bybee, Pagliuca, & Perkins, 1991; Bybee, Perkins, & Pagliuca, 1994) 16 Once a meaning of futurity is establis hed the construction starts to appear in contexts designating a future action. This causes the construction to go into direct competition with the Spanish synthet ic future tense like we see in 1. 12 and 1. 1 3 This competition, known as layering, is often an indication of grammaticalization. 1 .12 Sancho, cuanto quisieres, que yo no te ir a la mano, pero mira lo que vas a decir (CORDE, Cervantes Saavedra, Miguel de, 1615) Sancho, however much you should want, I will not go with you and hold your hand, but going to say 1.13 Buelue, Zara, por tu honor, mas diras que es ignorancia (CORDE, Annimo, 1605) Return, Sara, because of your honor, but you will say it is because of ignorance. T his variation can also indicate the gradual replacement of one form by another. Layering has been evidenced in a wide variety of research as a possible indication of a Corvaln, 1985). Development along the grammaticaliza tion cline can also involve phonological REDUCTION In the case of going to, we see that in 1. 1 4 the construction has phonologically reduced to gonna It is also through this phonological reduction that the 16 This specifically refers to what has be en discussed up to this point i.e. pragmatic inference whereby a movement/direction meaning grammaticalizes and acquires a future and intention meaning. The analogous paths of change with these two forms does differ in many ways like, for example, with red uction.

PAGE 39

39 evidence of semantic DIVERGENCE (from the origina l sense of movement/direction) is most conspicuous. As Bybee (2003) argues, the original form is not the structure that has undergone phonologically reduction Rather, it is the form that denotes intention and future that is the grammaticalized constructio n. This is evident because sentences that only imply direction cannot be used with gonna as in 1. 15 1.14 gonna help you later. 1.15 going to/*gonna the store. Phonological reduction as an effect of frequency has also been largely attested in t he literature. For more examples of this see Bybee (2001), Company (2006), and Tagliamonte (2004). Empirical support also identifies DECATEGORIALIZATION as another common characteristic of grammaticalization (Company, 2006; Torres, 2006; Cornillie, 2007). Torres Cacoullos (2006) illustrates this process in her diachronic study of the Spanish collocation a pesar de pesar originally a noun in Old Spanish, has grammaticalized and diverged, and the grammaticalizing form has lost its nomi nal trappings as an effect of frequency. This loss of certain attributes syntactic or morphological results in the inability to identify it as a full member of a major grammatica (Hopper & Traugott, 2003; 107). In the cas e of a pesar de the decategorialization of pesar as a noun can be credited to the high frequency of the construction a pesar de relative to the frequency of the lexical noun pesar and the subsequent loss of nominal trappings such as plural marking, and det erminer and adjective modification (Torres, 2006; 38). The fusion of these previously independent morphemes is the result of the re categorization of the construction as a

PAGE 40

40 single unit phonologically, semantically and syntactically (Bybee, 2006). As we will see below, decategorialization is particularly important as it is one of the processes where grammaticalization and subjectification overlap. What emerges from this discussion is that language use and thus frequency of certain structures are the locus of processes like grammaticalization. In the section that follows it will be seen that frequency is also a crucial factor in subjectification. Additionally, many of the same factors revealed to be central to grammaticalization are the same that produce the i deal environment for subjectification to ensue. 1. 2 3 Subjectification There are two different accounts of what subjectification entails. Those that follow Langacker (1999a, 297) define subjectification as a shift from a relatively objective construal of some entity to a more subjec tive one He explains an entity is construed subjectively to the extent that it functions as the subject of conception without itself (1999a: 297). This definition in fact works well with examples like tener q ue in 1.1 6 and dale in 1.1 7 where the speaker is essentially offstage and the statement is indeed subjective. 1.16 El enemigo tiene que estar precisamente dentro de este crculo ( Torrente Ballester, Gonzalo, 1972 : Corde ) The enemy must be precisely amo ng this group. (Based on given evidence) 1.17 Dale que dale, pero qu pesada eres. (Company, 2006) However, as Narrog (2005 a ) points out this does not always hold true since an example like sure can be subjective and the speaker is essentially onstage and conceived of, definition of subjectification. For

PAGE 41

41 the present paper a perspective that disallows the speaker to be onstage could possi bly exclude valid instances of subjectification. A more general account of subjectification is adopted by those who follow (1989) account where she identifies three tendencies in semantic pragmatic change, all of which represent different level s of subjectification. She describes the first tendency as a shift from meanings centered on the to a meaning that is more focused on the (1989: 34). She expla in s that the internal situation often e ntails evaluations and perceptions. As a construction continues to undergo subjectification, another change of meaning can be observed where there is a shift from the external or internal situation to those meanings based in the textu al and metalinguistic (1989: 35). She defines the textual or metalinguistic situation as a shift of the construction towards contexts where it has more of a speech act like function. She develops this idea by incorporating the third tendency where meanings tend to state/ (1989 : 35). Although this conceptualization of the differe nce as Langacker (1999b) claims, Traugott (198 9: 188) disagrees pointing out that, for her, subjectification can only be a diachronic phenomenon where ctification is synchronic Cornillie (2007) finds these two views com plementary illustrating that in such cases as the subjectification of Spanish prometer amenazar (1999b) view; and three diachronic stage (1989)

PAGE 42

42 most encompassing, and as a result will be the view adopted in the present paper. By adopting both a diachronic and synchroni c perspective on subjectification we can now attempt to identify the various points in subjectification as the forms develop. Several clearly identified processes have been acknowledged in subjectification, namely increased transparency, which involves dec ategorialization (Langacker, 1999; Torres, 2006; Company, 2006, Cornillie, 2007), the tendency to appear more with first and second person (Aaron & Torre s, 2005 and references therein) or, in the case of epistemic modality, the tendency to occur with third person animate subjects ( Pietrandrea 2005; Bybee and Fleischman,1985; Coates, 1983 ), and syntactical flexibility (Torres & Schwenter, 2004; Company, 2006). It has been particularly challenging for linguists to operationalize the various degrees of subjec tification (Traugott 1999, 1995; Torres & Schwenter 2004; and Company, 2004; Aaron & Torres, 2005), although there have been several attempts, most notably Company (2004), Torres & Schwenter (2004) and Aaron & Torres (2005). Torres and Schwenter (2004) ou tline a specific way to measure subjectification using the Spanish construction a pesar de identify several measures as ways to determine the evolution of subjectification, although their analysis is confined to co nnectives. Thus, although these measures are indeed quantifiable, they do not provide a general account of how to measure subjectification in language, but rather a very specific account of how to measure subjectification in connectives.

PAGE 43

43 Company (2004) co rrelates subjectification with syntax in her study on discourse markers in Spanish such as dizque, tate and dale She demonstrates how these forms same time, displaying syntactic isolation and cancellation of syntax. She clarifies that syntactic isolation happens when an originally rigid form becomes less rigid and is able to appear separate from the rest of the sentence, usually indicated by commas, as seen with ndale i n 1.18 She also emphasizes that a form in syntactic isolation can be a who which she would define as being stripped of sy or t he cancellation of syntax (2004: 14). 1.1 8 T cllate. Trenos algo, ndale (Carlos Fuent es, 1962 in Company, 2004) You shut She effectively operationalizes this process through the identification of various stages of subjectification as these verbs make their way towards a state of high subjectivity creati ng the shift into discourse markers. However, like Torres and Schwenter (2004) this operationalization is specific to discourse markers, which makes it difficult to apply to other domains. What both Torres and Schwenter (2004) and Company (2004) highlight nevertheless, is that there is a general dearth of research quantifying subjectification. Besides these specialized perspectives of subjectification related to discourse markers and connectors, there is also an example in the literature of the subjectif ication of grammaticalizing forms. Aaron and Torres (2005) measured the subjectification of salirse by employing a variationist method. They hypothesize that if salirse is indeed subjectifying, then it will likely occur more often with factors t hat reflected that

PAGE 44

44 process. To operationalize the process they chose three factors that might reflect subjectification: co occurrence of dative pronouns, grammatical person, and relationship to the speaker (in terms of distance). Through the uti lization of multivariate analyse s they were able to measure the effect of these factors and identify patterns of subjectivity in salirse Results indicated that this form is more likely to co occur with first person, as well as with referents close to the subject. This study appears to be the first and only to apply a variationist method to measure subjectification. At the same time, it does not negate that clear measures of subjectification are still lacking in the field. Consequently, the present analysis will be based less on the degree or phase of subjectification of these forms, but rather highlight some of the clearly identified characteristics of the process and correlate them with grammaticalization as well as with the constructions haber de, tener de, hay qu e and tener que 1. 2 4 Subjectification and G rammaticalization. In the literature there has been general acceptance that, in many cases, subjectification is part of grammaticalization (Bybee et al., 1991; Bybee et al. 1994; Hopper & Traugott, 2003, Brinto n & Traugott, 2005; Narrog, 2005 a ). This is especially accounted for in the postulation of universal paths of grammaticalization, particularly that of agent oriented modality (e.g. obligation) to epistemic, or speaker oriented, modality (Bybee et al. 1994, Narrog, 2005 a ). Since grammaticalization involves shifts toward more abstract, less referential, markers, the prime function of which is to represent the necessarily the case that subjectification is characteristic of grammaticalization (Brinton & Traugott, 2005; 108).

PAGE 45

45 As noted by some of the studies reviewed above, there are certain domains where these two processes overlap, primarily with decategorialization, pragmatic inferencin g and the general movement towards a more abstract meaning. This, like in grammaticalization, is evidenced with the rise in frequency of a construction. Take for example the Spanish construction a pesar de noun pesa r to a connective (Torres, 2006). With the same example Torres and Schwenter (2004) use this shift from noun to connective as an indication of subjectification. Similar examples can be found in Aaron and Torres (2005), Narrog (2005 a ), Cornillie (2007) and Nevertheless, as Traugott and Dasher (2002: 89 90) Essentially, subjectivity is ubi quitous in language and, as Fin egan (1995: 6) states, there is an communicating ready made content, but as an expression of self and, in part, its creation Therefore, it is difficult to confine subjectification to a sub process of grammaticalization when in fact it is present in many parts of language since expression of self is the ultimate goal (Fin egan, 1995). This can be seen in the subjectification of the Spanish form claro as it moves from an adjective to a discourse marker. Ocampo (2006) is quick to recognize similarities between grammaticalization and subjectification but also emphasizes some of the contradictory characteristics, namely the general path of grammaticalization from lexical to grammatical, w hich he specifies would indicate a movement towards syntax. Whereas he points out that in subjectification, at least in the subjectification of discourse

PAGE 46

46 markers, the path of these constructions is out of syntax and towards discourse ( 2006: 317). His study illustrates that subjectification can occur outside of grammaticalization and also raises some valid questions in drawing the line between the subjectification of discourse markers and grammaticalization. Nevertheless, these questions are beyond the scope of this paper since tener de, tener que haber que and haber de have not become discourse markers. In the present paper, as will be seen below, any subjectification that occurs is ultimately a small part of a bigger process: grammaticalization. We assume this not only due to the various examples of t hese constructions show of having gone through certain stages of grammaticalization, which will be discussed in more detail in section 1.4 but also because of the consensus in the literature that subjectificat ion within agent oriented modality, specifically with periphrastics, is related to grammaticalization (Traugott, 1989; Bybee et al. 1991; Bybee et al. 1994; Narrog, 2005). 1. 3 The Obligation System in Spanish As previously mentioned, there have been certai n universal paths of grammaticalization postulated by linguists such as Bybee et al. (1994), Bybee et al. (1991) and Hopper and Traugott (1993). These linguists posit that constructions denoting a notion of possession develop into obligation markers and ev entually acquire an epistemic or subjective meaning. The four constructions tener que, haber que, haber de and tener de all derive from such a source in that tener historically conveyed a notion of possession (for more on the histor ical development of these forms see Yllera, 1980) A semantic pragmatic shift in each of these collocations could imply grammaticalization and even subjectification if the polysemy includes epistemic modality. Furthermore, an apparent emancipation of these constructions from

PAGE 47

47 their source form could also be indicative of grammaticalization, while their decategorialization could designate both grammaticalization and subjectification. Finally, the fusion of independent morphemes to form collocations is also a possible indication of grammaticalization. Little work has been done on the grammaticalization of these expressions of obligation in Spanish. Taking each of these apparent changes into account, we will examine these constructions and determine whether the se forms are undergoing grammaticalization. In addition, we will look at the constructions and the ostensive shift towards a more subjective reading as possible corroboration that these forms are likewise undergoing subjectification. 1. 3 .1 The (N on)gramma ticalization of These Constructions One of the first measures of grammaticalization that we see in these constructions necessarily occurs prior to them becoming expressions of obligation. In order for each construction to grammaticalize from a construction expressing possession, semantic bleaching needs to occur where an obligation reading can be inferred from the context (Hopper and Tr a u gott, 19 93). We can see in examples 1.19 1. 21 that both a possession and an obligation reading can be inferred. Although in present day Spanish the verb 1.19 y yo soy el que tengo necesidad de ser regido ( Sahagn 1576 1577: Corde) who has the need to be 1.20 No tengo ms que dezir. (Sahagn, 1576 1577: Corde) 1.21 qu e no han menester ms de ver un no s qu y sin ms mirar ni enterarse de cierto en lo que era y cmo era ( Crdenas 1591: Corde) have more with more than seeing or finding out for sure about what it was and

PAGE 48

48 haber has lost its possession meaning, during the earlier centuries it was still used as a possession marker. Linguists (e.g. Cornillie, de Mulder, Van Hecke and Dieter Vermandere, 2009; Traugott & Dasher, 2002) suggest that when t he object is something that is not yet possessed semantic bleaching will occur. In the case of tener que there is disagreement as to whether the development of this obligation marker was the result of the aforementioned process of semantic bleaching (e.g Cornillie et al. 2009, Yllera, 1980) or whether it came about from the semi transitive examples of tener de (Cornillie, 2007: 228) in which the construction occurs in contexts with both the preposition de and the conjunction que as is seen in the example 1.22 given by Cornillie et al 17 1. 2 2 si el procurador fuere rebelde que non sea restituydo el seor del pleito maguer que diga que el procurador non tiene de que pagar. (Alfonso X. Siete partidas. 13 th century) if the attorney opposes the fact that the lawyer be refunded although he says that the attorney does not have of to pay. (i.e. something for which to pay) As these constructions start to occur in contexts where obligation may be inferred we also see each marker undergo decategorialization a nd a subsequent shift into agentive modality. We can clearly see the difference in examples 1.23 1.2 4 in comparison to 1.25 1.2 6 where both tener and haber, transitive verbs, have evolved into semi auxiliaries and lost some of the properties that would oth erwise categorize them as full verbs. For example, when these verbs occur in this construction they can no longer take a nominal complement. Additionally, the subject of the construction seems 17 Although Cornillie (2007) suggests this as a possible path of development of the obligation marker tener que, in Cornillie et al. (2009) he contradicts this claim by assuming that tener que developed through the semantic bleaching of contexts whe re it expressed possession.

PAGE 49

49 bound to the infinitive. That is, the subject of the two is alw ays the same 18 Both of these seem to point to decategorialization. This decategorialization, as was mentioned above, is not only a defining characteristic in grammaticalization but can also denote subjectification. As indicated in Torres (2006), decategori alization can often lead to autonomization. As the constructions lose their status as full verbs, we also see that the individual morphemes become more and more bound, indicating autonomization, and a possible divergence from the source form. 1.23 O Antig ono, tu as muchos amigos, & de aga tu as muchos grandes hombres (Fernndez de Heredia, Juan, 1379 1384) Oh Antigono, you have many friends, and in front of you, you have 1.24 Bien, bien (dije yo entonces) noticia tengo de estas fiest as; ( CORDE Gmez de Tejada, Cosme, 1636) Well, well (I then said) I have news of these celebrations. 1.25 Seor, con qu mandamiento te he de matar? ( CORDE Quirs, Francisco Bernardo de, 1656) Sir, with what order do I have to kill you? 1.26 Y o tengo d e ir con vos ( CORDE Caldern de la Barca, Pedro, 1629) I have to The first piece of evidence is the gradual disappearance of the source form haber as a notion of possession. This notion eventually becomes obsolete while the gram maticalizing construction haber de contin ues to be used Although tener has not become obsolete, we do see a continuous rise in frequency of the collocation tener que (see section 2.1.1 for frequency counts of this collocation) Fusion of the morphemes an d the infinitive is also noticeable with all the constructions as they begin to show a lack of syntactical flexibility in terms of what can 18 This could be the loss of agent control Company (2005) identifies as one of the stages in the subjectification of discourse markers.

PAGE 50

50 occur between the construction and the infinitive As we can see in 1 2 7 1.2 8 there was some flexibility early on, where clitics could be sandwiched between the collocation and the infinitive although even in the 17 th century these examples were already rare in the corpus. This is a strong indication that early on these units had already become obligatorily attached to an infinitive. 1.27 e de le responder con toda mayoria e superioridad (CORDE, Fernndez de Crdoba; 1625) I have to respond with superiority and the majority 1.28 Siendo esa vuestra opinin, ya no tengo que os decir (Caldern de la Barca, Pedro, 164 9) This being your (pl) opinion, I do not have to tell you (pl) anything. This layering of forms is another measure of grammaticalization and, as previously mentioned, frequently indicates the possible replacement of one form for another. This happens with tener de as its use steadily declines throughout the 18 th century until disappearing completely in the early 19 th century. Interestingly, haber de also appears to have declined albeit in more recent times according to the literature Cornillie (2007) points out that haber de is used more often used in writing than in speech. 1.29 de la doctrina cristiana que todos los nios han de (Annimo, 1626, CORDE) from the christian doctrine that all children have to 1.30 y o am ya no tengo qu e dudar, ni t tienes que (Caldern de la Barca, Pedro, 1635, CORDE ) have to have to know. 1.31 porque a riesgo de mi vida, tengo de saber quin sois. (Caldern de la Barca, Pedro,1635, CORDE ) because at the risk of my life, I have to know who you are.

PAGE 51

51 W ithout a comprehensive study however, it is hard to confirm or refute his claim. At this point we can posit that these forms are grammaticalizing. Now the question is whether there is also a n indication of subj ectification. 1. 3 .2 The ( N on)subjectification of These Constructions Previous resear ch on subjectification identifies several factors that may indicate the subjectification of these obligation markers. Based on the description of the d ecategorialization of each of these forms in the previous section, it is postulated that these constructions have entered into the subjectification process. This would indicate a semantic pragmatic shift towards epistemicity. In the case of deontic modalit y, the agent who expresses the obligation is usually not involved in the proposition (Bybee et al., 1994). Rather, the domain of obligation is based on social norms or physical necessity that is external to the speaker (Bybee et al. 1994; 201). As these fo rms strengthen pragmatically and the domains shift, they acquire an epistemic use whereby the attitude of the speaker towards the proposition is now central to the utterance. This would be a prime example of subjectification in that it shows the semantic r eallocation of a construction that is based on things external to the speaker and its eventual evolution towards the expression of the value of truth the speaker has towards the proposition (i.e. epistemicity) However, in order for a transformation like t his to arise, remember that high frequency of occurrence is indispensable. If we look at examples of haber de and tener de, the older and more common constructions in the earlier centuries, from the 17 th century we see that haber de has already acquired an epistemic meaning. In fact, this form has an epistemic notion as early as the 14 th century (1 .32 ) which is a strong indication that this verb came into Spanish (from Latin) already having gone through subjectification. In fact, as Cornillie

PAGE 52

52 (2007: 226) po ints out, due to the future reading of habere in Latin (cf. cantare habeo > cantar) it is possible that a future connotation precedes the deontic reading 19 Besides haber de we also see epistemicity expressed with tener de ( 1.3 3 ) early in the 17 th century although Corni llie (2007: 229) finds an example as early as the 16 th century. The fact that these constructions have decategorialized (e.g. no longer can take nominal complements) and both forms are comprised of a subjective function point to the possibi lity that they are undergoing subjectification. Additionally, t he continued layering of these forms could be an indication that they are still undergoing grammaticalization and subjectification. This study aims to determine whether these constructions cont inue to grammaticalize and whether or not they have entered into subjectification. 1.32 Sabis qul dellos ha de ser muerto? ( CORDE, Annimo, c 1400 1498) Do you (pl) know which one of them must 1.33 Mi esclaua tiene de ser, que es ley de Dios inuiolada, nunca vista dispensada ni con hombre ni muger (CORDE, Valdivielso, Jos de 1622) She must be my slave, which is the unviolated law of God, never seen dispensed nor with a man or a women. In addition to the aforementioned characteristi cs of subjectification, recall that Traugott (1989) identifies three tendencies she found to be common in constructions undergoing subjectification. In the same study she correlates these tendencies with the development of deontic (obligation) and epistemi c modality. She proposes a path of subjectification specific to obligation (table 1.1). Based on her postulation, each of the constructions in Spanish have already seen a shift from an which in this case is possession, to an which could be 19 This infers that, in order to get a clear representation of the subjectification path of ha ber de, it may be necessary to consider data from Latin.

PAGE 53

53 attributed to the shift from possessing a tangible object to possessing something more intangible, like a secret (Traugott, 1989:34). Based on table 1.1, we can see that each of these obligation markers is likely reflecting uses that represent the second or third tendency. Recall that the second tendency is identified as a shift from meanings centered on the external or internal situation to a meaning centered on the textual or metalinguistic situation (1989: 35) Traugott (1989:37) claims that, in regards to these modals, tendency II is defined by the shift of possession to obligation. This presumes then that these constructions have already undergone the shift from tendency I to tendency II and Table 1 1. Traugo printed with permission) Tendency I II III III stage main verb pre modal deontic weak epistemicity habitual prophetic/rel. future strong epistemicity all should have meanings associated with tendency I II, which would imply that each attitude towards the proposition (epistemicity). According to her path of development for these modals, each construction is somewhere betwee n weak and strong epistemicity (as can be seen in table 1.1). The question is, then, where in the subjectification process is each of these markers? One of the ways to determine this is to take a variationist approach to language change, which is discussed in more detail in the following section.

PAGE 54

54 1. 4 A Variation ist Perspective One way that linguists have been able to examine grammaticalization (and subjectification) within and across languages is through a variationist approach A variationist perspective l ooks at variation in language as structured and heterogeneous The idea that language is non uniform was first put forth by Weinreich, Labov & Herzog (1968) where they state that: a model of language which accommodates the facts of variable usage and its social and stylistic determinants not only leads to more adequate descriptions of linguistic competence, but also naturally yields a theory of language change that bypasses the fruitless paradoxes with which historical linguistics has been struggling for o ver half a century (1968; 99). Importantly, this structured heterogeneity indicates that the variant chosen by the speaker is not a random choice but, in fact, conditioned by certain social and linguistic factors. Historically, this type of variation has been largely ignored since other branches of linguistics primarily focus on those variables that are predictable by rule (Labov, 1972). This, in essence, disqualifies a variant that is not used categorically in one context or another. Through the adoption of this approach and the use of a standard statistical model that explains this variation, linguists have created a way to study these variable forms. 1. 4 .1 The Variationist Method Working within variation theory, Labov (1972, 1994, 2004) has established certain compulsory components to examining variation in the system. One of these is the identification of the linguistic or dependent variable. These are the forms speakers use that appear to be in free variation or competition (layering) as noted above. T hat is the class of variants among which speakers alternate in the expression of a given m eaning (Poplack & Tagliamonte, 1998: 88). For example, in the present

PAGE 55

55 study the linguistic variable is obligation which includes four vari ants: tener que, haber de, hay que, and tener de. These variables appear to have the same functions given the examples below (1. 34 1.37 ). However, from a variationist perspective we can see that this is not true. 1.34 Prndanme; crtenme la cabeza, que yo cumplo con mi oficio, y he de decir la verdad (Barrionuevo, Jernimo de, 1654 1658, Corde) Catch me, cut off my head but I do my j 1.35 Hermano, yo os tengo de decir la verdad; no s qu dicen de un indiano. (Vega Carpi o, Lope de, 1632: Corde) Brother, I have to tell you all the truth, I returning Spaniard who made his fortune in America 1.36 ay que decir aqui nada, porque tiene su lugar propio (Jimnez Patn, Bartolom, c 1604 1614 : Corde ) place. 1.37 Con esto no tengo que decir ms; harto siento lo dicho, y ms cansar a vuesa merced [usted]. (Gngora y Argote, Luis de, 1613 1626: Corde) Of this I Bayley (2002) also identifies the importance of the PRINCIPLE OF QUANTIT ATIVE MODELING which essentially allows for the close examination of the forms that a linguistic variable takes, and [the notation of] what features of the co ntext co occur with (2002: 118). Such analysis can be done through a statistical program and probabilistic data. As Bayley explains, these probabilities measure the effect each of the independent variables has on the linguistic variable. For the current study, this will be done through the use of Goldvarb (Cedergren & Sankoff, 1974, Sankoff, 1988), a statistical multivariate analysis program that takes the idea of structured heterogeneity (Weinreich et al., 1968) and quantifies it (T agliamonte, 2006).

PAGE 56

56 These types of statistical programs allow for probabilistic predictions of the conditioning effects of the independent variables on each of the obligation markers. That is, we can examine from a quantitative perspective t he surrounding linguistic context to identify which variables (linguistic or extralinguistic) co occur with each obligation construction (Bayley, 2002). In addition, Goldvarb provides the relative frequencies of occurrence of each of these markers in a giv en context, which also provides a general account of the status of these constructions. In addition to defining the linguistic variable and choosing independent variables it is also important to exclude any context that is categorical or near categorical w ith one of the variants. Labov (2004) describes these contexts as neutralizations. Excluding categorical categories allows for a clear analysis of the dependent variable and not the various other subclasses that these forms may comprise (Labov, 1972: 72). 1. 4 2 Comparative Method The adoption of the comparative method into variationist linguistics has provided sociolinguistics with a valuable tool for reconstructing languages, determining source languages in contact situations, comparing cross dialectal var ieties or two different time periods to determine the status of grammaticalizing forms, as well provide a way to look at variation from a more fine grained perspective. Although historically the comparative method was criticized for its assumption of unifo rmity in the source language, its adoption into variation theory has largely resolved this limitation. In addition, this method has led to a shift from focusing on surface forms and frequency distributions as ways to identify linguistic change to primarily focusing on the variable grammar and probability values. Through the comparison of these probability values across time periods, possible sources of change in the grammar can be identified. The identification of these

PAGE 57

57 sources can elucidate possible direct ions of change at a micro level and corr oborate assertions on the macro level thus substantiating universal claims of paths of development as well as providing new information on how these changes take effect. This is particularly pertinent to Spanish moda lity given that very little quantitative research has been conducted on this system and despite the recognition that modality systems in other languages are in flux. 1. 5 The Present Study Given the large amount of research that has been conducted on modal ity over the last several decades it is surprising that only a small amount of that research has been conducted on the Spanish modal system. Of the literature that does focus on modality in Spanish, very little of it recognizes semantic differences betwee n the aforementioned constructions (c.f. Rabadan, 2006). Furthermore, to my knowledge, there have not been any studies in Spanish that examine modality and its diachronic development using a quantitative variationist perspective. One may assume, however, t hat this can be accounted for by the research conducted on other languages given the recognition by linguists that modality tends to follow a universal path of change (Lyons, 1977; Coates, 1983; Bybee et al. 1994; Bybee & Fleischman, 1995; Palmer 2001). Ne vertheless, this evidence is widely based on studies that have (i) not taken a variationist sociolinguistic perspective (e.g. Lyons, 1977, Bybee et al., 1994), (ii) limited analyses to the comparison of relative frequencies (e.g. Myhill, 1995a; Cornillie, 2007) or (iii) did not use oral speech to consider possible sociolinguistic factors that have contributed to these changes (e.g. Bybee et al., 1994). As a result of the narrow scope that has been utilized thus far, little evidence has surfaced on how thes e obligation markers are used in both text and oral speech. This

PAGE 58

58 study seeks to fill this gap of knowledge by examining at the development of modality in Spanish both diachronically and synchronically through a comparative sociolinguistic approach (Poplack & Meechan, 1998; Poplack & Tagliamonte, 2001; Tagliamonte, 2002; Tagliamonte & Smith, 2006). The aim is to uncover the linguistic and extralinguistic factors that condition the use of one expression of obligation over another while at the same time substa ntiating universal claims of development in modality. In addition to answering this overarching question, a review of the questions central to this study are listed below: What are the semantic differences among these four obligation markers? Have these f orms continued to grammaticalize? If so, what is the path of grammaticalization of each marker? Have they also started to undergo subjectification? This dissertation will be organized as follows. C hapter 2 explains the methodologies used, while the third chapter reviews the results found in the multivariate analyses, and comparative analyses Chapter 4 is dedicated to analyzing the results of the multivariate analyses, the comparative analyses and the relative frequencies of these constructi ons in each ce ntury. T he final chapter presents the conclusions and implications of the study, as well as suggested directions for future work.

PAGE 59

59 CHAPTER 2 METHODOLOGY 2.1 Corpora T he region chosen for this dissertation was Mexico, since I have strong connections to a c ommunity in Guanajuato, Mexico, making it reasonable to conduct the sociolinguistic interviews there. The decision to limit my research to Mexico was primarily based on previous research on modality (Tagliamonte, 2002; Tagliamonte & Smith, 2006; Tagliamont in terms of the development of these modals thus making it necessary to limit this study to a specific region. Regrettably, the historic data available in Mexico are not confined to one region so the corpus consisting of literary works is representative of the country as a whole and not one particular dialect. This does not negate the number of different dialects that exist in Mexico, but simply exemplifies the dearth of regional documents avai lable for a diachronic study (Aaron, 2007). In order to capture both the development of these obligation markers through time and the reality of these constructions in modern Mexican Spanish three corpora were used: an oral corpus, a literary corpus and an other small digitalized corpus comprised of written non literary documents. Each corpus serves a unique purpose in this study but as a group the three corpora will give a clear picture of the development of obligation in Mexican Spanish. 2.1.1 Written Cor pora In order to get a clear picture of grammaticalization, a diachronic perspective was necessary The literary corpus includes Mexican literary prose from the 16th to the 20th

PAGE 60

60 centuries 1 The majority of these works were extracted from two digitalized co rpora. The first, Corpus Diacrnico del Espaol (CORDE), has a total word count of more than 250 million words but only works from Mexico were used The Mexican section includes 7,822,486 words. This corpus is comprised of plays, prose, poetry, legal docu ments historical accounts, religiou s works, and newspaper articles; however, only prose or historical works were used for this study. 2 The second corpus, Corpus de Referencia del Espaol Actual (CREA), includes the same range of written work as CORDE and Table 2 1 Frequency count of each variant per century in the written corpora 3 Century (word count) TQ N Normalized per 5,000 TD N Norm HD N Norm HQ N Norm Total 16 th (~2,339,244) 32 .11 89 .29 1091 3.58 21 .07 1233 17 th (~960,832) 40 .17 22 .10 852 3 .70 19 .08 933 18 th (~403,340) 39 .50 12 .15 256 3.28 9 .12 316 19 th (~789,217) 253 1.00 2 .01 687 .67 79 .32 1021 20 th (~3,501,511) 865 2.41 3 .01 354 .99 232 .65 1454 Total N 1563 128 3248 381 4957 has more than 160 million words but ; the tota l word count of the works from Mexico is 16,897,942. In addition, due to the lack of Mexican prose available in some of the earlier centuries, several works from the internet archive ( http://archive.org/detai ls/texts ) were also included Between the two corpora and the internet archive and after eliminating the non prose data, the total word count for the dataset in this study is 7,994,144. From 1 The 16th and 17th centuries do contain some historical and legal documents owing to the general dearth of written documents available during this period. There was an effort, nevertheless, to use a s few of these types of documents as possible. 2 It is possible, although not expected, that these constructions are used differently depending on the genre. Consequently, the study was limited to prose so as to eliminate this possible limitation. 3 The ab breviations found in the tables throughout this study represent each of the variants: TQ= tener que, HQ= haber que TD= tener de, HD= haber de. Additionally, each variant was normalized per 5,000 words as a way to make the frequency counts across century unamb iguous and are easier to compare.

PAGE 61

61 these words, a total of 5,055 occurrences were extracted, as ca n be seen in Table 2 1. 4 For a complete list of texts used in this study see Appendix A The other written corpus, Documentos Linguisticos de la Nueva Espana: Altiplano C entral (Company Company,1994; here inafter DLNE), is relatively small and consists of 320 documents from the 15 th through the 19 th centuries and include s nearly 260,000 words (Aaron, 2004: 589). Most of these documents are letters or accounts of incidents recorded by a scribe. Schneider (2002: 76) states that documents of this type often r The hope is that if there are differences among the use of obligation markers in formal (i.e. written) and vernacular speech, these two historical corpora (in addition to the synchronic analyses) will expose those differences. Every occurrence of the aforementioned obligation markers was extracted from this corpus and divided according to century 5 Frequency counts of the examples found in the DLNE corpus can be seen in Table 2 2. Table 2 2 Frequency c ount of each variant per century in the oral and DLNE corpora Century (word count) TQ N Normalized per 5,000 TD N Norm HD N Norm HQ N Norm Total 16 th (~83,372) 8 .25 6 .19 108 3.42 6 .19 128 17 th (~85,321) 16 .47 5 .15 107 3.16 9 .27 137 18 th (~71,68 7) 27 1.26 6 .28 53 2.47 1 .05 87 19 th (~19,767) 4 1.01 --11 2.79 1 .25 16 21 st spoken (189,840) 334 3.73 0 .00 8 .09 21 .23 363 Total N 389 17 287 38 734 2.1.2 Oral Corpus The spoken corpus serves two purposes. First, as mentioned before, it can provide a clearer picture of the development of these obligation markers, particularly in 4 The table only reflects a total of 4957 examples because 98 tokens were excluded from the written corpus, which is discussed in detail in section 2.2. 5 Although, in general, data were divided up according to century, i n some analyses the 17th and 18th centuries had to be combined given the lack of literature available during the 18th century.

PAGE 62

62 vernacular speech. This is because, as Torres Cacoullos and Walker (2009) point out, the lexical history of a form has been empirically substantiated through pat terns of use in the newer forms. Second, sociolinguistic interviews are used as a way to account for the possible sociolinguistic variables that may condition the use of each of the obligation markers. Including a corpus that allows for the analysis of soc iolinguistic variables is based on the general Labovian idea that, in order [t]o extract evidence for change in progress, we must separate the variation due to change from the variation due to social factors like sex, social class, social networks, and e thnicity, and from the variation due to internal factors like sentence stress, segmental environment, word or der, and phrase structure (1994: 26). Without sociolinguistic interviews, this separation would be impossible. The corpus is comprised of 26 sociol inguistic interviews with Mexican men and women from a small community in the state of Guanajuato, all of which were conducted in the summer of 2010. The total word count of th is corpus is 189,840 and there are 333 occurrences of obligation found in this c orpus (see Table 2 2). In order to obtain these data, however, two factors were taken into consideration : the speech community and participant selection 2.1.2.1 The speech community The speech community used in this study is located in a small town in th e state of Guanajuato, Mexico. This area was chosen (i) because of a longstanding friendship with families in the community and (ii) because this community could be considered a relic area. Poplack and Tagliamonte (2001) identify specific characteristics o f a relic area: (i) lengthy settlement history 6 (ii) minimal immigration of non members, (iii) historical 6 Although Poplack and Tagliamonte indicate length being approximately two centuries, this community was not in existence two cen turies ago. When the Agrarian Decree of January 1, 1915 passed, land was

PAGE 63

63 continuity of current informants with the original settles, (iv) geographic remoteness, (v) physical and psychological separation from adjacent popula tions and (vi) a strong sense of group identify (2001: 66). According to these criteria, this community qualifies as a relic area. This is particularly important given that relic areas commonly preserve older features of a language (Campbell, 1998; Poplack and Tagliamonte, 2001). This community is largely isolated located about two miles off a country road and more than ten miles away from the highway that leads to the two closest towns, Pnjamo and La There are other communities nearby, although these too appear to be part of the larger speech community. However, no interviews were conducted outside this community 2.1.2.2 Participant s election Since all interviews were conducted in one summer, it was imperative that I enter the target community already having a connection i.e. friend of friend method (Milroy, 1987); the participants must feel comfortable during an interview in order to produce a natural variety of Spanish (Labov, 1994). The community has a population of about 300 people but my association with several members in the community made my acceptance unproblematic. As a result, it was relatively easy finding members of the community who wanted to participate. In addition to my connections, I also spent the first three weeks of my time there familiarizing myself with the community. Once I began to interview participants, an effort was made to consult an equal amount of men and women of a range of ages, socioeconomic backgrounds, and religions. 7 distributed many Mexican families. It was during this time that the community was established. Prior to the establishment, most members of the community traveled around looking for wo rk. 7 It was thought that religion may be a possible conditioning factor given that Christian religions are often centered around what should and should not be done, according to the bible. Taking into consideration

PAGE 64

64 The age of the participants range from 18 to 80 years and had all spent the majority of their lives in this particularly community. All we re born in Mexico and Spanish i s their first and only language. M ost participants come from families who were part of the original settlement in the area which happened shortly after the constitution of 1917. The interviews were conducted with 26 residents, 14 women and 12 men using a digital recorder (see Appendix B for sample of interview questions). In an effort t o minimize the effect of the O BSERVER S P ARADOX (Labov, 1972), the recorder was always placed in an inconspicuous place. Additionally, I did not ask participants to be ; rather I asked them if they would with me about life in the commun ity. goings on in the community. Labov (1972) identifies these types of narratives as a way h. Participants were also given the choice to be interviewed outside, in their homes or at the home of the family with whom I was staying. Crucially, these are all ways to minimize the effects Besides creating quest ions that invoked narratives, I was also careful to design my questions in a way that would educe expressions of obligation, like in 2. 1 2. 1 Tena muchas responsabilidades en su niez? Explica. Did you have a lot of responsibilities as a child? Explai n. that obligation based on external forces (i.e. the word of God) is often deontic, it was thought that religious people may be use the obligation markers in different ways than their non christian counterparts. This was unable to be measured, however, because the vast majority of the community wa s Christian.

PAGE 65

65 This does not mean, however, that I was limited to these questions. Naturally, depending on the participant and the flow of the conversation, these questions varied greatly. 2.2 Data Collection Once the interviews were conducted and transcribed, all oc currences of the aforementioned obligation markers were extracted from the three corpora. With the CORDE, CREA and DLNE corpora, this was done via computer programs like Microsoft Excel and Word that allow for a careful extraction of every token. With the literary works not from CORDE or CREA, however, an exhaustive search for every occurrence was done by hand. Once every token was accounted for, I then verified that each obligation marker was part of the variable context. Owing to the tendency for these co nstructions to be polysemous, I excluded the following types of contexts: (i) cases where obligation/epistemicity were not expressed (2. 2 ); (ii) false starts during the oral interview (2.3) or (iii) inability to decipher what the participant was saying (2. 4). Table 2 3 shows the exclusions for tener de and tener que (T) and for haber de and haber que (Q) per century 2. 2 Deseo tengo de verle (Caldern de la Barca, Pedro, 1634; CORDE) 2. 3 Tenamos que pagar ci -, ten amos que pagar cinco a la semana o treinta al mes. (Roberto, 7/2010) We have to pay we have to pay five per week or thirty per month. 2.4 Pues le gusta haber que xxx. (Reinaldo, 7/2010) Well, he likes to have to xxx. Once these tokens were excl uded the total amount of occurrences was 5,691. This includes all written and oral corpora: 371 in the DLNE (Table 2 2), 363 in the oral corpus (Table 2 2) and 4957 in the written corpus (Table 2 1).These corpora will be

PAGE 66

66 analyzed separately and the results of each analysis will be discussed and compared in Chapter 4. Table 2 3 Exclusions of tener que /de (T) and haber de/que diachronically 16 th 17 th 18 th 19 th 20 th 21 st T H T H T H T H T H T H Expressed Possession -27 4 42 3 6 6 3 4 -Lexicalized False starts ----------2 1 Difficult to decipher ----------2 2 Total 27 4 42 3 6 6 4 4 3 2. 3 The Variable Context and Form Function Asymmetry A question one might ask is whether or not we can say that these forms are synonymous. That is, does each of these obligation markers essentially express the from a variationist perspective it is recognized that different forms like these obligation markers, can indeed be used for the same function (Tagliamonte, 2006). However, it is also important to acknowledge that each form likely has a range of functions, especially if they are undergoing change. There may only be overlap with a limited number of the functions and it is therefore important to define the context(s) where these functions overlap, since it is in thes e contexts that the variants co occur and the patterns of use of the forms can be discovered. Torres Cacoullos ( 2001) demonstrates this in her study on the grammaticalization of Spanish estar + gerund and andar + gerund. She illustrates how variation studies can in fact explain the patterns of use of these two forms, despite the asymmetry of form and function. She g oes on to explain that, based on the different histories of each of these constructions, there will always be contexts in which there is no overlap but, as they grammaticalize, they move towards becoming variants of the same variable. When

PAGE 67

67 they do start co mpeting for the same territory, the space is not necessarily homogonous, given the different histories of the forms (Aaron, 2010; Torres Cacoullos, 2001). However, according to the NEUTRALIZATION IN DISCOURSE H YPOTHESIS the meaning nuances that distinguis h each form may not always be considered by the speaker when making a choice, particularly in this shared space (Sankoff, 1988). We can assume then that obligation is no different and that, even though there are overlapping functions, each form likely has a bigger inventory of distinct functions. In the case of Spanish obligation, however, there have been limited studies that have focused on the semantic differences between these forms (c.f. Rabadan, 2006) Of the little research that has discussed these dif ferences, very few have empirically backed assumptions, and instead rely on intuition as a way to describe these differences (e.g. De Maeseneer, 1998). Using a variationist approach minimizes dependence on intuitions and, instead, operationalizes the possi ble differences in meaning among the forms. 2. 4 Circumscribing the Variable Context When trying to circumscribe the variable context, it is necessary to include the largest environment in which this variation occurs, in order to apply the PRINCIPLE OF ACC OUNTABILITY which states that every occurrence needs to be accounted for, including both the environments where the form occurs and where it could occur (Labov, 2008: 2) The variable context should additionally exclude instances where one of the linguist ic variables cannot be substituted with another (Poplack and Tagliamonte, 2001). This allows for a clear analysis of the dependent variable and not the various other functions that these forms may have (Labov, 1972: 72). Similar to defining the variable co ntext, not accounting for every possible occurrence can yield undependable

PAGE 68

68 results. Applying the principle of accountability, however, is especially difficult in morphosyntactic studies, as it is nearly impossible to identify and record all of the contexts where the dependent variable occurs and could occur. With the application of quantitative sociolinguistic analyses to morphosyntactic studies many linguists have argued for the validity of this type of analysis with non phonological variables (e.g. Labov, 1982; Winford 1993). Winford, for example, points out the need to use the principle of accountability in a way that still adheres to finding variables in the same context but not requiring every occurrence or non occurrence to be considered, given its nea r impossibility (Winford, 1993). For the purposes of this study, the variable context is obligation, as expressed by the aforementioned constructions. The variable context includes all the occurrences of these obligation constructions in the overlapping fu nctional space. Furthermore, it will necessarily exclude instances of the same constructions in contexts where one of the other constructions cannot be substituted or when obligation was not expressed. Thus, a construction like tener que would only compris e contexts where this form had an obligation meaning and not meanings such as possession. (see section 2.2 for additional exclusions). Once the envelope of variation has been established, a diachronic perspective of the development of these forms will be e xamined. This will be done in part using a variable rule analysis, which will be discussed in the next section. 2.5 Variable Rule Analysis Taking into consideration the three corpora used in this study there are 5691 occurrences of obligation markers which is an adequate amount of tokens for a multivariate analysis (Guy, 1988; Bayley, 2002). Provided that the variable context has been clearly defined (see sections 2.3 and 2.4), factors thought to condition the use of

PAGE 69

69 each of these markers were chosen. Each of these factors was based on previous research and will be discussed in detail in Sections 2.6.1 2.6.10. Once the factors were chosen each of these tokens was then coded in regards to its occurrence with the aforementioned markers in an excel spreadsheet (see Appendix B for list of factor groups and factors). Once all the data were coded they were analyzed quantitatively. For this process I used GoldVarb (Cedergren and Sankoff, 1974, Sankoff, 1988), a statistical multivariate analysis program that takes th e idea of structured heterogeneity (Weinreich et al., 1968) and quantifies it (Tagliamonte, 2006). This allows for probabilistic predictions of the conditioning effects of each of these obligation markers. That is, we can examine from a quantitative perspe ctive the surrounding linguistic context to identify which variables (linguistic or extralinguistic) condition each obligation marker (Bayley, 2002). The program also provides frequency distributions of each variable relative to the various factor groups a s well as the individual factors thus providing a well defined representation of the semantic environment of each of the obligation constructions. Importantly, these analyses do not rely solely on surface similarities. Rather this approach takes into accou nt that sometimes forms that appear to have similar functions on the surface do not necessarily signify that they are similar. Similarities are determined through a variable rule analysis, where the comparison of conditioning effects across time (real or a pparent) is considered. This is done through the examinations of the HIERARCHY OF CONSTRA INTS or the order of effect that the independent variables have on the dependent variable. This type of analysis allows for the measurement of various factors at one time and orders them in terms of the effect

PAGE 70

70 they have on the dependent variable thereby giving a fine grained representation of the system under examination and a basis with which to compare. This is in contrast to surface comparisons whereby assumptions a re made that likeness in form denotes similarity in function. Multivariate analysis allows me to test for each of the hypotheses (discussed in detail in sections 2.6.1 2.6.10), as well as see the general direction of change of each of these obligation cons tructions by comparing the results (hierarchy of effect) of these analyses across centuries or age groups (see section 2.6 for a discussion on how this is done). For example, with the factor group grammatical person and animacy, the results will first show whether the factor group is significant. If it is significant, we can then look at the probability values of each factor. These values will tell us if this significant factor favors (greater than .50) the obligation marker, or disfavors it (lower than .50 ) essentially showing the strength of association of each factor with the form (Bayley, 2002). If a factor group is not significant there are still conclusions that can be obtained. For example, if age is not a significant factor group, then we can postula te a variety of reasons why this is the case: (i) that these constructions undergo such gradual change that differences are too slow to be apparent between generations; (ii) that perhaps the age division was incorrectly defined and the insignificance is du e to this; or (iii) we could also look in the data for trends to see if there may be tendencies among age groups and draw a tentative conclusion. Crucially, significant results are important, but of equal importance are the conclusions we extract from the results.

PAGE 71

71 Results should corroborate or refute prior conclusions drawn in the literature and essentially position the findings in terms of theory. As Bayley (2002: 130) states history comparative sociolinguistic method is a technique that situates the results from a variable rule analysis in a way that clear conclusions can be drawn about the development of each form and the changes the forms incurred throughout history. 2.6 Comparative Sociolinguistic Method The comparative sociolinguistic method is a technique that has been adopted from diachronic historical linguistics as a way to examine s imilarities between dialects or varieties synchronically. In contrast to historical linguistics, these analyses do not rely solely on surface similarities. Rather this approach takes into account that sometimes forms that appear to have similar functions o n the surface do not necessarily signify that they are similar. Similarities are determined through a variable rule analysis where the comparison of conditioning effects (factors) across time (real or apparent) or across dialects is considered (e.g. Poplac k & Tagliamonte, 2001; Tagliamonte, 2002; Tagliamonte, Smith & Lawrence, 2005; Tagliamonte & Smith, 2006; Tagliamonte, 2006a, 2006b; Schwenter & Torres Cacoullos, 2008). One of the core principles of the comparative method is the UNIFORMITARIAN PRINCIPLE which states that the factors that effected language change in past centuries are likely to be similar to the factors that prompt language change in modern languages (Poplack & Tagliamonte, 2001 and references therein). Additionally, Labov (1972) indicates that synchronic variation is often a reflection of diachronic change so detecting changes in a modern system is essentially a window into the past. It has been pointed

PAGE 72

72 out in the literature that forms often retain connotations that are related to their or iginal meaning (e.g. Bybee et al. 1994; Poplack and Tagliamonte, 2001). Thus a synchronic study of a linguistic variable is likely to reveal certain patterns of use and semantic properties that can be connected to the past. Fundamentally, reconstructing th e language with synchronic principles complements the data already available by offering a new way to construct the history and development of a language. An important component of the comparative sociolinguistic method is looking for correspondences in th e variable context (Poplack and Tagliamonte 2001: 96). These correspondences are in reference to similarities found in the order of effect of the conditioning factors in two (or more) sources. That is, once the variable rule analysis has provided significa nt results the order of conditioning effects between the sources in this case the source is represented by the centuries -is compared (Tagliamonte, 2001: 732). When similarities are found in the conditioning effects of the forms we can deduce that these two sources are related, whereas in contrast, if no similarities are found, it is possible that the two sources are not related (Tagliamonte, 2002 and references therein). An example of this is in a study conducted by Poplack and Tagliamonte (2001) on five dialects of English, three of which are posited to be early varieties of African American English. In these three varieties they found identical conditioning effects thereby confirming their belief that these dialects derive from a similar source (2001: 2 32). In the present study, we know that the source language is the same, so this step is less informative in terms of results.

PAGE 73

73 Another important element of the comparative sociolinguistic method is the idea of a CONFLICT SITE (Poplack and Tagliamonte, 2001 forms which differs functionally and/or structurally and/or quantitatively across the variable rule analysis has been performed and correspondences have been verified, the conditioning effects of each source are compared. If there are differences in the order of the factors within each significant factor group then this is considered a conflict site. This is particularly important for the present study since these conflict sites are indications of changes occurring in the system. These conflict sites will help identify where on the cline of grammaticalization these constructions may be in each century (Poplack and Tagliamonte, 2001; Tag liamonte, 2002). Additionally, these conflict sites may pinpoint where in the grammar these changes are realized. There have been several variationist studies on modality that have employed the comparative sociolinguistic method. Tagliamonte (2004) uses th is method in an apparent time study she conducts on the modal system of Northern British English. Her results reveal several patterns, such as the diminishing use of must across age groups (p. 51). She also concludes that, in contrast to other dialects spo ken in the south, have to and have got to are still relatively stable in this variety, citing specialized functions as the primary reason for this distinction (2004: 51). These results are compared to other studies conducted on modality in other dialects s uch as Tagliamonte and Smith (2006) (2006) study on English in Northern England, similar results are found with the modal must which indicates a general move

PAGE 74

74 of the deontic modal to wards obsolescence in the British dialects. They also find that have to appears in contexts that were historically reserved for must while got to is associated with contexts that have indefinite reference (Tagliamonte & Smith, 2006:342). Compare these resu lts to the modal system in Toronto English where have to has essentially taken over the entire modal system. They claim that this is a strong indication that English modality is in flux and it appears as though t he Canadian English system has undergone the most The general lack of research on modality from a variationist perspective clearly illustrates the need for more studies like this. Applying the comparative method will provide a baseline for research on modality in other languages by showing us the path of change (e.g. Tagliamonte & Smith, 2006). If other studies mirror the paths found in this study, we can describe this path in more detail, perh aps including where in the grammar these changes begin to take place. 2. 7 Variables and Hypotheses In a usage based framework, variation in language is viewed as structured and ordered, yet heterogeneous as put forth in the principle of structured heteroge neity (Weinreich, Labov & Herzog, 1968). In order to study and analyze this variation in a reliable way, Labov (1972, 1994, 2004) has outlined certain fundamentals. One of these is the identification of the independent and dependent variables. The dependen t variable is the form or forms that are the focus of the study whereas the independent variables are the factors that may or may not have a conditioning effect on the use of the depende nt variable (Labov 1972, 2004).

PAGE 75

75 The independent variables chosen for t his study as possible conditioning factors were selected either because i) previous research on modality has identified them as pertinent or ii) research on universal paths of grammaticalization has categorized them as possible indicators of a grammaticali zed form. The independent variables that will be considered in this study are: type of verb, animacy, type of sentence, polarity, objects (direct, indirect), preceding clitic, tense aspect mood, temporal adverbs, lexical verb type, sex and age (see Appendi x C for a complete list). In sections 2.6.1 to 2.6.10, I will discuss the independent variables chosen for this study, as well as the hypotheses supporting each of these. 2. 7 .1 Verb Type Verb type is a common conditioning factor in morphosyntactic variatio n (e.g. Tagliamonte & Poplack, 1993; Silva Corvaln 1994; Casanova, 1999; Mayoral Hernndez, 2007; Cornillie, 2007) and therefore often considered in these types of studies. For the present study, verb type refers to the classification of the infinitive v erb that follows that follows the obligation marker. Many of these classifications are based on a study conducted by Cornillie (2007). Cornillie (2007: 219), who considers verb type in his study on evidential and epistemic modality in Spanish, looks at the frequency distribution of the deontic and epistemic uses of tener que and finds that it is overwhelmingly used as a deontic modal with action verbs (99.5%). He finds similar results with verbs of perception (100%) and mental activity verbs (96.1%). In fac t, the only verb type that falls below 90% range for deontic modals is the type Cornillie defines as copular verbs and other verbs referring to attribution (85.12%). This result implies that these copular verbs and verbs of attribution may be the access po int for expressing epistemicity for tener que A tendency for

PAGE 76

76 copular verbs to appear with epistemic modals will further support studies such as Pietrandrea (2005) and Coates (1983) who find that stative (e.g. copular) verbs are often found in epistemic co ntexts. It is important to recognize that Cornillie also had a verb type for mental activity and perception, which can easily overlap with verbs of attribution. Thus for the present study, verbs of attribution will be divided into separate clearly defined groups: verbs of perception, psychological/emotion verbs, copular verbs and communication verbs. If the statistics show similarities among these three factors, they will be collapsed into one group. Besides Cornillie (2007) verb type has not been consider ed as a possible factor in determining the use of one variant over another in other studies on obligation markers. I posit that verb type is crucial to this process, given that these collocations are dependent on the presence of a nonfinite verb. Therefore it is suspected that these non finite verbs are partially responsible in determining which variant will be employed. Owing to the frequency distribution of tener que in Cornillie, it is postulated that this construction is more likely to occur with actio ns verbs early on, but as it moves towards epistemicity, it should begin to appear more often with copular verbs and other stative verbs. Taking into consideration that haber de is so frequent in the earlier centuries, it may be that it is favored by the v erb types that encompass attribution early on but, as tener que increases in frequency, this may diminish. 2. 7 .2 Subject and Animacy Another common conditioning factor found in many variationist studies (e.g. Tagliamonte & Poplack, 1993; cy, 2007; Cornillie, 2007; Szmrecsanyi & Hinrichs, 2008; Meyerhoff, 2009) is grammatical person and animacy.

PAGE 77

77 This is particularly important when the variable under examination is a verbal construction, given that the subject is intricately bound to the ver b. In relation to obligation markers, Cornillie (2007: 219) measured the frequency of the different grammatical subjects and animacy with tener que comparing its use in epistemic and deontic contexts with each type of grammatical person (including animate and inanimate). In his data he finds that deontic tener que occurs with animate subjects 97.75% of the time. This is not surprising given that the subject of obligation markers is frequently animate (Tagliamonte, 2002). In addition, Coates (1983: 36) state generally true that examples with second person subjects are stronger (in terms of obligation) than those with first person subjects, while examples with first person subjects are usually stronger than those with third person subjects. Buildi ng on this, I postulate that as tener que advances along the grammaticalization path towards epistemicity it will be less and less favored by animate subjects, particularly in first and second persons since these are strongly associated with strong obligat ion. I expect that animacy, from a diachronic perspective, is an area where change occurs. This is based on patterns found in studies by linguists such as Coates (1983), Bybee and Fleischman (1995) and Pietrandrea (2005), who all suggest that contexts expr essing deonticity occur more often in first and second person subjects, whereas epistemic contexts are more likely to occur with third person subjects. Research also correlates animacy with high focus clauses, which tend to be less grammaticalized than the ir inanimate, low focus counterparts (Klein Andreu, 1991; Bybee et al., 1994 ; Diaz Campos, 2011 ). 8 8 Klein Andreu (1991) relates low focus contexts to relative and subordinate clauses whereas the high focus clauses are more likely to be main clauses. Other literature (e.g. Bybee et al. 1994) attributes low

PAGE 78

78 Taking this into consideration I postulate that, during the 15th and 16th centuries, it is likely that animate subjects will favor tener que predominantly f irst and second person animate subjects. This would mean that, as these constructions develop and move into epistemic contexts, they should shift from contexts with first and second person subjects into contexts with third person subjects. However, given t hat frequency plays an important part in this change, and tener que suffers the most dramatic shift in frequency across centuries, this construction will most likely be the one to undergo this shift in subject preference. That is, in the later centuries I believe that tener que will begin to also be favored by third person as it moves along the grammatical cline while its occurrence with animate subjects becomes less marked. Given the apparent advanced stage of the collocation haber de it is likely that th is construction is already favored by third person in the early centuries. In the later centuries, as tener que increases in frequency and starts to take over the territory of the other forms, we may see haber de lose ground in this area. 2. 7 .3 Sentence Ty pe Declarative or Interrogative sentences are the two sentence types that are included in the present study. Like the previous factors, sentence type is a factor that has been considered in a large number of variationist studies (e.g. Smith, Durham and For focus to clauses that provide bac kground information and high focus to clauses that provide new information. For the purpose of this study, these clauses will be identified primarily by there tendency to occur with the aforementioned factors (e.g. negation, stative verbs, (in)animate subj ects)

PAGE 79

79 Cacoullos, 2011). Although this factor group is often a conditioning factor in these studies, it has only been considered in a limited number of studies on modality. In a stu dy conducted on Toronto English obligation markers Tagliamonte and found that have to occurred almost categorically in interrogative sentences. As a result they were unable to include it in the multivariate analysis. Nevertheless, they ind icate interrogative contexts as a place of interest although they do not expound on why this is Furthermore, Coates (1983: 242) also observes that epistemic modals generally do not occur with interrogative constructions. Taking into consideration that the se epistemic contexts suggest a high degree of certainty in relation to the proposition, and an interrogative sentence would likely cast a degree of doubt on the proposition, this could be an explanation as to why these modals occur less often in interroga tive sentences when expressing epistemicity. Whether this is the case or not, we can postulate that, based on the frequency of haber de in the earlier centuries and the frequency of tener que in the later centuries, these constructions are less likely to o ccur in interrogative sentences at different points in history In the earlier centuries, haber de will likely have a disfavoring effect with interrogative sentences, but as tener que increases in frequency and likely continues to grammaticalize this disfa voring effect may shift towards tener que. In addition, the movement of haber de into epistemic context s could be an indication that the newer form, in this case tener que is grammaticalizing Aaron (2010: 30) indicates that when an older construction g eneralizes semantically, this can facilitate the movement of newer forms into the former territory of the older form. In this case, if haber de has moved into epistemic contexts, it will not be used with interrogatives in

PAGE 80

80 those contexts, thus giving tener que access to these types of sentences. Consequently, in the earlier centuries I posit that interrogative sentences will disfavor haber de while at the same time favoring tener que and the other variants. But as tener que expands in the la ter centuries th e other variant ( haber que ) will likely be used more oft en in interrogative sentences. 2. 7 .4 Polarity Polarity, which determines whether an item in an utterance is positive or negative, has long been considered to be a factor in linguistic change (e.g. Bol inger, 1977; Coates, 1983; Cacoullos, 2011). In terms of modality, negation also seems to play an important role. Many of the studies that examine polarity in changing modal systems look at the placement of the negation in relation to the modal marker (e.g. Plank, 1984; Fernndez de Castro, 1999). This, however, is not relevant to the present study, as the placement of the negation marker is fixed after the 16 th century, occurring directly b efore the construction (or the clitics). Negation is also mentioned when examining epistemicity. Coates (1983), for example, finds that, like with interrogative sentences, negation in English is limited in terms of its occurrence with epistemic contexts. T his limitation on negation in epistemic contexts is echoed in research conducted on Spanish modals by Solano Araya (1982: 17), although he is unclear about which epistemic modals can or cannot be negated. Coates (1983) and Plank (1984: 331) expand on this claim by illustrating that although English must is limited in negative epistemic contexts, it can be replaced with the modal This suppletive relationship between two epistemic modals has not mentioned in

PAGE 81

81 the literature on Spanish modals despite cla ims of the limitations on negation and epistemic expressions. Given these results, I also consider polarity as a factor group, postulating that contexts where negation occurs are likely to favor tener que early. As this construction increases in frequency and likely advances into more epistemic contexts in the later centuries, there should be a shift towards non negated contexts. For the purpose of this study, an example was considered to be negated if the obligation marker or the infinitive verb that follo wed was negated, as in (2. 5 ). Otherwise the example was coded as affirmative as in (2. 6 ) 2.5 yo no tuve que gritarles a ustedes (Fuentes, 1962; CORDE) I d 2. 6 Qu les hemos de hacer ?( Mendieta, 1604; CORDE) What do we have to do to them? 2. 7 .5 Objects Although none of the previous research on modality has included objects in their analyses, it has been examined in other areas. Much of this research has focused on phenomena such as lesmo (e.g. Choi, 1998; Tippits, 20 11 ) or clitic climbing (e.g. Myhill, 1988). These studies tend to examine the role of objects in terms of the placement of the object syntactically (which will be discussed in section 2.3.6) or whether the object is animate, inanimate, oblique etc., and th e role these objects have in the grammaticalization process. Research such as Klein Andreu (1991), Bybee et al. (1994), and Diaz Campos (2011) relates objects to the grammaticalization process by suggesting that, as certain constructions grammaticalize, th ey move from high focus clauses to low focus clauses. They correlate low focus clauses with more grammaticalized forms and identify these

PAGE 82

82 clauses through their association with negation, inanimate objects or subjects and stative verbs. Although the focus o f these aforementioned studies was not modality, it is still possible that a similar tendency is found in these constructions as they grammaticalize. In addition to previous research, it is also important to take the uniformitarian principle into account ( see section 2.6), which states that the factors that effected language change in past centuries are likely to be similar to the factors that prompt language change in modern languages (Poplack & Tagliamonte, 2001). Consequently, a study conducted on the Me dieval Spanish possession markers tener and haber may provide an answer to the development of these obligation markers Urrutia and Alvarez (1995:212) examine the variation of these two markers and find that haber is more often utilized with abstract objec ts, whereas tener appears more with concrete obj ects. As previously mentioned, t hese tendencies may have continued as these possessive verbs developed i nto obligation markers, so this is a pertinent factor group for consideration. If this is the case a sim ilar trend should be found in my data, whereby abstract objects are found with haber de and haber que while tener que and tener de will be the default forms occurring in all other contexts. As a result of the aforementioned studies, the animacy of these objects is considered as well as whether or not these objects are abstract or concrete. It is postulated that as each construction grammaticalizes and moves towards epistemicity, it will start to appear more often with inanimate objects essentially becomin g more and more associated with low focus clauses. Given the high frequency of the construction haber de in the earlier centuries, I posit that contexts with animate objects are more

PAGE 83

83 likely to be areas in the grammar for the newer constructions, like tener que, haber que and tener de, since these type of contexts are associated with high focus clauses. Conversely, I postulate that as these constructions grammaticalize, particularly tener que they will more likely occur in contexts with inanimate objects. F or this study, I considered whether animate (2. 7 ), inanimate objects (2. 8 ), obliques (2. 9 ) or the middle voice (2. 10 ) play a role in the choice of these constructions. In addition, objects were divided into two groups: direct and indirect objects. 2. 7 no te he de amar? (Fernndez de Lizardi, 1818; CORDE) 2. 8 tuvo que atravesar el ro (Heriberto, 1893; Internet Archive) he had to cross the river 2. 9 aquello tena que acabar las doce (Riva Palacio, 1896; CORDE ) That one 2. 10 tenan que retirarse a sus casas ( Fernndez de Lizardi, 1818; CORDE ) They had to withdraw to their houses 2. 7 .6 Preceding Clitic The next linguistic factor to be considered in the study is preceding clitic. Similar to the previous factor group, preceding clitic has not been considered in previous research on modality. Myhill (1998), however, looks at clitics and the semantic differences between preceding and attached clitics and concludes that the position of a clit ic largely depends on the status of the construction as a lexical or auxiliary verb. Given that these obligation markers are grammaticalizing and auxiliarihood is in fact a part of this process, it is possible that the form that has moved furthest along th e grammaticalization path i.e. haber de is likely to be favored by preceding clitics. This is assuming that the construction is being processed as a unit, so the clitic is attaching to the infinitive based on the status of the obligation marker (as a semi auxiliary) plus

PAGE 84

84 infinitive. 9 Any clitic that is not attached to the infinitive will be considered a preceding clitic. This includes clitics that come directly before the infinitive or attached to the obligation marker (2. 11 ) or before the obligation marker (2. 12 ). 2. 11 hanse de perder todo esto? (Annimo, 1541; CORDE) Do you (pl) have to lose all of this? 2. 12 se le haba de morir algn hijo o hija (Toribio de Benavente, 1536 1541; CORDE) one of his children had to have died. 2. 7 .7 Temporal Adver bs Considering that the most thorough research conducted on modality has been limited to the English modal system and that the approach taken by Bybee et al.(1994) is more focused on identifying cross linguistic tendencies, it is possible that there are de velopments within agent oriented modality that have not yet been substantiated. One possible inclination that has not been included in the research is the habitual reference of haber que and tener que in such examples as (2. 13 2. 14 ). Coates (1983: 54) ment ions it in her study of English modal auxiliaries pointing out that have to can be habitual in meaning while modals like have got to and must cannot. 2. 13 porque todos los das habamos de confesarnos (Gonzlez, 1999; CREA) because every day we had to co nfess. 9 Many studies have emphasized the difference between modal auxiliaries and periphrastics, highlighting that modal auxiliaries are more susceptible to change (Coates, 1983; Westney, 1995). For example, Westney (1995: 31) claims that modal auxiliaries ar e more likely to grammaticalize stating that the moda Consequently, criteria for determining the status of a modal verb as auxiliary or quasi modal i.e. periphrastic have been established, most notably Palmer (Westney, 1995, and references therein). If a verb does not meet these criteria then it is not a modal auxiliary but rather a periphrastic or lexical verb ( Westney,1995;15). Considering that these criteria are language specific to English it is questionable whether the claim Westney makes is valid or not in Spanish. Criteria for determining modal auxiliaries in English have been dubbed NICE properties which define the modal auxiliaries: negation, inversion, code, and emphatic affirm ation. For a more in depth discussion on English modal auxiliaries see Coates (1983), Palmer (1987), Westney (1995) or Krug (2000).

PAGE 85

85 2. 14 Todos los das tengo que meterme por ahi, a bailar. (Fuentes, 1958; CORDE) Every day I have to go there to dance. Westney (1995) also finds examples of habitual reference in his study of English modals. He correlates habitual action with future claiming that have to can be used present while with have got to and must the repeated actions are restricted to the future (1995; 134). The only mention of hab itual action in Bybee et al. (1994) is in reference to the past habitual use of would They claim that its use as a habitual past is a separate development from the modal would Crucially, they allude to the fact that this may be a possible lexical split f or the construction would (1994; 238). Taking this into consideration, and the fact that we see examples of habitual reference with these obligation constructions, it is possible that this is a common development of certain agent oriented constructions. Ex amples such as these should be acknowledged and examined in more detail. In addition to habitual action, temporal adverbs may also be able to account for movement of these obligation markers into epistemic contexts. For example, Blas Arroyo (2008) found th at the synthetic future, which is derived from an obligation marker (verb + habere ), is favored by non specific temporal adverbs. Although his study focuses on the future, because the synthetic future comes from an obligation marker and obligation, like fu ture, grammaticalizes into epistemic contexts it is possible that temporal adverbs can elucidate this movement into epistemicity. For the purpose of this study a temporal adverb will be any adverb that makes reference to time (2. 15 ) but each temporal adver b will then be classified as specific or

PAGE 86

86 non specific (for studies that examine the effect of specific and non specific adverbial markers, see Urrea & Gradoville, 2006; Aaron, 2006; Howe & Schwenter, 2008; Blas Arroyo, 2008) so as to elucidate possible ten dencies of habitual acti on, as in 2. 14 above. temporal adverbs as a possible conditioning factor, my hypothesis is centered on Blas 2. 15 para el viaje que tienes que hacer maana (Ibargengoitia, 1979; CORDE) for the trip you have to take tomorrow Aaron (2006) and Torres and Walker (2008) find that as a form moves into more epistemic contexts it is more likely to occur with temporal adverbs in future con texts, which they suggest helps to clearly distinguish epistemic meanings from the future meaning. I would postulate then, that a similar process is occurring with obligation. That is, as these constructions move away from obligation and intention and into purely epistemic contexts they will less likely occur with temporal adverbs thus creating a clear distinction between the two functional spaces. Due to the limited number of adverbs present in the data these will not be included in the multivariate analys is but they will be discussed in detail in Chapter 4. 2. 7 8 Tense, Aspect, Mood Another way to determine whether habitual action is part of the development of obligation is to look at aspect. However, in order to accurately code for aspect and avoid intera ction with temporal adverbs, it was necessary to include mood and tense in this factor group. Aside from habitual actions, tense may also play an important role in determining

PAGE 87

87 e xample, look at obligation in Toronto English and find that all non present tenses were categorically used with have to In their study, it was determined that have to had, in essence, taken over the function of obligation. If this is the case with one obl igation system, is it possible that tense has an effect on other systems? If this is the case and haber de is the more grammaticalized form and has moved towards more epistemic contexts, which of these forms has taken over? In her comparison of obligation marker use in several different canonical literary works De Maeseneer (1998) finds that tener que is the most frequent form used in Mexican Spanish. 10 Thus, it is likely that all non present forms will favor the use of tener que 2. 7 .9 Lexical Verb Type The lexical verb that follows the obligation construction is also considered as a factor group. Similar to some of the other factor groups discussed, lexical verb type has not been considered in any studies that have focused on modality. It has been, however, the focus of many variationist studies. Torres Cacoullos and Walker (2008: 5) point out that lexical type is of particular importance when looking at the grammaticalization of a collocation since these are often the source of language change in diachrony. They conclude that lexical type plays an important role in determining the distribution patterns of the English future. In addition, they illustrate that these collocations occur with certain lexical verbs and eventually become fixed expressions, ultimate ly contributing to the variation of these future forms. 10 She looked at two Latin American works, El general en su laberinto by Gabriel Garca Marquez and Como agua para chocolate by Laura Esquivel. She broke down the frequency counts of the various obligation markers in each of the works and tener que (n=124) occurred more seven times more often than the next most frequent marker, haber que (n=16)

PAGE 88

88 Poplack (2001: 411) looks at lexical type in her study on the French subjunctive and finds that it is constrained when co occurring with the subjunctive. Smith (2000) also finds that lexical verb type can have an effect on the variant in her study of negative do in Buckie Scots. Taking these previous studies into consideration, we can assume that lexical type may have an effect on one or more of the variants, whether it is through lexicalized expressio ns or highly constrained contexts. I would posit, like in verb type, that stative verbs, such as ser and estar tener are more likely to be favored by haber de in the later centuries, assuming that this form has moved into primarily e pistemic contexts. Although I expect to find a similar tendency among all stative verbs, there were enough examples of obligation markers occurring with ser, estar and tener to warrant separating these verbs from the rest of the stative verbs. Additionally Pietrandrea (2005) found that stative verbs often occur with epistemic future in Italian, particularly the verbs avere essere It is also postulated that in the earlier centuries, when tener que first comes into use as an obligation marker, it is not likely to be found with stative verbs (or copular verbs). This is because the form has just entered the functional space of obligation and may still have retention of the older form, possession. Therefore it may be limited to contexts in which both an obligation and a possession meaning can be gleaned thus limiting the lexical verb type with which it can appear. This is based on research conducted on the English obligation system and postulations that the construction have to as an obliga tion marker is derived from contexts such as I have a secret to tell you>I have to tell you a secre t (Brinton,1991; van der Gaaf, 1931).

PAGE 89

89 I also postulate that in the later centuries as tener que continues to grammaticalize, it is found with a greater varie ty of lexical types than hay que or haber de We see evidence of the expansion of tener que in Cornillie (2007: 219), who found that all verb types except copular verbs and verbs of attrition were used with tener que more than 90% of the time. Hay que alt hough also a quite frequent variant according to De Maeseneer (1998), lacks the liberty to appear in all grammatical persons, thus possibly limiting the number of verbs it can appear with. The results and tendencies found in this factor group will be dis cussed in Chapter 4 but this factor group will not be included in the variable analysis due to interaction with verb type. 2. 7 .10 Extralinguistic Variables As mentioned before data were collected on the age, sex, social class, and religious affiliation of each participant. The intention was to include each of these factors in the multivariate analysis. However, this is not possible with all of these categories. With religion, for example, due to the high number of Catholics in the community and the contrast ive low number of other religions (n=2) represented in the speech sample, there were not enough participants to consider religion as a factor in the multivariate analysis. A similar situation occurred with social class. Upon entering the community, I assum ed that the majority of the members seemed to belong to a lower social class. However, after a couple of weeks, it was evident that there was stratification in the social class are (i) the objective, economic measures of property ownership and the power and control it confers on its possessor, and (ii) the subjective measures of

PAGE 90

90 role in status in the community. The property owners are in a power position, given that the rest of the community depends on them for work during harvest months. This consequently creates a natural class boundary. 11 In order to consider this factor in the analyse s, however, there would have been several participants from each class. This was not possible due to the limited number of land owners who agreed to participate in the interviews. As a result, only two social variables were considered: age and sex. 2.7 .10. 1 Age. Age is an important variable in order to capture any change in progress in apparent time. Indeed apparent time is dependent on this dimension, where the element of time is created through the formation of age groups (Labov, 1972: 163). Although it is possible that distinctions among the different age groups could be due to age grading, i.e. linguistic changes that regularly occur during a certain period in life, these claims have rarely been substantiated (Bailey, 2002 and references therein). An em ic approach was used in order to divide the age groups (Eckert, 1997). That is, I subjectively devised groups in terms of life stages as a means to group speakers according to age. This resulted in two different age groups; young adulthood, and adulthood. This division seemed to be a natural division since most of the adults older than 45 years seemed to congregate together whereas the adults younger than 45 years congregated in separate areas of the community. 12 Although it would have been 11 There is one exception to this group. One female, who does not own land, is still considered to be in the middle lower class because she receives money every month from the U.S. government for the death of her husband. This makes her less dependent on the land owners, and neutralizes th e power relationship. 12 Of course there was also a group of people under the age of 18 that also congregated but I did not interview anyone younger than 18 so this group was not included in the analysis.

PAGE 91

91 advantageous to h ave another group, possibly separating middle aged participants from older participants due to life experience, there were not enough female and male participants to do this. 13 In regards to the use of the obligation markers, i t is posited that the younger generation will use tener qu e more given its rise in frequency across centuries. Additionally, I expect them to use haber de less often since Fernndez de Castro (1999) suggests that this construction has moved into formal and written contexts I postulat e that the younger generation will also use one (or more) of the obligation constructions in more contexts than the older generation. However, Tagliamonte (2002) showed no significant effect from age and sex in her study of obligation markers in northern B ritish English so it is also possible that age has no effect on variant choice. 2. 7 .10.2 Sex The other extralinguistic factor examined in this study is sex. In their studies on obligation markers in different varieties of English, Tagliamonte and Smith (20 06) and between men and women in terms of variant choice indicating that women tend to use newer forms. This supports patterns found in other sociolinguistic research such as Labov (1972). Based on this tend ency, it is posited that the choice of variant among women in this group will reflect a more advanced phase in the grammaticalization process of these markers in comparison to the men. Specifically, they will be more likely to use haber de in epistemic con texts, and tener que in a wider variety of contexts in comparison to their male counterparts. 13 An older and possibly separate group those older than 70 years rarely congregated in the streets like the other groups. Instead, they would remain in their homes and accept visitors. But, again, given the limited number of total participants, they were collapsed into the older adulthood group.

PAGE 92

92 2.8 Summary Taking into consideration that variation is structured and that these patterns can be attributed to certain l inguistic and social factors, both a sync hronic and a diachronic analysis of t his modality system allow s for explanations of patterns of modality in modern Spanish such as those observed in Bybee and Fleischman (1985), Marrano (1997) and Cornillie (2007). Furthermore, these results can also corr oborate such typologies postulated by Bybee et al. (1994) and Olbertz (1998) or explicate discrepancies among the functions outlined for these modals. Based on prior research, we can see that the English modal system is in flux and the research that is av ailable on Spanish suggests a similar situation. A comparative synchronic and diachronic analysis can confirm suspicions of a similar process occurring in Spanish. Similarly, results gathered can support or refute claims of diachronic universality. Conside ring the dearth of research that has been conducted on Spanish, this study add s to the existing research on Spanish modals of obligation, none of which examine variation within the system. Furthermore, it will add to our understanding of both the history a nd the modern day reality of Spanish grammar.

PAGE 93

93 CHAPTER 3 COMPETITION, VARIATI ON AND LANGUAGE CHAN GE 3.1 Grammaticalization of Changing Forms Up until this point this study has demonstrated how the aforementioned constructions have developed into obligation markers in similar ways : they all o riginated as verbs that expressed possession and, as their frequency of use increased, they developed into obligation markers. Research that ha s focused on these transformations ha s often categorized these changes as gra mmaticalization (Meillet, 1912 ; Hopper & Traugott, 2003, 2). In addition, investigation into the processes involved in this change has shown that these forms undergo a general semantic pragmatic shift as they move towards a more speaker oriented and thus s ubjective meaning, in a process known as subjectification (Traugott, 1989; Bybee et al., 1994; Tagliamonte, 2004; Narrog, 2005; As these forms begin to undergo grammaticalization and move towards mor e subjective meanings, processes such as decategorialization, phonetic reduction, reanalysis, semantic bleaching or emancipation from the source form are likely to occur Taking into consideration that grammaticalization is a gradual process, it is not req uired that a construction suffer each of these changes to be considered as having undergone grammaticalization. In fact, based on the inherent gradualness of grammaticalization, it is rare to have historical evidence showing a form going through each of th ese processes. What is crucial to the present paper, however, is to identify whether any of the four aforementioned obligation markers has undergone some of these processes. Given that linguists such as Bybee et al. (1991), Hopper and Traugott (1993) and Bybee et al.

PAGE 94

94 (1994) claim that the paths of grammaticalization of these types of modals are universal, it is also essential to identify whether these obligation markers follow this path and corroborate their claims. They posit that constructions denoting a notion of possession develop into obligation markers and eventually acquire an epistemic meaning. In Spanish, we already see evidence of this path within the obligation system. The four constructions haber que tener que haber de and tener de all derive from a source that historically c onveyed a notion of possession 1 As was discussed in section 1.4.1, there is evidence of an emancipation of the constructions haber que and haber de from their source form also indicative of grammaticalization T heir decat egorialization from transitive verbs to obligation markers also point to both grammaticalization and subjectification. We also see with the advent of these markers the fusion with independent morphemes ( de / que ) to form collocations which also indicates gr ammaticalization (Torres Cacoullos, 2006) Finally, a semantic pragmatic shift in each of these collocations from contexts where obligation is expressed to contexts where there is a more subjective meaning would imply grammaticalization Recall that in Ch apter 1, after a cursory examination of these four collocations, it was postulated that these forms have undergone both subjectification and grammaticalization and, with the exception of tener de continue to develop and change. This preliminary assessmen t suggested that Spanish modality develops like modality in other languages thus indicating that the changes found in the modal system in Spanish should follow the universal path postulated by Bybee et al. (1994) 1 For mo re discussion on the development of these forms see Chapter 1, section 1.4.

PAGE 95

95 If this is the case, the data should show the generalization of these constructions into new contexts which would be evident if we find certain patterns of variation possible decategorialization or replacement of older constructions by newer ones. Additionally, an increased fusion of the indepe ndent morphemes of the grammaticalizing constructions should also be seen in the data through a decreased amount of examples with elements such as clitics or negation markers occurring between the components of the collocation. If an increase in frequency is observed along with these possible patterns of change (for a more detailed description on the mechanisms associated with grammaticalization see section 1.3.1.1), it would verify that these forms are evolving through grammaticalization. Little work has b een done on the grammaticalization of these express ions of obligation in Spanish. Taking each of these apparent changes into account, we examine these constructions and determine whether these forms are undergoing grammaticalization. In addition, we look a t both the relative frequencies and the multivariate analyses of each of these constructions as possible corroboration that these forms are undergoing further grammaticalization These results should depict the path that each of these markers takes as they expand into new contexts as obligation marker s and eventually start to appear in epistemic contexts This chapter describe s the development of each of these markers by analyzing the tendencies of each construction over time. One of the ways in which these tendencies are identified is through frequency. There is a large amount of research that cites the importance of frequency in language change (e.g. Haiman, 1994 ; Bybee, 2003a; Torres Cacoullos, 2006 ). In fact, Bybee

PAGE 96

96 (2003a: 603) defines grammaticalization in relation to frequency by identifying it as the process by which a frequently used sequence of words or morphemes becomes automated 3.2 Frequency and Language Change There are two types of frequency to consider when lookin g at a grammaticalizing construction: token and type frequency. Token, or text, frequency is the overall frequency of a construction in the data T ype frequency on the other hand, refers to the variety of lexical elements that are used with a given constr uction. As was summarized in Table 2 1 in section 2.1.1 (repeated here in Table 3 1 for convenience) the token frequencies of these collocations reveal several patterns For example, tener que occurs .11 times for every 5 000 words in the 16 th century By the 20 th century this number jumps to 2.5 times for every 5 000 words, more than 22 times more frequent in this century. We also see a considerable increase in haber que from the 16 th century to the 20 th where in the earlier century, it occurred .07 tim es for every 5000 words but increased to .65 in the 20 th 2 Although this is not as sharp of an increase as tener que the number increase d nine fold which, again, is substantial. The opposite occurs with haber de in the 16 th century where it occurs 3.58 ti mes per 5000 words but decreases to .99 in the 20 th century. Although some of these differences may be attributed to dialect writing style and/or other social differences, these numbers are noticeably large enough to suggest 2 Compare this to Luna Traill (1980) who, in her study on the use of peri phrastic verbs in oral educated speech of Mexico City found that, of the obligation markers, haber de occur red in .4% of the data, haber que in 18.86% of the data and TQ in 51.01% (she also included deber (de), which accounted for 29.71% of the data). Additionally, she did not find any examples of tener que being used as an epistemic marker (p 195).

PAGE 97

97 that some change has taken place. Based on this change, we can assume that there are also changes in the type frequency since an increase in token frequency can often be the result of an increase in type frequency (Bybee, 2003a) Table 3 1 Absolute and relative frequency of each v ariant per century in the written corpora normalized per 5000 words Century (word count) TQ N Normalized per 5,000 TD N Norm HD N Norm HQ N Norm Total 16 th (~2,339,244) 32 .11 89 .29 1,091 3.58 21 .07 1,233 17 th (~960,832) 40 .17 22 .10 852 3.70 19 .08 933 18 th (~403,340) 39 .50 12 .15 256 3.28 9 .12 316 19 th (~789,217) 253 1.00 2 .01 687 .67 79 .32 1,021 20 th (~3,501,511) 865 2.41 3 .01 354 .99 232 .65 1,454 Total N 1,563 128 3,248 381 4,957 To measure the changes in type frequency it is nece ssary to look at the distribution patterns of these constructions within each factor group. As these markers grammaticalize, they will start to generalize and occur with more lexical items (Bybee, 2003a). For example, in a study on deontic and epistemic mo dality in German, Heine (1995: 26) found that deontic readings were more associated with action verbs, interrogative sentences, first and second person subjects, and past or present perfect tense. Given the small inventory of lexical items with which these deontic modals usually occur, their type frequency is relatively low. But as these modals start to increase in frequency they will start to generalize and be used in more contexts, namely those associated with epistemic readings. A similar trend is expect ed in the current study. In the earlier centuries the newer forms should have relatively low type frequencies but will generalize into new contexts as time progresses. A detailed look at each of these factor groups across centuries should uncover these gra dual changes. In each subsection below a table is provided to look at the frequency patterns of each variant across time within a particular factor

PAGE 98

98 group. Each century has two columns; one is labeled frequency of occurre nce of the variant in this particular context in comparison with the other three variants. The other column labeled represents the percentage of occurrences of the variant in the given context compared to the other variants. Each variant is di scussed separately in relation to the frequency of occurrence within each factor group. 3.2.1 Verb Type As we saw in Chapter 2 verb type can have an effect on the interpretation of a modal verb. For example, action verbs are more likely to be found in con texts with deontic modals than with epistemic ones (Heine, 1995, Silva Corvaln 1995; Cornillie, 2007). In addition, older more grammaticalized constructions (e.g. haber de ) are more likely to occur with stative verbs. This is confirmed by such studies as Coates (1983), Pietrandrea (2005); and Cornillie (2007). The patterns of frequency within this group type can be an additional tool to help illuminate the history of these constructions. In the following sections we partially reconstruct the patterns of u se of each of these obligation markers. 3.2.1.1 T ener que Table 3 5 reveals that as time passes tener que starts appearing with a wider variety of verb types. In the earlier centuries tener que is relatively limited in terms of contexts. Although the high est percentages of occurrence are found with perception verbs (7.1%) and communication verbs (17.3%), there is also a high number of examples with action verbs (n=13). As was suggested earlier this is an area in the grammar associated with deonticity and may be an entry point for the then new construction In fact, in terms of raw percentage tener que occurs the most often with

PAGE 99

99 action verbs in every century, despite this not being reflected in the overall relative frequency percentages. One area that has r elatively low overall frequency but is increasingly used with tener que is perception verbs. Cornillie (2007) found that when tener que appeared with a perception verb it was always in a deontic context. This association may be due to the historical associ ation of this construction with perception verbs. In the earlier centuries tener que does not appear in contexts with copular verbs and only minimal ly in contexts with psychological verbs Around the 17 th and 18 th centuries there is an increase in the fre quency of tener que in contexts with psychological verbs ( 14%) but that increase is practically halted during the 19 th century and up until the 20 th where we see a dramatic increase from 16% to 51.5%. Table 3 2 Frequency of tener que with the factor gr oup verb type 16 th 17 th /18 th 19 th 20 th N % N % N % N % Verb Type action 13 2.2 22 3.6 123 31.6 369 64.7 hacer 4 4 17 18.7 16 25.8 58 73.4 perception 1 7.1 11 42.3 7 33.3 62 82.7 communication 9 17.3 8 15.1 25 30.1 93 65.5 psychological 2 2.8 11 14.1 30 16.1 124 51.5 haber/tener 0 0 0 0 2 7.7 5 20 copular 0 0 1 .7 10 8.2 51 39.2 influence 0 0 0 0 2 33.3 7 87.5 other 3 2.2 9 4.1 38 30.4 96 52.2 A similar trend is seen with contexts containing copular verbs although the initial increas e is not seen until the 19 th century and i s minimal compared to that of psychological /emotion verbs. T ener que only accounts for 39% of the contexts containing copular verbs in the 20 th century. This i s relatively low compared to it s occurrence with all ot her verbs types (with the exception of the verbs haber and tener ). This result may imply that as this construction grammaticalizes and moves towards a

PAGE 100

100 more epistemic meaning, it will occur more an d more often with copular verbs (Pietrandrea 2005 ). 3.2.1. 2 H aber de The oldest construction in the group haber de, is used almost exclusively during the 16 th century when it comes to verb type (Table 3 3 ) It occurs in over 80% of the contexts regardless of the type of verb The only exception to this is with communication verbs. Based on the data, c ommunication verbs may have be en an entry point for the other obligation markers and thus were the source of the diminishing control haber de has over the system over the subsequent centuries This is particularly e vident with perception verbs. As haber de decreases in total percentage in this group tener que seems to take over these contexts. T his may be the advent of the spread of tener que into new contexts. Aside from this, haber de is used decreasingly across ti me in general. In fact, excluding contexts with the verbs haber or tener haber de is used less than 50% of the time regardless of the verb type during the 20 th century This is in stark contrast to the 16 th century An other notable increase occurs in th e period between the 16 th century and the 17 th and 18 th centuries with copular verbs. By the 17 th and 18 th centuries haber de has almost monopolized contexts with copular verbs (99.3%) This diminishes slightly however, in the 19 th century as tener que st arts to appear more often in these contexts. An even more dramatic change in occurrence with copulas is noted in the 20 th century as haber de decreases from 87.7% to 44.6%. This decrease is particularly important because the literature identifies stative v erbs a s verb types that commonly occur in epistemic contexts (e.g. Bybee et al. 1994; Pietrandrea, 2005; Aaron, 2006; Cornillie, 2007;) and copular verbs are prototypically the most common of the stative verbs

PAGE 101

101 (Pietrandrea, 2005). This implies that, by the 20 th century, tener que is not only taking over the obligation system but also starting to move into possible epistemic contexts that have historically been reserved for haber de A similar situation has occurred in the English obligation system. I n Toro nto English the obligation marker have to has virtually taken over the entire obligation system and in certain dialects of British English have to is being used in contexts histo rically reserved for epistemic must ( Tagliamonte & Smith 2006 ). Table 3 3 F requency of haber de with the factor group verb type 16 th 17 th /18 th 19 th 20 th N % N % N % N % Verb Type action 543 89.9 554 91.6 239 61.4 101 17.7 hacer 82 82.8 68 74.7 42 67.7 5 6.3 perception 12 85.7 14 53.8 10 47.6 9 12 communication 26 50 4 0 75.5 48 57.8 29 20.4 psychological 62 87.3 57 73.1 132 71.0 80 33.2 haber/tener 88 95.7 36 100 23 88.5 13 52 copular 146 96.1 134 99.3 107 87.7 58 44.6 influence 9 90 6 100 3 50 1 12.5 other 123 88.5 198 90.4 82 65.6 58 31.5 3.2.1.3 H aber que Although haber que appears as an obligation marker quite early in Spanish ( Yllera, 1980 ) it, like tener que tends to appear primarily in contexts with communication verbs during the 16th century (21.2%) as is illustrated in Table 3 4 Again, this furthe r support s the notion that communication verbs are an entry point for the se new er obligation markers. Other than the examples with communication verbs this marker is rarely used in the 16 th century But even the occurrences with communication verbs drop d uring the 17 th and 18 th centuries by 17%. However, as the use of haber que decreases in contexts with communication verbs, the construction spreads to new territory. This is evidenced by a 9% increase with psychological verbs and a 5 %

PAGE 102

102 increase with action verbs. Recall that an overall similar trend was found with tener que : communication verbs as the entry point, then a subsequent spread into new contexts. During the 19 th century haber que sees a small but evident increase of occurrence with all verb type s The found with tener que but it is still clear that this construction is being used in a wider variety of contexts. As it competes with the other obligation markers for territory in the 19 th century, it also bec omes increasingly used with verbs of perception This is based on the 15% increase seen in contexts with haber que between the 17 th and 18 th centuries and the 19 th century This trend changes, however, in the 20 th century as tener que begins to take over t he entire system. H aber que sees a 15% decrease in occurrence with perception verbs during this period Nevertheless, haber que continues to be used in the 20 th century but its use spreads into other contexts. In fact, with the exception of verbs of influe nce and perception there is an increase in the frequency of use of haber que with all types of verbs. Table 3 4 Relative f requency of haber que with the factor group verb type 16 th 17 th /18 th 19 th 20 th N % N % N % N % Verb Type action 5 .8 11 1.8 26 6.7 100 17.5 hacer 2 2 4 4.4 4 6.5 16 20.3 perception 0 0 0 0 4 19 3 4 communication 11 21.2 2 3.8 10 12 20 14.1 psychological 1 1.4 8 10.3 24 12.9 37 15.4 haber/tener 0 0 0 0 1 3.8 7 28 copular 1 .7 0 0 4 3.3 20 15.4 influence 0 0 0 0 1 16. 7 1 12.5 other 1 .7 3 1.4 5 4 29 15.8 3.2.1.4 T ener de As already mentioned tener de essentially falls out of use in the 19 th century (Table 3 5 ) This is evidenced by a steady decrease throughout the 16 th 17 th and 18 th

PAGE 103

103 centuries and the relatively limited number of occurrences of this construction in every century. Early on, during the 16 th century tener de does occur occasionally in contexts with all verb type s but is used most often with communication verbs (11.5%) and the verb hacer (11.1%). This quickly changes during the 17 th and 1 8 th centuries, where there is a shift towards using tener que and haber de with the verb hacer and also a large increase in the use of haber de with communication verbs. 3. 1 como chapuln, de aqu para all, aunque ya tengo de estarme quieto como quince aos. (Martn del Campo, 1976: CORDE ) like a locust, from here to there, even though I already have to stay still like a 15 year old. At the turn of the 19 th century haber que and tener que continue to compete with haber de but tener de becomes all but obsolete, occurring only in a few isolated cases (3. 1 ) This trend continues into the 20 th century suggesting that this form is no longer part of the obligation system. Table 3 5 Frequency of tener de with the facto r group Verb type 16 th 17 th /18 th 19 th 20 th N % N % N % N % Verb Type action 43 7.1 18 3 1 .3 0 0 hacer 11 11.1 2 2.2 0 0 0 0 perception 1 7.1 1 3.8 0 0 1 1.3 communication 6 11.5 3 5.7 0 0 0 0 psychological 6 8.5 2 2.6 0 0 0 0 haber/tener 4 4.3 0 0 0 0 0 0 copular 5 3.3 0 0 1 .8 1 .8 influence 1 10 0 0 0 0 0 0 other 12 8.6 9 4.1 0 0 1 .5 3.2.2 Grammatical Person and Animacy Grammatical person can be a strong indicator of the grammaticalization of these constructions. In separate stud ies of the English modal system Coates (1983) and P i etrandrea (2005) find that future epistemics are more likely to occur with stative verbs.

PAGE 104

104 Bybee and Fleischman ( 1995 ) find a similar pattern with modals and claim that they are more likely to be agent ori ented (i.e. deontic) when the subject is in first person and more likely to be epistem ic when it is in third person. This would imply that as these forms grammaticalize and move towards epistemic ity, they will more likely start to appear in contexts where the subject is in third person. Animacy also plays a role in this process. Recall that in his study on the evidential and epistemic readings of various Spanish semi auxiliaries Cornillie (2007:219) found that tener que occurs as a deontic construction (97 .75%) if an animate subject is present. Coates (1983) and Heine (1995) also suggest that animacy is a determining factor in whether a reading is epistemic or deontic. They suggest that inanimate subjects are more associated with epistemicity than animate s ubjects. Wrnsby (2004: 179) echoes this claim but extends it by adding that existential subject s are also associated with epistemicity. This seems to be partially true in Spanish where, regardless of the obligation marker used, an epistemic meaning can be gleaned, as in (3.2), but it is not an exclusively epistemic context Nevertheless, it does appear to be a context where either meaning can be extracted and thus a possible area in the grammar for identifying the gradual shift into epistemic contexts. Thu s, we can assume that as these forms grammaticalize over time, they will slowly begin to shift away from contexts in first and second person and move towards contexts with inanimate or existential third person subjects. 3 2 porque de aqu adelante no ha d e haber abdiencia en Panam. Otros: mandamos que ** (Annimo, 1544; CORDE) because from today going forward there must not be an audience in Panam

PAGE 105

105 3.2.2.1 H aber de Following a pattern similar to those of the earlier discussed factor groups, we can see in Table 3 6 that, during the 16 th century, haber de is the dominant form in every grammatical person except first (singular and plural) (36%) This could be described as the entry point for new forms in terms of grammatical person H aber de is, howeve r, increasingly used in first person through the 17 th 19 th centuries but this tendency is halted in the 20 th century where we see a decline of 40. 9% points when compared to the occurrences during the 19 th century. Knowing that epistemicity is often associat ed with third person inanimate subjects, it is surprising that haber de continues to be found less and less frequently in these contexts over the centuries, accounting for only 47 .9% of cases in the 20 th century This could be an indication that, because t his construction is moving into primarily epistemic contexts, as Cornillie (2007) suggests, tener que is expanding into its territory, even the territory is associated with epistemic meanings. This could be the beginning of the end of haber de as an obliga tion marker. Table 3 6 Frequency of haber de with the factor group grammatical person 16 th 17 th /18 th 19 th 20 th N % N % N % N % Grammatical Person 1 st 31 36 102 72.9 195 69.9 60 19 2 nd 110 97.3 62 91.2 72 84.7 51 38.6 3 rd animate 658 89.8 620 8 7.7 284 57.1 157 19 3 rd inanimate 240 97.6 284 96.6 130 86.7 80 47.9 animated objects 20 90.9 32 97 3 42.9 1 20 existential 32 97 7 100 2 100 5 71.4 Even though Wrnsby (2004) is focus ing on English when he determines that existential subjects are associated with epistemicity, we can see a somewhat similar

PAGE 106

106 trend in Spanish In the earlier centuries, existential subjects are associated almost exclusively with haber de This trend continues through the 19 th century. Even in the 20 th century the major ity of contexts with existential subjects are still used with haber de (71.4%) although there is a decline, indicating that this may be one of the last areas into which tener que will expand. 3.2.2.2 Tener de In the 15 th and 16 th centuries it is evident t hat tener de was principally used in contexts with first person subjects (Table 3 7 ). Consequently, we can conclude that this form was more connected with an obligation reading than an epistemic one This is expected given the relative youth of this form i n comparison to haber de The strong link to obligation is further supported by the minimal amount of occurrences of this marker with other grammatical persons. Nevertheless, tener de is found in contexts with all grammatical person s with animate or inanim ate subjects, which suggests that this form was perhaps grammaticalizing and moving towards a more subjective meaning before it Table 3 7 Frequency of tener de with the factor group grammatical person 16 th 17 th /18 th 19 th 20 th N % N % N % N % Gramma tical Person 1 st 43 50 22 15.7 1 .4 1 .3 2 nd 2 1.8 0 0 0 0 1 .8 3 rd animate 35 4.8 11 1.6 1 .2 1 .1 3 rd inanimate 6 2.4 2 .7 0 0 0 0 animated objects 2 9.1 0 0 0 0 0 0 existential 1 3 0 0 0 0 0 0 fell out of use. The marker appears to stop gr ammaticalizing, however in the 17 th and 18 th centuries where it occurs primarily in contexts with a first person subject. Those person subjects are predominantly associated with third person animate subjects, which is not associated with epistemicity

PAGE 107

107 to the same degree as third person inanimate subjects. B y the 19 th century even its use with first person subjects dwindles to a mere 4% of all contexts with first person subjects. The diminishing frequency and us e of this form is likely owed to the ever increasing use of tener que 3.2.2.3 Tener que As a new obligation marker we would expect tener que to be used primarily in contex t s with first or second person given that these are generally associated with obli gation. In the 16 th century we see that tener que accounts for 14% of all contexts with first person subjects and only 2.7% of both third person animate and inanimate subjects (Table 3 8 ). This is an expected result What this suggests is that this newer construction is entering the obligation system via contexts that the older forms have moved out of. This would imply, though, that haber de has already started to grammaticalize and move towards areas in the grammar that are associated with e pistemicity in the 16th century (e.g. contexts with third person inanimate subjects). In the 17 th and 18 th centuries there is an increase in the amount of times tener que occurs in third person animate subject contexts, which continues into the later centuries. We also see a decreased use of tener que with first person subjects during th e 17 th and 18 th centuries y the 20 th century this marker is the pre dominant construction used in contexts with first person subjects, account ing for 80.1% of all cases. What is equally apparent during th e 17 th and 18 th centuries is the increase of the use of tener que in contexts with second person subject s Given that Coates (1983) suggests that second person subjects encode the strongest obligation reading, we c an assume that tener que takes on a stronger obligation meaning early on. This continues

PAGE 108

108 all the way through the 20 th century where tener que accounts for 59.8% of all contexts with second person subjects. Table 3 8 Frequency of tener que with the factor group grammatical person 16 th 17 th /18 th 19 th 20 th N % N % N % N % Grammatical Person 1 st 12 14 16 11.4 81 29 253 80.1 2 nd 0 0 6 8.8 13 15.3 79 59.8 3 rd animate 20 2.7 48 6.8 137 27.6 442 53.4 3 rd inanimate 20 2.7 8 2.7 18 12 85 50.9 animate d objects 0 0 1 3 4 57.1 4 80 existential 0 0 0 0 0 0 2 28.6 All these tendencies imply that tener que has been established as an obligation marker and has essentially taken over the territory of both haber de and tener de by the 20 th century in terms of grammatical person and animacy We can also see that tener que is grammaticalizing and moving into epistemic contexts by the steady increase of cases of tener que occurring with third person subjects. By the 20 th century it accounts for over half of the cases with both animate and inanimate subjects. Although this suggests that this ma r ker has grammaticalized and is becoming more associated with epistemicity it is still used minimally with the epistemic related existential subject s where it accounts fo r 28% of all cases. This area in the grammar is still primarily associated with haber de 3.2.2.4 H aber que In terms of grammatical person, there are very limited examples where haber que (Table 3 9 ) is used with anything but non referential (i.e. third per son animate ) subjects. Of course, this is due to the fact that it can only occur with animate subjects and it is always conjugated in third person. Nevertheless, there are several isolated examples of this obligation marker being use with subjects other th an third person ( 3. 3).

PAGE 109

109 3. 3 Los podremos ayudar para que puedan hacer alguna carrera. Total, no hay que precipitarnos...( Azuela, 1985: CREA ) We will be able to help them so that they can run some race. Anyway, That being said, r esearch indicates that grammatical person and animacy are some of the grammatical components that measure or can be used to determine the age of deontic modals (e.g. Cornillie, 2007) Based on the limitations of this construction, it may be that this form can not move into epistemic contexts and thus is limited to deontic ity, as suggested by Cornillie, De Mulder and Vermandere (2009: 115): unlike haber de the impersonal construction hay que is r If this is the case, then th is should be reflected in the results. Table 3 9 Frequency haber que with the factor group grammatical person 16 th 17 th /18 th 19 th 20 th N % N % N % N % Grammatical Person 1 st 0 0 0 0 2 .7 2 .6 2 nd 1 .9 0 0 0 0 1 .8 3 rd animate 20 2.7 28 4 77 15 .4 229 27.6 3 rd inanimate 0 0 0 0 0 0 0 0 animated objects 0 0 0 0 0 0 0 0 existential 0 0 0 0 0 0 0 0 3.2.3 Sentence Type Coates (1983) and Bybee and Fleischman (1995) illustrate that in English there are limitations to epistemic modals occurring in interrogative sentences. Due to the limited amount of research done on the grammaticalization of obligation markers in Spanish, there is little evidence that a similar trend occurs in Spanish. A detailed look at each of these constructions may answer th is question. 3.2.3.1 Haber de The oldest of the four constructions haber de, show s tendencies to be the marker of choice in interrogative sentences up until the 19 th century (Table 3 10 ). Interrogative

PAGE 110

110 sentences, which only account for 6% of the data in the 16 th century, 10% of the data in the 17 th and 18 th centuries, and 18% of the data in the 19 th century occur with haber de in over 80% of the cases. There is a n overall decline in the number of interrogative sentences however, in the 20 th century wher e it accounts for only 6% of the data. During this century haber de only occurs with interrogative sentences 40% of the time, while tener que occurs 53% of the time. Thus we can conclude that this decrease in haber de in interrogative sentences is likely due to loss of territory to tener que It may also indicate the movement of haber de i nto more epistemic contexts to conclude that just based on this data. Table 3 10 Frequency haber de with the factor group type of sentence 16 th 17 th / 18 th 19 th 20 th N % N % N % N % Sentence Type declarative 1028 88.9 997 88.2 509 61.3 318 23.3 interrogative 63 82.9 110 92.4 177 93.7 36 40 3.2.3.2 T ener de As can be seen in Table (3 11 ), in the 16th century tener de accounted for a larger percen tage of interrogative sentences (17.1%) than declarative ones (6.6%) This may suggest that because haber de was grammaticalizing and thus being used in epistemic contexts, tener de was beginning to be the marker of choice in interrogative sentences, Tab le 3 11 Frequency tener de with the factor group type of sentence 16 th 17 th /18 th 19 th 20 th N % N % N % N % Sentence Type declarative 76 6.6 31 2.7 2 .2 2 .1 interrogative 13 17.1 4 3.4 0 0 1 1.1 since epistemicity would imply that haber de would be used in limited ways in these types of contexts (Coates, 1983) This is hard to substantiate, however since tener de

PAGE 111

111 falls out of use before it has the opportunity to completely replace haber de in these contexts. 3.2.3.3 T ener que This marker was ra rely used in interrogative sentences until the 20 th century at which point there is a large increase in the number of cases of tener que in these contexts (Table 3 1 2 ) This further suggests that interrogative sentences may not be indicators of grammatica lization in Spanish, since we would expect tener que to occur in the interrogative sentences early on and gradually, as it undergoes subjectification this would start to diminish. Table 3 1 2 Frequency tener que with the factor group type of sentence 16 th 17 th /18 th 19 th 20 th N % N % N % N % Sentence Type declarative 32 2.8 74 6.5 245 29.5 817 59.9 interrogative 0 0 5 4.2 8 4.2 48 53.3 3.2.3.4 H aber que Si milar to tener que haber que is seldom used in interrogative sentences from the earlier cen turies through the 20 th century (Table 3 1 3 ) Again this could indicate grammaticalization but this is difficult to determine given the relatively low number of interrogative sentences in general (n = 90 ) and the even lower number of haber que occurring i n interrogative contexts (n=5) Table 3 1 3 Frequency haber que with the factor group type of sentence 16 th 17 th /18 th 19 th 20 th N % N % N % N % Sentence Type declarative 21 1.8 28 2.5 75 9 227 16.6 interrogative 0 0 0 0 4 2.1 5 5.6

PAGE 112

112 3.2.4 Polarit y Recall that in section 2.6.4 that I discussed how negative polarity is less associated with epistemicity than with deonticity due to the limitations of negation in epistemic contexts (Coates, 1983; Palmer, 1990; Heine, 1995; Silva Corvaln 1995). Theref ore, the inclination of a construction to not occur with negative polarity is likely attributed to the movement of this construction into epistemic contexts. As can be seen in Table 3 1 4 haber de shows a tendency to be used increasingly less with negatio n as time passes This could reflect its movement into epistemic contexts and thus echo the claim s made by the aforementioned researchers We can assume, then, that contexts of negation may be a good entry point for new obligation markers T his appears to be the case with tener que and haber que They both appear to continually increase in frequency of occurrence in contexts with negative polarity. However, during the 19 th century tener que decreases by about 6%, while haber que increases by about the same amount. It appears, then, that at this point in time haber que seems to be taking over these contexts at a faster rate. B y the 20 th century the Table 3 1 4 Frequency of all constructions with the factor group polarity 16 th 17 th /18 th 19 th 20 th N % N % N % N % haber de positive 987 89.8 993 92.2 551 67.3 309 23.6 negative 104 77.6 114 66.3 135 67.2 45 30.8 tener de positive 80 7.3 34 3.2 2 .2 3 .2 negative 9 6.7 1 .6 0 0 0 0 tener que positive 21 1.9 41 3.8 221 27 799 61.1 negative 11 8.2 38 2 2.1 32 15.9 66 45.2 haber que positive 11 1 9 .8 45 5.5 197 15.1 negative 10 7.5 19 11 34 16.9 35 24 increase of haber que has been overshadowed by the dramatic increase of tener que in contexts with negation. There is also a sharp decline in frequenc y of haber de with negative polarity in the 20 th century. However, while tener que accounts for 45.2% of

PAGE 113

113 contexts with negation, there is still plenty of territory associated with the other two markers. Taking this into consideration, it may be that negati on is not a conditioning factor when determining which of these markers will be used. This is discussed in more detail in Chapter 4 3.2.5 Preceding Clitic In section 2.3.6 we discussed Myhill (1998) study, which indicated that clitic climbing occurs mor e often with auxiliary verbs rather than lexical verbs Although the constructions in the present study may not have reached auxiliarihood yet (Olbertz, 19 98 ), they are often referred to in the literature as both modal and semi auxiliaries. As these forms grammaticalize and move toward auxiliarihood there should be evidence of more clitic climbing. We would expect to see an increase of occurrences with haber de or in the later centuries with tener que in contexts with a preceding clitic. If we look at Tab le 3 1 5 we see that there is a decrease in the total percentage of preceding clitics found with haber de When we look at tener que we see that it increasingly occurs with p rece ding clitic s as time goes by although the difference in actual numbers betwee n preceding clitic a nd others is staggering (79 vs. 786 respectively ) Table 3 1 5 Frequency of all constructions with the factor group preceding clitic 16 th 17 th /18 th 19 th 20 th N % N % N % N % haber de yes 406 94.2 391 97.5 207 99 67 45.9 other 6 85 85.4 716 84.4 479 59.1 287 21.9 tener de yes 24 5.6 9 2.2 0 0 0 0 other 65 8.1 26 3.1 2 .2 3 .2 tener que yes 1 .2 1 .2 2 1.0 79 54.1 other 31 3.9 78 9.2 251 30.9 786 60.1 haber que yes 0 0 0 0 0 0 0 0 other 21 2.6 28 3.3 79 9.7 232 17.7 T hese results suggest that these obligation markers are still not fully grammaticalized and thus are not processed as auxiliaries yet. This is supported by

PAGE 114

114 Bybee et al. (1994:241) who suggest that agent oriented modality (i.e. obligation) is at the be g innin g stages of grammaticalization. T his is the other likely reason why preceding clitic s do not comprise the majority of the cases. 3.2.6 Direct Objects Bybee et al. (1994) and Diaz Campos (2011) echo similar claims as Klein Andreu (1991) in suggesting that one of the ways to identify semantic changes, particularly with constructions expressing modality is by looking at low or high focus clauses. In her study on the grammaticalization of the pluperfect indicative in Latin and the imperfect subjunctive in Sp anish, Klein Andreu (1991) identifies low focus clauses as being associated with the older form while high focus clauses was associated with the newer form. She defines high focus clauses as those that present new information and occur with animate subject s or objects She goes on to identify low focus clauses as those that are identified through the presence of stative verbs and negation and often occur ring in relative clauses. I f we apply the results of Klein Andreu (1991) to the case of the direct objec t we could posit that if it is animate the construction may be less grammaticalized since it is more likely to occur in high focus clauses I n the 15 18 th centuries then, haber de is expected to occur with inanimate objects more often, whereas the new con structions are more likely to be used with animate objects. As we move towards the 20 th century there should be a shift in tener que as it becomes more and more frequent, possibly indicating grammaticalization and therefore may be more likely to be found w ith inanimate objects.

PAGE 115

115 3.2. 6 .1 H aber de As previously mentioned we would expect haber de to occur more often with inanimate objects as time progresses since this is considered characteristic of a construction in a low focus clause In the 16 th century we see that haber de occurs more often with inanimate objects (n =247) than with any other type of objects. This trend continues until the 20 th century at which point the construction tener que experiences a dramatic increase in frequency (see Table 3 1 6 ) and haber de suffers a parallel drop. Nevertheless, there are still more overall cases of haber de with inanimate objects in the 20 th century than with any other variant, although tener que has only slightly less (92 vs. 88 respectively) Despite this, the percentage s show a decrease in the frequency of use of haber de with inanimate objects when compared to the other variants, particularly tener que This could be an early indication of tener que taking over the system and its movement from high focus clau ses to low focus clauses. Table 3 1 6 Frequency of haber de with the factor group direct objects 16 th 17 th /18 th 19 th 20 th N % N % N % N % Direct Objects oblique 122 87.1 157 84.9 51 45.5 50 17.2 copular 151 96.2 137 99.3 109 87.2 55 45.5 inanima te 247 89.5 322 85.4 197 63.5 92 17.6 animate 78 85.7 104 93.7 97 71.9 42 27.3 intransitive se 40 88.9 35 92.1 34 87.2 6 46.2 middle voice 227 96.2 172 96.6 90 69.2 23 46 other 226 78.5 180 81.1 108 63.9 86 28.4 There is also an increase in the p ercentage of occurrences in which haber de is found with copular complements This increase continues until the 19 th century where again, we start to see a decline. Similar increase s are seen during the same period with complements expressed through intra nsitive se (e.g. irse ), as well as with animate

PAGE 116

116 objects but like the copular complements the occurrences begin to decline by the 19 th century. 3.2.6 .2 T ener de During the 16 th century tener de appears to be used in all contexts (Table 3 17 ). In compar ison to the other markers, it occurs more often in contexts with animate objects, inanimate objects, contexts with intransitive se (e.g. irse ) or in contexts where a This tendency starts to change in the 17 th and 18 th centuries where there is a slight shift towards more occurrences in which an oblique complement is present. However, it persists in its occurrence with intransitive se Despite the shift towards oblique complements, the percentage of the cases where we find tener d e occur ring with oblique complements is lower than it was in the previous two centuries. Table 3 17 Frequency of tener de with the factor group direct objects 16 th 17 th /18 th 19 th 20 th N % N % N % N % Direct Objects oblique 7 5 7 3.8 0 0 1 .3 copu lar 5 3.2 0 0 1 .8 1 .8 inanimate 23 8.3 19 5 1 .3 1 .2 animate 11 12.1 3 2.7 0 0 0 0 intransitive se 4 8.9 2 5.3 0 0 0 0 middle voice 9 3.8 1 .6 0 0 0 0 other 30 10.4 3 1.4 0 0 0 0 This may not be categorized as a shift towards oblique complem ents, but rather one of the last strongholds of this form as it slowly starts to fall out of use. By the 19 th century tener de only occurs two times in all the data thus further illustrating its demise. 3.2.6 .3 T ener que As tener que starts to be used as an obligation marker during the 16th century it first starts to appear in contexts with out direct objects or with oblique complements (Table 3 18 ). There are, however, a fe w isolated cases where it occurs with intransitive

PAGE 117

117 se (n=1) or inanimate objects (n =3). This begins to change in the 17 th and 18 th centuries where the percentage of cases of tener que occurring with inanimate objects increases from 1.1% to 8.2%. Surprisingly many contexts where this construction is Table 3 18 Frequency of tener que with the factor group direct objects 16 th 17 th /18 th 19 th 20 th N % N % N % N % Direct Objects oblique 8 5.7 15 8.1 50 44.6 200 69 copular 0 0 1 .7 11 8.8 51 42.1 inanimate 3 1.1 31 8.2 92 29.7 323 61.8 animate 0 0 1 .9 30 22.2 88 57.1 intransitive se 1 2.2 1 2.6 4 10.3 7 53.8 middle voice 0 0 4 2.2 26 20 27 54 other 20 6.9 26 11.7 40 23.7 169 55.8 used with animate objects until the 19 th century. This suggests that, although animate objects may be associated with high focus cla uses and thus newer forms it may not be an entry point for new obligation markers. Instead these newer forms may first occur in contexts with objects. By the 20 th century tener que accounts for over half of all cases with or wit h out an object with t he exception of copular objects. This might indicate that contexts with copulas are one of the last areas in the grammar for obligation markers to acquire. 3.2. 6 4 H aber que Ha ber que and tener que appear to have similar use patterns during the 16 th century, although haber que is more often used in cases with oblique complements or contexts where a direct object is n o t present (Table 3 19 ) One difference between the two markers is that haber que is used with animate objects in the earlier centuries. This could indicate haber que earlier than tener que Or it may be that, because haber que had already come to be

PAGE 118

118 associ ated with these contexts, tener que entered the obligation system being used in other contexts. Table 3 19 Frequency of haber que with the factor group direct objects 16 th 17 th /18 th 19 th 20 th N % N % N % N % Direct Objects oblique 3 2.1 6 3.2 11 9.8 39 13.4 copular 1 .6 0 0 4 3.2 14 11.6 inanimate 3 1.1 5 1.3 20 6.5 107 20.5 animate 2 2.2 3 2.7 8 5.9 24 15.6 intransitive se 0 0 0 0 1 2.6 0 0 middle voice 0 0 1 .6 14 10.8 0 0 other 12 4.2 13 5.9 21 12.4 48 15.8 During the 17 th and 18 th c enturies there is little change in terms of haber que occurring with new types of objects, although there is a slight increase in the percentage of cases of haber que in the contexts in which it was already occurring in previous centuries In the 19 th cent ury this increase continues and we see an expansion of haber que into new contexts, like objects expressing middle voice (10.8%), inanimates (6.5%), There continues to be an increa se in the use of this marker with all types of objects with the exception of intransitive se and middle voice. Nevertheless, the percentage of occurrences of haber que in each of these cases is minimal in comparison to those of tener que 3.2.7 Indirect Ob jects Similar to direct objects, this factor group was primarily chosen to determine the age of the obligation markers and where in the grammaticalization process they are. I postulated that i ndirect objects can help identify whether these markers are in h igh or low focus clauses through animacy. In each of the data sets every example of an indirect object referred to an animate being. As a result each occurrence was marked

PAGE 119

119 as either having an indirect object or not. s of subjunctive forms, animate objects should be associated with the newer forms. Therefore, in the earlier years we expect to see haber de more often in contexts with indirect o bjects but there should be a de creasing number of cases in which it appears w ith indirect objects as time progresses. In addition, we would expect the presence of an indirect object to be a possible entry point for new forms. Just as expected, in the 16 th century haber de accounts for 87.8% of all contexts with indirect objects. T ener de is found in 8.5% of the data while tener que only accounts for 3.7% and haber que (Table 3 2 0 ) H aber de maintains its hold during the 17 th and 18 th centuries but starts to be used less (71.8%) in the 19 th centu ry as tener que and haber que start to be employed in its place By the 20 th century haber de only accounts for 2.4% of the cases with indirect objects as tener que becomes the dominant choice (57.8%) and haber que is found in 17.6% of Table 3 2 0 Frequen cy of all factors occurring with the factor group indirect objects 16 th 17 th /18 th 19 th 20 th N % N % N % N % haber de present 72 87.8 101 87.8 56 71.8 84 24.3 not present 1019 88.5 1006 88.7 630 66.9 270 24.4 tener de present 7 8.5 9 7.8 0 0 1 .3 not present 82 7.1 26 2.3 2 .2 2 .2 tener que present 3 3.7 4 3.5 19 24.4 200 57.8 not present 29 2.5 75 6.6 234 24.8 665 60 haber que present 0 0 1 .9 3 3.8 61 17.6 not present 21 1.8 27 2.4 76 8.1 171 15.4 cases. These results illustrate tha t haber de has lost its territory to the grammaticalizing tener que This could indicate that haber de has grammaticalized and is showing evidence of appearing in low focus clauses in the later centuries This claim is slightly diminished when an equally low percentage is found with haber de when an indirect object is not present thus suggesting that the general territory of haber de is being

PAGE 120

120 overtaken by tener que The increasing use of tener que whether an indirect object is present or not, demonstrate s that this form is being used as often in low focus clauses as in high focus clauses which seems to point to the grammaticalization of this construction. 3.2.8 Tense Aspect Mood As mentioned in section 2.6.8 tense mood and aspect play an important role in det er mining how much an obligation construction has grammaticalized. Recall that, i n his study of German modals Heine (1995:26) indicates that modals are more likely to express deonticity if they are in past or perfect tenses. Similar tendencies are fo und in studies conducted by Palmer ( 1990 ) and Tagliamonte ( 2007 ) on English modality. In their studies on Spanish modality, Sirbu Dumitrescu (1988) and Silva Corvaln (1995 : 90) claim that tener que when used as an obligation marker tends to o ccur in present tense Based on these findings we can hypothesize that the newer forms, especially in the earlier centuries, will be more likely to occur in present tense and, possibly, in the perfect tenses Heine (1995) identifies several important crit eria that help identify epistemic contexts. One of these criteria implies that the event can not have taken place yet. This would implicate future tense in epistemic contexts. In addition, he emphasizes the importance of probability stating that, although the event is non factual, there is a high probability that it will occur. Palmer (1990) and Coates (1995; 1983) echo this reference to non factuality. Of course, non factuality or probability is related to mood. Given these criteria for epistemicity, we wo uld assume that constructions occurring in conditional or subjunctive forms are more likely to be non factual and thus related to a more epistemic reading.

PAGE 121

121 Another important tendency in this factor group is habitual action. Both Coates (1983) and Westney ( 1995 ) identified habitual uses of English deontic modals in the ir descriptions of the modal system. Although they did not identify habitual action a s a possible point in the grammaticalization path, they did identify specific contexts in which these action s occur. For example, Westney (1995) found that the habitual action must refer to both the present and the future when referred to with have to wher e as with have got to and must the habitual action can only refer to the future. I s i t possible that this co nsistent reference to the future is an indicator of grammaticalization? This might be answered if we see regular patterns among these four const ructions. 3.2.8 .1 H aber de In the 16th century like in previous results, haber de accounts for the majority of the example s, regardless of tense, aspect or mood (Table 3 2 1 ). If we look at the tendencies of this marker across centuries it is difficult to see any clear patterns except for the steady decline of the use of haber de in every context. One possible patt ern may be the less dramatic decline of haber de in contexts where it occurs in present tense. However, even in this context it suffers a dramatic drop in the 20 th century. There is a slight increase from the 16 th centur y to the 17 th and 18 th centuries of haber de in imperfect (+3%), which is likely to be associated with habitual actions and may tell us something about the path of grammaticalization. We also see an increase of this construction in future tense which does fulfill one of the criteria outlin ed in Heine (1995) for epistemicity There is also an increas e in cases of haber de in subjunctive up until the 19 th century, which is associated with non factuality. Similarly related to non factuality is the conditional mood, a context in which haber de has managed to maintain control through the 20 th century. The fact that haber de still occurs a good

PAGE 122

122 percentage of the time in future tense (35.1%) and conditional (38.1%) suggests that this construction is associated with epistemicity. Table 3 2 1 Freque ncy of haber de occurring with the factor group tense aspect mood 16 th 17 th /18 th 19 th 20 th N % N % N % N % Tense Aspect Mood present 765 88.7 440 86.3 466 77.3 201 32.8 future 7 70 15 88.2 4 26.7 40 35.1 preterit 14 87.5 34 75.6 30 38.5 27 11.3 imperfect 234 91.4 561 94 167 66.5 39 14.4 present perfect 2 66.7 0 0 0 0 0 0 conditional 3 75 8 72.7 5 22.7 45 38.1 present subjunctive 10 55.6 11 68.8 6 46.2 0 0 imperfect subjunctive 36 94.7 7 58.3 8 57.1 1 5.9 future subjunctive 6 100 0 0 0 0 0 0 other 14 70 30 90.9 0 0 1 2.3 3.2.8 2 T ener de In absolute numbers t his construction is used the most in present tense in the 16 th century (n=74/89) This could partially be explained by the fact that third person represents 70% of all contexts during this century. If this is not the case, the tendency to occur more in present tense may limit this construction to deontic readings as Heine (1995) suggests This trend continues into the 17 th and 18 th centuries (n=24/32) before tener de falls out of use in the 19 th century. What is interesting about this is that present tense only accounts for 41% of the data during these two centuries, yet tener de still is found most often (75%) in the present tense further supporting claims made by Heine (1995). T his is likely another factor for the death of this form.

PAGE 123

123 3.2.8 .3 T ener que When looking at the frequency of this construction in the 16 th century we see the majority of the cases are in present tense (n=12) although it accounts for only 1.4% of the pre sent tense data (Table 3 2 2 ) In the 17 th and 18 th centuries tener que continues Table 3 2 2 Frequency of tener que occurring with the factor group tense aspect mood 16 th 17 th /18 th 19 th 20 th N % N % N % N % Tense Aspect Mood present 12 1.4 29 5.7 8 7 14.4 248 40.5 future 2 20 2 11.8 11 73.3 65 57 preterit 2 12.5 8 17.8 45 57.7 206 86.2 imperfect 6 2.3 26 4.4 59 23.5 188 69.6 present perfect 0 0 1 25 13 100 15 100 conditional 0 0 2 18.2 16 72.7 63 53.4 present subjunctive 5 27.8 4 25 5 38. 5 13 86.7 imperfect subjunctive 1 2.6 4 33.3 6 42.9 15 88.2 future subjunctive 0 0 0 0 0 0 0 0 pluperfect 0 0 0 0 0 0 7 100 conditional perfect 0 0 0 0 0 0 1 100 subjunctive pluperfect 0 0 0 0 0 0 1 100 other 4 20 3 9.1 11 100 0 0 to occur mo st ly in present tense even though it is used with a wider variety of other tense, aspect and mood markers during these centuries. This expansion continues into the 19 th century, particularly its occurrence in preteri t (37.7%), present perfect (100%) and f uture tense (73.3%). By the 20 th century the culmination of this trend is that tener que practically monopolizes all contexts, with the few exceptions being future tense, present tense and conditional Again, this clearly indicates that this form is quick ly taking over the obligation system and the few remaining areas future and conditional will likely take longer given their association with epistemicity and haber de Present tense also seems to be an area where there is still substantial variation, whi ch indicates

PAGE 124

124 that, despite the takeover by tener que the other constructions are still used in contexts with deontic readings. 3.2.8 .4 H aber que This construction seems to occur primarily in present tense and imperfect tense, although there are cases of it occurring in other forms as well. This pattern is evident through all centuries even into the 20 th century where haber que maintains a small but stable percentage of the contexts in present tense (26.3%) and imperfect (15.9%). On the one hand, if we con sider that the present tense is more associated with deontic readings, we could conclude that haber que has not moved into epistemic contexts yet. On the other hand, the fact that this construction is also associated with imperfect tenses could suggest tha t this construction is associated with habitual actions, which may point to movement along the cline of grammaticalization. 3.2.9 Temporal Adverbs This factor group was also chosen because of the tendencies found with the habitual use of obligation marker s in the English modal system by Coates (1983) and Westney ( 1995 ). Aside from using aspect markers like the imperfect tense, temporal adverbial clauses or adverbs can also indicate habitual actions (3. 4) Although there were a minimal amount of temporal a dverbs in general Table 3 2 3 shows the tendencies throughout the centuries. 3.4 se lo cont a mi confesor, porque todos los das habamos de confesarnos (Gonzlez, 1999: CREA) I told it to my confessor because we would have to confess every day. In addition to looking for the tendencies of these markers to appear in contexts that implied habitual activity, all adverbs referring to a future action were also examined.

PAGE 125

125 This is based on the reference to an action not yet completed as being one of the cr iteria identified by Heine (1995) as characteristic of epistemicity. As a result, it was important to investigate whether future temporal adverbs were part of the grammaticalization process. One way to do this would be to look to see if there is an increas e or decrease in the frequency of co occurrence of future adverbials for each marker. If there appear to be tendencies across the centuries that repeat with each marker, this could be another way to determine how far along the grammaticaliza tion path each construction is. Table 3 2 3 The occurrence of all obligation markers with t emporal adverbial Markers across centuries 16 th 17 th /18 th 19 th 20 th N % N % N % N % haber de habitual 15 83 13 86.6 11 68.7 5 17.8 future 42 97.6 47 95.9 18 62.1 19 25.6 t ener de habitual 2 11.1 1 6.6 0 0 0 0 future 0 0 2 4 0 0 0 0 tener que habitual 1 5.5 1 6.6 4 25 19 67.8 future 1 2.3 0 0 11 37.9 46 62.1 haber que habitual 0 0 0 0 1 6.2 4 14.2 future 0 0 0 0 0 0 9 12.1 Total temporal markers (including non fut ure/habitual) 77 100 68 150 Looking at Table 3 23 however, we see that there is no clear pattern that the constructions follow. As in the other factor groups, haber de accounts for most occurrences with both future and habitual markers during the 1 5 th and 16 th centuries but slowly loses it s ground as tener que becomes more and more frequent. By the 20 th century haber de only accounts for 17.8% of all contexts with habitual markers compared to tener que with 67.8%. A similar result is found with cont exts in which future markers occur with haber de accounting for 25.6% of the cases in the 20 th century whereas tener que accounts for 62.1%. Haber que

PAGE 126

126 temporal markers until the 20 th century with the exception of one case in the 19 th century. 3.3 Multivariate Analyses Although the results from the frequency data above illustrate that the obligation system in Mexican Spanish has experienced a shift over the last few centuries, they do not give a detailed account regarding whic h factor groups have the greatest effect on these changes, or where exactly in the grammar these changes are taking place. Multivariate analyses can provide this type of information. Aside from the numerous studies on changes that happen in modal systems across languages (e.g. Lyons, 1977, Bybee and Pagliuca, 1985, Traugott, 1989), the development of strong obligation in Spanish is largely unknown. Based on the results reported in the previous sections and the dramatic increase in frequency of tener que an d the subsequent decrease in frequency of haber de we might expect to see similar patterns of change in the conditioning factors across time. That is, the factors that influence the speaker to use one obligation marker over another may also fluctuate over time. The results from the frequency data hint at this, with the large fluctuations in type frequency from one century to another, but results from multivariate analyses can give a more detailed look as to when and where these shifts began. Given the low frequency of all of the variants except haber de during the earlier centuries, it was impossible to conduct multivariate analyses. I n effect, it was only possible to look at the variation of these construction s through a multivariate lens during the 19 th and 20 th centur ies. Even then, it was only possible to compare haber de and tener que during the 19 th century In the 20 th century the frequency of these f orms was

PAGE 127

127 slightly more balanced, so an analysis could be done to compare each of the remaining three constructions. Despite the limited number of multivariate analyses, the results distinguish certain patterns in the data that may have otherwise been difficult to ascertain. In fact, the most dramatic changes occurred between the 19 th and 20 th centuries, so the snapshot provided for these two time periods reveals the factors that constrain the use of each of these markers in a period of time where a shift appears to be occurring. A detailed look at the results of these analyses is offered Each subsectio n examines and presents the results of the binomial up and down multivariate analyses. The first three analyses have haber de as the application value, or the measuring source. In section 3.3.4 tener que is used as the source of measurement in order to det ermine the likelihood of a speaker choosing tener que or haber que Each subsection contains a table (see Tables 3 2 7 3 2 8 3 2 9 3. 30 ) detailing the results. The column marked N of variant gives the total number of the application variant in that parti cular context. The next column, labeled percentage of variant presents the relative frequency of occurrence of the application value in the given context. The column gives the probability values of the likelihood of the variant occurring in that particular context. A weight of .5 or above indicates that the context favors the use of the application, whereas a value of less than .5 would suggest a less favorable context. The further the number is away from .5, and keeping in mind its relative posi tion among other factors in its group, the stronger the favoring or disfavoring effect is. A weight that is close to .5 indicates that the factor neither favors nor disfavors the variant. Weights for factor groups that are rendered not significant by

PAGE 128

128 the b inomial up and down are placed in brackets. Finally, the column entitled total number of variant provides the total number of both variants that occurred in the given context. In addition to the multivariate analyses, a comparative analysis of haber de and tener que across centuries is examined in section 3.4. The comparative analyses give a cross century comparison of the conditioning factors of haber de and tener de to see if changes have occurred in terms of the magnitude of effect or the hierarchy of constraints (see section 2.5 for a detailed description on the hierarchy of constraints). The results from this comparison are presented in Table 3 31 Table 3 31 provides a list of both significant and non significant factor groups, in order of their ma gnitude of effect in the 19 th century. The column labeled provides a summary of which factor groups were altered in terms of the strength of effect from the 19 th century to the 20 th century. The two factor groups that show a change in order, direct objects and verb type, are examined in further detail in order to identify whether changes also occurred within the hierarchy of constraints. Table 3 3 2 which is provided for this portion of the comparative analyses has two columns labeled These are the probability values from each century. The order of these weights is compared. A change in order suggests changes in the underlying grammar of Mexican Spanish regarding these forms, which will be discussed in more detail below. 3.3. 1 H aber de vs. T ener que : 19 th Century Although there were enough examples of tener que to conduct a multivariate analysis, the type frequency of this form is still quite limited in the 19 th century. As a result, there were not enough examples to conside r type of sentence, preceding clitic or indirect objects in the analysis. T he fact that these groups could not be considered

PAGE 129

129 further illustrates the limited contexts within which tener que can occur at this time. That being said, a multivariate analysis wa s carried out with the remaining factor groups which can be seen in Table 3 2 4 There are three significant factor groups: direct objects, verb ty pe and polarity. The direct objects factor group has the strongest magnitude of effect whereas p olarity has the weakest. Within the direct object factor group it was necessary to combine the objects that were coded as middle voice, intransitive se or copular with the category labeled as other due to the limited number of examples and similar preliminary stat istical results in each of these groups. This factor slightly favo rs the use of haber de but this tendency i s tenuous at best. Both a nimate and inanimate objects show no tendency to favor or disfavor haber de whereas oblique complements tend to strongly disfavor the use of haber de Within the factor group v erb type there was also a need to combine groups given similar preliminary results and the limited number of examples. Copular verbs were combined with psychological/emotion verbs. The verb hacer was collapsed with action verbs Moreover, verbs of influence, perception verbs and the verb s haber/ tener were a few notable tendencies within this group. We do see that there is a slight tendency for p sy chological/emotion and copular verbs to favor the use of haber de This was also found in the results for type frequency. This reaffirms that tener que is still not used as often with these types of verbs, which indeed are the types of verbs most often ass ociated with an epistemic reading. There is also a tendency for communication verbs and action verbs to disfavor haber de This goes along with what Cornillie (2007: 219) suggests when he finds that

PAGE 130

130 these types of verbs infer a deontic reading when they oc cur with tener que Again, this is another indication that tener que has not expanded into contexts that include lexical item s associated with epistemicity. Table 3 2 4 Analysis from the 19th century: Linguistic factors conditioning haber de occurrence in comparison with tener que in CORDE. Probability of Input: 74.7 (.747) N =686 /939 Factor Group N of HD % of HD Weight % of data Direct Objects other 341 81 .56 45 (in)animate 294 71 .49 44 oblique 51 51 .29 11 Range 27 Verb Type pysch/emotion 239 86 .64 30 other 118 71 .47 18 action 281 67 .43 45 communication 48 66 .40 7 Range 24 Polarity negative 135 81 .59 18 positive 551 71 .48 82 Range 11 Grammatical person 1 st person 267 74 [.52] 38 other 419 73 [.49] 62 Range Log likelihood = 515.642 P value= .05 Total Chi square = 51.0837 Chi square/cell = 1.1610* The brackets [ ] indicate that this factor was not statistically significant The last significant factor group, polarity, shows tha t contexts where neg ation is present slightly favor haber de This could be an indication that despite the movement of haber de into more epistemic contexts, it has still managed to maintain control in certain contexts, negation being one of them. This ma y be one of the last areas into which tener que expands as it continues to increase in frequency. 3.3.2 H aber de vs. T ener que : 20 th C entury The multivariate analysis for these two variants in the 20 th century was more revealing compared to the 19 th centur y given the increased amount of occurrences and

PAGE 131

131 the ability to include every factor group in the analysis. As shown in Table 3 2 5 t he following factor groups were significant: verb type, grammatical person, preceding cliti c, direct objects, polarity, and type of sentence. Indirect o bjects were not found to be a significant factor group. Within the factor group verb type, which had the greatest magnitude of effect, it was necessary to collapse several factors. Verbs of perception and verbs of influence wer e combined i nto the group labeled as In addition all occurrences of hacer were combined with all action verbs. Results indicate that c opular verbs strongly favor the use of haber de (.76), which corroborates the results discussed in section 3.2.1. 2 and suggest that the association of copular verb s with epistemicity has perhaps slowed the expansion of tener que into this territory. We see a similar trend with psychological/emotion verbs, although they favor haber de to a lesser extent (.62) Action verbs disfavor the use of haber de thus support ing the claims that this construction is moving or has moved into primarily epistemic contexts, primarily because Heine (1995) and Cornillie (2007) associate action verbs with d eontic readings. Within the fact or group of animacy and grammatical person, contexts with a third person inanimate subject tend to favor the use of haber de Bybee and Fleischman ( 1995 ) Heine (1995) and Pietrandrea ( 2005) suggest that epistemic contexts are more often found with third p erson inanimate subjects. This association could have hindered the expansion of tener que into this territory. Contexts with second person subjects also tend to favor haber de slightly. Given that Coates (1983) has identified second person as the grammatic al person most associated with strong obligation, we can conclude that haber de still has strong ties to obligation readings However, based on the fact that

PAGE 132

132 contexts with first person subjects, another grammatical person strongly associated w ith obligatio n, disfavor the use of haber de it is also clear that haber de is losing ground The presence of a preceding clitic also seems to favor the use of haber de Given that linguists like Myhill (1988) associate clitic climbing with more grammaticalized forms we would expect that they are more likely found with the older form haber de which is indeed the case. Several scholars have also identifie d low focus clauses as associated with relative clauses, since they often supply background information (K lein And reu, 1991; Bybee et al., 1994). These clauses are often identified by negation and inanimate objects. We do not see a pattern either way within the factor group of direct objects with inanimate objects. There is a slight tendency for contexts with animate objects to favor the use of haber de but this is not strong. There is also a tendency for contexts with middle voice and intransitive se to favor the use of haber de (.66). T he nature of middle voice makes the object behave as both the subject and object, which may be why these contexts are associated with the older construction Thus i t could be suggested that middle voice is also associated with epistemicity although more research is needed to verify this We see that, within the factor group polarity contexts with negation slightly favor the use of haber de Although the aforementioned linguists associate negation with low focus it is also identified by Coates (1983), Westney ( 1995) and Silva Corval n ( 1995 ) as occur ring less often in epistemic contex ts given the limited contexts within which negation can occur with epistemic forms Taking this into consideration it is possible that haber de has grammaticalized and is thus associated with low focus clauses, which gives it an association with negation However, as it moves into more

PAGE 133

133 epistemic clauses this association should decline, which has not happened, thus indicating that this form has not fully grammaticalized. Table 3 2 5 Data from the 20th Century: Linguistic factors conditioning haber de occur rence in comparison with tener que in CORDE. Probability of Input:) 26.7 (.267), N = 354/1219 Factor Group N of haber de % of haber de Weight % of data Verb Type copular 71 56 .76 10 psych/emotion 80 39 .62 17 other 68 29 .51 19 communicatio n 29 24 .46 10 action 106 20 .39 44 Range 37 Grammatical Person 3 rd inanimate 85 49 .66 14 2 nd person 51 39 .57 11 3 rd animate 158 26 .49 49 1 st person 60 19 .39 26 Range 26 Preceding Clitic yes 67 46 .70 12 no 287 27 .4 7 88 Range 23 Direct Objects middle voice 29 46 .62 6 other 86 34 .59 22 animate 42 32 .57 12 inanimate 92 22 .46 37 oblique 50 20 .41 23 Range 21 Polarity negative 45 41 .66 9 positive 309 28 .48 91 Range 18 Type of Sentence interrogative 36 43 .66 7 declarative 318 28 .49 93 Range 17 Indirect Objects present 84 30 [.54] 23 not present 270 29 [.49] 77 Log likelihood = = 650.39 P value= .05 Total Chi square = 3 45.437 Chi square/cell =1.1326 *The brackets [ ] indicate that this factor was not statistically significant. The final significant factor group is the type of sentence. There is a slight tendency for interrogative sentences to favor haber de (.66). Coates (1983) Bybee and

PAGE 134

134 Fleischman (1995) suggest that epistemic contexts limit the possibility of interrogative sentences. The assumption was, then that if one of the markers had moved into primarily epistemic contexts it would less likely be found in i nterrogative sentences. Since this is not the case with haber de we assume that this construction still has a stronghold in contexts with deontic readings. 3.3.3 H aber de vs. Haber que : 20 th Century The vast majority of cases with haber que occur in third person and, if a clitic is present, it unfailingly occurs after the infinitive verb. As a result of these limitations, grammatical person and preceding clitic were excluded from the multivariate analyses In addition, due to the limited number of occurren ces of haber que in interrogative sentences type of sentence was also excluded. There are two significant factor groups, direct objects and verb type (Table 3 2 6 ) The results of d irect objects, which had the greatest effect on the choice of variants, sho w that contexts with inanimate objects disfavor the use of haber de Given that a similar trend was found when comparing haber de with tener que we can assume that these types of contexts are areas in the grammar into which the newer constructions are exp anding. There is also a slight tendency for the objects labeled as other which includes copular objects, middle voice and intransitive se to favor the use of haber de This is likely due to the stronghold haber de seems to have on contexts with copular verbs, and verbs used with middle voice. This is somewhat corroborated in the statistics of the second significant factor group, verb type. Within verb type, there were several factors that had to be combined. Verbs of communication, influence, perceptio n and the verbs haber and tener were combined with the group labeled as other Contexts with the verb hacer were also collapsed with

PAGE 135

135 action verbs while contexts with copular verbs were combined with p sychological/emotion verbs. In the results we see th at copular and psychological/emotion verbs slightly favor the use of haber de which is similar to what was found in the comparison between haber de and tener que This further supports the idea that haber de has a stronghold in this area of the grammar. T here is also a tendency for action verbs to disfavor the use of haber de signifying the possible development of haber de into non deontic associated contexts. Table 3 2 6 Data from the 20 th century: Linguistic factors conditioning haber de occurrence in comparison with haber que in CORDE. Probability of Input: 61.3 (.613) N =354 /586 Factor Group N of HD % of HD Weight % of data Direct Objects other 170 73 .62 40 animate 42 64 .55 11 oblique 50 56 .45 15 inanimate 92 46 .37 34 Range 25 Verb Type pysch/emotion 138 71 .58 33 other 110 65 .55 29 action 106 48 .39 38 Range 18 Polarity positive 309 61 [.51] 86 negative 45 56 [.45] 14 Indirect Objects not present 270 58 [.50] 75 present 84 58 [.49] 25 Log likelihood = 368.426 P value= .05 Total Chi square = 44.6078 Chi square/cell = .9913 *The brackets [ ] indicate that this factor was not statistically significant. 3.3.4 T ener que vs. Haber que : 20 th Century Similar to the previous analysis, gra mmatical person and preceding clitic are not considered in the multivariate analysis. In addition, due to the limited number of examples of haber que in interrogative sentences, this group was also excluded. The significant factor groups were polarity and direct objects (Table 3 27 )

PAGE 136

136 Table 3 27 Data from the 20 th century: Linguistic factors conditioning tener que occurrence in comparison with haber que in CORDE. Probability of Input: 79.4 (.794), N =865 /1097 Factor Group N of HD % of HD Weight % of data Polarity positive 799 80 .52 91 negative 66 65 .32 9 Range 20 Direct Object oblique 200 84 .59 22 other 254 80 .52 29 inanimate 411 76 .45 49 Range 14 Verb Type other 263 82 [.55] 29 action 427 79 [.50] 50 psy ch/emotion 175 75 [.44] 21 Indirect Objects not present 665 80 [.51] 76 present 200 77 [.46] 24 Log likelihood = 553.905 P value= .05 Total Chi square = 41.67 Chi square/cell = 1.157 *The brackets [ ] indicate that this factor was not statistic ally significant. Within the factor group polarity there is a strong tendency for contexts where negation is expressed to disfavor the use of tener que In the pr evious analyses, we also saw that these same contexts disfavored the use of haber de This s uggests that this is an area in the grammar into which haber que was able to generalize The other significant factor group shows a slight tendency for contexts with oblique complements to favor the use of tener que The factor labeled as other which i nclude s middle voice intransitive se the other. This is probably because n either one of these two constructions is strongly associated with these types of objects yet since they are still more likely to be used with haber de Contexts with inanimate objects slightly disfavor the use of tener que although the tendency is too slight to draw any concrete conclusions.

PAGE 137

137 3.4 Comparative Analysis Based on the fact that the only common multivariate analyses possible in both centuries were the comparisons of haber de and tener que these were the only analyses that could subsequently be compared using the sociolinguistic comparative method (see section 2.6 for details on this method). As a reminder, a shift in the order of effect of significant factor groups across centuries suggests a possible change in progress. If in addition to the shift in significant factor group order, there is also a change in the constraint hierarchy within the factor groups, this also indicates a change and suggests where in the grammar the change is probably occurring. In the case of the present study a change in the constraint hierarchy will indicate where in the grammar the changes are taking place. The results of this comparison ar e seen in Table 3 28 Factor groups that were significant in the 20 th century but were excluded in the 19 th century were not included in this comparative analysis. Table 3 28 Variation of haber de and tener que across two centuries Factor Group 19 th Centu ry Direction of effect 20 th century Direct Objects significant different significant Verb Types significant different significant Polarity significant same significant Grammatical Person not significant -significant As we can see, there is a shift in the order of significant factor groups from the 19 th to the 20 th centuries. In the 19 th century, direct objects had the greatest magnitude of effect whereas in the 20 th century, verb type has the greatest effect. This suggests possible changes in progre ss and warrants further examination of the constraint hierarchies of each of these two factor groups. An examination of the constraint hierarchies of each factor group exposes several important changes. Within the factor group verb type, there is a change in the order of

PAGE 138

138 action verbs and communication verbs. This could be an area in the grammar where change is occurring. Although the details are more ambiguous in the direct object factor group, it appears as if there has been no change in the order of cons traints. This would suggest that the change that is occurring is centered on verb type. As these forms grammaticalize and expand it is likely that there will be a shift in the conditioning effects. This Table (3 29 ) demonstrates that these shifts, particul arly when looking at haber de and tener que are likely to happen within these two factor groups. Table 3 29 Analysis of constrain hierarchies of haber de vs. tener que across the 19th and 20th centuries. 19 th Century Factor Group Weight 20 th Century Fac tor Group Weight Direct Objects Direct Objects middle/copular/ other .56 middle .62 other .59 inanimate animate .49 animate .57 inanimate .46 oblique .29 oblique .41 Range 27 Range 21 Verb Type Verb Type Copular psych/emo tion .64 copular .76 pysch/emotion .62 other .47 other .51 action .43 action .39 communication .40 communication .46 Range 24 37 In sum, the results from the written corpus suggest that specific factors are associated with the older, m ore frequent forms. This suggests that these factors (e.g. copular verbs, existential subjects, preceding clitics) are likely connected to movement towards epistemic contexts. Accordingly in the earlier centuries, the newer constructions occur often with c ertain factors (e.g. communication verbs, oblique objects), which indicates that these factors may help determine the age of each of these constructions. These results will be discussed in more detail in Chapter 4.

PAGE 139

139 3.5 DLNE and Oral Corpus In general, the results of the oral corpus suggest an ongoing change is occurring in Mexican Spanish deontic modality, as can be seen in Table 3 3 3 Speakers overwhelmingly select tener que as the preferred obligation marker in natural speech (92%). The findings also sho w an overall tendency for these obligation markers to not appear in future tense (.3%) suggesting future action may already be encoded in the meaning of the construction, which in effect could indicate that this marker may have advanced along the grammatic alization cline towards more epistemic contexts. However, considering that the future tense (both periphrastic and synthetic futures were considered) only occurred one time in all of the data (3.55), it is difficult to conclude that this is the result of a dvancement of these markers along the grammaticalization cline. This could be explained by the nature of obligation to imply future meaning through intention, which would mean that a future marker would be unnecessary because future is already implied (Tr augott, 1989; Bybee et al.,1994; Aaron, 2007). It could also be a reflection of the type of questions that the participants were asked. Most questions were narrative in nature and often involved talking about stories about past experiences. Whatever the re ason, it is difficult to make any conclusive statements about the absence of the future tense in the corpus. 3.5 Si estar con mi famil ia porque digo en un momento si pasara el examen voy a tener que salir (Roberto, 8/2010) Yes, be with my family beca use if I do pass my exams I am going to have to l eave. Within verb type the only variation that is seen occurs with verbs that fall into the other category where there are 6 examples of haber que (10.3%) and 2 of haber de (3.4%) as can be seen in Table 3 3 0 H aber de is not common but, of the 8 times it

PAGE 140

140 occurs, 3 are with psychological/emotional verbs, 1 with copular verbs and 1 with the verb tener This means 5 of the 8 occurrences (63%) appear in contexts associated with epistemicity. However, when thes e contexts are compared to the total number of occurrences of psychosocial/emotional verbs, copular verbs or with the verb tener, haber de only accounts for a minimal percentage (8%) of these cases. H aber que does not occur with psychological/emotional ver bs nor with copular, which again suggests that this form may not be grammaticalizing. Table 3 3 0 Data from the 21 st Century: Frequency of occurrence of each obligation marker within factor group verb type HD TQ HY Total N % N % N % N % Verb Type ac tion 1 .5 176 94.6 9 4.8 186 51.5 hacer 0 0 30 100 0 0 30 8.3 perception 0 0 2 66.7 1 33.3 3 .8 communication 1 4.3 18 78.3 4 17.4 23 6.4 psychological 3 33.3 6 66.7 0 0 9 2.5 haber/tener 1 3.8 24 92.3 1 3.8 26 7.2 copular 1 4.3 22 95.7 0 0 23 6.4 influence 0 0 6 100 0 0 6 1.7 other 1 1.8 48 87.3 6 10.9 55 15.2 What is equally worth mentioning is that haber que is relatively absent in both the DLNE and the oral corpus. It only accounts for 5.8% of the data in the oral corpus. A similar per centage is seen in most of the centuries in the DLNE as well: 19 th century (5.3%), 18 th century (0%), 17 th century (6.8%) and in the 16 th century (2.5%). Based on these data haber que seems stagnant in terms of frequency. This corroborates previous results which indicate that this form does not appear to be moving towards epistemicity. Nevertheless, it is important to note that in both the spoken and the DLNE corpora, the context typically includes personal narratives, which could essentially create a bias towards the other constructions since haber que is used with non referential subjects the majority of the time.

PAGE 141

141 Within the group of grammatical person and animacy we see that 100% of the cases with existential subjects occur with tener que (Table 3 3 1 ). This is a departure from the results found in the written corpus. This is an indication that obligation is grammaticalizing at a quicker pace in oral speech (which is expected as writing is more conservative) and thus tener que has already taken over the s ystem and is moving into epistemic contexts and slowly taking over that territory as well. A cursory glance at the DLNE corpus provides little support, considering that the only examples that contain existential subjects occur in the 16 th and 17 th centurie s. Also worth noting are the three cases of haber que referring to first person (3.56) identified as non third person through the use of the intransitive se or some reflexive object pronoun. We can see in example 3 56 that the object nos is expressing the middle voice. That is, the object is the same as the subject. In order for this to be the case, however, the construction hay que cannot be impersonal since nos refers to the first person plural. Taking that into consideration, this example clearly illustr ates that hay que is not exclusively used as an impersonal expression. Nevertheless, examples of this construction in first person are relatively rare compared to the total number of examples in first person (3.1%) but it does seem to parallel a similar pa ttern found in the written corpus. 3. 6 Haba un enfermo que no haba medicamento y hay que cooperarnos entre todos y darle a esa persona. (Marisol, 7/2010) If there was a sick person and no medi cine, we had to cooperate among ourselves and provide for this person. Given that it occurs only in the 19 th and 20 th centuries in the written and the oral data and accounts for 3 of the 21 cases (14%) of haber que in the oral corpus, this could point to the beginning of this form moving into new contexts with different

PAGE 142

142 grammatical persons thus opening the door towards expansion and grammaticalization. In terms of sociolinguistic variables there was very little difference between the choices of variants, which was expected. For example younger speakers used habe r de and haber que with the same frequency as the older speakers. Similar results were found when comparing male to female speakers. What this suggests is that, if a change has occurred and tener que has virtually taken over the system, this change already had taken place and had been adopted into this community by the time the younger generation began speaking. Regardless of the reason, this community seems to rely heavily on tener que as an obligation marker. Table 3 3 1 Data from the 21 st Century: Frequ ency of occurrence of each obligation marker according to factor group HD TQ HY Total N % N % N % N % Grammatical Person 1 st person 1 .8 123 96.9 3 2.4 127 35.2 2 nd person 1 2.1 47 97.9 0 0 48 13.3 3 rd person animate 5 3 142 86.1 18 10.9 165 45.7 3 rd person inanimate 1 12.5 7 87.5 0 0 8 2.2 existential subject 0 0 13 100 0 0 13 3.6 The fact that tener que accounts for the majority of the cases also echoes reports of similar results found in the English obligation system. In their study on To ronto have to has taken over the entire obligation system. Given that such results are being found in other languages, this offers support to the idea that results found in this oral corpus may in fact refle ct the status of the obligation system in modern Mexican Spanish. Taking into consideration that 62% of the examples in the oral corpus are in present tense (Table 3 3 2 ), we would assume that all of these constructions would be employed since Silva Corval n (1995), Heine (1995) and others suggest that present

PAGE 143

143 tense is associated with a deontic reading. This is not the case, however, since tener que is the marker of choice. This further supports the hypothesis that tener que has taken over the obligation sy stem. This expansion indicates that this form has grammaticalized and spread into new contexts in a relatively short amount of time possibly within 100 years. Another tendency found in the data is that 23% of the total examples are in the imperfect tense. As previously mentioned, one of the ways habitual actions are identified is through aspect. T ener que accounts for all but 2 of the cases of imperfect verb form. If we compare this to the 18 th and 19 th centuries from the DLNE corpus, where imperfect accou nts for 50% of the cases and tener que is only found in 23% (N=11) of those cases, we can see that there appears to be a shift. This could be another identifiable way to determine if an obligation marker is grammaticalizing. This will be discussed in more detail in Chapter 4. Table 3 3 2 Data from the 21 st Century: Frequency of occurrence of each obligation marker within factor group tense aspect mood (TAM) HD TQ HY Total N % N % N % N % TAM present 6 2.4 228 89.4 21 8.2 255 70.6 preterit 0 0 14 100 0 0 14 3.9 future 0 0 1 100 0 0 1 .3 imperfect 2 2.4 82 97.6 0 0 84 23.3 conditional 0 0 1 100 0 0 1 .3 subjunctive 0 0 3 100 0 0 3 .8 other 0 0 3 100 0 0 3 .8 Due to the small number of examples in both of these corpora, it was difficult to d raw any concrete conclusions about the tendencies found. It is neverthless important to examine these data, given that the tendencies found in both corpora seem to follow similar patterns found in the written corpora. This is further support that these con structions were grammaticalizing and, in fact, continue along this path. As a result of

PAGE 144

144 the minimal data available only the data extracted from the oral corpus was discussed in detail, although reference to the DLNE corpus was made throughout the section to corroborate (or not) possible tendencies. 3.6 Summary This chapter has covered the changing tendencies of each of these obligation markers. Through a careful examination of text and type frequency, we verified that the obligation system in Spanish has indeed undergone a shift since the 16 th century. The results suggest certain patterns of change, that may help corroborate the claims of universal paths of grammaticalization in modality. The discussion of these tendencies is examined in more detail in Cha pter 4. In addition to analyzing patterns of frequency, multivariate analyses were conducted for the 19 th and 20 th centuries. The results further implicate a change in the system, particularly within the factor group verb type. These data provided a snapsh ot of the patterns of use among three obligation markers in the 20 th century. The multivariate analyses for each variant further suggest that tener que is taking over obligation. We see a similar trend in the oral and DLNE data. The oral data are almost e xclusively comprised of examples of tener que Despite the patterns of use identified in the written data, this corpus offers a different picture of the obligation system, showing minimal use of haber que decreasing use of haber de and a takeover by tener que Although there may be alternate explanations for these results, which will be discussed in Chapter 4 and Chapter 5, the results from the DLNE corpus seem to suggest a similar pattern. Also, the minimal number of examples in both the DLNE and oral cor pus, and the subsequent restrictions on the conclusions that can be drawn from these two corpora, are also discussed in detail in Chapter 4.

PAGE 145

145 CHAPTER 4 DATA ANALYSIS Based on the overall frequency patterns of these four obligation markers, the system appe ars to change in a way similar to that of obligation systems in other languages. That is, diachronically these forms appear to undergo what Hopper and is essentially a n ew way to say the same thing. Bybee (2003) explains that this often happens when the older forms become so frequent that they lose their intensity and force. As a result, new, innovative forms enter the system and begin to compete with the older forms (Hop per and Traugott, 2003: 125). I n F igure 4.1 we see that, in the early centuries haber de is the primary obligation marker used. This continues through the 17 th and 18 th centuries although by this point tener que has entered the system and is slowly start ing to expand. The expansion of tener que correlates with the loss of territory of haber de both in overall frequency and within the various semantic contexts. For example, a s tener que is used more and more often with stative verbs haber de starts to be used less and less in the same context This suggests that these two constructions may be ultimately filling similar semantic roles albeit at different periods in time. The question that remains though, is how tener de and haber que play a part. At a qui ck glance we can see that the increase in frequency of use of tener que and haber que also correspond to the decrease in frequency of tener de S o the question is not whether tener de and haber que were part of this longitudinal renewal or not, but rather replace haber de This was answered by looking at the development of each of these

PAGE 146

146 m behind the rise and fall of each construction (Hopper & Traugott, 2003: 71 ). Figure 4 1. Frequency of variants across centuries In addition to understanding the story behind each of these constructions, the results reported in C hapter 3 also shed light on the path of grammaticalization of obligation markers in Spanish. Essent ially, each of these forms reflects a different stage of grammaticalization. H aber de is the oldest form dating back to the 12 th century and well established a s the marker of choice by the 13 th century (Yllera, 1980:99). H aber que came into the system a li ttle later and around the 14 th century bec ame limited paradigmatically occurring only with 3 rd person impersonal subjects (Cornillie, 2007:227). T ener de according to Yllera (1980: 110 116), was used as a deontic construction as early as the 13 th centur y. Tener que also appeared around the 13 th century as an obligation marker although it becomes more frequent around the 16 th century (Yllera, 1980: 117). Taking the age of each of these markers into consideration and the difference between them in terms o f when they were used most frequently, the

PAGE 147

147 stages of each of these constructions throughout the centuries reveal a relatively clear path of change. The following sections will r eview the results presented in C hapter 3 and situate these obligation markers i n their respective paths of grammaticalization. The discussion will begin with section 4.2, which will review the various phases in grammaticalization pertinent to the present study. Section 4.3 will expand on this path by re examining how subjectification plays a role in the development of these constructions. Then, in the sections that follow, the cumulative results of each marker will be discussed and the various factors that place these markers along points in the grammaticalization path will be identif ied. Section 4.4 will include a general discussion about what the tendencies of each o f these constructions represent when compared against the other variants. 4. 1 Identifying the Path of Change One of the primary goals of this study was to answer several central questions in regards to the aforementioned markers of obligation. The se questions were presented in C hapter 1 but are listed here again for review. What are the semantic differences among these four obligation markers? What are the factors condit ioning the use of each of these obligation markers. Have these forms continued to grammaticalize? If so, what is the path of grammaticalization of each marker? Have they also started to undergo subjectification? Does the path of change of these markers co rroborate claims made for universal paths of grammaticalization in modality? I have shown in this study that these questions can be answered by examining the path that each of these markers take s between the 16 th and 21 st centuries. In order to do this, however, it was paramount to operationalize the relevant stages in

PAGE 148

148 grammaticalization and subjectification. By doing this, the extent to which these forms have developed and advanced along this path could be measured. 4. 1 .1 Grammaticalization Recall that e ach of these constructions has already been identified as having gone through some of the processes identified in grammaticalization. Grammaticalization, the gradual process whereby a construction becomes less specific and moves towards more grammatical, a bstract meanings is identified by several characteristics. To review, we know that these constructions evolved from the transitive verbs expressing possession, haber and tener into semi auxiliaries expressing obligation. They have lost many of the propert ies that were associated with the transitive possession verb, which in grammaticalization is identified as decategorialization. We are also confident in assuming that these verbs have become bound to the preposition de or the conjunction que, which is ano ther stage in grammaticalization In the data, in fact, there were not any examples of these constructions occurring with a clitic or any other element between the verb and de/que after the 16 th century thus indicating a high level of fusion between these two elements These constructions became even more fused as time progressed After the 16 th examples of a word such a clitic, occurring between the constructions and the infinitive. It wa s determined that this is further support that these forms have already entered the grammaticalization process, even as obligation markers. As these verbs became more and more fused to de/que we expected to see autonomization take place, which often includes the divergence of these newer construc tions from the source form. We demonstrated that there was early evidence for this, as was discussed in C hapter 1. Both tener que/de and haber de/que are no longer

PAGE 149

149 associated with the notion of possession. In fact, the possession meaning has been all but l ost with the verb haber in the written and spoken corpora Again, this was a clear indication that these forms had already started to grammaticalize. Another important indicator of grammaticalization is the presence of layering or when two or more forms ap pear to compete for the same territory another possible sign that grammaticalization is taking place. We know that these forms were in direct competition by the 16 th century because they emerge d in what appear ed to be similar contexts, such as in the exam ples seen here in (4 1 4 4 ) 4. 1 no tengo que decir ms de remitirme a lo que ellos dijeron ( Corts, Hernn 1524: CORDE ) say anything more than refer ring to what they said 4. 2 De las dems casas de este signo no hay que dezir ms de lo que est dicho atrs. (Sahagn, Fray Bernardino de, 1576 1577 : CORDE) about houses of this 4. 3 quien tengo de dezir verdad y lo que siento ( Moya y Contreras, Pedro ,1575: CORDE) to whom I have to tell the t ell the truth and what I feel 4. 4 que ninguna falta haze en lo que ha de decir ( Sahagn, Fray Bernardino de 1576 1577: CORDE) no thing is lacking in what I have to say In order to further trace this path, some of the processes that supported the notion of movement along this path were operationalized. For example, retention of source semantic content was one of the processes in grammaticalization operationalized through verb type (Bybee et al., 1994, Torres Cacoullos, 2011). This was done to see what kind of verbs with which each of these constructions is associated. It was expected that the infinitive verbs that follow these obligation constructions may in fact be similar to

PAGE 150

150 those verbs associated with these constructi ons prior to their movement into the obligation system. In addition to retaining lexical content, these constructions also generalize as they grammaticalize. This generalization, or s emantic bleaching is easily measured and analyzed through relative freq uencies. As these constructions are used more frequently, they begin to occur in new contexts. Semantic bleaching can be seen in every factor group but it is particularly noticeable with the verb, both in type and lexicon The data show that generalization in meaning correlates with movement towards more subjective readings. As strong obligation markers expand into new territories, they can often infer an intention or future meaning (Cornillie, 2007; Bybee & Fleischman, 1995; Bybee et al., 1994, Fleischman, 1982) prior to moving into epistemic contexts. Taking this into consideration, it was determined that i ntention needed to be operationalized as another possible way to determine the age of these markers. Since this has been operationalized via temporal ad verbs we can expect that the beginning stages of this inference will likely occur with adverbs or adverbial clauses referencing the future (Blas As Hopper and Traugott (2003:82) mention however, the i ntention meaning must occur often in order for it to have an effect. As a result, we expected that an increase in occurrence with temporal adverbs will likely indicate movement towards epistemicity. Inference, another important process in grammaticalizatio n was also operationalized through several different factor groups : sentence type, polarity, verb type and grammatical person. In regards to sentence type, as was discussed earlier, if

PAGE 151

151 the sentence was a declarative sentence, it was less likely to infer e pistemicity since there are limitations when it comes to epistemicity and interrogative sentences Corvaln 1995; Coates, 1983). Thus the forms that occur less often in interrogative sentences are more l ikely to infer epistemicity. Polarity another factor dealing with inference, has been found to occur in epistemic contexts in limited cases (Olbertz, 1998; Heine, 1995; Silva Corvaln 1995; Palmer, 1990; Coates, 1983). This would imply that the construc tions that do not occur with negation are more likely to have epistemic meanings. In fact, Olbertz (1998:405) goes so far as to say that strong truth commitment can be e xpressed in positive terms and suggests that sentences such as 4. 5 are not possi ble. 4. 5 No tiene que ser fcil para ti. (Olbertz, 1998: 405) It does not have to be easy for you. Grammatical person and animacy, also a factor group associated with inference, is an important element in determining whether a context can infer epi stemicity. As Pietrandrea (2005), Bybee and Fleischman (1985) and Coates (1983) mention, epistemicity is often associated with third person subjects. Cornillie (2007), Heine (1995) and Coates (1983) also suggest that animacy plays a role in determining whe ther a context will have an epistemic reading as well. Finally, Wrnsby (2004) identifies existential subjects as being uniquely epistemic. We posited, then, that the constructions associated with third person, inanimate subjects are more likely to infer e pistemicity whereas constructions that are employed more often with first and second person subjects will be less likely to infer epistemicity (Krug, 2000).

PAGE 152

152 Mood is yet another factor that may infer epistemicity. Given that linguists such as Coates (1983) Palmer (1990) and Heine (1995) categorize an episte mic proposition as non factual ( where the probability of the proposition occurring is high ), we can connect n on factuality to mood based on its non factual status Therefore, constructions commonly occur ring in conditional or subjunctive mood are more apt to infer epistemicity given the scale of non factuality inherent to these types of morphological markers. The last factor group associated with inference is verb type. We know that obligation markers are more likely to express deonticity when they occur with communication verbs or action verbs (Cornillie, 2007). Conversely, the constructions are more likely to infer epistemicity if they occur with stative verbs. The assumption was, then, that as these obl igation markers become more and more grammaticalized they would occur more often with stative verbs, and less often with action verbs and communication verbs. Another process identified in research on grammaticalization is the movement of the grammaticali zing form from high focus clauses into low focus clauses. Based on previous research, low focus clauses are typically associated with more grammaticalized forms (e.g. Diaz Campos, 2011 ; Bybee et al., 1994 ; Klein Andreu, 1991 ) Based on this research, low and high focus clauses were deemed an important area in the grammar to help determine how much the obligation markers have grammaticalized. T his process was operationalized throug h object pronouns (via animacy), polarity and verb type. We know that low foc us clauses are associated with negation, inanimate subjects or objects and stative verbs. As a result, we postulated

PAGE 153

153 that, as these obligation markers increase in frequency, the occurrence with the aforementioned factors will be indicators of movement alon g the grammaticalization path. On the other hand, if some of these constructions are not used in contexts with the previously mentioned factors, this could be an indicator that the construction is not grammaticalizing. The grammaticalization of obligation markers also includes the shift of these constructions from an agent oriented reading to a more speaker oriented reading. This sub process of grammaticalization was also considered when examining the results of each marker. 4. 1 .2 Subjectification An integr al part of the grammaticalization path of obligation markers is movement towards more speaker oriented contexts, also known as subjectification. Accordingly determining the extent of subjectification that each of these forms has undergone, particularly in terms of weak or strong epistemicity is paramount. Recall that Traugott (1989) identifies three tendencies (see C hapter 1) she thinks are part of the subjectificiation process She proposes a path of subjectification specific to agent oriented modals ( e. g. obligation). That path is shown in Table 1 1 with each tendency (1 3) noted above where she believes it occurs. Langacker (1999a, b) defines subjectification in a similar to Traugott (1989) but focuses on synchrony and the attenuation of the modal in te rms of the amount of control the agent has over the proposition. Both the stages identified by Traugott (1989) as well as the synchronic attenuation found in Langacker (1999a,b) have been operationalized in a number of different studies, as mentioned in C hapter 1 Unfortunately the large majority of these studies were not relevant to the variants under examination in this study.

PAGE 154

154 Given that so little research has attempted to operationalize subjectification, several factors were chosen that were thought t o reflect this process. The factor groups associated with subjectification are i) tense and aspect, ii) temporal adverbs, iii) verb type and iv) grammatical person/animacy These factors primarily centered on the stages identified by Traugott (1989) althou gh the characteristics of weak epistemicity Traugott (1989) posits that many deontic constructions have already passed through two stages of subjectific ation when they undergo the shift from being main verb s to deontic verb s Considering that each of these markers has already made that shift, we concluded that they have already passed through the first two stages of subjectification and started the path towards the third stage of s ubjectification Traugott (1989: 39 40) labels t he first phase of the third tendency as weak epistemicity, which correlates with what Bybee et al. (1994) refer to as intention. Traugott (1989: 40) identifies weak epistemicity through markers that indicate a typical or habitual action or implicate prophetic or relative future. She describes relative future but nothing spe cific like, for example, later and claims that the relative future shows that a construction ha s not fully s ubjectified. I deduce d that any specificity given to the time period means that the form has advanced along the subjectification path. This wa s operationalize d through temporal adverbs. If there was reference to the future with a specific time adverbial m arker we assume d that this form ha d entered into strong epistemicity whereas if the future marker was non specific we assumed weak epistemicity We can also look at tense as a way to measure movement into weak or strong epistemicity since Bybee (1987:5 as

PAGE 155

155 cited in Traugott, 1989) suggest s that present tense may be a strong indication of movement into future tense. Habitual actions were measured through the incorporation of aspect and temporal adverbs We presumed that the use of a temporal adverb connotin g a habitual action would likely place the construction in the weak epistemicity phase In regards to aspect, we know that Traugott (1989) postulates that a deontic obligation marker referring to a typical or habitual action is associated with weak epistem icity which would imply that constructions commonly found in imperfect or progressive are less likely to infer strong epistemicity The other factors that are associated with subjectification verb type and grammatical person follow the same tendencies t hat were discussed in 4.1. That is, as these constructions undergo subjectification they are more likely to occur with stative verbs. In addition, we also postulated that contexts with third person inanimate subjects, as well as existential subjects, are m ore likely to have epistemic and thus subjective meanings than those in first and second person. 4. 2 Data Analysis In order to determine the path of change for the obligation system in Spanish, a thorough examination of the tendencies is paramount. Only t hen can we draw more general conclusions and identify the possible evidence that implicates where on the grammaticalization path each of these constructions is. In the following sub sections, we analyze the various tendencies presented in C hapter 3 in each factor group. These tendencies are connected with the aforementioned evidence that has been shown to account for stages in the grammaticalization process. Additionally, tendencies that are found to be unique to a particular marker or not identified in pr evious research are also

PAGE 156

156 discussed. These patterns may show that Spanish has a distinct path of grammaticalization for obligation markers. 4. 2 .1 Tener de When there are many choices in a language for a specific function, like obligation, sometimes the ch oice of forms is reduced such as in the case of tener de This often happens because the other constructions start to generalize and move into new relevant when we talk about the obl igation system in Spanish and of the disappearance of tener de although sometimes as Hopper and Trau gott (2003:115 old forms may continue to coexist therefore specialization does not necessarily entail the elimination of alternatives, bu t may be manifested simply as textual preferences, conditioned by semantic types, sociolinguistic contexts, discourse genres, and other fact We see that in the case of haber que haber de and tener que they continue to coexist In the case of tener de however, it was eliminated. As F igure 4. 1 above indicates, tener de essentially stops being used as a marker of obligation around the end of the 18 th century. Although there are an isolated number of cases of this marker occurring in the later centuries, these are few and far between. The big question then, is why did this happen? One of the first signs of trouble for this construction is found within the factor group verb type. Although in general, tener de was used occasionally with all verb types, in the 16 th century it was primarily associated with the verb hacer and with communication verbs. R emember that tener de was frequently found with the verbs fablar librar th century (Yllera, 1980: 113) so communication verbs and action verbs may have been the entry point for this construction. Between the 16 th and

PAGE 157

157 the 18 th centuries haber de expand s and essentially takes over contexts with communication verb s accounting for 76% of these contexts, essentially w iping out any h old tener de may have had in this area. In addition, because tener que is presumed to be associated with tener de (according to Cornillie, 2007), communication verbs and action verbs were also the likely entry point for tener que and, ultimately, more comp etition for tener de H aber que also started to appear with communication verbs early on so between the three constructions tener de quickly los t the association with communication verbs. During the 16 th century tener de was often used in contexts with ps ychological and emotion verbs, an area in the grammar where epistemicity could possibly be inferred. This was a potential opportunity for tener de to expand into more subjective contexts. However, like with communication verbs, both tener que and haber que used these same contexts as vehicles for their expansions. As a result, tener de l oses all these initial associations by the end of the 18 th century. In sum, the factor group verb type appears to be an area in the grammar where tener de lost a lot of grou nd quickly. We see that in terms of grammatical person, tener de was strongly associated with deontic contexts in the 16 th century accounting for 50% of all contexts with first person subjects. In addition, it was also frequently employed with third per son animate subjects, where it accounted for 39% of all tener de occurrences Both these results suggest that tener de was well established as a deontic modal. But by the 17 th and 18 th centuries we see that haber de increased by more than 35% in contexts w ith first person subjec ts. In fact, both tener que and tener de lost territory to haber de during this period although tener que eventually gain ed it back. E ven though tener de

PAGE 158

158 d id not gain this territory back, it d id manage to retain some of its associat ion with these contexts as the second most frequent construction used with first person subjects during the 17 th and 18 th centuries. This indicates that this area in the grammar was likely one of the last areas to lose its association with tener de before this construction fell out of use. The other important trend regarding tener de during these centuries is the expansion of tener que into contexts with third person subjects This development essentially separated tener de from the only other grammatical p erson with which it was associated. And to exacerbate the departure of tener de haber que was limited to contexts with third person animate subjects during these early centuries and also experienced a rise in frequency, thus helping take over the little t erritory left for tener de We observed a similar pattern in the factor group sentence type where haber de increased in frequency with both interrogative and declarative sentences while the new forms haber que and tener que also experience d increases altho ugh they were minimal T ener de Despite the fact that tener de did not expand during the 17 th and 18 th centuries and that the majority of the contexts in which it was found were primarily associate d with deonticity (e.g first person subjects communication and action ( hacer ) verbs ) there were still signs that it may expand into territories associated with epistemicity. Early on, it did occur often with psychological and emotion verbs and occasional ly with copular verbs. We also saw that tener de was used most with verbs in present tense between the 16 th (83%) and the 18 th centuries (75 %) and typically associated more with positive

PAGE 159

159 polarity than negative polarity. These are all contexts that may have inferred epistemicity as we see in 4 6 4. 6 que la tierra adonde se da el cacao tiene de ser muy buena; y porque este cacao es comida (Toribio de Benavente, 1536 1541: CORDE) food. This evidence led us to believe that this construction would continue to grammaticalize. The one thing this marker did not do as time progressed, however, is increase in frequency and spread into new contexts. I n fact, i t was quite the opposite. In the 16 th century it occurred with specific verbs regularly like comer (n=9), ser (n=5) or the verb hacer (n=11) among others. By the 17 th and 18 th century it rarely occurred with the same verb more than once, with the exception of the verbs acabar (n=3), dar (n= 5) and pasar (n= 3). In addition, it dropped dramatically in frequency. In the 16 th century it accounted for 7% of all data, but by the 17 th and 18 th centuries it only accounted for 3% of all data. So despite the patterns this marker shows towards being as sociated with weak epistemicity (e.g. contexts in present tense, imperfect and 11% of habitual temporal markers), its decrease in frequency and failure to spread into new contexts did not allow this to happen. One possible reason to explain why tener de a ppears in so many contexts early on could be based on the similarities this construction shared with haber de Yllera (1980: 111) suggests that tener de came into being as a result of the substitution of tener for aver in contexts historically reser ved for aver. It is possible that this substitution may have spilled over into the periphrastic uses of haber de as well. If this were the case, then it would explain why tener de was employed in so many different contexts from the beginning.

PAGE 160

160 Yllera (1980 :97) also mentions that, in old Spanish, constructions with verb+ de+ infinitive often indicated purpose or finality. She credits this formulaic construction for the quick rise in frequency of haber de It could also be another possible reason why tener de w as found in so many contexts, despite being a new construction If we consider that haber de was already associated with finality and purpose this may have easily associated the construction with intention thus propelling this form into contexts implying f uture actions. As a way to differentiate between intention and obligation perhaps another construction, not containing de, was needed. The need for an obligation marker removed from the finality or purpose feature of the de construction could also explain the appearance of tener que and the rise in frequency of haber que This idea supports th e theory proposed by Cornillie (2007) who suggests that tener que came from tener de and not from the relative clause tener +object+ que+ verb. Another important aspect to consider when we look at the disappearance of this construction is to look beyond the variable context. Given that this marker still retained possession readings in certain contexts, as illustrated in example 4. 7 it was important to determine whether t he decrease in frequency of tener de as an obligation marker was paralleled by a corresponding increase in frequency by the possession construction tener de 4. 7 Deseos tengo de verlos enfadados (Fernndez de Lizardi,1818: CORDE) I have the desire to s ee you (pl) annoyed. As we can see in Figure 4.2, there were 16 examples of this construction in the data being used to express possession (15% of total data) in the 16 th century. In the 17 th and 18 th centuries there were 52 examples (60% of total data) and in the 19 th century there were 10 instances (83%). The rise in frequency of cases where this construction

PAGE 161

161 expressed possession correlates to the decline in frequency of the construction being used as an obligation marker. Taking this correlation into c onsideration, it is likely that the contexts where possession was expressed also played a principle role in the decline and eventual disappearance of tener de Figure 4 2 Comparison of tener de as an possession marker and as an obligation marker across centuries 4. 2 .2 Tener que One of the first patterns we observe with the construction tener que is that it start ed out being used with a variety of verbs, but was especially used with communication verbs, which might suggest that it may be closely related t o tener de Recall that t his is based on research conducted by Y llera ( 1980: 113 ) who finds that tener de is most often found with the verb fablar in the earlier centuries Again, this observation offers further support that tener que may be derived from t ener de since it was found in similar contexts at this point in time. If it is the case that tener que entered the language on the tail of tener de it may explain why we also see that this construction start s to appear in contexts that could

PAGE 162

162 possibly inf er epistemicity. I n the 16 th century for example, it ha d already started to be used with psychological and emotion verbs T his trend continued into the 17 th and 18 th centuries where it accounted for 14% of all contexts with psychological and emotion verbs Likewise, within the factor group sentence type tener que is not associated with interrogative sentences until the 20 th century, meaning that it often occurred in contexts that linguists (e.g. Coates, 1983; Corvaln, 1995) typically associated with epis temicity in the research It was also noted that as early as the 17 th and 18 th centuries, tener que occur red more often with inanimate subjects then with animate subjects, again suggesting that this form was occurring in contexts with possible epistemic r eadings from early on. R ecall that R uiz de Mendoza Ibez and Prez Hernndez (200 4 : 106) suggest that tener que is more associated with internal obligation. They also suggest that this association disappears in the past tense but claim it is preserved in the future tense (Ruiz de Mendoza Ibez and Prez Hernndez 2003: 110). Given that internal obligation is associated with epistemicity, we deduce d that when tener que occurs in past tense it is less likely to be epistemic, whereas when it occurs in the fu ture or present tense it is more likely to be epistemic. This is supported by Solano Araya (1982: 17) who suggests that tense If we look at the data, we see that in the 16 th century tener que occurred in the preterit in about 13% of the cases. As time progressed this number steadily increase d until the 20 th century at which point tener que accounted fo r 86% of all cases in preterit We see a similar increase in frequency of tener que appearin g in future tense as well. It

PAGE 163

163 appears that, as time progressed, tener que bec ame increasingly used in both deontic and epistemic contexts. We know that Traugott (1989) also finds that tense plays a role in subjectification although she separates relative f uture from deictic future. When we look at the factor group tense a spect mood we see that tener que account ed for 70% of all contexts in imperfect in the 20 th century. Considering that the imperfect often indicate s habitual action s this was identified as corresponding with weak epistemicity ( Traugott 1989). This tendency was corroborated by the percentage of habitual temporal markers that occur red with tener que during the same century (68%). Both the frequency of occurrence of tener que in the imperfect along with the high number of temporal markers indicating habitual actions are clear indicators that this form was associated with weak epistemic during the 20 th century. We also see that this construction account ed for 57% of the contexts in future ten se during the 20 th century. Traugott (1989) identifies this as a nother possible characteristic of subjectification. T ener que accounted for 68% of the temporal future markers during the 20 th century of which about half (n=24) were non specific and the oth er half (n=22) wer e specific. Considering that Tr a u gott (1989) regards deictic future as an indication of strong epistemicity, we can conclude that tener que had already moved into contexts associated with strong epistemic readings in the 20 th century Wh en tener que began to expand into all of the territories in the 19 th century haber de managed to maintain its association with preceding clitics accounting for 99% of all contexts with a preceding clitic during the 19 th century This seems to be o ne of th e areas that was a stronghold for haber de This is corroborated by the multivariate

PAGE 164

164 analyses where, in both centuries preceding clitic still strongly favors the use of haber de However, b y the 20 th century tener que did start to appear in these conte xts at a rapid rate. In the span of just one century tener que accounted for 54% of the cases with preceding clitics Based on the claims made by Myhill ( 1988 ) and Gutirrez (20 10 ) who suggest that clitic climbing occurs more often with auxiliaries than with other verbs coupled with the observation that tener que comprised 54% of the cases with a preceding clitic during the 20 th century we can deduce that this marker progressed further along the grammaticalization path during this century Many of these te ndencies are substantiated by the multivariate analyses. For example, in both the 19 th and 20 th centuries, when compared to haber de we see that tener que continues to be linked to communication and action verbs by the disfavoring effect these two factors have with haber de We also observe that, although tener que was used with psychological and emotion verbs early on, this factor still favors the use of haber de supporting the claim that tener que is still in the earli er stages of subjectification. We d o see that, in the 20 th century, interrogative sentences slightly favor the use of haber de This could be one of the areas that allowed tener que to be associated with epistemicity in this later century. We also see in the 20 th century that there is a sli ght tendency for inanimate objects to disfavor the use of tener que Again, like with interrogative sentences, this is an opportunity for tener que to occur in contexts that infer epistemicity. Overall, there is an observable renewal taking place when we look at the frequency patterns of tener que in comparison to haber de T ener que appears to be used in the same contexts as haber de just at a later time period. In fact, according to

PAGE 165

165 the spoken data from the 21 st century tener que has literally taken ove r the obligation system This raises the question as to whether haber de will continue on the grammaticalization path or, like tener de fall out of use. As Fernndez de Castro (1999) mentions, haber de is already primarily limited to written or formal reg isters which points to the possible disappearance of this construction. The de s cent of haber de is discussed in more detail in the next section 4. 2 .3 Haber de One of the first tendencies we notice about haber de is that its overall frequency decreases ov er time. As has been mentioned on several occasions haber de had control over the entire obligation system during the 16 th century This is seen in every factor group. However, as time progresses we see a general decrease in frequency across the board. W ithin the verb type factor group haber de continued to regularly occur with the psychological and emotion verbs and the verbs haber and tener throughout the centuries. Considering that these verbs are all perfect areas in which epistemic readings can be i nferred, we assume d that this construction was moving towards epistemicity. This was corroborated by the fact that haber de was also most likely to occur with third person inanimate subjects and existential subjects. Interrogative sentences, however, d id n ot support this assumption. During the 20 th century haber de was often used in interrogative sentences (40%) whereas it wa s only used with declarative sentences 23% of the time. We also observe that when negation was expressed haber de was used 31% and wa s used only 24% of the time when negation was not expressed. Considering that negation is associated with low focus clauses, but it is not

PAGE 166

166 associated with epistemicity, we might say that this result indicated that haber de was in an intermediate stage of gr ammaticalization and moving towards epistemicity I f we look at the results of the factor group polarity, we see that haber de was associated with low focus clauses Klein Andreu (1991) mentions that stative verbs are also indicative of low focus clauses, an area in which haber de has had control since the 16 th century. This was also supporte d by the fact that haber de represented more of the total percentage of contexts where negation was present (31%) tha n those where it wa s not (24%). Again, although ne gation corresponds to low focus clauses, it is not associa ted strongly with epistemicity. We see that within the factor group indirect objects haber de occurred as much with an indirect object as it did without ( 24%). In turn where direct objects were rep resented haber de occurred with animate objects 27% of the time and inanimate objects only 18% of the time. Although this suggested that this form was still primarily relegated to high focus clauses we know that epistemicity is as sociated with stative ver bs and, logically, stative objects, an area where haber de account ed for 46% of the data. Overall, we see that haber de did have characteristics (negation, stative verbs, no preference with IO) of occurring in low focus clauses. Although it has been well documented in the literature that haber de expresses future (e.g. Cornillie, 2007; Yllera, 1980), we see that, based on its occurrence with temporal adverbs referencing the future, this tendency has been declining. In the 16 th century haber de accounted fo r 98% of all contexts with future adverbials Du ring the 20 th century however, it only account ed for 26%. Although, of the 19 future adverbial markers with which it occurred 47% (n=9) of those adverbs make reference to specific

PAGE 167

167 time periods This indicat e s that this construction was still associated with and/or moving towards strong epistemicity during the 20 th century This was corroborated by tendencies found in the tense aspect mood factor group. H aber de occur red in 38% of a ll cases when conditional mood wa s employed and 35% when future wa s used. This only strengthen ed the association with epistemicity. We also observe that it accounted for 33% of the contexts in present tense, which has also been found to be associated with future reference ( Bybee,19 87; Traugott, 1989 ). I n addition to relative future, habitual actions are associated with weak epistemicity as well We see that in the 20 th century haber de account ed for 14% of those cases, which was more than a 50% decrease from the previous century, th us indicating movement from weak to strong epistemicity. The multivariate analyses echo what the frequency data tell us. For example, in the 19 th century, when compared to tener que we observe that direct objects were a significant factor group and ha d th e highest magnitude of effect followed closely by verb type. Both of the factor groups point ed to haber de moving into epistemic contexts. haber de Th is is expected since this factor included copular objects and these types of objects are typically associated with epistemicity. We see the same pattern with verb type where copular verbs and psychological/emotion verbs favor ed the use of haber de 1 1 And we al so have the same confusing result with negation, where it favored the use of HD which associated it with low focus clauses but disassociated it with epistemic contexts according to (authors).

PAGE 168

168 In the 20 th century t he same analysis show ed even more robust results and supporting evidence that haber de was the more likely construction to be used in epistemic contexts. Verb type was the factor group with the greatest effect and, like in the 19 th cen tury copulas and psy chological/ emotion verbs still favor ed haber de We also see that grammatical person was significant in this century and, not surprisingly, we see that the third person inanimate subjects favor ed the use of haber de again associating this construction wit h epistemicity. Contexts with preceding clitics strongly favor ed haber de which pointed to haber de being a more grammaticalized construction than the other constructions In addition, contexts with copular objects favor ed haber de as d id contexts with ne gation present. An interesting result that we did not see in the 19 th century analysis was that interrogative sentences favor ed the use of haber de This may be an indication that, even though this form was moving into epis temic contexts, it still retained its deontic reading as well. When we compare haber de with haber que the results, once again, indicate that copular objects and copular/psychological/stative verbs favor the use of haber de 4. 2 .4 Haber que One of the first indicators of language change, as suggested above, is semantic bleaching. We see semantic bleaching occurring with haber que particularly in the 19 th and 20 th centuries, when it started to occur with a wider variety of verb types. Taking a more detailed look at the data, we see that th is constru ction occurred mostly with the verb decir in the 16 th century ( n= 9/21), which illustrates the limitations of this marker at that point in time. This also illustrates that this form retained some of its lexical roots given that Yllera (1980:109) f ind s that haber que, prior to being used an obligation marker, often occurred with the verbs decir and fablar (among others). But, even as

PAGE 169

169 early as the 17 th century, haber que had start ed to spread into new contexts. This was illustrated by the number of n ew verbs with which t his construction was appearing during this century This is similar in grammatical person where, in the 19 th century we start to see contexts where there are first and second person subjects. One of the primary tendencies we see wit h this construction in regards to grammatical person is its expansion into new territories. In fact, this expansion is still rare enough in the data that calling it a tendency may be too strong. Nevertheless, there are examples in all of the data sets of h aber que being used in a context that implies a subject ot her than an impersonal one. Being confined to contexts with impersonal subjects along with the inability for haber que to occur with preceding clitics le d us to believe that this construction may be doomed to deonticity. This was further supported by studies such as Cornillie et al. (2009) who highlight the inability of haber que to occur in any context but deontic in their data That being said, we did find examples of this marker in the corpora i n contexts with first or second person subjects (ex. 3. 54 repeated here in 4. 8 ) suggesting that this limitation could possibly change. Although haber que did not occur in contexts inferring epistemicity in the data, (ultimately supporting c laims made by Cornillie et al., 2009) a cursory glanc e at the corpus CREA produced a context where epistemicity can be inferred (4. 9 ) with haber que 4. 8 Los podremos ayudar para que puedan hacer alguna carrera. Total, no hay que precipitarnos... (Azuela, 1985: CREA) We will be able to help them so that they can run some race. Anyway, 4. 9 pero la verdad es que hay que ser un hombre rarsimo para echarle la culpa de todo a tu abuelito. (Bryce Echenique, Alfredo; 1986 : CREA ) but the truth is t hat you have to be a strange person to blame everything on your grandfather.

PAGE 170

170 Taking this into consideration along with the results demonstrating that haber que has been used in contexts with subjects other than 3 rd person impersonal, this may be an area of expansion for the construction. H aber que has been used rarely in interrogative sentences throughout the centuries. Knowing that epistemicity is rarely expressed in interrogative sentences, this tendency points to haber que occurring in contexts that could be associated with epistemicity The tendency for haber que to not be employed in interrogative sentences in conjunction with its proclivity to spread into new contexts as time progresses hints at the possible grammaticalization of this form and mo vement towards epistemic contexts. This was not corroborated, however, by the results found with the factor group negation. H aber que account ed for a higher percentage of occurrences in contexts where negation occurs than those where it does not in every c entury. Despite the fact that haber que occurred in more contexts with positive polarity when considering absolute frequency, the overall p ercentages illustrated that this construction wa s used more often with negation. In regards to tense aspect mood th e results indicate d that this construction occurred primarily in present and imperfect tense during most centuries. There were also cases of it occurring in conditional and present subjunctive early on, and these did increase during the 20 th century. We al so observe that in the 19 th century haber que started to occur in the past tense, which is not associated with epistemicity. However, s ince imperfect is associated with weak epistemicity (Traugott, 1989) and present tense is often linked to future tense (Traugott, 1989) we concluded that haber que might infer weak epistemicity in these contexts This is supported by the expansion of haber que

PAGE 171

171 into new territories during the last couple centuries, particularly into those associated with epistemicity. This is further corroborated by examining the occurrence of this construction with temporal adverbs Between the 16 th and the 19 th centuries haber que did not appear with habitual or future temporal markers. In the 20 th century however, it appeared with both. The habitual markers suggest connections to weak epistemicity while the future markers suggest an association with strong epistemicity provided they refer to a specific point in time Upon further examination of these future markers we see that of the ni ne markers that refer to future, seven (78%) refer to a non specific time in the future, as can be seen in 4. 10 which occurs with the non specific marker luego This implies that this marker is more likely associated with weak epistemicity than st rong epistemicity. 4. 10 y luego haba que tumbar la puerta porque no iba a caber (Elizondo Elizondo, Ricardo, 1987, CREA) and later one had to knock down the door because it wa s not going to fit. Another interesting result that was seen in the analy sis of haber que is the relative frequency patterns in the different corpora. In the written corpus the rate of haber que increased almost 10 times the rate of occurrence between the 16 th and the 20 th centuries. It only represented approximately 2% of all obligation markers during the 16 th through the 17 th centuries but had jumped to 16% by the 20 th century. This was different in the DLNE corpus where haber que accounted for about 5% of the data through every century and only 6% of the data in the spoken co rpus. This difference seems to highlight a possible preference of haber que in the written corpora. If this is indeed the case, the possibility of this marker grammaticalizing will likely diminish.

PAGE 172

172 In the multivariate analysis with haber de we see that an imate objects favor ed haber de as opposed to haber que during the 20 th century. This again supports a possible semantic shift for haber que towards more low focus clauses. This is further supported in the multivariate analysis with tener que where inanimat e objects and negative polarity disfavor ed the use of tener que Both of these, indicators of low focus, suggest again that haber que may be moving into contexts where weak epistemicity can be inferred. Nevertheless, we also know that Coates (1983), Westne y (1983) and Silva Corvaln (1995) mention that there are limitations to the type of contexts in which epistemicity can be expressed when negation is present. We can assume then that, although haber que is moving in the direction of strong epistemicity, it is still in the nascent stages. In the oral corpus we see that haber que is only employed occasionally (6%) and typically occurs with communication verbs and perception verbs. Although we can not directly compare the oral data to the written data, we can still assume that the oral data is likely years ahead of the written data If this is indeed that case, haber que may never move into epistemic contexts as it appears to be stagnant in the spoken data. That being said, some of the same patterns that were seen in the written corpus appear in the spoken corpus as well. For example, we see examples of haber que with subjects other than third person impersonal and all the exam ples of haber que are in present tense 4. 3 General Discussion Taking all the aforem entioned conclusions into consideration there are some observable overall trends that we see in the system that may indicate general trends of obligation markers in Spanish. Additionally these trends offer insight and support to

PAGE 173

173 postulations put forth abo ut universal paths of grammaticalization and the movement of obligation markers towards more speaker oriented subjective contexts. One of the notable trends we see with tener de tener que and haber que is that all these constructions enter the obligation system being associated with communication verbs. Historically, two of these constructions tended to occur with the communication verb fablar before they came to mean obligation (Yllera, 1980) Although we do not know if this is also the way haber de ente red the system we can conclude that, in modern Spanish, obligation markers enter the system via communication verbs. This may be the beginning of the path. We also see a similar pattern with action verbs in the earlier centuries, a lthough by the 17 th and 18 th centuries these markers begin to appear in other contexts on a regular basis This may be an indication that, as these constructions move towards epistemicity, they associate less with communication and action verbs. We see some change happening in thi s area in the comparative analysis between tener que and haber de There is a change in the constraint hierarchy between communication and action verbs with verb type. Although tener que has clearly not shifted away from its lexical origins and the tendenc y to appear with these types of verbs, the comparative analyses suggest an impending change. Regardless, we can still postulate that the path from possession to obligation is partially facilitated through action and communication verbs. We also see a simi lar pattern with grammatical person where the entry point seems to occur in contexts with both first and second person animate subjects, although given the limitations in which haber que can occur conclusion about whether or n o t these two areas are truly access points We could also

PAGE 174

174 observe that these constructions were typically found in imperfect or present tense. Although this is not represented in the overall percentage during the 16 th century, we can see this trend in the absolute frequency of tener que tener de and haber que What is also evident in the data is that these constructions likely entered the language in declarative sentences without indirect objects. The rest of the factor groups do not show general trends d uring this period thus we can conclude that while each construction was used in distinct ways they also entered into the obligation system via a similar path One of the early signs of the grammaticalization of these constructions aside from generalizi ng and spreading into new contexts, is the tendency to occur with stative verbs, particularly those expressing a psychological state or emo tion. This is evidenced by tener de tener que and haber que early on and as time progresses, haber que and tener de cease to appear in these contexts offering further support for the ongoing grammaticalization of tener que and the apparent halt in grammaticalization of haber que It may be that psychological and emotion verbs are one of the first contexts in which th ese markers have the opportunity to infer epistemicity. Another observable genera l trend found in the data is the general dearth of examples of the markers with negation as these constructions develo p H aber de domi nates these contexts until the 19 th cent ury. This suggests that, despite its limitations in epistemic contexts negation may in fact be an indication of grammaticalization. If this is indeed the case, it would explain the tendency of haber que to not a ppear in contexts with negation and back up t he claim that the development of

PAGE 175

175 this construction has slowed since the 16 th century. What this may point to is another way to determine the age of these forms. It may be that these markers only become associated with negation when they begin to move int o low focus clauses. As they continue to develop and start to become associated with epistemicity, they once again move away from negation. We see this with haber de in the 20 th century where it drops dramatically in frequency in contexts with negation. In addition, in the multivariate analyses polarity is a significant factor group in the 19 th and 20 th centuries when comparing haber de and tener que thus suggesting that this is an area that conditions the choice of marker The results in the multivariate a nalysis between haber de and haber que in the 20 th century suggest a similar conclusion Based on the assumption that haber que did not grammaticalize much between the 16 th and 19 th centuries, we wou focus clauses where it would be associated with negation. Conversely haber de seems to have grammaticalized a great deal since the 16 th century given the aforementioned indications of its movement into epistemic contexts Thus w e would not expect haber de to occur often with n egation given the limitations of negation and epistemicity mentioned by Coates (1983) Westney ( 1995) and Silva Corvaln ( 1995 ) In sum, both of these markers should show little preference when it comes to polarity since they both tend to associate with po sitive polarity. The multivariate analysis support s these assumptions illustrating that when negation is present there is no preference between these two markers.

PAGE 176

176 Th e results indicate that these constructions are used in both negative and positive polarit y contexts more or less equally and therefore there is no preference. We could postulate that this is an indication that haber que is slowly moving into low focus clauses so we see that, occasionally, it is used with negation. Similarly, we can claim that haber de is slowly moving into epistemic clauses and thus is associated less and less with negation. The claim that negation is an intermediate stage of grammaticalization is further supported w hen we look at the multivariate analysis between tener que an d haber que We see that negation strongly favors the use of haber que Assuming that haber que is slowly moving into low focus clauses, this result substantiates that movement. But additionally, and based on earlier discussed results, we know that tener q ue is showing indications of moving into strong epistemicity (e.g. stative verbs, existential subjects). We would assume then that as tener que moves into strong epistemicity, it will be used less and less with negation, which is reflected in this multiva riate analysis. In sum, we can postulate that these constructions start out not being associated with negation. They develop, move into low focus clauses and begin to occur with negation. As they continue to grammaticalize and move into epistemic contexts they stop occurring with negation. Thus negation is an intermediary stage of the grammaticalization of these obligation markers. Another general trend we see in the data is the tendency for these newer constructions to retain and maintain some of their pr e obligation lexical features. Recall that Yllera (1980) found that tener de and haber que were commonly associated with the verb fablar. We saw that each of these constructions along with tener que entered

PAGE 177

177 the obligation system via these types of verbs. We also observe that in the 20 th century there were still tendencies for haber que and tener que to occur with these types of verbs. This may be an area in the grammar that these constructions will maintain until a new obligation marker enters the syste m. As these markers grammaticalize and move into low focus clauses and epistemic contexts once again we see evidence that low focus clauses may be associated with the intermediate stages of grammaticalization (like with negation). In both the multivariat e analyses and the comparative analysis, we see that directs objects are considered a significant factor group. Recall that low focus clauses are often identified by the occurrence with negation and inanimate objects. When we look at direct objects in more detail we see some obvious tendencies. Of course, one of the first tendencies we observe is that in the earlier centuries the newer constructions are used most often focus clauses occ urs in the 17 th and 18 th centuries with the construction tener que as it starts to increasingly be used with inanimate objects. Haber que follows suit in the 20 th century where we see a 15% increase in contexts with inanimate subjects. I n the 20 th century, as tener que expanded into new territory we found contexts with inanimate objects disfavor ed haber de suggesting that at this point, tener que was associated with these contexts. This indicates that the occurrence with inanimate objects is an indicatio n of movement into low focus clauses but that low focus clauses are more of an intermediary stage that occurs as these markers move into epistemicity. We see a similar pattern with oblique complements where, as tener que generalizes and spreads into new c ontexts, it begins to occur more often with obliques.

PAGE 178

178 In fact, that multivariate analysis from the 19 th century indicated that oblique complements strongly disfavor ed the use of haber de when compared with tener que A similar result is found in the analys is between haber de and haber que in the 20 th century, where both obliques and inanimate objects disfavor the use of haber de supporting the claim that as these markers develop, these two contexts are one of the first they acquire Although the reason fo r occurring with inanimate objects can be explained by the movement from high focus clauses into low focus clauses, the reason why oblique objects are also an area associated with this path is unclear. Finally, based on the tendency for haber de (and late r tener que ) to occur with copular objects, we can also conclude that contexts with these types of objects are one of the last areas these markers acquire as they move towards epistemicity. This may be why this factor group has a change in effect in the co mparative analyses. T ener que may be edging towards more contexts with copular objects and away from contexts with oblique objects. We can thus postulate a path in terms of occurrence with objects, which appears to be oblique>inanimate>copular. Other areas in the grammar that seem to occur more towards the end of the obligation path are found in various factor groups. For example, we see that haber de continues to occur in contexts with copular verbs into the 20 th century and, although the new er markers app ear with other statives (e.g. psychological/ emotion verbs ), copulas appear to be among the last contexts in which they occur before they move into epistemic contexts. This would support claims made by Pietrandrea (2005); Cornillie, 2007; and Coates (1983)

PAGE 179

179 The placement of a clitic also seems to be an area that could indicate the age of these construction s H aber de occurs in 99% of the contexts with a preceding clitic as late as the 19 th th century that tener que starts to ap pear in these contexts and, even so, the contexts with a preceding clitic still strongly favor the use of haber de This is a clear indication that the occurrence with a preceding clitic is an advanced stage of the grammaticalization process of these oblig ation markers. This corroborates research conducted on clitic climbing that suggests that clitic climbing occurs more often with auxiliary or grammaticalized constructions ( Gutirrez, 2008 ; Myhill, 1988). Finally, another important general trend we see in the data is found in the multivariate analysis between haber de and tener que The results suggest that this system is not only undergoing change, but that it is in the process of longitudinal renewal The fact that in the 20 th century the multivariate analysis between haber de and tener que produces results in which every factor group except indirect objects is significant points to a shift in the system across the board. So despite the path that we identified above, it is also important to recognize t hat tener que has gone through the grammaticalization process at a faster pace than haber que and has essentially caught up to where haber de was in the 16 th century by the 20 th century. As tener que continues to move into the territory of haber de we woul d expect a similar trend to occur, which is exactly what we see in the spoken corpus. As we mentioned in sect ion 3.5 the results from the spo ken corpus indicate that tener que has essentially taken over the obligation system. Taking into consideration th at there have not been studies that look at the diachronic development of obligation

PAGE 180

180 markers in Spanish, it is difficult to determine the lag time, if any between the written and spoken corpora. Research that has compared written and spoken corpora has ha d varying results (e.g. Biber, 1995). Based on the relative frequencies of these constructions, however, it appears that the lag seems to vary from century to century. If we compare the relative frequencies of the DLNE corpus to the written corpus, the 16 th century written data appear to be about 200 years behind the DLNE corpus. This gap diminishes, however, in the 19 th century where the normalized frequencies are similar across both corpora, with the exception of haber de The 20 th century we see a stead y increase of tener que in the written corpora but the DLNE corpus ends during the 19 th century so there is no data with which to compare the written corpus. The oral corpus, as we mentioned, shows a sharp increase in the use of tener que and a correspondi ng sharp decrease in haber de while haber que seems to have leveled off (when compared to the 19 th century DLNE). Taking these tendencies into consideration, it is difficult to determine the lag time but it is likely to be no more than 100 years based on the comparisons of earlier centuries. Regardless of the size of the gap in time, the spoken data clearly show that tener que has taken over the system. This, again, supports the idea that obligation in Spanish has undergone longitudinal renewal and tener que has essentially replaced haber de in most contexts.

PAGE 181

181 CHAPTER 5 FUTURE DIRECTIONS AN D CONCLUSIONS 5.1 Future Directions This paper has provided a detailed account of the development of obligation in Mexican Spanish O ne of the principle results is th at tener que has taken over the obligation system over the last century. This shift from haber de to tener que is primarily restricted to the 19 th and 20 th century, although semantic generalization does begin to occur as early as the 16 th century. Neverthe less, an important direction in which to go in the future would be to include more spoken data. Comparing these markers across dialects and generations is likely to give a clearer picture of the obligation system today. Recall that Labov (1972) indicates t hat synchronic variation is often a reflection of diachronic change so detecting changes in a modern system is essentially a window into the past. In addition to including more oral data, it is also necessary to examine more written data as well This is particularly important if the spoken corpora do in fact reflect a system that has been taken over by tener que as the results from the present study suggest Collecting additional written data from the 19 th to the 21 st centuries would further substantiate tendencies found in the present study and allow for comparisons across dialectal regions in order to account for and trace these changes. Besides accounting for changes in the system examining spoken and written data from more recent centuries also allo w for the examination of sociolinguistic factors. Although age and sex did not show any differences in terms of the choice of obligation marker in the present study research on other obligation ma r kers have shown differences in use across generations, sex and education level (e.g. Tagliamonte and

PAGE 182

182 Smith, 2006; Blas Arroyo, 2010 ). The inclusion of other social factors, such as education level, may provide a clearer picture of why and how some of the changes in the system occurred. Including more written and spoken data would also likely provide enough data to apply the comparative method (Tagliamonte & Poplack, 1988; Tagliamonte, 2006) Taking into account the limited number of occurrences of tener que tener de and haber que in the earlier centuries, it was impossible to incorporate the comparative method as a measure of change. This is in part because haber de accounted for more than 88% of all occurrences during the 16 th century. A larger corpus, however, may contain a sufficient number of examples of the other three obligation markers to apply the comparative method. Recall that the adoption of the comparative method into variationist studies is a valuable tool for determining the source of change establish ing the status of grammati calizing forms, as well providing a way to look at variation with a more fine grained perspective. This is done by comparing cross dialectal varieties or different time periods and identifying areas in the grammar that may be possibl e sources of c hange The discovery of these so urces can elucidate possible directions of change at a micro level and corroborate assertions on the macro level In this case comparative analyses could further help in substantiating universal claims of paths of development as well as provide new informa tion on how these changes take effect. A future study should consider including the obligation marker deber (de) as a possible variant. Although this variant does not express strong obligation, which was the focus of the present study, incorporating it int o the analysis may shed some light on changes that are occurring in the system as these markers move towards epistemicity

PAGE 183

183 This suggestion is based primarily on results from qualitative studies on obligation systems in other languages. For example, Myhill (1995: 160) found that the English obligation system underwent a general shift from using modals that were more focused on speaker control and societal norms like we see in the example 5 1 to newer markers that f interactiv e factors such as mutual cooperation, emotional appeals, advice, apologies or threats like we see in example 5 2 Although he recognizes that there are exceptions in his data, he believes that this is a general trend in the history of obligation markers 5 1 my daughte r shall marry you ( Myhill, 1995: 160) 5 2 got to help me (emotional plea) (Myhill, 1995: 160) Smit h (2003 on the gradual She posits that b ecause it is prototypically subjective and insistent, sometimes authoritarian sounding, root MUST is likely to be increasingly av oided in a culture where overt markers of power or hierarchy are much less in In order to consider whether or not a si milar trend is occurring in Spanish, an extensive examination of all obligation markers, including those that have been identified as being related to social norms in the literature (e.g. deber (de) in Molina Plaza 2005) is required. Traugott (1989) trie s to connect changes in deontic modals to the social motivations behind obligation. She claims that any obligation that is based in morals, reason, law or divine decrees is enforced by some outside force and therefore cannot be considered subjective (1989: 39) contrary to what Smith (200 3 ) is claiming If we assume that obligation based in morals, reason etc. are more authoritarian, due to the

PAGE 184

184 outside force, then we may be able to relate authoritarian with less grammaticalized obligation markers. If that i s indeed the case, we can deduce that the movement from agent oriented deonticity to subjective, epistemic meanings is also a paralleled by a social shift from more authoritarian to less authoritarian (or what Myhill calls more interactive obligation, 1995 ). Nevertheless, a more inclusive study would be useful to corroborate this. In sum, a ny further examination of this system over a time period of more than a century will provide a more reliable overview of the obligation system 5.2 General Summary This study looked at the development of obligation in Mexican Spanish from the 16 th through the 21 st century. Considering the dearth of variationist research available on Spanish modals this study was an important step in determining whether the obligation sys tem in Spanish is undergoing a change. There were several approaches adopted in an effort to reveal both the path of change of each of these markers, as well as to corroborate (or not) the postulations of universal paths of change in deontic modality. As a result of the narrow scope that has been utilized thus far, little evidence has surfaced on how obligation markers are used in Spanish in both text and oral speech. This study sought to fill this gap of knowledge by looking at the development of modality in Spanish both diachronically and synchronically by adopting a variationist and comparative sociolinguistic approach (Poplack and Meechan, 1998; Poplack and Tagliamonte, 2001; Tagliamonte, 2002; Tagliamonte and Smith, 2006) In addition, an examination o f the relative frequencies of these four obligation markers across time in a variety of contexts was also employed The aim was to uncover the linguistic and

PAGE 185

185 extralinguistic factors that condition the use of one expression of obligation over another while at the same time substantiating universal claims of change in modality. The variationist approach was adopted in part because a fundamental element of grammaticalization is variation. The areas in grammar where variation is found are ideal for examining h ow change and thus grammaticalization occurs. In the present study the variation was particularly noticeable in the later centuries when haber que and tener que increased in frequency. T hrough variable rule analysis the s imilarities and differences among these constructions were identified. We found that, contrary to the existing literature on obligation markers in Spanish, these constructions are used in distinct contexts and are conditioned by different factors. In addition to the variationist method, t he comparative method was also employed. Again, through a variable rule analysis t he comparative sociolinguistic method i s used as a way to examine similarities and differences in the language across time essentially offering a new detailed way to construc t the development of the obligation system. As a result of the comparative analysis we saw that, between the 19 th and 20 th centuries, verb type and direct objects were areas in the grammar where changes were taking place, particularly with haber de and ten er que This study was the first to look at obligation in Spanish from a both diachronic and variationist perspective. The results of this study not only illustrated that this system is in flux but it also contributed to the existing literature on the gram maticalization of modals across languages. In regards to these constructions, we see that each marker does indeed have its own semantic space and, although these forms grammaticalize and move towards epistemicity, they are in distinct parts of this path an d thus fulfill different

PAGE 186

186 semantic roles. Throughout the centuries, several factors have played a role in determining which marker will be used in the various contexts. The factors are also indicators of the movement of these obligation markers along the gr ammaticalization path. In spite of the fact that many of these factors have been mentioned in the previous literature, very few have been included in quantitative research on obligation makers (cf. Cornillie, 2007). I was able to incorporate the claims mad e in the literature and operationalize these claims into factors and successfully use these as measures of subjectification. Similar to work done on obligation in English (e.g. Tagliamonte & Smith 2006 ; 2007 ), this study offers evide nce supporting the claims of a universal path of grammaticalization, specifically the path proposed by linguists such as Bybee et al. (1994), Hopper and Traugott (2003) among others. In particular, evidence was provided for processes such as semantic weake ning, decategorialization and general movement into contexts associated with weak epistemicity and strong epistemicity. The data substantiates assertions that obligation systems undergo longitudinal renewal and that new markers come into the system because the older Previous literature on Spanish obligation markers rarely recognizes the semantic differences among these obligation markers or attributes the differences in meaning to psychologi cal motives (e.g. De Maeseneer, 1998; Sirbu Dumitrescu D, 1988 ). We now have a concrete example of the various factors that condition the use of each of these obligation markers, which is evidence against the idea that these forms essentially mean the same thing. It also illustrates that the factors that condition the use of each

PAGE 187

187 marker are not entirely dependent on the psychological motives of the speaker. The results of this study also go hand in hand with other, similar studies on obligation (Tagliamonte patterns. Through the adoption of comparative analyses during the 19 th and 20 th centuries, this study uncovered patterns of change in the system that illustrated the movement of constructi ons into and out of the semantic space of obligation. We can observe where in the grammar these specific changes are occurring through the use of the comparative method, which to my knowledge has not been done on obligation markers in Spanish before. This type of analysis has provided a baseline from which to build on for future research on obligation markers in both Spanish and other languages. It illustrates how both a diachronic and synchronic variationist examination of the obligation system can expose the path of obligation markers within Spanish, and perhaps across languages. The importance of frequency was also clearly noted in this study. A construction may occur in contexts that could implicate possible movement into new contexts, but without an in crease in frequency of use, this movement is unlikely. Recall that change like grammaticalization occurs in high frequency forms. We saw an example of this with the construction tener de which appeared in contexts associated with weak epistemicity as earl y as the 16 th century. Despite this apparent association, the frequency of this form decreased over time and, eventually, this form fell out of use. This example offers further support for the usage based model where we see that usage and frequency are imp ortant factors in language change.

PAGE 188

188 Finally, this study also predicted a possible path of grammaticalization for obligation markers in Spanish, where these constructions enter the language via declarative sentences containing communication or action verbs, typically in present tense or imperfect. Further grammaticalization is shown to be characterized by increasing occurrence with psychological and emotion verbs, habitual temporal markers and contexts containing negation. As the frequency in these new conte xts rises, they begin to be associated with weak epistemicity. If the grammaticalization continues the construction begins to appear more often with stative verbs, and less often when negation is present. There is also an increase of occurrence with tempor al future markers and inanimate subjects and objects. This is the stage that is identified as strong epistemicity by Traugott (1989). What appears to be one of the last stages of grammaticalization of these obligation markers as they move into epistemic c ontexts is the occurrence in contexts with existential subjects and the presence of preceding clitics. We did not see examples of counter examples of the principle of unidirectionality. All these constructions appeared to go along this path in the same dir ection thus supporting claims for unidirectionality made by linguists such as Haspelmath (2004), Taeymans (2004) and Hopper and Traugott (2003). While not all these constructions have grammaticalized to this last stage, we see that haber de and tener que a re going in that direction and, as they enter into epistemicity, it will likely open this space up for a new obligation marker to take.

PAGE 189

189 APPENDIX A LIST OF TEXTS USED I N WRITTEN CORPORA Work Author Word Count Century Corpus Carta del Cabildo de la ciuda d de Mrida al Rey Don Felipe Annimo 987 16 CORDE Carta annual del P. Francisco Ramrez Annimo 4289 16 CORDE Carta de Don Luis de Velasco, virey de la Nueva Espaa, al Emperador Don Carlos, recordndole la indicacin que, en su Real nombre, le hizo el Secretario Francisco de Eraso, al conferrsele aquel vireynato, de permutarlo, a los tres aos de servicio. Velasco, Luis de 825 16 CORDE Carta de Fray Pedro de Gante al Emperador D. Carlos Gante, Fray Pedro de 4777 16 CORDE Carta de la ciudad de Michoa can Annimo 1488 16 CORDE Carta de naturales de la provincia de Tlascala al Rey Don Felipe II, suplicando les concediera exenciones, ttulos y privilegios en remuneracin de los servicios de sus antepasados al tiempo de la conquista. Annimo 2184 16 CORDE Carta del Cabildo de la ciudad de Mrida al Rey Don Felipe II, haciendo patente la necesidad que tenia aquella tierra de defensa contra los franceses luteranos, recomendando la gobernacion de Don Luis Cspedes. Annimo 987 16 CORDE Carta del clrigo Ped ro de Logroo al Rey Don Felipe II Logroo, Pedro de 1120 16 CORDE Carta del doctor Diego Quixada, alcalde de Mrida de Yucatn, S.M., dando cuenta de algunas medidas de buen gobierno y administracion de justicia, y consultando la adopcion de otras. Qui jada, Diego 4157 16 CORDE Carta del licenciado Alonso Zuazo al padre Fray Luis de Figueroa, prior de la Mejorada. Zuazo, Alonso 3768 16 CORDE Carta del licenciado Francisco Ceynos Ceynos, Francisco 2224 16 CORDE Carta del licenciado Marcos de Aguilar Ag uilar, Marcos de 2628 16 CORDE Carta del padre fray Jernimo de Mendieta Mendieta, Fray Jernimo de 12,918 16 CORDE Carta del virey de la Nueva Espaa Don Martn Enrquez Enrquez, Martn 2875 16 CORDE Carta relacion del arzobispo de Mexico D. Pedro de Moya y Contreras, remitiendo al Rey Don Felipe II reservados informes personales del clero de su dicesis. Moya y Contreras, Pedro de 8472 16 CORDE Contrato adicional entre Juan Cromberger y Juan Pablos Annimo 591 16 CORDE Demanda de Ceballos en nombre de Pnfilo de Narvez Ceballos, Hernando de 3013 16 CORDE Desposorio espiritual entre el pastor Pedro y la Iglesia mexicana Prez Ramrez, Juan 3875 16 CORDE Fragmento de la visita hecha a don Antonio de Annimo 27,143 16 CORDE

PAGE 190

190 Mendoza Historia verdader a de la conquista de la Nueva Espaa Daz del Castillo, Bernal 138,000 16 website Informe al rey por el cabildo eclesistico de Guadalajara Annimo 9820 16 CORDE Instruccin civil y militar a Francisco Corts Cortez, Hernn 2265 16 website Los mayas de Yucatn Landa, Diego de 20,087 16 CORDE Memorial de Fr. Bartolom de las Casas Casas, Fray Bartolom de las 839 16 CORDE Memorial de fray Bartolom de las Casas al Consejo de Indias Casas, Fray Bartolom de las 1421 16 CORDE Origen de los mexicanos Ann imo 9896 16 CORDE Parecer de Alonso del Castillo Castillo, Alonso del 462 16 CORDE Parecer de Fray Domingo de Betanzos Betanzos, Fray Domingo de 2997 16 CORDE Probanza hecha en la Villa Segura de la Frontera Annimo 3632 16 CORDE Real cdula dirigida a Juan Gutirrez Caldern Annimo 393 16 CORDE Relacin hecha por el seor Andrs de Tapia, sobre la conquista de Mxico. Tapia, Andrs de 17,855 16 CORDE Revelacin sobre la reincidencia en sus idolatras Feria, Fray Pedro 4494 16 CORDE Segunda carta d el doctor Ceynos Ceynos, Francisco 2696 16 CORDE Tercera relacin annima de la jornada que hizo Nuo de Guzman Annimo 8986 16 CORDE Apologtica historia sumaria Casas, Fray Bartolom de las 577,295 16 CORDE Captulo de carta del rey al virrey de Nueva Espaa Annimo 273 16 CORDE Carta al Rey Don Felipe II, del virey de la Nueva Espaa, Marqus de Villamanrique, dando cuenta del arribo del corsario ingls Francisco Drakc, al puerto de la Habana. Marqus de Villamanrique 2468 16 CORDE Carta annua del P Francisco Ramrez Annimo 4289 16 CORDE Carta de Diego de Ocaa Ocaa, Diego de 5745 16 CORDE Carta de Don Luis de Velasco, virey de la Nueva Espaa, al Emperador Don Carlos,, recordndole la indicacion que, en su Real nombre, le hizo el Secretario Fra ncisco de Eraso, al conferrsele aquel vireynato, de permutarlo, a los tres aos de servicio. Velasco, Luis de 825 16 CORDE Carta de Fray Angel de Valencia, custodio, y otros religiosos de la rden de San Francisco, proponiendo los medios necesarios para doctrinar los indios del Nuevo Reino de Galicia y de la provincia de Mechoacan. Valencia, Fray ngel de 6391 16 CORDE Carta de Fray Domingo de Betanzos Betanzos, Fray Domingo de 1258 16 CORDE Carta de Fray Jacobo de Tastera, y de otros religiosos de la rden de San Francisco, al Emperador D. Carlos, dndole cuenta del estado de sus misiones y de la buena disposicin de los indios. Tastera, Fray Jacobo de 1879 16 CORDE Carta de Fray Martin de Valencia, y otros misioneros al emperador. Valencia, Fray Mart n de 1099 16 CORDE Carta de Fray Miguel Navarro, y otros religiosos de Navarro, Fray 781 16 CORDE

PAGE 191

191 la orden de San Francisco, al Real Consejo de las Yndias recomendando al licenciado Muoz. Miguel Carta de Jernimo Lpez al emperador. Lpez, Jernimo 555 6 16 CORDE Carta del arzobispo de Mexico D. Pedro de Moya y Contreras al presidente de los Reales consejos de Indias y Hacienda, sobre los conflictos ocasionados por la representacion de un entrms, y otros sucesos. Moya y Contreras, Pedro 7839 16 CORDE Carta del contador Rodrigo de Albornoz Albornoz, Rodrigo de 12,400 16 CORDE Carta del ejrcito de Corts al emperador Annimo 3297 16 CORDE Carta del obispo de Mexico, Fray Juan de Zumarraga Juan de Samano, secretario de S.M., hacindole presente algun as necesidades de sus diocesanos, y rogndole que apoyara su proyecto de edificacion de colegios y monasterios para jvenes de ambos sexos. Zumrraga, Fray Juan de 4399 16 CORDE Carta del obispo de Yucatn Fray Francisco de Toral, al adelantado de la Flor ida, Pedro Menndez de Avils, acusando el recibo de otra suya, anuncindole el pronto envo de bastimentos y dndole consejos para su buen gobierno. Toral, Fray Francisco de 1275 16 CORDE Carta del padre fray Jernimo de Mendieta Mendieta, Fray Jernimo de 12,918 16 CORDE Carta del virey de la Nueva Espaa Conde de Corua 3806 16 CORDE Carta del virey de la Nueva Espaa Enrquez, Martn 3977 16 CORDE Carta indita de Hernn Corts Cortez, Hernan 5961 16 CORDE Cartas de relacines Hernn Cortez 246,000 16 website Contrato celebrado entre Juan Cromberger Annimo 702 16 CORDE Contrato de compaa celebrado entre Juan Cromberger y Juan Annimo 2298 16 CORDE Crnica de la Nueva Espaa Cervantes de Salazar, Francisco 400,000 16 website Cuarta relacin an nima de la jornada que hizo Nuo de Guzm Annimo 9177 16 CORDE Daos e inconvenientes que se derivan, en lo temporal y en lo espiritual, de la ignorancia de los idiomas aborgenes. Molina, fray Alonso de 760 16 CORDE Expediente tramitado en Len de Nic aragua Annimo 2403 16 CORDE Genealoga y linaje de los Seores de Nueva Espaa Annimo 6873 16 CORDE Historia de los Indios de la Nueva Espaa Fray Toribio de Benavente 97,329 16 CORDE Historia de los mexicanos por sus pinturas Annimo 13,001 16 CORDE Historia general de las cosas de Nueva Espaa Sahagn, Fray Bernardino de 368,588 16 CORDE Historia verdadera de la conquista de la Nueva Espaa Daz del Castillo, Bernal 48,000 16 CORDE Informe al rey por el cabildo eclesistico de Guadalajara, acerca de las cosas de aquel reino. Annimo 9820 16 CORDE Instruccin civil y militar a Francisco Corts, para la expedicin de la costa de Colima. Corts 2265 16 CORDE Instrucciones al virrey de la Nueva Espaa para que se divulgue el conocimiento de la lengua espaola Annimo 328 16 CORDE

PAGE 192

192 entre los indios, a fen de evitar errores en la cristianizacin; tratndose asunto tan importante con la audiencia, obispo y misioneros. Leyes de los indios de Nueva Espaa, Anhuac o Mxico. Alcobiz, Fray Andrs de 2222 16 CORDE Leyes y ordenanzas Annimo 9098 16 CORDE Lo que pas con Cristbal de Tapia acerca de no admitirle por gobernador, con los procuradores de Mxico y dems poblaciones, y los de Hernn Corts. Annimo 4633 16 CORDE Memoria de lo acaecido en esta ciu dad despus que el gobernador Hernando Corts sali della, que fue a los doce das del mes de octubre de mil e quinientos e veinte e cinco aos. Annimo 5228 16 CORDE Memorial de don Alonso de Zurita Zurita, Alonso de 4232 16 CORDE Memorial del obispo Fr ay Bartolom de las Casas y Fray Domingo de Santo Toms. Casas, Fray Bartolom de las 2244 16 CORDE Ordenaciones generales dadas por el padre Diego de Avellaneda por las que deban regirse las provincias mexicanas de la Compaa de Jess. Annimo 366 16 C ORDE Ordenanzas militares y civiles mandadas pregonar por don Hernando Corts en Tlaxcala, al tiempo de partirse para poner cerco a Mxico. Annimo 2435 16 CORDE Parecer de don Sebastin Ramrez de Fuenleal, obispo de Santo Domingo, y presidente de la re al audiencia de Nueva Espaa. Ramrez de Fuenleal, Sebastin 8928 16 CORDE Primera parte de los problemas y secretos maravillosos de las Indias Crdenas, Juan de 88,739 16 CORDE Primera relacin annima de la jornada que hizo Nuo de Guzmn a la Nueva Ga licia Annimo 3312 16 CORDE Proceso de Pedro de Ocharte Annimo 22,780 16 CORDE Real Cdula a los Oficiales de la Casa de la Contratacin de Sevilla. Annimo 313 16 CORDE Real Cdula al arzobispo de Mxico pidiendo informes sobre las caractersticas de la tierra. Annimo 1049 16 CORDE Real Cdula dirigida a Juan Cromberger, por la que se le concede privilegio para la impresin de la Doctrina en lengua de indios de Mechoacn. Annimo 448 16 CORDE Real Cdula y cuestionario para la formacin de descripci ones geogrficas eclesisticas del arzobispado de Mxico. Annimo 1411 16 CORDE Relacin de la conquista de los Teules chichimecas que dio Juan de Smano. Smano, Juan de 11,332 16 CORDE Relacin de la entrada de Nuo de Guzmn, que dio Garca del Pilar, su intrprete. Pilar, Garca del 4949 16 CORDE Segunda relacin annima de la jornada que hizo Nuo de Guzmn Annimo 4331 16 CORDE Historia eclesistica indiana Mendieta, Fray Jernimo 334,813 17 CORDE Denuncia del indio tarasco Antonio Joan Apatzi po r haber sido solicitado por un fraile franciscano. Annimo 1085 17 CORDE Fragmentos de una historia de la Nueva Galicia Tello 39,077 17 CORDE Historia de la Nueva Mxico Villagr, Gaspar 78,687 17 CORDE Historia eclesistica indiana Mendieta, Fray Jern imo 334,813 17 CORDE Infortunios de Alonso Ramrez Sigenza y 17,793 17 CORDE

PAGE 193

193 Gngora Ordenanzas de los maestros de escuelas de la ciudad de Mxico Annimo 1182 17 CORDE Los sirgueros de la Virgen sin original pecado Bramn 18,585 17 CORDE El pastor de nochebuena Palafox y Mendoza 33,489 17 CORDE Fragmento de una Real Cdula Annimo 806 17 CORDE Historia de la nacin chichimeca Ixtlilxochitl 100,502 17 CORDE Carta autgrafa de Juan Bruno Eusebio de Palma, tirador de oro y plata, denunciando a un soli citante Palma, Juan Bruno Eusebio de 487 18 CORDE Historia Antigua de Mxico Clavijero, Francisco Javier 334,302 18 CORDE La portentosa vida de la muerte Bolaos, Fray Joaqun 65,525 18 CORDE Memoria sobre la construccin de sumideros para purificar la atmsfera Basadre, Vicente 3026 18 CORDE El Zarco, episodio de la vida mexicana en 1861 63 Altamirano, Ignacio Manuel 51,187 19 CORDE Gil Gmez, el insurgente novela histrica Daz Covarrubias, Juan 81,934 19 CORDE La Quijotita y su prima Fernndez de L izardi 172,696 19 CORDE Memoria sobre la necesidad y utilidades de la construccin de un camino carretero desde Veracruz a Mxico. Austria, Jos Donato de 7008 19 CORDE Clemencia Altamirano, Ignacio Manuel 51,118 19 CORDE Cuentos del General Riva Pala cio 31,086 19 CORDE Historia de Chucho el Ninfo Facundo (Jos Toms de Cullar) 67,761 19 CORDE La bola Rabasa, Emilio 38,889 19 CORDE La gran ciencia Rabasa, Emilio 39,827 19 CORDE La Linterna Mgica Jos Toms de Cullar 35,900 19 internet archive Noches tristes y da alegre Fernndez de Lizardi 25,726 19 CORDE Oracin patritica en la primera fiesta conmemorativa de la MXICO Barquera, Juan Wenceslao 3570 19 CORDE Suprema Ley Gamboa, Federico 114,215 19 CORDE Tomchic Fras Heriberto 68,300 19 internet archive Cosas de cualquier familia Medina 52,310 20 CREA Dalia Martin del Campo 35,420 20 CORDE Duerme Boullosa 30,215 20 CREA El apando Revueltas 10,570 20 CORDE El canto de la grilla Rubn 47,527 20 CORDE El error de la luna Aguilar Cam n, Hctor 61368 20 CORDE

PAGE 194

194 Imposibilidad de los cuervos Padilla 28,676 20 CREA La casa de las mil vrgenes Azuela, A 98,104 20 CREA La frontera increble Revueltas 1961 20 CORDE La muerte de Artemio Cruz Fuentes, Carlos 95,060 20 CORDE La regin ms tra nsparente Fuentes, Carlos 145,067 20 CORDE Por vivir en quinto patio Alatriste 62,798 20 CREA Sinfona pastoral Revueltas 8159 20 CORDE Enfermera, doctora o santa? Olivera Figueroa 75,376 20 CREA Andanzas del indio Vicente Alonso Montao Hurtado 24 ,799 20 CREA Arrncame la vida ngeles 75,001 20 CREA Como agua para chocolate Esquivel, L 53,874 20 CREA Cristbal Nonato Fuentes 196,455 20 CORDE De Los Altos Chao Ebergenyi 149,939 20 CREA Dios en la tierra Revueltas 2126 20 CORDE Dos crmenes Iba rgengoitia, Jorge 53,934 20 CREA El batallador Chvez Jr., Gilberto 136,959 20 CREA El callado dolor de los tzotziles Rubn 34,751 20 CORDE El error de la luna Aguilar Camn, Hctor 61,368 20 CREA El lenguaje de nadie Revueltas 3358 20 CORDE El perro de la escribana o Las Piedecasas Mendoza 24,644 20 CREA El tamao del infierno Azuela, A 141,269 20 CORDE Fbrica de conciencias descompuestas Gerardo 40,924 20 CREA Hegel y yo... Revueltas 3859 20 CORDE Jess el bisabuelo y otros relatos Espinosa 43, 129 20 CREA Juegos florales Pitol 59,357 20 CREA La caricia rota Aguilera, Nuri 32,136 20 CREA La lucirnaga Azuela, Mariano 39,624 20 CORDE Los de abajo Azuela, Mariano 34,644 20 CORDE Morir en el Golfo Aguilar Camn, Hctor 83,261 20 CREA Palinuro de Mxico Paso 276,537 20 CREA Que la carne es hierba Campos, Marco 24,270 20 CREA Quin como Dios Gonzlez 136,658 20 CREA Setenta veces siete Elizondo 75,357 20 CREA Un grito desesperado. Novela de superacin personal Cauhtmoc 49,588 20 CREA Una pi ata llena de memoria Leyva 101,865 20 CREA

PAGE 195

195 APPENDIX B SAMPLE QUESTIONS FOR SOCIOLINGUISTIC INTE RVIEW 1. Demografa : Cmo se llama? Cul es su direccin? Hace cunto tiempo que vive all? Dnde naci? Ha vivido en otros lugares? Dnde? Dnde naci eron y crecieron sus padres? sus abuelos? su marido/esposa? Trabaja? Dnde? A qu se dedican sus padres? su marido/esposa? Cuntos aos de escuela pudo terminar? Cul fue el primer trabajo que tuvo al salir de la escuela? Me puede contar algo so bre su casa/apartamento? Qu tipo de casa es? Habla otras lenguas adems del espaol? 2. El barrio Este barrio parece interesante. Hace cuanto tiempo que vive aqu? Qu tipo de gente vive en esta calle? en esta rea? Por qu se mudaron aqu sus padr es? (o usted) Queda lejos del trabajo de su padre/suyo? Cmo ha cambiado el barrio en su vida? Se siente que el barrio sea tan seguro como era cuando era nio/a? por qu o por qu no? Es un barrio donde la gente se habla? Conoce Ud. a sus vecino s? Cmo son? Se dice que hoy en da todos estn demasiado ocupados para pasar por las casas de los vecinos para charlar, qu cree Ud.? Por qu cree que (no) ha cambiado? Hay que gente aqu que conoce tan bien que no le necesita anunciar su visita? Ha y que gente de aqu que pasa por su casa para charlar? Hay que un lugar en esta vecindad donde la gente se rene? Dnde se puede ir para un refresco o un t/caf por la tarde? Hay que un bar que frecuenta? Dnde se rene la gente fuera de casa? Hay que gente con quienes quiere pasar ms tiempo, pero no puede? Por qu no la ve tanto ya? Cree que la comunidad/barrio puede ser ms unida/o? Cmo? Cul es la cosa que le gusta ms del barrio? Cules cosas le hace sentir bien/mal de su barrio? 3. Cos tumbres sociales Hay que gente por aqu que no son familiares y con quienes pasa mucho tiempo? Viven cerca? Por dnde? Qu hacen cuando se renen? Sale con amigos? Qu hacen?

PAGE 196

196 Va a bailes? Qu tipo de entretenimiento hay que por aqu? Qu tip o de baile hace? Cmo se viste la gente cuando va? 4. Familia Sabe hace cunto tiempo que vive su familia aqu en este pueblo/estado/pas? De dnde venan? Recuerda cuando lleg a este pueblo/estado/pas? Cmo era cuando era nio/a? Era travieso? Qu tipo de cosas traviesas haca? Le castigaron? Quines? Se le culparon por cosas que no haba hecho? Sus padres imponan un toque de queda? Qu pas si no lo obedeci? Sus padres ya tenan ideas de lo que queran que Ud. fuera cuando se crec a? Tiene hermanos? Cuntos? Cmo era su relacin con ellos? Cmo era ser el mayor/menor/en el medio de los hermanos? Cree que sus hermanos pudieron hacer cosas que sus padres no le permitieron hacer a Ud.? Puede darme un ejemplo? Pasaba mucho ti empo con sus __________? abuelos, tos y tas, primos etc. por qu o por qu no? Iban de vacaciones mucho? Adnde? Cmo llegaron? Haba una vez en que alguien en su familia le avergonz por algo que haba dicho o hecho? Qu pas? Cmo reaccion ? Qu puede decir ahora retrospectivamente? 5. Trabajo Cmo fue su primer trabajo? Cuntos aos tena cuando empez a trabajar? Recuerda cunto gan? Recuerda cmo quera gastar ese dinero? A qu se dedicaban sus padres? A qu se dedica Ud.? Qu le gustara hacer como trabajo? 6. La comida Se dice que mucha gente tiene su mejor comida el domingo, era as en su familia? Qu tipo de comida coman en su niez? Mucha gente recuerda ciertos platos que hacan sus madres/padres, se acuerda de un plat o especial que haca su madre/padre?

PAGE 197

197 Qu coma su _______________? abuelo, abuela, padre, madre Cocina Ud.? Coce? Qu tipo de comida le gusta cocer/cocinar? La comida de hoy es diferente de la comida de su niez? Cmo? Por qu? 7. Medicina popular Iba la gente mucho al doctor cuando era nio/a? Iba Ud. cuando estaba enfermo/a? Se dice que la gente dependa mucho de la medicina y los remedios populares, recuerda ese tiempo? Recuerda un remedio particular? Me acuerdo que mi abuelo tomaba mucho ________ (ajo, miel, limn, etc.), Ha escuchado de ese remedio? Qu hace cuando se enferma? Qu hace para evitar los resfriados o la gripe? 8. Jvenes/padres Se dice que ser joven hoy en da es muy diferente que serlo cuando era joven Ud., est de acue rdo? Cul es la diferencia? Por qu? Puede comparar que haca para divertirse cuando era joven con las cosas que hacen los jvenes de hoy para diversin. Se dice que los padres de hoy son ms estrictos, est de acuerdo? Qu no le gustaba de sus padr es, hermanos? Gastabas bromas a sus hermanos? Qu fue la mejor broma que gast? La ms chistosa? La peor?

PAGE 198

198 APPENDIX C COMPLETE LIST OF FAC TOR GROUPS AND FACTO RS Verb Class Psychological /emotional (e.g The verbs haber/tener The verb hacer Animacy/grammatical person 1 st person animate ( e.g. tenemos que ponernos al da. (TQ) ) 2 nd person animate (e.g. has de morir en Guayaquil (HD)) 3 rd person animate (e .g. ya ves que mi mam siempre tiene que ganar (TQ) ) 3 rd person inanimate ( mi angustia hubo de alcanzar su punto ms alto anguish must have reached its maximum point ) other (entities that have been given a nimate characteristics for example: Puebla tendra que darle al Presidente la recepcin President the reception (TQ); Mastretta CREA ) Type of Sentence I nterrogative (e.g. Tuviste que pasar por all, por Colima? You had to pass through (TQ)) D eclarative ( Hay que dar el apoyo a l. h (HQ) ) Polarity Negative ( Yo no tena que pasar la noche fuera de mi casa. spend (TQ) ) Positive ( Ten amos que ir a una pltica (TQ)) Direct Objects Animate ( tena que pagar el rico. (TQ)) Inanimate ( tiene que pagar uno renta (TQ) ) Oblique ( tengo que ir con una persona que me hable who talks Middle Voice (e.g. y si te cuesta tanto se tiene que traer. you have to (TQ) ) Other Indirect Objects Present ( tena que dar un respecto a nuestros padres had to give respect to our Not present ( tienes que ganar mi confianza Preceding clitic

PAGE 199

199 Present ( Usted me tiene que ensear. ) Not present ( tenamos que salir ms temprano ed to have to leave (TQ) ) Tense aspect mood For example, (TQ) ) imperfect: ( no haba qu irnos a veces (HQ) ) imperfect subjunctive: (s pasara el examen voy a tener que Temporal adverbs For example, specific: ( ahora ya tengo que cerrar los ojos Fuentes, CORDE) non specific: ( quiz muy pronto tengamos que salir del pas soon other: ( y tuvo que pensar Lexical verb type For example: seguir: ( Hay que seguir adelante tener: ( ya has de tener dinerito Sex Male Female Age 20 45 years greater than 45 years

PAGE 200

200 LIST OF REFERENCES A ARON J ESSI E LANA 2004. The gendered use of salirse in Mexican Spanish: S i me sala yo con las amigas, se enojaba Language in Society 33(5). 585 607. A ARON J ESSI E LANA 2006. Me voy a tener que ir yendo: A Corpus Based Study of the Grammaticization of the ir a + INF Construction in Spanish. Selected Proceedings of the 9th Hispanic Linguistics Symposium, ed. by Nuria Sagarra and Almeida Jacqueline Toribio, 263 72. Somerville, MA: Cascadilla Proceedings Project. A ARON J ESSICA E LANA 2007. Variation and change in spanish future temporal expression: Rates, constraints, and grammaticization Albuquerque, NM: UNM dissertation. A ARON J ESSI E LANA 2010. Pushing the envelope: Looking beyond the variable context. Language Variation and Change 22 1. 1 36. A ARON J ESSI E LANA AND R ENA T ORRES C ACOULLOS 2005. Q uantitative measures of subjectification: A variationist study of Spanish salirse. Cognitive Linguist ics 16 4.607 633. A LAMILLO A SILLA R EIG 2009 Cross dialectal variation in propositional anaphora: Null objects and propositional lo in Mexican and Peninsular Spanish. Language Variation & Change, 21( 3 ) 381 415 A MMON U. 2004. Sociolinguistics: An interna tional handbook of the science of language and society Berlin: Walter de Gruyter. A SH S HARON 2002. Social Class. T he handbook of language variation and change ed. by J.K. Chambers, Peter Trudgill, and Natalie Schilling Estes, 402 422. Malden, MA: Blackw ell. A XELROD M ELISSA 1999. Lexis, grammar, and grammatical change: The koyukon classifier prefixes. Functionalism and formalism in linguistics: Volume ii: Case studies ed. By In Darnell, Michael, Moravcsik, Edith, Newmeyer, Frederick,Noonan, Michael, & K. Wheatley 39 58. Amsterdam: John Benjamins. B AILEY G UY 2002. Real and Apparent Time. T he handbook of language variation and change ed. by J.K. Chambers, Peter Trudgill, and Natalie Schilling Estes, 312 332. Malden, MA: Blackwell. B ARLOW M ICHAEL 200 0. Usage, Blends and Grammar. Usage based models of language ed. by Michael Barlow, and Suzanne Kemmer, 315 346. Stanf ord, Calif: CSLI Publications, Center for the Study of Language and Information.

PAGE 201

201 B ARLOW M ICHAEL AND S UZANNE K EMMER 2000. Usage based mo dels of language. Stanford, Calif: CSLI Publications, Center for the Study of Language and Information. B AYLEY R OBERT 2002. The quantitative paradigm. T he handbook of language variation and change ed. by J.K. Chambers, Peter Trudgill, and Natalie Schilli ng Estes, 117 141. Malden, MA: Blackwell. B IBER D OUGLAS 2000. Investigating Language Use through Corpus Based Analyses of Association Patterns. Usage based models of language ed. by Michael Barlow, and Suzanne Kemmer, 287 314. Stanford, Calif: CSLI Publ ications, Center for the Study of Language and Information. B LACKWELL 1987 Observing and analysing natural language: A critical account of sociolinguistic method Oxford, OX, UK: Blackwell. B LAS A RROYO J OS L UIS 2008. The variable expression of future t ense in Peninsular Spanish: The present (and future) of inflectional forms in the Spanish spoken in a bilingual region. Language Variation and Change 20.85 126. B LAS A RROYO J OS L UIS 2010. Confluencia de normas sociolingsticas en un hecho de variaci n sintctica: Factores sociales en la seleccin de la variante "deber de" + infinitivo (vs. "deber") en un corpus de espaol peninsular Hispania 93(4).624 649 B OLINGER D WIGHT 1977. Meaning and form London: Longman. B RINTON L AUREL J 1991. The Origin E nglish. The Workshop on Verbal Periphrases Amsterdam. Online: http://faculty.arts.ubc.ca/lbrinton/ B RINTON AND T RAUGOTT 2003. Lexicalization and language change. Cambridge, England: Cambridge University Press. B RINTON AND T RAUGOTT 2005 Lexicalization and grammaticalization all over again. Historical Linguistics 2005: Selected papers fro the 17 th international conference on historical linguistics, ed. by J oseph C. Salmons and Shannon Dubenion Smith, 3 31. Amsterdam Netherlands: John Benjamins Publishing Co. B YBEE J OAN L. 1987. The semantic development of past tense modals in English and other languages. Modality in grammar and discourse ed. by Joan Bybee and Suzanne Fleischman, 503 517. Amsterdam, Netherlands: John Benjamins Publishing Co. B YBEE J OAN L 1998. A functionalist approach to grammar and its evolution. Evolution of Communication 2 2.249 278. B YBEE J OAN L. 2001. Phonology and language use Camb ridge, England: Cambridge University Press.

PAGE 202

202 B YBEE J OAN 2003a. Mechanisms of change in grammaticization: The role of frequency. The handbook of historical linguistics ed. by Brian D. Joseph and Richard D. Janda, 602 23. Oxford: Blackwell B YBEE J OAN L. 20 03 b Cognitive processes in grammaticalization. The new psychology of language: Cognitive and functional approaches to language structure ed. by Michael Tomasello, 2.145 167. Mahwah, NJ: Lawrence Erlbaum Associates. B YBEE J OAN L. 2006. From usage to gram mar: The mind's response to repetition. Language 82 4.711 733. B YBEE J OAN L., AND P AGLIUCA ,W ILLIAM .1985. Crosslinguistic comparison and the development of grammatical meaning. Historical semantics: Historical word formation ed. by Jacek Fisiak, 59 83. Berl in: Mouton de Gruyter. B YBEE J OAN L., R EVERE D. P ERKINS and W ILLIAM P AGLIUCA 1994. The evolution of grammar: Tense, aspect, and modality in the languages of the world Chicago: University of Chicago Press. B YBEE J OAN & S ANDRA T HOMPSON 2000. Three fr equency effects in syntax Berkeley Linguistics Society 23 65 85. B YBEE J OAN AND S UZANNE F LEISCHMAN 1995. Modality in grammar and discourse: An introductory essay. In Modality in grammar and discourse. ed. by Joan Bybee, Suzanne Fleischman, 1 14. Amste rdam, The Netherlands: John Benjamins Publishing Co. C AMPBELL L YLE 1998. Historical linguistics: an introduction Edinburgh: Edinburgh University Press. C ASANOVA S EUMA L OURDES 1999. El sujeto en cataln coloquial. Revista Espaola de Lingstica 29( 1 ) 105 131. C EDERGREN H ENRIETTA J., and D AVID S ANKOFF 1974. Variable rules: Performance as a statistical reflection of competence. Language 50 2.333 355. C HOI J INNY 1998. Languages in contact: a morphosyntactic analysis of Paraguayan Spanish from a histo rical and sociolinguistic perspective Washington, D.C.: Georgetown University dissertation C OATES J ENNIFER 1983. The Semantics of the modal auxiliaries London: Croom Helm. C OMPANY C OMPANY C ONCEPCION 1994. documentos lingsticos de la Nueva Espaa: A ltiplano central. Mxico: UNAM C OMPANY C OMPANY C ONCEPCION 2004. Grammaticalization or degrammaticalization? the reanalysis and subjectivization of verbs as discourse markers in the history of spanish. Revista de Filologia Espanola 84(1).29 66.

PAGE 203

203 C OMPANY C O MPANY C ONCEPCION 2006. Subjectification of verbs into discourse markers: Semantic pragmatic change only? Belgian Journal of Linguistics 20.97 121. C ORNILLIE B ERT 2007. Evidentiality and episte mic modality in S panish (semi )auxiliaries. A cognitive funct ional approach Berlin : Mouton de Gruyter. C ORNILLIE B ERT D E M ULDER W ALTER V AN H ECKE T INE AND D IETER V ERMANDERE 2009. Modals in Romance Languages Modals in the Languages of Europe. A reference work ed. by Bjrn Hansen., Ferdinand de Haan Berlin, 5 11 561. Berlin: Mouton de Gruyter. D AVIES M ARK 1995. Analyzing syntactic variation with computer based corpora: the case of modern Spanish clitic climbing. Hispania 78( 2 ) 370 380. D E M AESENEER R ITA (1998). On some problems related to periphrases of ob ligation. [Sobre algunos problemas relacionados con las perifrasis obligativas] Linguistica Antverpiensia 32 39 53. D C AMPOS M ANUEL 2011. The handbook of Hispanic sociolinguistics Malden, MA: Wiley Blackwell. E CKART P ENELOPE 1997. Age as a Sociolinguistic Variable. In The handbook of sociolinguistics ed. by Florian Coulmas,151 168. Oxford, UK: Blackwell Publish ers. E RMAN B RITT AND B EATRICE W ARREN 2000 The idiom principle and the open choice principle. Text 20.29 62. F ENTE G R AFAEL J F A LVAREZ AND L OPE G. F 1983. verbales Madrid: EDI 6 F ERNANDEZ D E C ASTRO F ELIX 1999 Madrid: Gredos F INEGAN E DWARD 1995. Subjectivity and subjectivisation: an introduction. Subj ectivity and subectivisation: Linguistic perspectives ed, by D.Stein and S. Wright 1 15. Cambridge: Camb ridge University Press. F LEISCHMAN S UZANNE 1982. The future in thought and language: Diachronic evidence from Romance. Cambridge, England: Cambridge University Press G IVON T ALMY 1984. Syntax. A functional typological introduction Vol. 1. Amsterdam/P hiladelphia: John Benjamins Publishing Co G T ORREGO L EONARDO 1988. Madrid: Arco/Libros.

PAGE 204

204 G UTIRREZ M ANUEL J. 2010 La media subida de clticos en el espaol de Houston. Nueva Revista de Filologa Hispnica 58( 2 ). 645 662 G UY G REGORY 1988. Advanced VARBRUL analysis. Linguistic Contac t and Change: Proceedings of the Sixteenth Annual Conference on New Ways of Analyzing Variation, ed. by Kathleen Ferrara, Becky Brown, Keith Walters and John Baugh, 124 36. Austin, TX: University of T exa s Department of Linguistics. H AIMAN J OHN (1994). Ritualization and the development of language. Per spectives on grammaticalization ed. by William Pagliuca, 3 28 Amsterdam: John Benjamins Publishing Co. H ASPELMATH M ARTIN 2004. Does linguistic explanati on pre suppose linguistic description? Studies in Language 28 ( 3 ). 554 579 (special issue on linguistic evidence, guest edited by Martina Penke and Anette Rosenbach) H EINE B ERND 1995. Agent Oriented vs. Epistemic Modality: Some Observations on German Modal s Modality in grammar and discourse ed. by Joan L. Bybee and Suzanne Fleischman 17 55. Amsterdam: J ohn Benjamins. H EINE BERND ULRIKE CLAUDI AND FRIEDERIKE HNNE MEYER 1991. Grammaticalization: A conceptual framework. Chicago: University of Chicago P ress. H ENGEVELD K EES 1988 Illocution, Mood and modality in a functional g rammar of Spanish. Journal of Semantics 6.227 269 H OPPER P AUL J., A ND E LIZABETH C LOSS T RAUGOTT 1993. Grammaticalization Cambridge textbooks in linguistics. Cambridg e,England : Cam bridge University Press. H OWE C HAD & S CHWENTER S COTT A. 2008 Variable constraints on past reference in dialects of Spanish. Selected Proceedings of the 4th Workshop on Spanish Sociolinguistic s ed. by Maurice Westmoreland & Juan Antonio Thomas, 100 108 Somerville, MA: Cascadilla Proceedings Project. H YMAN L ARRY : 1984. Form and Su bstance in Language Universals. Explanations for Language Universals ed. Brian Butterworth, Bernard Comrie, and sten Dahl. Berlin: Mouton. 67 85. I NTERNET A RCHIVE (2002a). Ab out the Internet Archive. Online : http://www.archive.org/about/about.php K ASTOVSKY D IETER 1991. Historical English syntax Berlin: Mouton de Gruyter. K IEFER F ERENC 1987. On defining modality. Fo lia Linguistica XXI 67 94

PAGE 205

205 K IM M INJU 2009. The intersection of the perfective and imperfective domains: A corpus based study of the grammaticalization of Korean aspectual markers. Studies in Language, 33(1), 175 214. K LEIN A NDREU F LORA 1991. Losing g round: A discourse pragmatic solution to the history of ra in Spanish. Discourse pragmatics and the verb: the evidence from Romance ed. by Suzanne Fleischman & Linda R.Waugh, 164 78. London: Routledge K RUG M ANFRED 2000. Emerging English modals: A co rpus based study of grammaticalization Berlin, New York: Mouton de Gruyter. K URYLOWICZ J ERZY 1976[1965]. The evolution of grammatical categories. repr. in Jerzy Kurylowicz, Esquisses linguistiques Vol. 2, 38 54. Mnchen: Fink. L ABOV W ILLIAM 1966. Th e social stratification of English in New York City Washington: Center for Applied Linguistics. L ABOV W ILLIAM 1972. Sociolinguistic Patterns Philadelphia: The University of Pennsylvania Press. L ABOV W ILLIAM 1994. Principles of linguistic change vol. 1 : Internal factors Oxford, UK: Blackwell. L ABOV W ILLIAM 2008. Quantitative reasoning in linguistics. http://www.ling.upenn.edu/~wlabov/Papers/QRL.pdf L ANGACKER R ONALD 2000. A Dynamic Us age Based Model. Usage based models of language ed. by Michael Barlow, and Suzanne Kemmer, 1 64. Stanford, Calif: CSLI Publications, Center for the Study of Language and Information. L AVENDERA B. 1978. Where does the sociolinguistic variable stop? Languag e in Society 7.171 182 L EHMANN W. P., & M ALKIEL Y AKOV 1982 Perspectives on historical linguistics Amsterdam Netherlands : J ohn Benjamins Publishing Co. L EVEY S TEPHEN 2012. standard spoken English: a perspective from var iationist sociolinguistics Languag e and Educatio n 1 17 L UNA T RAILL E LIZABETH ; 1980: Sintaxis de los verboides en el habla culta de la ciudad de Mxico Mxico: UNAM. L UNN P ATRICIA 1995. Modality in grammar and discourse: An introductory essay. In Moda lity in grammar and discourse. ed. by Joan Bybee, Suzanne Fleischman, 1 14. Amsterdam, Netherlands: John Benjamins Publishing Co. L YONS J OHN 1977. Semantics Cambridge: Cambridge University Press.

PAGE 206

206 M ARRANO A NN M ARIE 1997. The syntax of modality: A comp arative study of epistemic and root modal verbs in Spanish and English Georgetown WA: Georgetown University dissertation. M ARTINEZ D IAZ E VA 2003. perifrsticas de obligacin. Algn fenmeno d e variacin en el espaol de Catalua. Interlingstica 14. 681 694 M ATISOFF J AMES 1991. Areal and universal dimensions of grammaticalization in Lahu A pproaches to Grammaticalization vol. II ed. by Elisabeth C. Traugott and Bernd B. Heine, 383 454. Amste rdam, Netherlands: John Benjamins Publishing Co. M AYORAL H ERNANDEZ R OBERTO 2007. A variation study of verb types and subject position: verbs of light and sound emission Romance linguistics 2006: S elected papers from the 36th linguistic symposium on Roma nce Languages (LSRL), N ew Brunswick, March April 2006, 219 233 ed. by Jos Camacho, Nydia Flores Ferrn, Liliana Snchez, Viviane Dprez, and Mara Jos Cabrera, Amsterdam, Netherlands, John Benjamins Publishing Co. M EILLET A NTOINE tion de s formes grammaticales. Scientia (Riv ista di Scie nza) 12(26).6. Reprinted in Meillet 1958: 130 48 M ELIA J. (2003). Modality Montreal: McGill Queen's University Press. M EYERHOFF M IRIAM 2009. Animacy in Bislama? Using quantitative methods to evaluate tr ansfer of a substrate feature. Variation in indigenous minority languages ed. by James N. Stanford and Dennis R. Preston, 369 396. Amsterdam, Netherlands, John Benjamins Publishing Co. M ILROY L ESLEY 1987. Language and social networks. (2nd edition). Ox ford: Basil M ILROY L ESLEY and J AMES M ILROY 1992. Social network social class: Toward an integrated sociolinguistic model. Language in Society 21.1 26. M YHILL J OHN 1988 a The grammaticalization of auxiliaries: Spanish clitic climbing. Berkeley Linguist ics Society 14.352 363. M YHILL J OHN 1988 b Variation in S panish clitic climbing. Georgetown University Round Table on Languages and Linguistics 227 250 M YHILL J OHN 1995. Change and continuity in the functions of the American English modals. Linguistics 33.157 211. M YHILL ,J OHN 1996. The development of the American English strong obligation system American Speech ,339 388. N ARROG H EIKO 2005a. Modality, mood, and change of modal meanings: A new perspective. Cognitive Linguistics 16.677 731.

PAGE 207

207 N ARROG H E IKO 2005b. On defining modality again. Language Sciences 27(2) 165 192. N ORDLINGER R ACHEL AND E LIZABETH C LOSS T RAUGOTT 1997. Scope and the development of epistemic modality: evidence from ought to. English Language and Linguistics 12.295 317. O CAMPO F R ANCISCO 2006. Movement towards discourse is not grammaticali zation: t he evolution of claro from adjective to discourse particle in spoken Spanish. Selected proceedings of the 9th Hispanic Linguistics Symposium ed. by In Nuria Sagarra and Almeida Jacquelin e Toribio, 308 319 Sommerville, MA: Cascadilla Online: http://www.lingref.com/cpp/hls/9/paper1388.pdf O LBERTZ H ELLA 1998. Verbal periphrases in a functional grammar of Spanish Berlin: Mouton de Gruyter. P ALMER F RANK .R, 1979. Modali ty and the English modals London England : Longman P ALMER F RANK R. 2001. Mood and modality Cambridge, England: Cambridge U niversity Press. P AOLILLO J OHN 2002 Analyzing linguistic variation: Statistical mode ls and methods Stanford, CA: Center for the Study of Language and Information. P IERREHUMBERT J ANET 2001. Exemplar dynamics:Word frequency, lenition and contrast. Frequency and the emergence of linguistic structure ed. by Joan Bybee and Paul Hopper 137 157 Amsterdam Netherlands : John Benjamins Publishing Co P IETRANDREA P AOLA 2005. Epistemic modality: functional properties and the Italian system Amsterdam Netherlands : John Benjamins Pub lishing Co P LANK F RANS 1984. The modals story retold. Stud ies in Language 8 3.305 364. P OPLACK S HANA 2001. Variability, frequency and productivity in the irrealis domain of French. Frequency and the emergence of linguistic structure, 405 428. ed. by Joan Bybee and Paul Hopper, 405 28. P OPLACK S., AND E LISABETE M ALVAR 2007. Elucidating the transition pe riod in linguistic change: The expression of the future in Brazilian Portuguese. Probus 19(1). 121 169. P OPLACK S HANA AND M ARJORY M EECHAN 1998. How languages fit together in c odemixing. International Journal of Bi lingualism 2(2).127 38. P OPLACK S HANA and S ALI T AGLIAMONTE 2001. African American English in the diaspora Malden, MA: Blackwell.

PAGE 208

208 R ABADAN R OSA 2006. Modality and modal verbs in contrast: Mapping out a translationally relevant approach English Spanish. Languages in Contrast 6 2.261 306. R EAL A CADEMIA E SPAOLA : Banco de datos (CORDE) Corpus diacrnico del espaol. online: http://www.rae.es R EAL A CADEMIA E SPAOLA : Banco de datos (CREA) Corpus de referencia del espaol actua l. online: http://www.rae.es R OSSARI C ORINNE C LAUDIA R ICCI AND A DRIANA S PIRIDON 2009. Grammaticalization and pragmatics: Facts, approaches, theoretical issues Bingley, UK: Emerald Group Publishing. R UIZ DE M ENDOZA I BE Z F RANCISCO J OS & L ORENA P REZ H ERNNDEZ 2004. High level modal me tonymies in English and Spanish Jezikoslovlje 4; Special issue: How Universal are Conceptual Metonymies? ed by Klaus Uwe Panther and Linda L. Thornburg 103 120. Osijek, Croatia: Unive rsity J. Strossmayer. S ANKOFF D AVID 1988. Sociolinguistics and syntactic variation. Linguistics: the Cambridge Survey 140 61 ed. by Frederick J. Newmeyer Cambridge ,England : Cambridge University Press. S CHNEIDER E DGAR 2002. Investigating Variation an d Change in Written Documents T he handbook of language variation and change ed. by J.K. Chambers, Peter Trudgill, and Natalie Schilling Estes, 549 572. Malden, MA: Blackwell. S CHWENTER S COTT A. AND R ENA T ORRES C ACOULLOS 2008. Defaults and indeterminac y Language Variation and Change 20 1 39 S KOTAREK B EATA 1996. The category of modality in Spanish. [Categora de la modalidad en espaol] Studia Romanica Posnaniensi 21 .91 114. S ILVA C ORVAL N C ARMEN 1985. Modality and Semantic Change Historical Semantics. Historical Word Formation ed. by Jacek Fisiak, 547 572. Berlin: Mouton. S ILVA C ORVALN C ARMEN 1994. Language contact and change: Spanish in Los Angeles. Oxford, England: Oxford Univer sity Press. S ILVA C ORVAL N C ARMEN Modality in grammar and discourse ed. by Joan L. Bybee and Suzanne Fleischman, 67 106. Amsterdam Netherlands : J ohn Benjamins Publishing Co S IRBU D UMITRESCU D OMNITA 1988. Contribucin al estudio de la semntica de los verbos modales en espaol con ejemplos del habla de Madrid. Hispania 71( 1 ). 139 148.

PAGE 209

209 S MITH J ENNIFER 2000 Synchrony and diachrony in the evolution of English: evidence from Scotland. Unpublished PhD thesis, University of York. S MITH J ENNIFER M ERCEDES D URHAM & L IANE F ORTUNE 2007. Variation & Change 19( 1 ). 63 99. S MITH N ICHOLAS 2003 Changes in th e modals and semi modals of strong obligation and epistemic necessity in recent British English Modality in contemporary English ed. by Roberta Facchinetti, M anfred Krug and F rank R obert Palmer, 241 266. Berlin & New York: Mouton de Gruyter. S OLANO A R AYA J M IGUEL 1984. Modality in Spanish: an account of mood Lawrence, KA: University of Kansas dissertation S ZMRECSANYI B ENEDIKT AND L ARS H INRICHS 2008. Probabilistic determinants of genitive variation in spoken and written English. The dynamics o f linguistic variation: Corpus evidence of English past and present. ed.by Teritu Nevalainen, Irma Taavitsainen and Pivi Pahta, 291 310. Amsterdam; Netherlands: John Benjamins T AEYMANS M ARTINE 2004. DARE and NEED in British and American Present day Engl ish: 1960s 1990s. New Perspectives on English Historical Linguistics. Vol. 1: Syntax and Morphology. An investigation into the emergence and development of the verb need ed by Christian Kay, Simon Horobin and Jeremy J. Smith, 215 228. Amsterdam: Benjamin s. T AGLIAMONTE S ALI 2002. Comparative Sociolinguistics. T he handbook of language variation and change ed. by J.K. Chambers, Peter Trudgill, and Natalie Schilling Estes, 729 763. Malden, MA: Blackwell. T AGLIAMONTE S ALI 2004. Have to, gotta, must: Gramma ticalisation, variation and specialization in English deontic modality. Corpus approaches to grammaticalization in English ed. by Hans Lindquist and Christian Mair, 33 55. Amsterdam; Netherlands: John Benjamins. T AGLIAMONTE S ALI 2006 a Analysing socioli nguistic variation Cambridge, UK: Cambridge University Press. T AGLIAMONTE S ALI 2006 b So cool, right?: Canadian English entering the 21 st century. The Canadian Journal of Linguistics. 51( 2 ). 309 331. T AGLIAMONTE S ALI and A LEXANDRA D' ARCY 2007. The Mod als of Obligation/Necessity in Canadian Perspective. English World Wide: A Journal of Varieties of English 28 1.47 87.

PAGE 210

210 T AGLIAMONTE S ALI and J ENNIFER S MITH 2006. Layering, change and a twist of fate: Deontic modality in dialects of English. Diachronica 2 32.341 380 T AGLIAMONTE S ALI J ENNIFER S MITH AND H ELEN L AWRENCE 2005. No taming the vernacular! Insights from the relatives in northern Britain. Language Variation and Change 17 75 112. T AGLIAMONTE S ALI AND S HANA P OPLACK 1988. How black english past g ot to the present: Evidence from samana. Language in Society 17(4).513 533. T AGLIAMONTE S ALI AND S HANA P OPLACK 1993 The zero marked verb: Testing the C reole hypothesis. Journal of Pidgin and Creole Languages 8. 171 206. T AGLIAMONTE S ALI AND S HANA P OPLA CK 2000. The grammaticization of going to in (African American) English. Language Variation and Change 11:315 342 T IPPETS I AN 2011. Differential object marking: Quantitative evidence for underlying hierarchical constraints across Spanish dialects. Selec ted Proceedings of the 13 th Hispanic Linguistics Symposium, ed. by Ortiz Lpez 107 117. Somerville, MA: Cascadilla Proceedings Project. T ORRES C ACOULLOS R ENA C. 1999. Grammaticization, synchronic variation, and language contact: A study of Spanish progre ssive ndo constructions Albuquerque, New Mexico: University of New Mexico dissertation. T ORRES C ACOULLOS R ENA 2001. From lexical to grammatical to social meaning. Language in Society 30:443 478. T ORRES C ACOULLOS R ENA 2006. Relative Frequency in the Grammaticization of Collocations: Nominal to c oncessive a pesar de Hispanic Linguistics Symposium ed. by Timothy L. Face and Carol A. Klee, 8.37 49. Somerville, MA: Cascadilla Proceedings Project. T ORRES CACOULLOS R ENA 2011. Grammaticalization and th e Neturalization in Discourse Hypothesis The Handbook of Hispanic Sociolinguistics ed. by i n M anuel Daz Campos 148 167. Oxford, England: Blackwell. Publishing Ltd. T ORRES C ACOULLOS AND S COTT A. S CHWENTER 2004. Towards an operation notion of subjectifica tion. Presented at the 31st Annual Meeting of the Berkeley Linguistics Society California. T ORRES C ACOULLOS AND J AMES A. W ALKER 2009 On the persistence of grammar in discourse formulas: a variationist study of that Linguistics 47 1 43. T RAUGOTT E LIZAB ETH C LOSS 1989. On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65 1. 31 55.

PAGE 211

211 T RAUGOTT E LIZABETH C LOSS 1995. Subjectification in grammaticalisation. Subjectivity and subjectification, ed. by Dieter Stein and Susan Wright, 31 54. Cambridge England : Cambridge University Press. T RA UGOTT E LIZABETH AND R ICHARD D ASHER 2002. Regularity in Semantic Change. Cambridge: Cambridge University Press. T UCKER T HOMAS G. 1985. Etymological dictionary of Latin Ch icago IL : Ares Publishers, Inc. U RREA E VELYN D URN AND M ICHAEL G RADOVILLE 2006. Variation in the Future Tense of New Mexican Spanish. Albuquerque, NM: University of New Mexico. U RRUTIA H ERNAN AND M ANUELA A LVAREZ 1995. Haber and tener with the meanin g of possession in medieval spanish. RLA, Revista de linguistica teorica y aplicada 33. 209 21 V AN DER A UWERA J. AND V.A. P LUNGIAN 1998. Linguistic Typology 2. 79 124. V AN DER G AAF W ILLEM 1931. Beon and habban connected with an i nflected infinitive. English studies 13. 176 188. W RSNBY A NNA 2004. Constraints on the interpretation of epistemic modality in English and Swedish. English modality in perspective. Genre analysis and contrastive studies, ed. by Roberta Facchinetti and Fra nk Palmer, 163 190 New York, NY: Peter Lang Publishing. W EINREICH U RIEL W ILLIAM L ABOV and M ARVIN H ERZOG 1968. Empirical foundations for a theory of language change. Directions for h istorical l inguistics: A Symposium ed. by Windfred P. Lehmann and Yako v Malkiel, 95 188. Austin, TX: University of Texas Press. W ESTNEY P AUL 1995. Modals and periphrastics in English: An investigation into the semantic correspondence between certain English modal verbs and their periphrastic equivalents Tbingen: Max Nie meyer Verlag. W INFORD D ONALD 1993. Variation in the use of perfect have in Trinidadian English: A problem of ca tegorial andsemantic mismatch. Language Variation and Change 5 141 187. Y LLERA F ERNANDEZ A LICIA 1980. medievales Universidad de Zaragoza.

PAGE 212

212 BIOGRAPHICAL SKETCH Katherine Honea, raised in Fallon, NV, moved to California in 1996 to study In 2001 she gradu ated with a Bachelor of Science in b usiness a dministration with an emphasis in international marketing from California State University, Sacramento She then worked for a couple years in her field as a sales representative before embarking on a year long t rip abroad in Chile. After spending a year in South America, and taking several prerequisite courses in California, she started a graduate program in Spanish at California State University, Sacramento It was during this time she spent a semester and a sum mer studying in Mexico and Central America. After graduating in 2007 Katherine decided to pursue a doctoral degree in Spanish at the University of Florida which she received in August 2012 While at the University of Florida, she traveled to Mexico to do field work. She also took a group of students to Santander, Spain for a six week long study abroad course as well as had the opportunity to develop and teach several upper division content courses She has presented her research at several national and int ernational conferences.