Uncontracted negatives and negative contractions in contemporary English : a corpus based study

Universidade de Santiago de Compostela Facultade de Filoloxía Departamento de Filoloxía Inglesa Uncontracted Negatives and Negative Contractions in Contemporary English: A Corpus-based Study María del Pilar Castillo González 2007 Universidade de Santiago de Compostela Facultade de Filoloxía Departamento de Filoloxía Inglesa Uncontracted Negatives and Negative Contractions in Contemporary English: A Corpus-based Study Tesis Doctoral realizada por María del Pilar Castillo González dirigida por la Dra. María José López Couso Autora Visto Bueno de la Directora A AC CK KN NO OW WL LE ED DG GE EM ME EN NT TS S First of all, I would like to thank my supervisor, Dr. María José López Couso, for her patience, her support and her valuable comments during the realisation of this dissertation. I hope she is proud of this thesis because it is also the fruit of her teaching. I am also indebted to other members of staff at the Department of English of the University of Santiago de Compostela, especially to Dr. Ignacio Palacios Martínez and Dr. Belén Méndez Naya, for their always useful suggestions and comments on earlier versions of this study, which have helped me to improve my work. Also thanks to all the members of the examining board for accepting to judge the thesis and for the constructive feedback I will receive from them. Needless to say, all shortcomings in the dissertation are my own responsibility. I also wish to thank all my friends who give me their unfailing support and, day after day, trust in everything I do. In a very especial way, thanks to those with whom I have shared not only good but also bad moments, namely Blanca Abel Vázquez, Begoña Bravo López, Cristina Amaya Silva, Milagros Chao Castro, Cristina Díaz López, Mary Luz Díaz Prieto, Gabriel Dorado Barrios, Manolo González Calderón, Marga Lendoiro Esmorís, Vanesa Rodríguez Juíz, Cristina Suárez Gómez and Sonia Velille Álvarez. Last, I owe a very special debt of gratitude to all my family, especially to someone who, unfortunately, is no longer with us, but who I am sure would be very proud of this work. Thanks Mum. Research for this thesis has been funded by the Xunta de Galicia (grant numbers PGIDIT02PXIC20402PN, PGIDIT05PXIC20401PN and 2006/XA133) and the grant Patricia Shaw from the Spanish Association for English and American Studies (AEDEAN). All these grants are hereby gratefully acknowledged. C CO ON NT TE EN NT TS S Pag ABBREVIATIONS AND SYMBOLS……………………......…….………………..……i LIST OF TABLES……………………………….………………………..…………….vii LIST OF FIGURES…………………………….………..………………………..……xiii I. INTRODUCTION…………………..…………………………………….……….….1 II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH….…....…..….9 1. Some Preliminary Considerations………...…….…….……..….………………...…9 2. A Note on the History of Clausal Negation and Negative Contractions in English………..…..……………………………………...………..14 3. Uncontracted Forms vs. Operator Contractions vs. Not-contractions in PDE.…………………………………………………………….………….…...26 3.1. Full Forms vs. Contracted Forms……………………....……..…………….26 3.1.1. Contractions and Text-Types…………………….….…..…………..27 3.1.2. Contractions and Geographical Variation……….…..…..…….…….32 3.1.3. Contractions and Social Stratification……..….......…………………34 3.1.4. Contractions and Gender and Age Distinctions……..……....………34 3.1.5. Contractions and Structural Factors…………..……....……...……...36 3.2. Operator Contractions vs. Not-contractions……..….………...……………40 3.2.1. Theoretical Considerations…………..….…………………………..41 3.2.2. Some Important Constraints……………….….…………………….50 3.2.2.1. Knockout Contexts for OpeC……………....………………50 3.2.2.2. Knockout Contexts for NotC……….……….……………...54 3.2.3. Factors Conditioning the Selection between NotC and OpeC….......57 3.2.3.1. OpeC/NotC and Text-Types………….……....…………….57 3.2.3.2. OpeC/NotC and Geographical Variation…...…….…..….…59 3.2.3.3. OpeC/NotC and Gender………….….……………...…..…..63 3.2.3.4. OpeC/NotC and Structural Factors……………..……….….64 3.2.3.5. OpeC/NotC and Phonological Factors…………..………….67 4. Survey of the Forms Allowing both OpeC and NotC……….……….……...…….67 4.1. Be……………….…..………………………………………………………..68 4.2. Have…………....…………………………..……………………………...…74 4.3. Will…………………….…………………..………………………………....75 4.4. Would……….….……………………………..……………………………...77 5. Summary…………….……….…………………..…………………………..……78 III. ANALYSIS OF THE CORPUS……..……….…………....…………..…………...81 1. Description of the Corpus Used in the Present Study…………….……………….81 1.1. Written Corpora….………………….………………………..……………...82 1.2. Spoken Corpora…………...…….……………………..…………………….90 1.3. Some Problems Found with the Corpus Used………….…….……………...95 2. Negative Forms Excluded from the Present Study……….…...…………………102 2.1. Subclausal Negation……….…..………………………………....…………103 2.2. Neutralised Forms……….…..…………………………………..………….112 2.3. Negative Forms of Shall…….………...……………………………………115 2.4. Had Better and Would Rather……….….……………….………………….120 2.5. Other Exclusions……….…………………….……………………..………122 3. Negation in the Corpus……..……………...………………………..……..……123 3.1. General Data……...……………….………….………..………………...…124 3.2. Knockout Contexts…..………………….……………………..…………...127 3.2.1. Knockout Contexts for OpeC……...……...……....…………….….128 3.2.2. Knockout Contexts for NotC …………………...……..………...…141 3.2.3. Other Knockout Contexts………...……….…..…..………………..143 3.3. UncNs, OpeCs and NotCs as True Variants……………...………………...148 3.4. Analysis of the Written Corpora…………...……….……………..………..155 3.4.1. The Lancaster-Oslo-Bergen Corpus of British English (LOB)….....155 3.4.2. The Freiburg-Lob Corpus of British English (FLOB)…….…..…...184 3.4.3. The Brown University Corpus of American English (BROWN).......202 3.4.4. The Freiburg Brown Corpus of American English (FROWN)….…218 3.4.5. The Australian Corpus of English (ACE)……….…...…………….235 3.4.6. The Wellington Written Corpus (WWC)….…….…….……………251 3.4.7. Summary……..…….…………………..…………….…………….265 3.5. Analysis of the Spoken Corpora………….….…………………….…….…266 3.5.1. The London-Lund Corpus (LLC)………..……….…….…………..266 3.5.2. The Corpus of Spoken Professional American English (CSPAE).....281 3.5.3. The Wellington Spoken Corpus (WSC)…...……..…....……………296 3.5.4. Summary………....…………………………………………...…….314 3.6. Comparisons across Corpora…………….……………………..…………..315 3.6.1. Changes over Time: from the 1960s to the 1990s.……………..…..315 3.6.1.1. Diachronic Comparisons between the Written British English Corpora: LOB vs. FLOB………..….……..316 LIST OF TABLES vii L LI IS ST T O OF F T TA AB BL LE ES S Page Table 1. Types of operator negation in PDE……....….....……....…….………..13 Table 2. Biber’s (1987) frequencies of the use of contractions in AmE and in BrE in different text-types....……….…...............….....28 Table 3. Morpheme order correlations with verb and adposition o rder in Greenberg’s (1963) sample...…….….…..……….....………...44 Table 4. Percentages of contractions according to text-type in Biber et al. (1999)…………………………………..……….……...59 Table 5. Percentages of contractions of is and are in Biber et al. (1999)..…......64 Table 6. D istribution of words in each text-type in the six written corpora….…89 Table 7. D istribution of LLC text-types and word targets...….……...…..……...91 Table 8. Distribution of the selected CSPAE text-types and word targets….......92 Table 9. Distribution of WSC text-types and word targets.....…..........….….......94 Table 10. Summarised information of the nine corpora used in the present study……………………….………………………….95 Table 11. N egative shall in the corpus.…….…......…....……..….....................118 Table 12. N egation of had better and would rather...............…......…..............122 Table 13. Total number of examples in each corpora.................…...................124 Table 14. Total number of tokens f ound in the corpora according to type of operator……………………………….……………..……...126 Table 15. Distribution of KO contexts in the different corpora.......……..........147 Table 16. Number of examples in each corpora (KO contexts excluded)….…148 Table 17. Total number of examples (KO contexts excluded) according to the type of operator………………….….................….150 LIST OF TABLES viii Table 18. Distribution of tokens according to type of subject.......……............154 Table 19. U ncN/OpeC/NotC distribution in LOB according to text-type…......157 Table 20. D istribution of occurrences with the be-operator in LOB………......158 Table 21. Individual forms of the be-operator in LOB…………………….….160 Table 22. L exical be vs. auxiliary be in LOB…..…...….……………..….……167 Table 23. OpeCs and NotCs with lexical be and auxiliary be in LOB.….....….168 Table 24 Negative forms with the have-operator in LOB………………….....169 Table 25. Individual forms of the have-operator in LOB…………..……….....171 Table 26. D istribution of negative forms with the will-operator in LOB……...174 Table 27. T he operator would in LOB…………………………………...….....176 Table 28. Distribution of negative forms according to type of subject and operator in LOB………………...…..…...…..……………….…182 Table 29. UncN/OpeC/NotC examples according to text-type in FLOB…......185 Table 30. Distribution of occurrences with the be-operator in FLOB…...........187 Table 31. Individual forms of the be-operator in FLOB...………………...…..189 Table 32. Lexical be and auxiliary be in FLOB…………..……………….…..190 Table 33. OpeCs and NotCs with lexical be and auxiliary be in FLOB…...….191 Table 34. Negative instances with the have-operator in FLOB.…………...….192 Table 35. Individual forms of the have-operator in FLOB………………........194 Table 36. Occurrences with the will-operator in FLOB….…………..………..197 Table 37. Negative examples with the would-operator in FLOB………..........199 Table 38. Distribution of negative forms according to type o f subject and operator in FLOB……..……....………..……….…...201 Table 39. Instances according to text-type in BROWN………………...….......204 LIST OF TABLES ix Table 40. Examples with the be-operator in BROWN.…………………..........205 Table 41. Individual forms of the be-operator in BROWN…………………....206 Table 42. Lexical be and auxiliary be in BROWN…………….……………....208 Table 43. OpeCs and NotCs with lexical be and auxiliary be in BROWN...….209 Table 44. Negative examples with the have-operator in BROWN……….…....209 Table 45. Individual forms of the have-operator in BROWN………………....211 Table 46. Distribution of examples with the will-operator in BROWN…….....214 Table 47. Examples with the would-operator in BROWN………………….....215 Table 48. Distribution of negative forms according to type of subject and operator in BROWN……….……….……..………....217 Table 49. Full forms and contractions according to text-type in FROWN…....220 Table 50. Occurrences with the be-operator in FROWN……………….……..221 Table 51. Distribution of individual forms of the be-operator in FROWN…....222 Table 52. Lexical be and auxiliary be in FROWN………………………....….225 Table 53. OpeCs and NotCs with lexical be and auxiliary be in FROWN...….225 Table 54. Examples with the have-operator in FROWN…………………........226 Table 55. Individual forms of the have-operator in FROWN………………....227 Table 56. Negative variants with the will-operator in FROWN……….....…....230 Table 57. The operator would in FROWN………………………………….....232 Table 58. Distribution of negative forms according to type of subject a nd operator in FROWN……..…...….…………….….….……...….233 Table 59. UncNs and contractions according to text-type in ACE…………....236 Table 60. E xamples with the be-operator in ACE…..………………...….…....237 Table 61. Individual forms of the be-operator in ACE….………………...…..238 LIST OF TABLES x Table 62. Occurrences with lexical be and auxiliary be in ACE……..….…....239 Table 63. OpeCs and NotCs with lexical be and auxiliary be in ACE……..….240 Table 64. Negative forms with the have-operator in ACE…....………….…....241 Table 65. Individual forms of the have-operator in ACE………………….......242 Table 66. Negative variants with the will-operator in ACE….…………...…...245 Table 67. E xamples with the would-operator in ACE……………………...….246 Table 68. Distribution of examples according to type o f subject and operator in ACE…….…………………...….…….….…….…...248 Table 69. Tokens according to text-type in WWC……………………….…....251 Table 70. Occurrences with the be-operator in WWC…………………............253 Table 71. Individual forms of the be-operator in WWC………………….…....254 Table 72. Lexical be and auxiliary be in WWC……….…………..…….……..255 Table 73. Contracted forms with lexical be and auxiliary be in WWC….…….255 Table 74. Examples with the have-operator in WWC…………………....…....256 Table 75. Individual forms of the have-operator in WWC.………………...….258 Table 76. N egative tokens with the will-operator in WWC…..…………...…...260 Table 77. Instances with the would-operator in WWC.…………………....…..261 Table 78. Distribution of examples according to type of subject and operator in WWC……...……………….……….….…….….….263 Table 79. Full forms and contractions according to text-type in LLC…….......267 Table 80. Examples with the be-operator in LLC……..………………...….....269 Table 81. Individual forms of the be-operator in LLC………..………....….....270 Table 82. Lexical be and auxiliary be in LLC……...…..…….………….…....272 Table 83. Contracted forms with lexical be and auxiliary be in LLC………....272 LIST OF TABLES xi Table 84. Occurrences with the have-operator in LLC……..…………......…..273 Table 85. Individual forms of the have-operator in LLC………..………...…..274 Table 86. Tokens with the will-operator in LLC………..……………..…........277 Table 87. Instances with the would-operator in LLC.……..……………..........278 Table 88. Distribution o f negative forms according to type of subject and operator in LLC……..…………………..…..……….……....…280 Table 89. Negative examples according to text-type in CSPAE………....…....283 Table 90. Instances with the be-operator in CSPAE………..……………........284 Table 91. Individual forms of the be-operator in CSPAE.……..………….…..285 Table 92. Lexical be and auxiliary be in CSPAE……….……………....…….287 Table 93. OpeCs and NotCs with lexical be and auxiliary be in CSPAE……..287 Table 94. The operator have in CSPAE………..………………………...........288 Table 95. Individual forms of the have-operator in CSPAE……..………...….289 Table 96. Examples with the will-operator in CSPAE……..………………….292 Table 97. Occurrences with the would-operator in CSPAE……..………...…..293 Table 98. Distribution of examples according to type of subject and operator in CSPAE……………………...……..…………….…294 Table 99. Negative examples according to text-type in WSC……....…...….....297 Table 100. T he operator be in WSC….…….…….……………………...…....299 Table 101. Individual forms of the be-operator in WSC……….…………….301 Table 102. Lexical be and auxiliary be in WSC……..…......……………..….302 Table 103. OpeCs and NotCs with lexical be and auxiliary be in WSC......…302 Table 104. Instances with the have-operator in WSC………..………...…......303 Table 105. Individual forms of the have-operator in WSC….....……………..304 LIST OF TABLES xii Table 106. Examples with the will-operator in WSC……....….……...….......309 Table 107. Occurrences with the would-operator in WSC……..……….........310 Table 108. Distribution of examples according to type of subject and operator in WSC………………………..….……..………..…313 Table 109. UncNs in categories J, L and P in LOB and FLOB…….......….…316 Table 110. OpeCs in categories K, P, L, E and J in LOB and FLOB….…......318 Table 111. N otCs in categories B, C, K, L and R in LOB and FLOB….….....318 Table 112. Distribution of examples according to type of subject and operator in the written BrE corpora (LOB/FLOB)…...…..….360 Table 113. Full forms and contractions with the operators have, w il l and would in FROWN and CSPAE....……………………..…365 Table 114. UncN/OpeC/NotC variants in the BrE corpora according to four general textual categories……..………......……………...…379 LIST OF FIGURES xiii L LI IS ST T O OF F F FI IG GU UR RE ES S Page Figure 1. Distribution of negative forms in Kjellmer’s (1997) analysis of the LOB corpus.….....…...……..………...….…….…...…...30 Figure 2. Percentage of contractions depending on age (from Petyt 1978)...….....35 Figure 3. Percentage of NotCs in BrE texts in Hiller’s (1987) study.….…….......61 Figure 4. Percentage of NotCs in AmE texts in Hiller’s (1987) study....…….......61 Figure 5. Number of contractions of is in mid position in Hiller’s (1987) study...63 Figure 6. Number of contractions of is in positions other than medial in Hiller’s (1987) study...………………….......…..............64 Figure 7. Percentage of NotCs in Hiller’s (1987) analysis of the SEU corpus…...69 Figure 8. Percentage of OpeCs in Hiller’s (1987) analysis of the SEU corpus......69 Figure 9. Degree of formality of the texts included in the written corpora...….....84 Figure 10. N egative examples found in the corpus......….………….…….….....124 Figure 11. UncN/OpeC/NotC in the corpus.......………………….…....…....….125 Figure 12. Negative forms in the corpus (KO contexts excluded)……….…….149 Figure 13. UncNs, OpeCs and NotCs as true variants in the corpus………..….149 Figure 14. UncN/OpeC/NotC distribution in LOB………..................................155 Figure 15. UncN/OpeC/NotC distribution in FLOB……...…….........…............184 Figure 16. UncN/OpeC/NotC distribution in BROWN…….....….......................203 Figure 17. UncN/OpeC/NotC distribution in FROWN…………........................219 Figure 18. UncN/OpeC/NotC distribution in ACE………..…..................…......235 Figure 19. U ncN/OpeC/NotC distribution in WWC………..…................….......251 Figure 20. U ncN/OpeC/NotC distribution in LLC……..…………..….….…….266 LIST OF FIGURES xiv Figure 21. UncN/OpeC/NotC distribution in CSPAE………….......…...............282 Figure 22. UncN/OpeC/NotC distribution in WSC………..…...........….............296 Figure 23. Full forms and contractions with individual forms of the be-operator in LOB and FLOB………..………….…..………320 Figure 24. Full forms and contractions with individual forms of the have-operator in LOB and FLOB……….....….…...…………322 Figure 25. Contractions with be in BROWN and FROWN......……..…..............327 Figure 26. Full forms and contractions with individual forms of the be-operator in BROWN and FROWN………...………………328 Figure 27. Full forms and contractions with individual forms of the have-operator in BROWN and FROWN….………...…...……330 Figure 28. Degree of contraction in written texts according to dialect………....334 Figure 29. Distribution of negative occurrences in the six written corpora.........334 Figure 30. Negative occurrences with the be-operator in the six w ritten corpora.…............……………….....................................…..337 Figure 31. Instances with the have-operator in the six written corpora.........…..340 Figure 32. Negative occurrences with the will-operator in the six w ritten corpora……………………………………….……………...343 Figure 33. Negative examples with the would-operator in the six written corpora.….....……….......................................…...344 Figure 34. Distribution of negative occurrences in the three spoken corpora.....347 Figure 35. Degree of contraction in spoken texts according to dialect………....348 Figure 36. The be-operator in the three spoken corpora.…...............…........…..348 Figure 37. Occurrences with the have-operator in the three spoken corpora…..350 Figure 38. Instances with the will-operator in the three spoken corpora.….…...352 LIST OF FIGURES xv Figure 39. The operator would-operator in the three spoken corpora....…..…....352 Figure 40. Distribution of negative forms in the BrE corpora.....…..…......…....355 Figure 41 Full forms and contractions with individual forms of the be-operator in LOB, FLOB and LLC….…………...…………356 Figure 42 Full forms and contractions with individual forms of the have-operator in LOB, FLOB and LLC…....………....………358 Figure 43. Negative forms in the AmE corpora....………………..…..…...........362 Figure 44. OpeCs and NotCs in the AmE corpora....…………………...............362 Figure 45. Full forms and contractions with individual forms of the be-operator in FROWN and CSPAE………….…...……….…363 Figure 46. Full forms and contractions with individual forms of the have-operator in FROWN and CSPAE…...……….……….…366 Figure 47. Distribution of negative contractions and uncontracte d negatives in WWC and WSC.….…….…………………………..…..367 Figure 48. D istribution of contractions in the NzE corpora........……..…….......368 Figure 49. Full forms and contractions with individual forms of the be-operator in WWC and WSC……...….…...........…..………369 Figure 50. Full forms and contractions with individual forms of the have-operator in WWC and WSC….………...….........…….…371 I. INTRODUCTION 6 cases of subclausal negation (Section III.2.1.), neutralised forms (Section III.2.2.), negative forms of shall (Section III.2.3.) and had better/would rather (Section III.2.4.) and other exclusions not included in any of the preceding groups (Section III.2.5.). The central section of the second part of the study is devoted to the detailed analysis of the data drawn from the nine selected corpora (Section III.3.). First of all, I provide an overview of the data obtained from each of the corpora as regards the use of uncontracted negatives versus negative contractions. However, a number of examples have to be examined on their own, since interchangeability between the three alternatives of negation is not always possible. In other words, the three forms to negate are not true variants in all cases. The aim of Section III.3.2. is precisely that of examining these so-called knockout contexts. It is worth noting here that, in most of the earlier studies on the topic, these contexts which do not allow free interchangeability between the three variants at issue are not distinguished from those cases in which the variants are true alternatives. Once knockout environments have been excluded, Sections III.3.4. and III.3.5. analyse the selection of full forms, OpeCs and NotCs as true variants in each of the nine selected corpora individually. Since the distribution of the variant forms is not random, but may be determined by different conditioning factors, my analysis will pay special attention to such determinants of variation as the following: I. INTRODUCTION 7 (a) Dialect: BrE, AmE, AusE and NzE. (b) Date of compilation of the texts: 1960s versus 1990s for both BrE and AmE. (c) Text-type: the selection of uncontracted negatives and negative contractions not only in written vs. spoken material, but also taking into account the different degrees of formality of the texts included in the nine selected corpora. (d) Type of operator: be, have, will and would. (e) Type of subject: variation between full forms and contractions depending on whether the subject of the clause is a pronoun (simple or complex), a noun phrase (simple, complex or compound), an existential there or a clause. A comparison of the results obtained from the analysis of the different corpora is offered in Section III.3.6. Such a comparison allows us to investigate the differences and similarities among corpora according to the following criteria: (a) Diachronic comparisons from the 1960s to the 1990s in both written BrE (Section III.3.6.1.1.) and written AmE (Section III.3.6.1.2.). (b) Dialectal comparisons between all the written corpora (Section III.3.6.2.1.) and all the spoken corpora (Section III.3.6.2.2.). I. INTRODUCTION 8 (c) Comparisons regarding medium, i.e. written BrE vs. spoken BrE (Section III.3.6.3.1.), written vs. spoken AmE (Section III.3.6.3.2.) and written vs. spoken NzE (Section III.3.6.3.3.). In turn, in Section III.3.7., my data are compared with those obtained in related studies, such as those by Biber (1987), Hiller (1987), Kjellmer (1998) and Biber et al. (1999). Finally, chapter IV closes the present piece of research by providing the reader with a number of concluding remarks on the topic of negative contractions versus uncontracted negatives in contemporary written and spoken English. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 9 I II I. . N NE EG GA AT TI IO ON N A AN ND D N NE EG GA AT TI IV VE E C CO ON NT TR RA AC CT TI IO ON NS S I IN N E EN NG GL LI IS SH H 1. Some Preliminary Considerations Most languages over the world have a two-term system of polarity, i.e. positive versus negative polarity. This study will be concerned with the latter type. In English, negative polarity can be divided into two main different classes: clausal and subclausal negation.1 The former implies that kind of negation which affects the polarity of the clause as a whole, while in the latter the negation affects only a single word or phrase without making the whole clause negative. In accordance with this distinction, example (1b) below is the negative counterpart of (1a) and it is, thus, an example of clausal negation, where the negation affects the whole clause. By contrast, in (1c) the scope of negation is restricted to the adjective untidy; (1c) constitutes, therefore, an example of subclausal negation, the positive counterpart of (1d). (1) a. The room is tidy b. The room is not tidy c. The room is untidy d. The room is not untidy 1 These labels are used by scholars such as Huddleston (1995: 419ff) or Huddleston & Pullum (2002: 789ff), among others. Other scholars prefer labels such as ‘sentence vs. word negation’ (Klima 1964: 247), ‘nonaffixal negation vs. affixal negation’ or ‘sentence negation vs. constituent negation’ (Tottie 1991: 49; Mazzon 2004: 2) and ‘clause negation vs. local negation’ (Quirk et al. 1985: 775; Biber et al. 1999: 175). It should also be noted that Quirk et al. (1985: 775) and Mazzon (2004: 97) distinguish a third type of negation, namely ‘predication negation’, as in the following example: They may 'not go swimming [‘They are allowed not to go swimming’] (1985: 797) which is “a minor type applying only after certain auxiliaries, in which the predication is negated.” Cf. also example (43b) in Section 3.2.2.2. below. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 10 Following Klima (1964: 262ff), Quirk et al. (1985: 777f) and Huddleston (1995: 419ff), among others, two main criteria can be used to distinguish between clausal and subclausal negation: (a) While clausal negation is followed by positive tag questions, as in (2a) below, subclausal negation is followed by negative tag questions, as in (2b). (2) a. The room is not tidy, is it? b. The room is untidy, isn’t it? (b) Clausal negation is followed by negative elements such as nor/neither or either in coordinate tags, while subclausal negation is followed by positive elements such as so or too (cf. (3) below). (3) a. The room is not tidy and nor/neither is his house b. The room is untidy and so is his house c. The room is not tidy and his house either d. The room is untidy and his house too Examples (3a) and (3c) are clear instances of clausal negation, since the links in the coordinated tag are nor/neither and either respectively. By contrast, in examples (3b) and (3d), the coordinated tags exhibit so and too respectively, thus being instances of subclausal negation. Of these two types of negation, the present study will be restricted to the analysis of clausal negation. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 11 As seen in example (1b) above, clausal negation is basically expressed in English by means of the negative particle not placed after the first operator.2 Clausal negation can also be conveyed through the negative items no or never, as in examples (4a-b) below. (4) a. There is no sugar b. He has never studied French The different operators which can be followed in PDE by the negative particle not are: all the forms of the operators be, have and do (am, is, are, was, were, have, has, had, do, does and did), will, would, shall, should, can, could, must, may, might, dare, need, ought to and used to. The combination of these operators and the negator not without resort to contractions will be termed here ‘uncontracted negatives’ or ‘full forms’.3 Alternatively, negative clauses in English can show contracted forms of two different types: (a) ‘Operator contractions’,4 where, the operator is reduced and attached to the preceding word, while the negator not appears in its full form. As a consequence of this, the stress is on the negator, not on the operator. Some examples of operator contraction are the following: ’s not (for both is not 2 In the present study the term ‘operator’ will be used to refer to all types of auxiliary verbs and to the verbs be and have also functioning as lexical verbs. 3 The terms ‘uncontracted negatives’ and ‘full forms’ are, by far, the most common in the literature on the topic. However, Jacobson (1980: 50), Huddleston (1995: 420) and Huddleston & Pullum (2002: 799) refer to uncontracted negatives as ‘main-verb negation’, ‘analytic negation’ and ‘analytic primary negation’ respectively. 4 Biber et al. (1999: 165) use this label in their grammar. Alternative terms are: ‘auxiliary reduction’ (Zwicky 1970: 327), ‘auxiliary-(verb) contraction’ (Pullum & Wilson 1977: 743; Dillard 1980: 386; Jacobson 1980: 50; Anderwald 2002: 72; Tagliamonte & Smith 2002: 251; Yaeger-Dror, Hall-Lew & Deckert 2002: 83), ‘auxiliary cliticisation’ (López-Couso 2006a), ‘Not forms’ (Hiller 1987: 534ff), ‘verb contraction’ (Kjellmer 1998: 155; Westergren 1998: 34ff). Quirk et al. use the term ‘AuxC’ (1985: 1595ff) besides ‘contracted verb form’ (Quirk et al. 1985: 141f). The latter label is also used by scholars such as Bolinger (1972: chapter 5) or Leech & Svartvik (1975: 246). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 12 and has not), ’d not (for both had not and would not) or ’ll not (for both will not and shall not). (b) ‘Not-contractions’,5 where, the negative particle not is reduced to n’t, losing its stress and being fused with the operator.6 Instances of this kind of contraction are aren’t, haven’t or hadn’t. Table 1 below shows the distribution of uncontracted negatives, operator contractions and not-contractions in PDE for the operators mentioned above. Notice that some of them do not allow the three alternative ways of negation. As can be seen, the not-contraction type is, in principle, allowed by the 16 operators, while operator contraction is only admitted by the operators be (in its present forms), have, will, would and shall. This distinction proves to be particularly relevant to the present study, since our analysis will be restricted to those operators which allow interchangeability between the three kinds of clausal negation (cf. Section III.3.3. below). 5 The term ‘not-contraction’ is used by scholars such as Ali (1970: 73); Zwicky (1970: 327); Quirk et al. (1985: 1595ff); Kjellmer (1998: 155); Westergren (1996: 5ff, 1998: 34ff); Biber et al. (1999: 165); Yaeger-Dror, Hall-Lew & Deckert (2002: 83). Other labels used to refer to this type of contraction are: ‘contracted negative’ (Leech & Svartvik 1975: 246; Anderwald 2002: 72), ‘contracted negation’ (Culicover 1976: 129; Dillard 1980: 386), ‘negative contraction’ (Pullum & Wilson 1977: 743; Tagliamonte & Smith 2002: 251), ‘N’T forms’ (Hiller 1987: 534ff), ‘inflectional negation’ (Huddleston 1995: 420), ‘synthetic primary negation’ (Huddleston & Pullum 2002: 799) and ‘negative cliticisation’ (López-Couso 2006a). As shall be seen in Section III.2.4., Jacobson (1980: 50) uses the term ‘modal negation’ to refer to NotC with the semimodals had better and would rather. 6 The fact that negators are attracted to all kinds of operators is what Labov (1972: 777) calls ‘negative attraction’ (‘Negattrac’). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 13 Table 1. Types of operator negation in PDE7 OPERATORS UncN OpeC NotC BE PRESENT 1st p.sg. am not 3rd p.sg. is not 2nd p./1st & 3rd p.pl. are not PAST 1st & 3rd p.sg. was not 2nd p./1st & 3rd p.pl. were not ’m not ’s not ’re not aren’t/ain’t/amn’t isn’t/(ain’t) aren’t/(ain’t) wasn’t weren’t HAVE PRESENT 3rd p.sg. has not 1st & 2nd p./3rd p.pl. have not PAST All persons had not ’s not ’ve not ’d not hasn’t /(ain’t) haven’t/(ain’t) hadn’t DO PRESENT 3rd p.sg. does not 1st & 2nd p./3rd p.pl. do not PAST All persons did not doesn’t/(ain’t) don’t/(ain’t) didn’t/(ain’t) WILL All persons will not ’ll not won’t WOULD All persons would not ’d not wouldn’t SHALL All persons shall not ’ll not shan’t8 SHOULD All persons should not shouldn’t CAN All persons cannot can’t COULD All persons could not couldn’t MUST All persons must not mustn’t MAY All persons may not (mayn’t) MIGHT All persons might not (mightn’t) DARE All persons dare not daren’t NEED All persons need not needn’t OUGHT TO All persons ought not to oughtn’t to USED TO All persons used not to (usen’t) to 7 Forms between brackets can be used to some degree in English, although they are not accepted by all grammarians. 8 It is generally acknowledged (e.g. Quirk et al. 1985: 122) that the form shan’t for shall not is now virtually non-existent in AmE, while in BrE it is becoming rare. This is so because, even in affirmative clauses, shall is restricted to very few expressions. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 14 2. A Note on the History of Clausal Negation and Negative Contractions in English The English system of negation has undergone considerable changes from OE to PDE. These changes are normally classified according to the different periods in the history of the language or according to the different stages of what has been called ‘the negative cycle’ (cf. Mazzon 1993: 114, 2004: 5 or Anderwald 2002: 19). As Schwenter (2006) states, the term ‘cycle’ in this context was apparently first used by Östen Dahl in his 1979 article “Typology of sentence negation,” although the idea had already been formulated by Jespersen (1917: 4) in the following terms: The history of negative expressions in various languages makes us witness the following curious fluctuation: the original negative adverb is first weakened, then found insufficient and therefore strengthened, generally through some additional word, and this in its turn may be felt as the negative proper and may then in course of time be subject to the same development as the original word. According to Beukema & Mišeska (1995: 123), I will distinguish five different stages in the history of clausal negation in English. STAGE 1 The first stage corresponds to the OE period. At that time clausal negation was expressed by the negative particle ne, which was placed immediately before the auxiliary verb, if there was one, otherwise immediately before the lexical verb, as in (5a-b) below. In accordance with this, it can be said that negation in OE presents the typical pattern of OV languages, since the negator is placed before the finite verb (cf. Ohkado 1989: 1). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 15 (5) a. ic ne can secgan ‘I cannot say’ b. ic ne secge ‘I do not say’ As regards negative contractions, in OE the negative particle ne could be fused with the verbs habban ‘have’ and some forms of bēon/wesan ‘be’, together with a number of verbs belonging to the preterite-present group, namely agan ‘owe’, willan ‘will’ and witan ‘know’ (cf. Denison 1993: 449, among others). The phonological precondition for such contractions was that the form following ne began with a vowel, /h/ or /w/9 (cf. Hogg 1992: 187ff, among others). The most frequent OE verbal contractions are the following: nabban (< ne + habban ‘have’) nad (< ne + had) nagan (< ne + agan ‘owe’) nam (< ne + am) nart (< ne + art) nas (< ne + was or ne + has)10 nere(n) (< ne + were(n)) nillan (< ne + willan ‘will’) nis (< ne + is) nolde (< ne + wolde ‘would’) nytan (< ne + witan ‘know’) not (< ne + wot ‘knew’) 9 The rule, however, did not apply to other verbs beginning with a vowel, /h/ or /w/. For instance, the verb unnan ‘grant’ did not allow negative contraction with ne (cf. Warner 1993: 151). Similarly, as Levin (1958: 493) points out, the contraction *neorðan (< ne + weorðan) is not recorded in OE. 10 The form nas, derived from the fusion between ne and has, is only mentioned by Jespersen (1917: 12). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 22 STAGE 4 In the eModE period, some additional changes took place in the system of clausal negation. The first of these was supported by syntactic factors. As seen above, at Stage 3, the word order verb plus negator not was established as the norm. However, in eModE, the reverse order reappeared in the system, i.e. there was a growing tendency to place not before the finite verb (cf. Jespersen 1917: 13, 1954: 428, or Mazzon 2004: 6, among many others), with the exception of be and have, with which not continued to be placed postverbally (cf. Denison 1993: 451). In this way, the structure of English negation changed again to the typical pattern of an OV language, with the negator before the finite verb (cf. Ohkado 1989: 6ff). Given that in eModE times the pattern VO was completely established in the English language, a new change was needed to re-establish the VO pattern in the system of negation. This favoured the emergence and generalisation of the socalled ‘do-support’ in negative constructions in those cases in which no auxiliary verb was present (cf example (12a) below). (12) a. I do not say The generalisation of the ‘dummy auxiliary’ do took place in the course of the sixteenth century and it became increasingly common from the seventeenth century onwards (cf. Berndt 1989: 168ff, among others). However, the old pattern, i.e. finite verb plus not remained in the colloquial language until the late eighteenth century, specially with some high frequency verbs such as know, say, mistake, matter, and so on. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 23 As regards contractions in this period, the negator not, losing its stress and tone, came to be fused with the operator to form contractions of the type shown in (12b) below, that is, the NotC variant. (12) b. I don’t say Following Jespersen (1917: 117), most linguists agree that this type of negative contraction emerged in speech around 1600 and in writing around the year 1650 (cf. Strang 1970: 151f, Barber 1976: 254 or Haugland 1995: 170f, among others). Nevertheless, Denison (1993: 309) and the evidence provided by the OED (s.vv. shall and will v. 1, 6b) show that certain forms of contractions of an operator with a following not are already recorded from the fifteenth century (cf. ex. (13) below). (13) c1420 Chron. Vilod. 2147 How […] Sathanas Dude hurre Þere lette wt alle his myút, Þat he shulnot haue come to Þat ioyfulle place ‘How […] Satanas made her let out with all his might there, that he shouldn’t have come to that joyful place’ (OED, s.v. shall v. A, II, 9b) However, forms like shulnot in example (13) above are considered by scholars such as Brainerd (1993: 177) or Warner (2005: 275f) as cases of portmanteau forms rather than instances of NotCs proper. According to Brainerd (1993: 181) the development of such forms took place during the ME period: Assuming a stress on the first syllable, it is not difficult to see that the vowel might weaken, metathesize with n and be lost some time after 1400 and, […] before 1630, when the first concrete evidences of n’t began to appear. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 24 Brainerd (1993: 176ff) also demonstrates that negative contractions in drama appeared in the early 1620s and that they were introduced “from regionally or socially nonstandard versions of the language” (1993: 184) for comic effect in the speech of bucolic characters. However, such contractions were still regarded at that time as “not quite respectable.” He also shows that the first clear evidence of the reduction of not into n’t dates from 1621 (1993: 183). This reduction was a consequence of the weakening of the vowel /o/ in unstressed position, when not was joined to the preceding operator. At this stage of the language, contractions of the operator and the subject also occurred to form the OpeC variant. Contracted forms of this kind appeared in writing a few years earlier than the NotC type. According to the OED, the OpeC of will, which is the earliest form recorded, dates from 1610 (cf. OED s. vv. will v. 1, 5b), as shown in example (14) below. However, scholars such as Jespersen (1917: 117) mention that the first occurrences of the OpeC in affirmative clauses date from the late sixteenth century, in Shakespeare’s writings. (14) 1610 Heywood Gold Age i. i I’le not kill my part (OED, s.v. will v.1 5b) STAGE 5 As already mentioned for the previous stage, from the seventeenth century onwards, the only productive pattern of negation in the language has been that of operator plus not (cf. Denison 1998: 92ff). Since the beginning of lModE the use of contractions with not, as in example (12) above, has been increasing, though in the eighteenth century contractions were still considered unsuitable for writing, II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 25 being described in grammars and spelling books of the time as improper, vulgar, colloquial, inelegant, barbarous, harsh and ungrammatical (cf. Sundby et al. 1991: 162ff). As Sundby et al. (1991: 161) mention […] our grammarians are generally sceptical of contracted forms, they do not reject them indiscriminately. Their acceptability is often seen to depend on genre and style, thus concessions are made, although reluctantly, to their use in poetry and familiar style. Moreover, taking into account the individual forms, not all of them behave in the same way, since the contracted forms I’m and I’ll seem to be preferred to their uncontracted variants I am and I will, respectively (cf. Sundby et al. 1991: 161). Nevertheless, it was in the course of the nineteenth century that the use of contractions became generally accepted (cf. Brainerd 1993: 191). This is corroborated by López-Couso’s (2006a) article, where she analyses the BrE data included in the four 50-year subperiods corresponding to the eighteenth and nineteenth centuries in the ARCHER corpus, comprising different kinds of texts, such as drama, fictional dialogue, medical research articles, personal letters, and so on. She comes to the conclusion that full forms predominate over contractions in all text-types analysed, although contracted forms have been increasing in frequency in the course of time, their rise being “especially noticeable in the second half of the nineteenth century” (2006a). She also mentions that contractions are particularly frequent in those written texts more closely related to the spoken language, such as drama and fiction. As regards the distribution of the two kinds of contractions, the data in ARCHER also show that there is a clear predominance of OpeCs over NotCs in all the text-types analysed with the operator be, while with the have-operator, the reverse holds true. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 26 In PDE, the use of contractions tends to be associated with the spoken language. In writing, although their frequency has increased considerably, the use of contractions is not well accepted by all grammarians. As Mazzon (2004: 105) mentions, “they are still normally avoided in the higher registers.” However, the selection of contracted forms may, in principle, be “left largely to the caprices of individual authors and printers” (Potter 1969: 143). 3. Uncontracted Forms vs. Operator Contractions vs. Not-contractions in PDE The aim of this section is to provide some information about different intrasystemic and extrasystemic factors which, either alone or in combination,18 determine the selection of full negative forms vs. contracted ones, on the one hand (cf. 3.1. below), and on the other, the choice between NotC and OpeC (cf. 3.2. below) for those operators which allow the two types of contractions outlined in Section 1 above. 3.1. Full Forms vs. Contracted Forms Uncontracted and contracted negative forms are not randomly distributed in PDE. As seen in Section 2 above, their choice is determined, in principle, by a number of factors, such as text-type, dialect, social stratification, gender and age distinctions and various structural factors. Moreover, the selection between full forms and contractions is also related to questions of emphasis. Thus, according to 18 As Rydén (1979: 12, fn 2) states, “factors usually operate in combination.” II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 27 Bolinger (1972: 116f) and Jespersen (1954: 437), the uncontracted negative is much more emphatic than the contracted counterparts. In view of this, example (15a) is more emphatic than either (15b) or (15c). (15) a. I would not do that b. I’d not do that c. I wouldn’t do that Huddleston & Pullum (2002: 800) also mention the relevance of emphasis for the choice between the three variants under consideration. They claim that, in many conversational contexts, full forms sound unnatural and, thus, contractions seem to be preferable (cf. Section 3.1.1. below). However, in order to put emphasis on the negative particle not, then, uncontracted negation is selected instead. In the following sections those features which condition the selection of uncontracted forms and contracted ones will be analysed in detail. 3.1.1. Contractions and Text-Types The selection of uncontracted and contracted negative forms depends, to a large extent, on the type of text in which they appear. Scholars such as Fries (1940: 8) or Forsheden (1983: 36) claim that the frequency of contractions is higher in spoken texts, such as Conversation, than in written ones. Thus, as mentioned in Section 2 above, in contemporary English, contractions are considered unsuitable for writing and are specially used in colloquial speech. Linguists such as Biber (1988: 243), among others, have explained their frequent use in conversation as a consequence of fast and easy production. The degree of formality of a given text II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 28 also conditions the selection of contractions or uncontracted forms, since the former are more commonly used in informal registers (cf. Leech & Svartvik 1975: 207; Quirk et al. 1985: 123ff; Kjellmer 1998: 160f; Yaeger-Dror, Hall-Lew & Deckert 2002: 80ff; Castillo-González 2003: 677ff). Biber (1987) studied the use of negative contractions in different types of texts taken from the Lancaster-Oslo-Bergen Corpus (LOB) and the BROWN Corpus. These corpora contain written BrE and AmE texts, respectively, from the 1960s, but with different degrees of formality. For his analysis, Biber selected nine categories out of the 15 contained in these corpora. I have included the data he obtained from his analysis in Table 2 below. Table 2. Biber’s (1987) frequencies of the use of contractions in AmE and in BrE in different text-types19 BRITISH ENGLISH TYPE OF TEXT AMERICAN ENGLISH 1.8 0.8 1.4 2.6 0.9 0.0 0.1 11.2 19.0 Press Editorials Hobbies Popular lore Belle lettres Official documents Academic prose General fiction Romantic fiction 21.1 17.1 15.3 16.6 14.9 8.3 7.0 28.6 46.1 The most immediate insight to be gained from the data in Table 2 is that AmE makes use of negative contractions to a greater extent than BrE in all text-types analysed, “apparently because of greater attention to grammatical prescriptions by 19 It should be noted that Biber (1987) uses mean frequencies. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 29 British writers” (Biber 1988: 243) (cf. Section 3.1.2. below). Moreover, both in BrE and in AmE, contractions are very scarcely found in Official documents and Academic prose. However, AmE uses more negative contractions in texts of this kind than BrE, negative contractions in BrE Official documents and Academic prose being practically non-existent. By contrast, Romantic fiction favours the use of negative contractions, not only in AmE, where the frequency of contractions is the highest among all the different texts included in Table 2 above, but also in BrE. Biber’s findings are corroborated, in general, by the data obtained in Castillo-González (2003: 678ff) for the operator be in AmE and BrE from the 1960s, since the former variety favours the use of contractions to a greater extent than BrE. In this study, the data are taken from all categories of the LOB and BROWN corpora. As in Biber’s analysis, the most formal text-types, such as Miscellaneous (Cat H) or Learned and scientific writings (Cat J), are those in which contractions occur less frequently, while the most informal categories favour the occurrence of contracted forms. However, in Popular lore (Cat F), General fiction (Cat K) and Romance and love story (Cat P), Castillo-González’s data for be differ from those of Biber’s, since contractions are more frequent in BrE than in AmE in these text-types. Kjellmer (1998) also investigated the use of contractions in written BrE texts from the 1960s, as represented in the LOB corpus. In Kjellmer’s data, II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 30 uncontracted forms are preferred to contracted ones, as illustrated in my Figure 1 below (Kjellmer 1998: 180). Figure 1. Distribution of negative forms in Kjellmer’s (1998) analysis of the LOB corpus20 He also mentions that contractions are practically non-existent in the most formal categories (Categories H and J) (only one example of NotC in Cat H and four instances of OpeCs vs. 16 occurrences of NotC in Cat J), while, in the fictional text-types (from category K to R), they are rather frequently used, with the exception of Science fiction (Cat M), where the number of contracted forms is lower than in the other fictional categories. In this respect, as Kjellmer (1998: 171) mentions, Cat M “has close affinities” with the Learned and scientific writing category (Cat J). As Gunn (2006) states, “[s]cience fiction works differently from other written categories” due to the difficulty to read texts of this kind. The language used in such texts is rather technical, as shown in example (16) below. (16) monopole magnet mining operations in the outer asteroid belt of Delta Cygni (cited by Gunn 2006; taken from Samuel R. Delany’s MLA presentation) 20 The normalised frequencies given in brackets are mine. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 31 In the non-fictional category F (Popular lore), by contrast, the number of contracted forms, namely NotCs, is the highest among the most formal categories. Once again, the degree of formality of the texts proves to be a crucial factor to be borne in mind when studying the variation between different types of negative forms.21 Biber et al. (1999) have also analysed the use of negative contractions in a selection of texts from the LSWE Corpus, including different registers, such as Conversation, Fiction, News and Academic prose. These authors come to the conclusion that uncontracted negatives are virtually the only choice in Academic prose, while NotCs are more frequently used firstly in Conversation, secondly in Fiction and thirdly in News (1999: 159ff). Similarly, OpeCs are most likely to occur not only in Conversation, but also in written registers with some connotations of the spoken style, such as Fiction writings. As Biber et al. (1999: 1129) explain, “the common occurrence of contractions in fiction and (to a lesser extent) in news can be largely explained by the direct reporting of spoken discourse in those registers.” Huddleston & Pullum (2002: 91; 800) also mention the choice between contracted and uncontracted negative forms in relation to the type of text. They claim that “contractions are felt to be informal, and are generally avoided in the most formal styles, especially in writing” (2002: 21), not being used, therefore, in “solemn contexts or in some kinds of written (especially published) language” 21 A detailed analysis of the selection between OpeCs and NotCs in Kjellmer (1998)’s study in relation to text-type will be provided in Section 3.2.3.1. below. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 38 (1998: 158) mentions, “auxiliaries are much more frequent than lexical verbs” and therefore, more likely to be affected by reductive processes. Kjellmer (1998: 158) also adds that “the more frequent an auxiliary is, the more often it is contracted.” This accounts for the proportional differences of contracted forms with different verbs. Thus, for instance, in Kjellmer’s data, the auxiliary be, which is the most frequent, is contracted 23% of all its occurrences, while contracted forms with have represent only 14% of the relevant forms. Nevertheless, the general preference for contractions with be and have as auxiliaries discussed by Kjellmer (1998) is not accepted by all scholars. Thus, Philips & Reynolds (1987) maintain that with be contractions are more common when it functions as a main verb than when it is an auxiliary. Moreover, in Biber et al.’s (1999: 1129) data “contractions of be are more strongly associated with the copula and progressive constructions than with passive.”24 Finally Westergren’s (1998: 148ff) data for the forms are not, is not, was not and were not also show that the proportion of contractions with be as a copular verb is higher than with be as an auxiliary. 4. Another important factor which is said to condition the selection of one negative form over its competitors is the type of subject used in the clause: while pronoun-subjects favour the occurrence of contractions, as in (19a) 24 It should be noted, however, that Biber et al.’s (1999: 1129) data apply only to OpeCs of be and have in both affirmative and negative forms. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 39 below, full forms are preferred with more complex subject types, as in (19b) (cf. Greenbaum 1977: 99; Kjellmer 1998: 161; Biber et al. 1999: 1129, among many others). Once again, such a distinction is related to the notion of frequency discussed above: since pronominal subjects tend to be far more numerous than nominal subjects, contractions tend to be favoured with subjects of the former kind.25 (19) a. You wouldn’t do that b. Your brother and your sister would not do that However, the preference for contractions with pronominal subjects does not hold true with all pronouns. Kjellmer (1998: 161f) demonstrates in his study that contractions are more frequent after certain pronouns, for example I and you (41% and 49%, respectively), while with other forms, such as they the frequency of contractions is considerably lower (only 9%). In this connection, Krug’s (1998: 294) string frequency, that is, the frequency with which two or more items co-occur in a text, also plays an important role in the selection of contractions. In his analysis of the London Lund Corpus (LLC) and the spoken component of the Bank of English Corpus (BEC), Krug (1998: 294) observed that potentially contractible forms such as I have, which are far more numerous than combinations such as where have, are those which favour the use of contractions to a greater extent (77.9% in LLC and 82.8% in BEC for the 25 The higher frequency of contractions with pronominal subjects also holds for the lModE period, as shown in López-Couso (2006a). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 40 former string vs. 20% and 16.7% for the latter).26 Krug then comes to the conclusion that “the more frequent a given string is in either database, the higher is its contraction ratio” (1998: 294).27 Moreover, Biber et al.’s findings show that the frequency of contracted forms after pronominal subjects also depends on the type of texts, since in Fiction and News the proportion of contractions with subjects of this kind amounts to 20%, while in Conversation it represents only 10%. 3.2. Operator Contractions vs. Not-contractions Of the two types of negative contractions available for operators in PDE, NotC appears to be more frequent than OpeC (cf. Quirk et al. 1985: 123). As shall be seen in Section 4 below, NotCs are generally favoured with those operators which allow both types of contractions, except for the different forms of the verb be, with which OpeCs seem to predominate (cf., among others, Forsheden 1983: 27f; Quirk et al. 1985: 1596f; Biber et al. 1999: 166, 1129ff; Kortmann 2003: 70f). Thus, (20) a. I haven’t rather than I’ve not b. I won’t rather than I’ll not c. I wouldn’t rather than I’d not BUT d. They’re not rather than They aren’t 26 Notice that Krug’s (1998) analysis is not done only with negative contractions but also with positive, that is, he studies the use of the contracted forms such as ’ve, ’s, ’m or ’re with the preceding subject. 27 Other scholars who have studied the effects of frequency on PDE contractions are, among others, Biber et al. (1999: 166f) and Bybee (2003: 617). López-Couso (2006a; and 2007) studies frequency effects in earlier stages of the language. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 41 In my analysis of the selection between the two contracted types, the following topics will be discussed: (a) relevant theoretical considerations (cf. Section 3.2.1.), (b) the different constraints which block the selection of one of the two contracted types (cf. Section 3.2.2.) and, finally, (c) the factors which condition the choice of one of the two contracted forms (cf. Section 3.2.3). 3.2.1. Theoretical Considerations The aforesaid general preference for NotCs in English can be explained as a result of the cross-linguistic predominance of suffixation over prefixation. Most languages of the world seem to have a basic word order, that is, the normal order of elements in an ‘unmarked’ clause. This basic word order is established according to the way in which the three major components of clause structure, subject, verb and object, are organised. This notion was introduced by Greenberg in 1963, in his search for those properties which are common to all human languages by comparing data from a number of them. Greenberg established a series of linguistic universals based on some correlations between basic word order patterns and the order of elements within the NP and the VP, including the order of adjectives and relative clauses with respect to nouns, of auxiliaries relative to main verbs, of prepositions relative to nouns, as well as the placement of negative particles in relation to the finite verb and the use of affixes with respect to stems. According to Greenberg (1963), there are six possible types of languages: SVO, SOV, VSO, VOS, OSV and OVS, which can be simplified into VO and OV languages taking the order of verb and object as the central II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 42 criterion. As regards the position of negative items, negative particles tend to precede the finite verb in OV languages, while in VO languages they tend to be placed after the finite verb. For years linguists have studied the frequency of affixes in different languages and most of them agree that there is a greater tendency for suffixation over prefixation in the languages of the world. The first scholar who explored the predominance of suffixes over prefixes was Greenberg in his 1957 article “Order of Affixing: A Study in General Linguistics.” Later the topic has been examined in detail by linguists such as Cutler, Hawkins & Gilligan (1985), Hall (1988) or Hawkins & Cutler (1988). The tendency for suffixation had already been observed by Sapir (1921: 67), who wrote that “of the three types of affixing — the use of prefixes, suffixes and infixes — suffixing is much the commonest” (reference from Bybee, Pagliuca & Perkins 1990: 1). In order to prove that this is so, Cutler and his associates distinguish between lexical material, i.e. those lexemes which have a meaning of their own and cannot be divided into smaller units, and grammatical material, that is, those lexemes which need to be joined to other lexemes in order to have a meaning. Cutler and his colleagues argue that the optimal position for grammatical categories is after the lexical category which they modify. As a consequence of this, the process of affixation is considered to be phonological in nature, because grammatical categories involved in the process lose their independence, their stress and even, in certain cases, some phonological material (consonants or vowels). This weakening of grammatical categories II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 43 makes possible their fusion with other elements in the clause. For instance, operators can fuse with unstressed pronouns, yielding suffixes, as in (21) below. (21) They have > They’ve As already mentioned, grammatical categories tend to be fused with the lexical categories which precede them, and not with those which follow them. This might account for the overall tendency for suffixes over prefixes. However, the distribution of prefixes and suffixes in a given language is related to the typological group to which the language in question belongs and is determined by two principles, following Cutler, Hawkins & Gilligan (1985: 727): 1. First principle: while prefixes occur productively only in languages where prepositions precede noun phrases (Pr + NP), i.e. VO languages, suffixes, by contrast, occur in languages where NPs precede postpositions (NP + Po), i.e. OV languages. 2. Second principle: the use of suffixing is favoured in both VO and OV languages. In his 1963 study of 30 languages, Greenberg observed that there is a correlation between basic word order and morpheme order, as summarised in Table 3 below (Cutler, Hawkins & Gilligan 1985: 728). These data confirm the aforementioned overall preference for suffixation. In this respect, Cutler, Hawkins & Gilligan (1985: 723ff) offer an explanation based on the fact that words, like sentences, are processed from left to right, and argue that such a preference may II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 44 be explained because of the importance of the stem, which is considered the most ‘salient’ element of the word and is, therefore, placed before the grammatical elements. In this connection, Bybee (1985: 36f) argues that the combination between stem and affix represents a coherent conceptual unit, since affixation is a process by which a stem and a grammatical element come to be treated as a single word. Table 3. Morpheme order correlations with verb and adposition order in Greenberg’s (1963) sample (from Cutler, Hawkins & Gilligan 1985: 728) Prefix only Both Suffix only Overall morpheme order and verb position VO OV 1 0 16 1 2 10 Overall morpheme order and adposition order Pr + NP NP + Po 1 0 15 2 0 12 The correlations with word order which appear to be possible between lexical and grammatical material are the following: Possessives tend to be attached to Nouns, Negation is attached to Verbs (operators), Aspect affixes, Voice affixes and also Person-marking affixes are attached to Verbs. Therefore, the fact that negation is attached to verbs may go a long way towards explaining the preference for NotCs over OpeCs mentioned above. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 45 However, an obvious question arises here, namely, is the reduced form n’t really an affix? This is a controversial issue, which will be discussed in more detail in the following paragraphs. The reduced form n’t has been characterized in the literature on the topic in different ways. Thus, for instance, Quirk et al. (1989: 374) or Tottie (1991: 8), among others, refer to n’t as a clitic. Alternatively, contracted not has been considered an affix by scholars such as Zwicky & Pullum (1982: 14) or Huddleston & Pullum (2002: 91, 801, 1610). A clitic is normally defined as: an item which exhibits behaviour intermediate between that of a word and that of and affix. Typically, a clitic has the phonological form of a separate word, but cannot be stressed and is obliged to occupy a particular position in the sentence in which it is phonological bound to an adjoining word (Trask 1996: 46f). In English, for instance, auxiliary verbs like am, is, are, have, has, had, will, would and shall are considered clitics, since their reduced forms, ’m, ’s, ’re, ’ve, ’s, ’d, ’ll, ’d and ’ll respectively, are attached to a preceding word, their host. In contrast with this characterization of clitics, affixes are usually defined as those morphemes which are added or attached to roots to produce word-forms (i.e. inflectional affixes), for instance, the -s of the third person singular added to English verbs in the present indicative, or to produce a new lexeme (i.e. derivational affixes), for example, the affix -ly added to some adjectives to form adverbs (cf. Trask 1996: 46f). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 46 Although clitics and affixes are, therefore, types of bound morphemes which can be attached to other morphemes, each class has a number of distinctive features of its own. The following paragraphs will be devoted to the main differences between clitics and affixes. As will be seen below, while the combinability between clitics and lexemes is governed by syntactic considerations, the conditions governing the combinability between affixes and stems are of a quite different sort, namely morphological and/or lexical. Following Zwicky & Pullum (1982: 3ff), four basic criteria are used here to distinguish between clitics and affixes: 1. Affixes are attached to lexical categories, such as verbs, nouns and adjectives. Thus, they exhibit a high degree of selection with respect to their stems. On the contrary, clitics present a low degree of selection, since they can be attached to all kinds of words, as shown in (22) and (23) below.28 (22) The person I was talking to’s going to be angry with me (preposition) (23) The ball you hit’s just broken my dining room window (verb) 2. Clitics tend to appear in the same position as their corresponding nonreduced counterparts, while affixes do not. Compare (24) with (22) above. (24) The person I was talking to is going to be angry with me This criterion, however, is not valid for all languages, since some languages have specific clitic positions. For instance, in West Greenlandic 28 Examples taken from Zwicky & Pullum (1982: 5). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 47 (example (25) below) clitics follow the first sentential constituent (Bauer 1988: 100). (25) tassa·gunq kangirlussuatsia·kkut umiar·passuit ilummukaa·pput that is·clitic (place name)·case-marker boat·many move-inwards·3rd person indicative ‘Well, it is said many boats came inland by way of K (fjord)’ 3. Clitics can join freely to words already containing clitics, while affixes cannot. See, in this connection, example (26), taken from Zwicky & Pullum (1982: 7), where the clitic ’ve is attached to the clitic ’d. (26) I’d’ve done it if you’d asked me 4. Morphophonological and semantic idiosyncrasies are characteristic of affixes but not of clitics. This means that hosts are unaffected by clitic groups. For instance, I would like and I’d like have the same meaning and the pronoun I is unaffected by the clitic ’d, i.e. it shows the same form in both cases. By contrast, the English affix for the plural -(e)s, for example, changes both the meaning and also the form of the stem to which it is attached (e.g. book vs. books). Once the different criteria for the identification of clitics and affixes have been stated, it is time to turn to the case of the contracted negator n’t. As mentioned above, some linguists, such as Quirk et al. (1989: 374) or Tottie (1991: 8), view n’t as a simple clitic, since it derives from the full form not. In this respect, the contraction of not into n’t is analogous to the contractions of II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 54 (40) a. But two-storey brick house, built with red clay dug from a nearby creek, is not the farm's original homestead which was on the banks of the Helena River. (ACE E07b) b. But Bush said he is “not prepared to buy into those statistics.” (FROWN A10 207-209) 3.2.2.2. Knockout Contexts for NotC In spite of the overall predominance of NotC over OpeC mentioned in Section 3.2. above, NotC is also limited by a number of restrictions. Together with the KO contexts already mentioned for NotC, namely inverted conditionals (cf. example (35c) above) and typographic blockings (cf. (40b)), the following are the most important constraints on the formation of NotC discussed in the literature on the topic: (a) NotC can only occur with operators, in contrast to OpeC which can be attached to words of any category (prepositions, nouns, lexical verbs and so on) (cf. Zwicky & Pullum 1982: 4f). Consider in this respect example (22) above, repeated here as (41) for convenience, and (42) below. (41) The person I was talking to’s not going to be angry with me (preposition) (42) a. I won’t accept that proposition (operator) b. *I acceptn’t that proposition (finite verb) In example (41), the contraction of the be-operator, ’s, is attached to a preposition, yielding a completely grammatical sentence. By contrast, contracted n’t cannot be attached to a preposition or to a main verb, as in (42b). It can only be contracted with a preceding operator, as in (42a). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 55 (b) NotC is only possible when the negative particle not modifies the operator and not when it modifies the following VP (cf. also Huddleston & Pullum 2002: 805). Therefore, we would expect NotC in a sentence like (43a), but not in one like (43b) (examples taken from Radford 1988: 67): (43a) The President <could not> ratify the treaty (43b) The president could <not ratify the treaty>34 In example (43a) the operator could and the negative particle not form a single syntactic unit, thus allowing NotC. By contrast, in example (43b) not is attached to the main verb ratify and, therefore, cannot be fused with could. This implies a difference in meaning with respect to the affirmative sentence in (44) below. The meaning of example (43a) is: ‘It would not be possible for the President to ratify the treaty’, while (43b) means: ‘It would be possible for the President not to ratify the treaty’. 34 As seen in footnote 1 (Section II.1. above), this is called ‘predication negation’ by Quirk et al. (1985: 775). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 56 (44) The President could ratify the treaty Thus, the negative counterpart of (44) is that found in example (43a), not the one in example (43b). (c) In those cases in which the VP contains more than one operator, not can only be contracted with the first one (cf. Culicover 1976: 121; Quirk et al. 1989: 374). For instance, not can contract in (45a), but not in (45b), although the contexts for contraction seem identical. (45) a. They have not been working = They haven’t been working. b. They must have not been working ≠ *They must haven’t been working. In other words, although the combination have not is present in both cases, contraction is only possible in the first one, where have is the first operator within the VP. (d) NotC is not possible either when an element is inserted between the operator and the negator, thus giving way to a split verb phrase, as in example (46) below, taken from Castillo-González (2003: 676). (46) The nature of the junction with the overlying Purbeck has been much discussed since then, and is still not settled. (LOB J11 37-J11 39) II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 57 3.2.3. Factors Conditioning the Selection between OpeC and NotC Once the different constraints for the selection of one of the two contracted types have been analysed, it is time to see whether, in the remaining cases, OpeCs and NotCs are randomly distributed or not. The answer to this question seems to be in the negative. For instance, avoidance of ambiguity may condition the selection between the two negative contracted variants. As seen in Table 1 above, forms like ’s, ’d or ’ll can be the contractions for is or has, had or would and will or shall, respectively. Thus, in such cases, NotCs are preferred to OpeCs in order to avoid potential ambiguity (cf. Kjellmer 1998: 181f). As will be mentioned in the following sections, others factors, such as text-type, geographical variation, gender or structural factors, also play an important role in the selection between the two variants at issue. 3.2.3.1. OpeC/NotC and Text-types The selection between OpeC and NotC in different text-types has been analysed in several corpus-based studies, among them Kjellmer (1998) and Biber et al. (1999). As mentioned in Section 3.1.1. above, Kjellmer (1998) investigates the use of OpeCs and NotCs in all categories of the LOB corpus. As shown in Figure 1 above, NotCs clearly predominate over OpeCs in Kjellmer’s material. Nevertheless, when dealing with individual forms, he concludes that the predominance of NotCs over OpeCs is evident with all operators except be, with II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 58 which the latter type is preferred (1998: 181).35 A comparison of the data in his Tables 3 (1998: 160) and 12 (1998: 171) shows that NotCs predominate over OpeCs only in categories B (Press editorial), C (Press review), E (Skills, trades and hobbies), G (Belles lettres), H (Miscellaneous) and J (Learned and scientific writings).36 Kjellmer (1998) also notes that the frequency of NotCs is higher in the fictional categories (from Cat K to Cat R), in most of them well over 50%, than in the non-fictional text-types (from A to J), where the proportion of NotCs is lower than 32%. As seen in Section 3.1.1. above, Biber et al. (1999) have also investigated the use of OpeCs and NotCs in a selection of texts from the LSWE Corpus, concluding that both NotCs and OpeCs are most likely to occur in those categories related to the spoken register, such as Conversation and Fiction. Concerning differences between text-types, Biber’s data (his table A.8) have been included in Table 4 below.37 As shown in this table, NotCs are more frequent than OpeCs in the four kinds of texts analysed with all operators except be. With this operator the situation is reversed in favour of OpeCs in Conversation, Fiction and News, whereas the distribution of the two variants under consideration is similar in Academic texts, where contractions represent less than 2.5%. Besides, other interesting results from Biber et al.’s (1999: 1132) study are that NotCs with have, will and would are rare in Academic texts, while they are very commonly used in 35 Cf. also Hiller’s (1987) and Section 4.1. below. 36 It must be noted here that the data for Kjellmer’s OpeC contain both negative examples and affirmative VPs where the operator is contracted. 37 Table 4 includes only Biber et al.’s data for OpeC and NotC, the form ain’t and the uncontracted variants being excluded. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 59 Conversation (95% for have and will and 100% for would) and rather frequent in Fiction (55%, 80% and 60 %, respectively). Table 4. Percentages of contractions according to text-type in Biber et al. (1999) TEXT-TYPE OPERATORS OpeCs NotCs Be 70% 10% Have 5% 95% Will 5% 95% Conversation Would Less than 2.5% 100% Be 45% 25% Have Less than 2.5% 55% Will 5% 80% Fiction Would Less than 2.5% 60% Be 20% 10% Have Less than 2.5% 30% Will 5% 45% News Would Less than 2.5% 30% Be Less than 2.5% Less than 2.5% Have Less than 2.5% 5% Will Less than 2.5% 5% Academic texts Would Less than 2.5% 5% 3.2.3.2. OpeC/NotC and Geographical Variation As in the selection between full forms and contractions (cf. Section 3.1.2. above), geographical variation also conditions the choice between OpeCs and NotCs. Some scholars who have paid attention to the selection between variants regarding dialect are Hiller (1987) and Castillo-González (2003). In what follows, the results from these studies will be analysed in more detail. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 60 Hiller (1987) selected for his analysis the Survey of English Usage Corpus (SEU). This corpus is a collection not only of written BrE texts, but also of spontaneously spoken BrE ones. The texts chosen by Hiller from the SEU corpus comprise three groups of spoken categories: 1. S.1, S.2 and S.3 (S.1-3). These three categories comprise 30 texts of about 5,000 words each of informally spoken and surreptitiously recorded conversations. The social relations of the speakers involved range from ‘intimate’ (S.1) to ‘equal’ (S.2) and ‘distant’ (S.3). The register of this first category is informal. 2. S.5. It comprises spontaneously and publicly spoken texts such as radio discussions. It contains about 45,000 words distributed through nine texts. The register of this group is intermediate. 3. S.11 and S.12. As the previous text group, S.11-12 comprise publicly spoken texts but rather formal, i.e. sermons, speeches and trials. It contains about 55,000 words and the register is formal. Besides the texts selected from the SEU Corpus, Hiller (1987) also included in his study some written BrE and AmE texts: (a) A number of plays by the British author J. Osborne, namely Look Back in Anger, Inadmissible Evidence and The Entertainer. (b) A number of plays by American authors: - T. Williams’ Glass Menagerie and A Streetcar Named Desire. - A. Miller’s Death of a Salesman and All My Sons. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 61 - Th. Wilder’s Long Christmas Dinner and Our Town. From the data in all these BrE and AmE texts, Hiller concludes that both dialects show an overall preference for NotCs over OpeCs. However, while the proportion of NotCs with had, would and will are similar in the BrE and AmE samples, the number of NotCs used with am, are, is, has and have are clearly different, as my Figures 3 and 4 below illustrate. In AmE texts, NotCs with be are far more numerous than in BrE, while with the present forms of have, NotCs are more frequent in BrE texts than in AmE ones. Figure 3. Percentage of NotCs in BrE texts in Hiller’s (1987) study 0 20 40 60 80 100 ain't aren't isn't haven't hasn't hadn't won't wouldn't Figure 4. Percentage of NotCs in AmE texts in Hiller’s (1987) study 0 20406080100 ain't aren't isn't haven't hasn't hadn't won't wouldn't II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 62 The results obtained in my earlier study of the negative forms of the beoperator in written texts from the 1960s (cf. Castillo-González 2003: 677ff) confirm Hiller’s conclusion that AmE favours the use of NotCs to a greater extent than BrE. My comparative analysis of the LOB and BROWN data also allowed me to maintain that the selection between NotCs and OpeCs with the operator be is a clear dialectal marker: while in AmE NotCs constitute the predominant contracted type (13.95% for NotCs vs. 10.99% for OpeCs), OpeCs are preferred in BrE (7.34% for NotCs vs. 11.9% for OpeCs). Besides the distinction between AmE and BrE dialects, the geographical variation between northern and southern areas, in both Britain and the United States, has also proved to be an important factor which conditions the choice between OpeC and NotC. Thus, according to Trudgill (1984: 33) or Hughes & Trudgill (1996: 20f), among others, there is a general preference for OpeCs in the dialects of the North of England and in Scotland, while NotCs are preferred in the dialects of the South, except with the be-operator, with which southern speakers tend to favour the OpeC type (cf. Hughes & Trudgill 1996: 21 and Section 4.1. below). However, other scholars, such as Swan (1980: 159), affirm that in the South, OpeC and NotC with be “are probably equally frequent.” In their analysis of the variation between OpeC and NotC in British dialects, Tagliamonte & Smith (2002: 268ff) claim that, with the operators have and would, there is no distinction between geographical areas, “north and south are parallel in having” NotC. However, with the operator will the situation is variable, since the preference for OpeCs over NotCs is only found in three northern places, while in other northern II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 63 and southern areas analysed by them, NotCs are preferred (2002: 269). As for the operator be, they come to the conclusion that OpeCs are more common than NotCs in all areas, except for the form is in Maryport and Tiverton, where NotCs are preferred. Finally, according to Yaeger-Dror, Hall-Lew & Deckert (2002: 107ff), the North/South distinction for the selection of contracted types also holds true for AmE. Nevertheless, in this variety of English, OpeCs are favoured in the South, while NotCs predominate in the North (2002: 107 and 109). 3.2.3.3. OpeC/NotC and Gender Gender is another determinant of variation which can condition the choice of contracted types. Hiller (1987: 547) investigated the effects of position in the clause on the selection between OpeCs and NotCs in relation to this variable. He found that both women and men behave almost identically with regard to the use of the two types of contractions with is after pronouns and in mid position, OpeCs clearly predominating over NotCs (cf. Figures 5a-b below). As regards the occurrences of OpeCs and NotCs with is in other positions, by contrast, both men and women tend to use more NotCs, as illustrated in Figures 6a-b below, though the proportion of NotCs is higher in males than in females. Figure 5. Number of contractions of is in mid position in Hiller’s (1987) study a. MALES b. FEMALES II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 70 However, as seen in Section 3.2.3.2. above, according to Swan (1980: 159), OpeCs and NotCs with be in the South of Britain “are probably equally frequent.” Hazen (1996: 110) provides a phonotactic explanation for the preference of be for OpeCs; according to him, forms such as he’s not sound more natural than forms like he isn’t. Anderwald (2002: 92), in turn, argues that copular be has a very low semantic content, and can thus be reduced without disturbing the meaning of the sentence. Thus, OpeC is preferred with this operator to NotC. Another peculiarity of be with regard to negative contractions is that it is defective in a very peculiar way, both in its past tense and in its first person singular present. In the past tense, the paradigm of be lacks OpeCs, that is, there are no contractions for were and was with the subject.41 On the contrary, were not and was not can be fused, giving weren’t and wasn’t respectively.42 In the first person singular present, the verb be is also defective in SE, since there is no way of merging the verb and the negator, giving a NotC. However, non-standard varieties of English have a number of different NotCs for am not (cf. Chesire 1982: 54ff; Francis 1985: 145f or Anderwald 2002: 86, among others). Although there is no universally accepted colloquial form for it, typical NotCs for am not are ain’t (with different pronunciations, [eint, єnt, ∂nt]), amn’t and aren’t, depending on dialectal areas (cf. Section 3.1.2. above). The lack of a 41 Nevertheless, in some areas of the United States — e.g. Pennsylvania and the rural South — a form I’s is very frequently heard as a contraction of I was (cf. Tucker 1966: 76f). 42 Jespersen (1917: 22) mentions the existence of forms such as wa’nt, wa’n’t or warn’t in earlier English for the contraction of was not. He explains that these forms were frequently used by authors like Defoe or Howells, as in I warrant you were frighted, wa’n’t you or we wa’n’t ragged. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 71 NotC for the first person singular present indicative with be, at least in SE, could offer a plausible explanation for the aforementioned preference of OpeCs over NotCs with this operator. In other words, the stigmatised non-standard use of NotC with the first person singular may exert some influence on the whole paradigm (cf. Hughes & Trudgill 1996: 21 or Anderwald 2002: 91; also Section 3.2.3.1. above). The origin of the form ain’t is far from being clear. What follows is a summary of the two most important theories on its genesis which can be found in the literature on the topic: 1. Linguists such as Raven (1941: 59), among others, affirm that ain’t is a direct phonological derivation of aren’t. They argue that aren’t became arnt in the course of eighteenth century, which later lost its /r/ and, thus, a compensatory lengthening of the vowel yielded /æ:/. This eModE phoneme developed into the present-day diphthong /ei/ and aren’t became ain’t. 2. A different development for the origin of ain’t has been proposed by Jespersen (1917: 118ff) and Stevens (1954: 199). According to these authors, ain’t goes back to am not, which suffered a process of syncope of the vowel /o/, due to its weak character, hence becoming amn’t. Later on there was an assimilation of the first nasal, giving ann’t, and, finally, a lengthening of the vowel took place, resulting in ain’t: Am not > amn’t > ann’t > ain’t II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 72 Besides its function as the contracted form for am not, ain’t also occurs without being marked for person and number in some dialects of English around the world and may function as (cf. Cheshire 1981: 365f; 1982: 51 and Table 1 above): 1. The present tense negative form of be as copula or main verb. Thus, it can appear instead of aren’t, isn’t and am not, as in (50): (50) a. We ain’t your friend b. He ain’t your friend c. I ain’t your friend 2. The whole present tense negative forms of the auxiliary be (cf. (51) below). (51) a. You ain’t coming soon b. He ain’t coming soon c. I ain’t coming soon 3. The present tense negative form of the auxiliary have, thus replacing haven’t or hasn’t, but not of have as a full verb (cf. (52a-c) below). It is believed that, in such cases, ain’t derives from hain’t, which suffered a process of /h/-dropping (cf. Malstrom 1963: 285). (52) a. You ain’t found it b. He ain’t found it c. *We ain’t our lunch at one II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 73 4. In some varieties of Black English, ain’t is used for don’t, doesn’t and didn’t,43 as in: (53) a. I ain’t go to the cinema every week b. She ain’t go to the cinema every week c. I ain’t go to the cinema yesterday The widespread use of ain’t just discussed constitutes clear evidence of the tendency towards the simplification of linguistic systems. Its extension throughout the be-paradigm and the have-paradigm may have taken place by analogy with other negative contracted forms, such as won’t or wouldn’t, which have a single form throughout the whole paradigm (cf. Chesire 1981: 367). Another important feature of the form ain’t is that it is more frequently used in AmE than in BrE (cf. Section 3.1.2. above), while in Scotland and in Irish English the form amn’t is preferred (cf., among others, Jørgensen 1979: 40; Quirk et al. 1985: 129; Huddleston & Pullum 2002: 1611f). Besides, according to Huddleston & Pullum (2002: 1611), the use of ain’t is also a distinctive social marker, since, in BrE, it is used by working-class speakers, but not in academic discourse, while in AmE it is used by educated speakers not only in informal style, but also in writing. 43 Ain’t for didn’t is quoted in Chesire (1981: 366, 1982: 51) and in the Merriam-Webster Collegiate Dictionary (1999). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 74 4.2. Have The second operator analysed here is have in all its forms, namely have, has and had. In PDE it is possible to find negative sentences with have according to one of the following forms: UncNs (have not, has not, had not), OpeCs (’ve not, ’s not, ’d not) and NotCs (haven’t, hasn’t hadn’t). The different OpeC of the forms have, has and had, namely ’ve, ’s and ’d, seem to have first occurred in the eighteenth century (Pyles & Algeo 1993: 204), while the NotC forms appeared one century earlier, as mentioned in Section 2. In contrast to the operator be (cf. Section 4.1. above), NotCs are more common than OpeCs with the have-operator (cf. Forsheden 1983: 27f). One of the most obvious reasons for such PDE preference with have is that ’s and ’d are also contractions for other operator forms, such as the third person singular present indicative of the verb be (is) and would respectively. Besides representing is or has, the contraction ’s may occasionally stand for does in informal style (cf. Quirk et al. 1985: 132), as in the following example: (54) What’s it matter? Therefore, as seen in Section 3.2.3. above, in order to avoid potential problems of ambiguity, speakers tend to use hasn’t, haven’t and hadn’t more frequently than their OpeC counterparts. Another important characteristic of the operator have is that both types of contractions are more likely to be favoured with have as an auxiliary than with have as a lexical verb (cf. Quirk et al. 1985: 123; Sinclair 1990: 453; Kjellmer 1998: 158, 172; Biber et al. 1999: 1129, among others; also Section 3.1.5 above). II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 75 Thus, for instance, the contracted form haven’t is more likely to occur in example (55a) below than in (55b). (55) a. We haven’t found any money b. We haven’t any money As mentioned in Section 3.1.5., this preference for contractions with have as an auxiliary is probably related to a question of frequency, since auxiliaries are more frequent than lexical verbs, and, therefore, contraction is more likely to occur in the former function (cf. Kjellmer 1998: 158; Krug 1998: 288ff; 2003: 9ff). Quirk et al. (1985: 131f) also mention that in sentences such as (55b) above, the contracted form with have as a main verb, despite being “the traditional construction in BrE” is becoming “somewhat uncommon, particularly in the past tense” and is used in the most formal English. This construction tends to be replaced by a structure with the operator do,44 as in the following example: (56) We don’t have any money Moreover, in informal BrE, the have got construction (cf. example (57) below) seems to be preferred when have functions as a main verb. (57) We haven’t got any money 4.3. Will As already mentioned, will is one of those operators which can be contracted both with the subject and with the negator. In spite of this, will, as well as would (cf. 44 In AmE the have-operator functioning as a lexical verb is normally negated with the operator do. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 76 4.4. below) and have (cf. 4.2. above), favours NotC (cf. Forsheden 1983: 27f Hughes & Trudgill 1996: 21 or Kjellmer 1998: 181, among many others). However, the form ’ll not is more frequently used than won’t in some varieties of the English language, such as Scottish English or Northern English (cf. Quirk et al. 1985: 123; Lass 1987: 261f), where there is a general predominance of OpeC over NotC (cf. Section 3.2.3.2. above). Will has allowed contractions throughout the history of the English language. As mentioned in Section 2 above, in the OE and ME periods will contracted with the pre-verbal negator ne (ne + willan > nillan). From eModE onwards, will could also be contracted with the subject to form an OpeC, although at this stage contractions of this kind occurred only with subjects ending in a vowel, such as Ile (for I’ll) or youle (for you’ll) (cf. Pyles & Algeo 1993: 203f). Therefore, in eModE the OpeCs with will were usually written -le, in contrast to PDE ’ll. A notable peculiarity of will is that it shows a vocalic alternation between the affirmative form will and its NotC counterpart won’t, where a form such as *willn’t would be expected. A few words seem in order concerning the origin of this negative form. In OE the operator will had two tense stems: one for the present, namely wil-, and another for the past, namely wol- (cf. Lass 1992: 141). Most scholars (Strang 1970: 151f, Brainerd 1993: 181 and Pyles & Algeo 1993: 203f, among others) agree that the contracted form won’t derives from wolplus not (> wonnot) II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 77 and they give the sixteenth century as the point of departure for this contraction. All of them argue that, at that time, there was a transfer, first witnessed in the Midlands dialects, from the past stem wol to the present. From the Midlands the form wol spread to the rest of the dialects. Since then, wol has been used only in negative contractions. However, the OED gives evidence that the origin of won’t may be found in will + not (> winnot) rather than in wol + not. Consider, in this respect, the following example, dated 1420, with the form wynnot, which antedates by about one century the contracted forms given by Strang (1970: 151f) and other scholars as the point of departure for the NotC with will: (58) c1420 Liber Cocorum (1862) 45 Зif Þay ben harde and wynnot alye (s.v. will v.1, A, 6b)45 In view of this, it can be argued that the origin of the contraction won’t is not so clear as it may seem at first sight, although wol + not appears to be the most likely source from which won’t derives. 4.4. Would As in the case of the operators analysed so far, would also allows the two different types of negative contraction available in SE, though OpeC “is rarely found with would + not” (Biber et al. 1999: 1128). A plausible explanation for this is, once again, a question of ambiguity, since, as mentioned in Sections 3.2.3. and 4.3. above, the contracted form ’d stands for both had and would. 45 For forms like wynnot, see Section 2, Stage 4 above. II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 78 Originally, would was the past tense of will and had the same functions in the past as its corresponding present form. According to the OED, the first evidence of the OpeC form of would dates from the late sixteenth century (see example (59a) below), but unlike PDE, would was contracted as ’ld, a form which remained until the early eighteenth century when the contraction ’d was adopted (Pyles & Algeo 1993: 204) (cf. example (59b) below). (59) a. 1591 Shakes. Two Gent. iv. iii. 3 Ther’s some great matter she’ld employ me in. (OED s.v. will v.1, A, 10c) b. 1712 Steele Spect. No. 326 35 My Request to you is, that...you’d speedily afford us your Assistance. (OED s.v. will v.1, A, 10c) Thus, the OpeC with would first occurred in eModE times. Little by little, however, the NotC with would, which appeared in the seventeenth century as wou’not (Brainerd 1993: 180), became more and more frequent, until it eventually became the preferred contracted form for English speakers since the eighteenth century. 5. Summary This chapter has been devoted to an overview of clausal negation and negative contractions in English. The starting point for my discussion has been the distinction between clausal negation, i.e. that in which the negator not affects the whole clause, and subclausal negation, where only one part of the clause is affected by the negator. The establishment of such a distinction was followed by the identification of three different alternative ways of operator negation, namely II. NEGATION AND NEGATIVE CONTRACTIONS IN ENGLISH 79 uncontracted negatives (UncN), operator contractions (OpeC) and Notcontractions (NotC) (Section 1). In Section 2, the evolution of negation and negative contractions throughout the history of the English language has been delineated. The remaining sections have been concerned with the selection of full forms and negative contractions. In Section 3, I discussed some of the factors which condition, on the one hand, the choice between uncontracted negatives and negative contractions (Section 3.1.), and, on the other, the overall preference for NotCs over OpeCs (Section 3.2.). Factors such as text-type, geographical variation, social stratification, gender and age distinctions and various structural factors have been discussed in these sections in connection with the selection of variants. Finally, in Section 4, I have provided a brief description of each of the four operators allowing the three alternatives of negation under study: be, have will and would. In this way, the present chapter has paved the way for the corpus-based study on the variation between uncontracted negatives and negative contractions in contemporary written and spoken English presented in the following pages. III. ANALYSIS OF THE CORPUS 86 (62) A01 0010 The Fulton County Grand Jury said Friday an investigation A01 0020 of Atlanta’s recent primary election produced “no evidence” that A01 0030 any irregularities took place. The jury further said in termend A01 0040 presentments that the City Executive Committee, which had over-all A01 0050 charge of the election, “deserves the praise and thanks of the A01 0060 City of Atlanta” for the manner in which the election was conducted. (BROWN A) (63) A01 1 <#FROWN:A01\><h_><p_>After 35 straight veto victories, intense A01 2 lobbying fails president with election in offing<p/> A01 3 <p_>By Elaine S. Povich<p/> A01 4 <p_>CHICAGO TRIBUNE<p/><h/> A01 5 <p_>WASHINGTON - Despite intense White House lobbying, Congress has A01 6 voted to override the veto of a cable television regulation bill, A01 7 dealing President Bush the first veto defeat of his presidency just A01 8 four weeks before the election.<p/> (FROWN A) (D) The Australian Corpus of English (ACE ) This corpus comprises a collection of written AusE texts. It consists of 500 samples containing 2,000 words each. At first sight, ACE is similar to LOB, FLOB, BROWN and FROWN, but, in fact, there are some important differences, III. ANALYSIS OF THE CORPUS 87 not only as regards its internal structure, but also as regards the date of composition of the texts included and even its coding system. As seen above, the 500 samples of LOB, FLOB, BROWN and FROWN are distributed among 15 categories. However, in ACE the number of categories amounts to 17. In addition to those already present in the other corpora, two new categories are included in ACE, namely Historical fiction (Cat S) and Women’s fiction (Cat W). Furthermore, while the texts in the BrE and the AmE corpora date from the 1960s and the 1990s, ACE consists of material from the 1980s. Such a difference may be of considerable interest in intercomparisons between the six written corpora analysed in the present piece of research. Another important difference between ACE and the rest of the corpora mentioned so far is its coding system. While both the BrE and the AmE corpora show codes preceding all lines (cf. examples (60) to (63) above), ACE only gives information about the type of text and the different subsamples within each texttype at the very beginning of each extract, as example (64) below clearly illustrates: (64) <Section> <X> A: PRESS REPORTAGE </X> <X> NOTE: THIS SECTION KEYED FROM HARDCOPY} </X> <sample><X> A01 </X> <X> The Australian </X> <X> 2007 words </X> <subsample><X> A01a </X> <X> The Australian - 28 October 1986 </X> III. ANALYSIS OF THE CORPUS 88 <h> Gala Opening for extension to Qld Govt’s DP centre </h> THE Premier of Queensland, Sir Joh BjelkePetersen, didn’t disappoint the crowd at the opening of the $20 million extension to the State Government Computer Centre in Brisbane last week. (E) The Wellington Written Corpus (WWC) This corpus contains a selection of written NzE texts dated from 1986 to 1990. These texts are divided into 500 samples of about 2,000 words each, like the other written corpora mentioned so far. However, its internal structure differs from that of the LOB, FLOB, BROWN and FROWN corpora, since the 500 samples of the WWC are distributed among 10 categories (A-K), instead of the 15 categories present in the other corpora. Such a difference is related to the dearth of fictional material in NzE from the late 1980s. As can be read in the manual to the corpus: there is not the abundance or variety of genres in New Zealand writing to allow us to follow strictly the subclassification used in the LOB corpus, (…). Because not enough material was published in 1986 to fill this section, collection was thrown open to all years up to 1990. Books were collected using whatever came to hand through a search of shelves in most local libraries. A measure of the success of this method is that only 12 works of fiction were missed out of those listed in the 1987 New Zealand Books in Print index, when that became available. As in the previous written corpora, all the texts are prefaced by a number of identification codes which provide us with useful information about text-type. Moreover, within the texts themselves, there is a limited amount of symbols, such as |, which indicates a new paragraph or a new line, or ^, which stands for a new sentence, as example (65) shows: III. ANALYSIS OF THE CORPUS 89 (65) A01 001 **[001 TEXT A01**] A01 002 *<*'*4Slap in Face**' On Pay for Police*> A01 003 *<*5Police Reporter*> A01 004 |^*4Police industrial action looms closer with the A01 005 Government giving notice yesterday that it will stick to its 5 A01 006 per cent basic salary rise offer. Table 6 below shows the distribution of the total number of words included in the six written corpora used in the present study according to texttype. Table 6. Distribution of words in each text-type in the six written corpora LOB FLOB BROWN FROWN ACE WWC CAT A 88,000 88,000 88,000 88,000 88,000 88,000 CAT B 54,000 54,000 54,000 54,000 54,000 54,000 CAT C 34,000 34,000 34,000 34,000 34,000 34,000 CAT D 34,000 34,000 34,000 34,000 34,000 34,000 CAT E 76,000 76,000 72,000 72,000 76,000 76,000 CAT F 88,000 88,000 96,000 96,000 88,000 88,000 CAT G 154,000 154,000 150,000 150,000 154,000 154,000 CAT H 60,000 60,000 60,000 60,000 60,000 60,000 CAT J 160,000 160,000 160,000 160,000 160,000 160,000 CAT K 58,000 58,000 58,000 58,000 58,000 252,000 CAT L 48,000 48,000 48,000 48,000 30,000 - CAT M 12,000 12,000 12,000 12,000 14,000 - CAT N 58,000 58,000 58,000 58,000 16,000 - CAT P 58,000 58,000 58,000 58,000 30,000 - CAT R 18,000 18,000 18,000 18,000 30,000 - TOTAL 1,000,000 1,000,000 1,000,000 1,000,000 926,000 1,000,000 For my analysis, I have opted to include all the categories in LOB, FLOB, BROWN, FROWN and WWC, together with categories A-R in the ACE corpus. I III. ANALYSIS OF THE CORPUS 90 have left aside categories S and W, which appear only in ACE and, therefore, do not offer the possibility of establishing comparisons with the remaining corpora.49 Thus, the total number of words analysed from written texts in the present piece of research amounts to about 5,926,000 words in all. 1.2. Spoken Corpora (A) The London-Lund Corpus (LLC) The LLC corpus contains a selection of one hundred spoken texts of BrE compiled from 1960 to 1990. Its texts are divided into 12 categories, which represent different degrees of formality: from highly formal speech, such as physics demonstration (included under category S.10) or university lectures (included under S.12), to very informal speech, such as conversation between personal friends (included under category S.7). The total number of words in this corpus amounts to 500,000. The distribution of the texts is given in Table 7 below. All the texts are prefaced by a number of identification codes which provide information about text-type and speakers. Moreover, in all texts coded identification and other symbols are present, as shown in example (66) below. (66) 3 1a 1 10 1 1 a 20 1come in . come in - - ah good morning / 3 1a 1 20 1 1 A 11 1^good m\orning# / 3 1a 1 30 1 1 a 20 1you`re Mrs Finney / 3 1a 1 40 1 1 A 11 1^y=es# / 3 1a 1 50 1 1 A 11 1I ^\am# / 49 The total number of words selected from ACE is about 926,000. III. ANALYSIS OF THE CORPUS 91 Table 7. Distribution of LLC text-types and word targets TEXT-TYPE WORD TARGET S.1 Conversation between equals S.2 Conversation between equals S.3 Conversation between disparates 175,00050 S.4 Conversations between intimates and equals 35,000 S.5 Conversations51 65,000 S.6 Non-surreptitious conversations between disparates 45,000 S.7 Surreptitious telephone conversations between personal friends 15,000 S.8 Surreptitious telephone conversations between business associates 20,000 S.9 Surreptitious telephone conversations between disparates 25,000 S.10 Spontaneous commentary52 55,000 S.11 Spontaneous oration53 30,000 S.12 Prepared but unscripted oration54 35,000 TOTAL 500,000 (B) The Corpus of Spoken Professional American English (CSPAE) This corpus includes a selection of different types of conversations of AmE dated from 1994 to 1998. It is divided into four main categories: Math committee meeting, which contains five subcategories (5/97, 6/97, 7/97, 8/97 and 8A/97), Reading committee meeting, which comprises three subcategories (6A/97, 6B/97, 7/97), University of North Carolina meeting, divided into three subcategories (95, 50 Although in the manual of the LLC the number of words of S.1, S.2 and S.3 are grouped together (175,000 words in all), I have calculated the approximate number of words of each of the three categories. Thus, S.1 comprises 65,000 words, S.2 comprises 73,000 and S.3 comprises 37,000 words. 51 Category S.5 is divided into 13 texts: 1-7 are non-surreptitious public conversations between equals (radio discussions), 8-11 are non-surreptitious private conversations between equals, whereas 12 and 13 correspond to a committee meeting and an academic meeting, respectively. 52 This category includes the following: commentary on sports (1-4), state funeral (5), radio commentary (6-7), wild life, physics demonstrations, biology demonstration and cookery demonstration (8-11). 53 Category S.11 contains 6 texts: legal cross examination (1), dinner speech (2), radio ‘My Word’ (3), recordings in the House of Commons (4-5) and House of Lords debate (6). 54 Category S.12 is divided into 7 texts: sermons (1), University lectures (2), cases in court (3-4), political speech (5), popular lecture (6) and foundation oration (7). III. ANALYSIS OF THE CORPUS 92 96 and 97), and White House, which only includes “press conference transcripts from the White House” (Barlow 2000: 7) and is divided into six subcategories (94, 95, 96A, 96B, 97A and 97B). As can be noticed, despite being an oral corpus, the CSPAE comprises more formal texts than the other spoken corpora used in the present piece of research. Such a difference will prove relevant to the variation between the three alternatives of negation under analysis. As mentioned above, the total number of words of the CSPAE amounts to over 2,000,000. Nevertheless, in order to establish comparisons with the other corpora used in this study, I have decided to include in the analysis a selection of texts from each category, which amounts to 1,000,000 running words in all. My random selection comprises the following: subcategory 6/97 from Math committee meeting, subcategories 6A/97 and 6B/97 from Reading committee meeting, the three subcategories from University of North Carolina meeting, and, finally, subcategories 95, 97A and 97B from White House (cf. Table 8 below). Table 8. Distribution of the selected CSPAE text-types and word targets TEXT-TYPE WORD TARGET Math Com 6/97 110,000 Read Com 6A/96 100,000 Read Com 6B/97 140,000 North Carol 95 51,000 North Carol 96 45,000 North Carol 97 66,000 WH 95 100,000 WH 97A 180,000 WH 97B 208,000 TOTAL 1,000,000 III. ANALYSIS OF THE CORPUS 93 Another important difference between this corpus and the ones described so far is the scarcity of codes and other symbols. As a matter of fact, the only information provided in the texts is the identification of the speaker, which appears at the beginning of each paragraph, as shown in example (67) below, taken from WH 97a. (67) (White House press briefing by MIKE MCCURRY January 6, 1997) <SP>MCCURRY:</SP> Well, welcome to the White House briefing room in this new year, for our first briefing of the new year. Anticipating news, of which I have none, we shall go to your questions -- unless you want to know about the President’s call to Chancellor Kohl. <SP>VOICE:</SP> Sure. Yes. <SP>MCCURRY:</SP> Those of you who have followed Chancellor Kohl's travels know that he met very recently with President Yeltsin, so President Clinton took that opportunity to have a half-hour conversation with Chancellor Kohl today, a very wide ranging conversation that obviously focused on relations between the Russian Federation and the West, but touched on other subjects as well. Such a neat appearance of the CSPAE, in contrast to the other spoken corpora, is due to the fact that typical features of the spoken language, such as the use of tags (e.g. well, okay, Mmm-hmm) or pauses, have been edited out in order “to make the transcript much easier to read” (Barlow 2000: 5). (C) The Wellington Spoken Corpus (WSC) The WSC corpus contains a selection of different spoken NzE texts compiled from 1988 to 1994. The total number of words amounts to 1,000,000, distributed into 15 categories, which represent different degrees of formality: from highly formal III. ANALYSIS OF THE CORPUS 94 speech, such as Parliamentary debate (DGU), to very informal speech, such as Conversation (DPC) (cf. Table 9 below). Table 9. Distribution of WSC text-types and word targets TEXT-TYPE WORD TARGET DGB Radio talkback 80,000 DGI Broadcast interview 80,000 DGU Parliamentary debate 20,000 DGZ Transactions and meetings 100,000 DPC Conversation 500,000 DPF Telephone conversations 70,000 DPH Oral history interview 20,000 DPP Social dialect interview 30,000 MUC Sports commentary 20,000 MUJ Judge’s summation 4,000 MUL Lecture 28,000 MUS Teacher monologue 12,000 MSN Broadcast news 24,000 MST Broadcast monologue 10,000 MSW Broadcast weather 2,000 TOTAL 1,000,000 In all texts coded identification and other symbols are present, as shown in example (68) below. (68) <&>Wellington Corpus of Spoken NzE Version One</&> <&>Copyright 1998 School of Linguistics & Applied Language Studies</&> <&>Victoria University of Wellington</&> <&>side two</&> <&>0:24</&> <WSC#DGB004:0005:HS> so what HAPPENS if <.>you're</.> if you’re made redundant <,> through no fault of your own III. ANALYSIS OF THE CORPUS 95 <WSC#DGB004:0010:Z1> <.>they</.> they have a labour shortage in fact <&>clears throat and laughs</&> Table 10 below summarises the main features of the nine corpora used in my study. Table 10. Summarised information of the nine corpora used in the present study Medium Dialect Total number of words analysed Date of composition of texts LOB Written BrE 1,000,000 1960s FLOB Written BrE 1,000,000 1990s BROWN Written AmE 1,000,000 1960s FROWN Written AmE 1,000,000 1990s ACE Written AusE 926,000 1980s WWC Written NzE 1,000,000 1980s LLC Spoken BrE 500,000 1960s-1990s CSPAE Spoken AmE 1,000,000 1990s WSC Spoken NzE 1,000,000 1980s 1.3. Some Problems Found with the Corpus Used The first step in my analysis of the data from the nine corpora described in the preceding sections has been the localisation of the different variants of negation at issue. Given the size of the corpus, the computer program WordSmith Tools version 3.0 has been used in order to identify the different negative forms of the operators under investigation. The use of such tools proves to be a time-saving and labour-saving method of data collection. It must be acknowledged, however, that the automatic search of variants unavoidably makes the researchers confront III. ANALYSIS OF THE CORPUS 102 2. Negative Forms Excluded from the Present Study As mentioned in Sections II.1. and II. 4., this study is devoted to those operators which accept the three possibilities of clausal negation, i.e. the forms selected are: am not, are not, is not, ’m not, ’re not, ’s not (corresponding both to is and to has), ain’t, aren’t, isn’t, have not, has not, had not, ’ve not, ’d not (referring both to had and to would), haven’t, hasn’t, hadn’t, will not, ’ll not (both for will and for shall), won’t, shall not, shan’t, would not and wouldn’t. However, not all the occurrences of these forms recorded in the corpus have been included in the total count. Thus, for example, in (79) and (80) below, the sequences in bold type cannot be counted as occurrences of negation in the corpus, since in the former example will is not an operator but a noun, meaning ‘wish,’ and in the latter is not is part of an editorial comment and not of the spoken text. (79) She had forgotten how it felt to be touched and found desirable, to want with a will not one’s own. (FROWN K05 12-14) (80) a. it’s upstairs or next door mm <&>AM talks through next two turns and is not transcribed</&>. (WSC#DPC073:0225:JG) b. where’s my phone <&>This section is not transcribed as it is a conversation between A and the daughter</&>. (WSC#DPC169:0825:X2) 55 The number of examples excluded from the total count amounts to 953. In what follows, I will deal individually with the different types of exclusions. A first group is related to subclausal negation (Section 2.1. below); Section 2.2., in turn, 55 In examples of this kind all codes have been removed with the exception of those which are markers of editorial comments. III. ANALYSIS OF THE CORPUS 103 is concerned with some neutralised forms; Sections 2.3. and 2.4. will be devoted to the negative forms of the operator shall and to those had better and would rather, respectively, and finally, other exclusions will be analysed in Section 2.5. 2.1. Subclausal Negation One of the problems found in the analysis of the corpus data is whether to classify some examples as cases of clausal negation, and, thus, included in the study, or, by contrast, as cases of subclausal negation, where the negative particle not does not affect the whole clause but only part of it. This section will be concerned with those examples in which negation occurs precisely at subclausal level. The patterns discussed in this section include the following: (a) examples in which the negative particle not does not follow the first operator, but the second one or even the main verb; (b) instances in which the operator and the negator are separated by intervening elements, such as adverbs; (c) examples in which not takes part in coordinating correlatives; and (d) other instances of subclausal negation which cannot be included in any of the preceding groups. The total number of examples of subclausal negation in the corpus amounts to 710. (a) The negator does not follow the first operator (9 examples). Examples (81) to (87) below illustrate a type of construction in which the VP contains more than one operator, and the negator not follows the second one. (81) There would then have not been necessary the thousands of years which the ages span, and none of the “tares” would have challenged III. ANALYSIS OF THE CORPUS 104 by true seed and occupied so much of their territory. (LOB D12 3538) (82) When he remembered that he might have not signed the check, Mercer made out another for the same amount, instructing the bank to destroy the other especially if he had happened to have absentmindedly signed both of them. (BROWN G39 0730-0760) (83) They may, indeed, have not liked what they saw but they only had themselves to blame. (ACE G59) (84) No magistrate with a trace of human feeling could have not responded to the mournful look on the dog's face nor would he have dared to have failed to respond to the reaction of everyone in the court. (WWC G29 126-130) (85) oh damn should have not posted that cheque to the um tauranga district council without the parking tick in it. (WSC#DPF051:0115:GB) (86) As I said earlier, I said I don’t want to suggest there may have not been some new information, but I think it was generally consistent with what we have told you in the past. (CSPAE WH 97A) (87) I just must have not quite gotten it right. (CSPAE WH 97B) As already seen in Section II.3.2.2.2., when one or more operators are present within the VP, not can only be contracted with the first one (cf. Culicover 1976: 121; Quirk et al. 1989: 374). Therefore, in the examples above there is no possible way to contract the negator, since not is said to be attached to the following word, thus constituting a clear example of subclausal negation. Similarly, in (88) and (89) below, the negative particle not does not occur after the operator, but follows have as a main verb. Thus, once again, the negator takes part in the complement that follows it and it is not part of the VP. These two examples are, therefore, also instances of subclausal negation. III. ANALYSIS OF THE CORPUS 105 (88) those of and the majority of maoris i’m not sure this is just a generalisation here would have would have or have not that backup. (WSC#DGI 167:0155:IM) (89) The box was on the floor by the bed, and next to it the pile of unused answering tapes, all methodically labelled in Philip’s neat handwriting so that Kate should have had not trouble in picking out the right one. (FLOB L08 142-145) The constitutions of the VPs in examples (81) to (89) are shown in (81a) to (89a) below. (81a) would [...] Operator have Operator not Negator been Main verb (82a) might Operator have Operator not Negator signed Main verb (83a) may [...] Operator have Operator not Negator liked Main verb (84a) could Operator have Operator not Negator responded Main verb (85a) should Operator have Operator not Negator posted Main verb (86a) may Operator have Operator not Negator been Main verb (87a) must Operator have Operator not[...] Negator gotten Main verb (88a) would [...] Operator have Main verb not Negator (89a) should Operator have Operator had main verb not Negator (b) Pattern operator + adverbial + not (478 instances). This group of examples comprises those instances in which the operator and the negative particle not are separated by different adverbial forms, which according to Huddleston & Pullum (2002: 806), serve “to dissociate the not from the auxiliary.” Consider in this respect the examples under (90) below: III. ANALYSIS OF THE CORPUS 106 (90) a. They are, unfortunately, still not high enough. (LOB E34 79) b. Men are simply not women, not girls, not boys, and certainly not poofters. (WWC F15 072-073) c. Reliance is therefore not to be placed upon the archaeological particulars in an oral poem; no-one today would hope to discover the unmistakable ruins of Heorot or the palace of Priam. (BROWN J67 1270-1290) d. If you’re going to generate reports that are useful, because just having one number is really not very useful. (CSPAE Read Com 6A/97) Examples (90a) and (90b) above constitute clear cases of subclausal negation, where not affects only one part of the clause. Examples (90c) and (90d), on the other hand, are more difficult to classify, since not can be interpreted as negating either the whole clause or just one part of it. If the speaker makes a pause between the adverb and the negator, the latter can be seen as being attached to the following words and not to the preceding operator. In such a case, a negative tag can be added, which indicates that the example involves subclausal negation (cf. examples (91a-b) below. (91) a. Reliance is therefore not to be placed upon the archaeological particulars in an oral poem […] isn’t it? b. […] because just having one number is really not very useful, isn’t it? By contrast, if there is no such pause, the negator can be interpreted as affecting the operator, and the tag is positive, in which case negation is at clause level (cf. examples under (92) below). III. ANALYSIS OF THE CORPUS 107 (92) a. Reliance is therefore not to be placed upon the archaeological particulars in an oral poem […] is it? b. […] because just having one number is really not very useful, is it? In the case of those spoken corpora which mark the presence of emphasis on the negator in some way as, for instance, example (93) below56 (cf. also Section 3.2.1.), it is relatively easy to opt for either clausal or subclausal negation. (93) somebody who drafts a tax is simply ’not aware of (LLC S1 13 12410140 1 1 B 11 - 13 12410150 1 1 B 11) However, most of the examples found with an adverbial between the operator and the negator are difficult to classify as belonging to one or the other type, and, thus, two different analyses could, in principle, be proposed for (90c-d) and similar instances. In view of this, I elaborated a questionnaire in order to check a number of examples of this kind with some native speakers (cf. Appendix A). I selected six informants at random, who were asked to: (a) paraphrase the sentences, (b) give a possible question tag, (c) translate them into Spanish, and (d) add a clause with neither. Four of my informants considered that all the sentences provided were negated at clause level, while the remaining two assigned some of the examples to the clause level and others to subclausal level. Thus, in view of such ambiguity and following Huddleston & Pullum (2002: 806), I have decided not to include in my study all those examples which contain an adverbial between the operator and the negator, as my earlier instances (90) and (91) above. 56 The emphasis on the negator in sentence (93) is marked by the apostrophe before the negator, which indicates that not is attached to the following word. Thus, example (93) is a case of subclausal negation. III. ANALYSIS OF THE CORPUS 108 Nevertheless, I am well aware that at least some of the instances of this kind are liable to be interpreted in a different way. (c) Coordinating correlatives (cf. Quirk et al. 1985: 611) (219 examples). The third group of examples included in this section comprises, on the one hand, those instances in which the negator takes part in a coordinated structure of the type not … but, and, on the other, those cases where not is followed by only/just/merely which can also be coordinated with but (also/even). Illustrative examples of the former construction are given in (94) to (97) below. (94) I really don’t have an answer, John, on how you can do both easily, other than the fact that I would like to see some balance so that it’s just not a multiple choice and short answer, but also some constructed response. (CSPAE Math Com 6/97) (95) And this point is, of course, not that Fischer is a better sociologist than Anderson but that he has an enormous advantage over Anderson in being able to generate his own data rather than having to rely on whatever relics might happen to be extant. (FLOB J30 126-130) (96) Finally, the crux of this matter is surely not wages, but spending power. (LOB B09 71-72) (97) Functionalism as a sociological credo is, therefore, not a direct consequence of observations, but rather an indirect consequence of philosophical inference J51 0370 and judgment. (BROWN J51 034003770) In these examples, what is negated is/are the word(s) following not, and not the word(s) preceding the negator. Therefore, examples of this kind are considered cases of subclausal negation. III. ANALYSIS OF THE CORPUS 109 As regards not followed by only/just/merely, it is generally acknowledged that, in such combinations, the negation is to be interpreted as subclausal (cf. Quirk et al. 1985: 611). The sequence not + only is normally followed by the conjunction but or but also/even, thus forming a coordinating structure, as shown in (98) below. (98) From this perspective general ability ('calibre') is not only a confused and unscientific concept but also a biased and ideologically oppressive one. (WWC G54 161-164) Both words (not only) tend to form a single structure and they may be moved together to another position, without altering the polarity and the meaning of the clause. For instance, not + only may be placed in preverbal position, so that the negator cannot be fused with the operator, as shown in example (98a) below. (98) a. From this perspective not only is general ability (‘calibre’) a confused and unscientific concept but also a biased and ideologically oppressive one. Example (98) above should, therefore, be analysed in the following terms: (98) b. From this perspective general ability (‘calibre’) is <not only> a confused and unscientific concept but also a biased and ideologically oppressive one. and not as in (98c): (98) c. From this perspective general ability (‘calibre’) <is not> only a confused and unscientific concept but also a biased and ideologically oppressive one. III. ANALYSIS OF THE CORPUS 110 The assignment of instances like (98) to the domain of subclausal negation implies that, in principle, NotCs are not allowed. Nevertheless, I have recorded six examples in my corpus with the operator be in which the NotC occurs in a not only structure. These are examples (99) to (104) below. (99) Mr Kinnock isn’t only against further modernisation of the unions. (FLOB B14 85-86) (100) Sad to say, it isn’t only in the UK that paper proof of investments is disappearing. (FLOB F21 180-181) (101) The group isn’t only for people who have had breakdowns. (WWC A41 243) (102) it isn’t only lords (LLC S5 5 22 3870 1 1 a 11) (103) That this isn’t only about basketball tickets. (CSPAE North Carol 95) (104) they seem quite keen to foster golf for the girls for some reason but it isn’t it isn’t in fact only our school and it isn’t only girls but they’ve offered free coaching to anybody who’s interested between nine thirty and eleven on sunday mornings about twice a month (WSC#DPC170:1000:JU) In such instances, the negator is fused with the operator, thus indicating that not belongs to the operator and not to the following phrase. Notice also that in these cases the expression not only cannot be moved to another position (cf. example (99a), which is an ungrammatical sentence). (99) a. * isn’t only Mr Kinnock against further modernisation of the unions. All this evidence seems to indicate that, in examples (99) to (104), the negation functions at the clause level. Nevertheless, I have decided to exclude these six III. ANALYSIS OF THE CORPUS 111 examples from the total count, since they represent a very low proportion out of the total of not only structures in the corpus (3.17%, six out of 189 examples). I believe, however, that the occurrence of NotC in not only constructions may be an innovative feature which may gain ground and may eventually become more commonly used, since all the examples recorded belong to the most recent corpora. (d) Other instances of subclausal negation. (4 instances) Examples (105) to (107) below have also been excluded from the present study. As is well-known, contractions are only possible with finite verb forms. The verb form in these examples is, however, non-finite, more specifically an infinitive preceded by to, so that contractions are not allowed. (105) All the national organizations that are connected in one way or another to the National Test Panel get invitations are encouraged to have not just a national representative, but a local and a state folks participate. (CSPAE Math Com 6/97) (106) And then, most importantly, even after this draft is given, the document is given to General Administration, I encourage all of you to keep in mind Stirling’s role as a member of the System-wide committee for any input you want to have, not just about this document, but, too, directly to General Administration about what they do with this and the other 15 they’ll receive. (CSPAE North Carol 96) (107) It was just simply the building seemed to have not kept up with the vitality of the institution. (CSPAE North Carol 95) III. ANALYSIS OF THE CORPUS 118 Table 11. Negative shall in the corpus UncNs NotCs TOTAL LOB 33 5 38 FLOB 14 3 17 BROWN 18 1 19 FROWN 3 1 4 ACE 8 - 8 WWC 8 1 9 LLC 9 9 18 CSPAE 2 - 2 WSC 1 - 1 TOTAL 96 20 116 The most immediate insight to be drawn from the above table is that shall is used in all dialects and both in written and in spoken texts. As seen in Table 11, the UncN form is the predominant option (82.76%). Nevertheless, 20 examples with shan’t have been recorded in the corpus (see examples under (116) below), all of them with a first person (singular or plural) subject. This contrasts with what scholars such as Quirk et al. (1985: 122) affirm about the non-existence of the form shan’t in AmE and its sporadic occurrence in BrE (cf. footnote 8 above). Judging from the data in Table 11, the form shan’t is apparently decreasing, since the proportion of examples in the corpora from the 1960s is higher than that from the most recent corpora. However, the most striking feature from the data in this table is the relatively large number of occurrences recorded in spoken BrE. Nevertheless, all the examples with shan’t recorded in the LLC corpus appear in texts dated from 1964 to 1976, thus testifying to the decrease in the use of the form from the 1960s onwards. III. ANALYSIS OF THE CORPUS 119 (116) a. We shan’t starve overnight anyway. (LOB K17 105) b. If it’s just a professional relationship you want, then it’s a pity, because I shan’t give up. (FLOB P06 142-143) c. A Clerfayt may moon on about the face of Death in the cockpit; a Portago could say, as he did say to me, “If I die tomorrow, still I have had twenty-eight wonderful years; but I shan’t die tomorrow; I’ll live to be 105.” (BROWN C16 0570-0610) d. “I wonder whether I shan’t apply for an LST command in the Pacific,” I said moodily. (FROWN K04 136-137) e. “We shan’t be there for long, the girls and Richard and I, but my older sons are well settled.” (WWC K07 199-200) f. We shan’t see him again. (LLC S4 5 85 7290 1 1 a 11) Three of the examples with UncNs found in the LOB corpus merit special consideration. These are given under (117) below. (117) a. And if thou wilt make an Altar of stone, thou shalt not build it of hewn stone, for if thou lift up thy sword upon it, thou hast polluted it.” (LOB D04 3-5) b. Thou shalt not steal. Thou shalt not kill. Thou shalt not commit adultery. (LOB D16 181-182) c. ‘I am a lifelong vegetarian’ ‘I believe in the biblical injunction “thou shalt not kill”.’ (LOB G25 5-6) In these examples the operator shall appears in a very archaic form, shalt, typical of the ME and eModE periods for the second person singular in connection with the personal pronoun thou. Some remnants of shalt can still be found in very conservative examples in PDE, mainly in religious texts, as in (117a-b) above (Cat D). In the case of example (117c), although the form shalt occurs in a text III. ANALYSIS OF THE CORPUS 120 which belongs to Belles lettres, biography and essays (Cat G), it appears within a quotation taken from the Bible. 2.4. Had Better and Would Rather The fourth group of examples excluded from the overall count comprises the occurrences of negated had better and would rather, two expressions usually considered as semi-modals (cf. Biber et al. 1999: 73),58 since they do not fulfil all the typical characteristics of modal verbs. In principle, the two semi-modal expressions had better and would rather share with other operators the three different ways of marking negation: UncN (cf. (118) below), OpeC (cf. (119) below) and NotC (cf. (120) below).59 (118) I would rather not go into the details in Cyprus. (CSPAE North Carol 97) (119) “He’d better not!” exclaimed Wakington-Snell with surprising vehemence. (ACE L07) (120) “Wouldn't you rather be flying one of those?” he shouted across the engine roar. (FLOB N01 164-165) Grammarians like the Evanses (1957: 205) or Quirk et al. (1985: 141f), among others, maintain that both had better and would rather prefer uncontracted negation, since they are treated as a single idea and should not be split. As regards the use of NotCs with these two semi-modals, Quirk et al. (1985: 141f) and 58 Alternative labels are ‘quasi-modals’ and ‘periphrastic modals’ (Biber et al. 1999: 484) or ‘modal idioms’ (Quirk et al. 1985: 141). 59 Jacobson (1980: 50) uses the labels ‘modal negation’ and ‘main-verb negation’ to refer to NotCs and uncontracted negatives respectively with had better and would rather. III. ANALYSIS OF THE CORPUS 121 Denison (1998: 173) affirm that they are restricted to occur in interrogative negative clauses, as in (120) above. Quirk et al. (1985: 141) also mention that sometimes the choice of had better not and hadn’t better or would rather not and wouldn’t rather is associated with different meanings, as the examples below demonstrate. (121) a. Had we better not go? [= ‘Would it be advisable if we didn’t go?’] b. Hadn’t we better go? [= ‘I think we had better go; don’t you agree?’] By contrast, Palmer (1974: 160) cites “‘I hadn’t better go’ without further comment and without any mention of the more generally acceptable ‘I had better not go’” (quoted from Jacobson 1980: 50). Similarly, Jespersen (1954: 183) comments that the negative is attracted to had in the expression had better. In this connection, Jacobson (1980: 50) suggests that such an attraction may have been induced by regular questions: ‘Hadn’t you better wait?’. Finally, Palmer (1990: 167) maintains that had better and would rather only have UncNs and NotCs. However, the NotC example he provides (cf. example (122) below) is, once again, an interrogative negative. (122) Who wouldn’t rather die in a ditch than a pool? The overall number of examples with had better and would rather found in the nine corpora analysed amounts to 29, distributed as shown in Table 12 below. III. ANALYSIS OF THE CORPUS 122 Table 12. Negation of had better and would rather UncNs OpeCs NotCs Total Had better 2 7 3 12 Would rather 5 11 1 17 TOTAL 7 18 4 29 As can be seen, OpeCs predominate over NotCs and uncontracted forms in the corpus. Furthermore, a detailed examination of the examples of NotCs with these two semi-modals confirms the aforementioned restriction of this negative form to occur in interrogative negative clauses (cf. ex. (120) above), since all my examples of NotCs with these operators are found in interrogative negatives. This seems to suggest that the three alternatives of negation with had better and would rather cannot be regarded as true variants. In view of this, all examples with had better and would rather have been excluded from the total count. 2.5. Other Exclusions This subsection is devoted to a number of examples not included in any of the preceding groups, more specifically to those cases in which the codification system used in the corpora under study interrupts the reading of some instances, as in the examples under (123) below. (123) a. okay okay your salary your pay’s <.>not</.> you’re not paid till next week is it. (WSC#DPF048:0160:LB) b. we <.>haven’t</.> <.>h</.> i don’t think you’ve seen our c b o’s for that that’s why computer starts printing. (WSC#DGZ071:0125:WL) III. ANALYSIS OF THE CORPUS 123 c. (107) i am <.>not</.> you didn’t let me finish <,>. (WSC#DPC013:0200:TM) Here, the negator not, in the case of examples (123a) and (123c), and the NotC haven’t, in the case of example (123b), are in between two codes. The first one (<.>) indicates the beginning of incomplete words, while the second (</.>) implies the end of incomplete words. Therefore, the word between such codes is an incomplete word. In view of this, all instances of this kind, 96 examples in all, have been excluded from the count. 3. Negation in the Corpus Once the examples discussed in the preceding section have been excluded from the count, it is time to analyse the remaining instances. The general data is provided in Section 3.1. In turn, Section 3.2. is devoted to those contexts which inhibit the selection of one of the three alternatives of negation at issue. The analysis of UncNs, OpeCs and NotCs as true variants is presented in Section 3.3., while the individual description of the data in each of the corpora used in the present study is provided in Sections 3.4. (written corpora) and 3.5. (spoken corpora). Comparisons among the different corpora are drawn in Section 3.6. Finally, Section 3.7. is devoted to comparisons of the data in the present piece of research with related studies. III. ANALYSIS OF THE CORPUS 124 3.1. General Data The total number of examples found in the corpora amounts to 23,752. These are distributed as shown in Figure 10 and Table 13 below. Figure 10. Negative examples found in the corpus 0 1000 2000 3000 4000 5000 6000 LOB FLOB BROWN FROWN ACE WWC LLC CSPAE WSC Table 13. Total number of examples in each corpora TOTAL LOB 2,087 (8.79%) FLOB 2,002 (8.43%) BROWN 1,921 (8.09%) FROWN 2,124 (8.94%) ACE 1,281 (5.39%) WWC 2,033 (8.56%) LLC 2,807 (11.82%) CSPAE 4,321 (18.19%) WSC 5,176 (21.79%) TOTAL 23,752 The distribution of these 23,752 occurrences as regards the three alternative ways of negation considered in this study is given in Figure 11 below. III. ANALYSIS OF THE CORPUS 125 Figure 11. UncN/OpeC/NotC in the corpus The most immediate conclusions to be gained from this figure are, on the one hand, that contracted forms are used to a greater extent than uncontracted ones and, on the other, that, as concerns contractions, NotCs are preferred to OpeCs. The detailed distribution of these forms in the different corpora analysed according to the type of operator is given in Table 14 below. The data in this table evince that medium is an important factor in the variation between uncontracted forms and negative contractions. As can be seen, while in the spoken corpora negative contractions are clearly favoured over uncontracted forms with all operators, in the written corpora contractions are slightly more frequent than uncontracted forms only with will (except in FLOB), with have and would in FROWN and with would in WWC. As for contractions, it should be noted that NotCs are more commonly used than OpeCs with all operators in all corpora, with the exception of the be-operator in FLOB, in WWC and in all the spoken corpora (LLC, CSPAE and WSC). This means that, in written BrE texts from the 1960s and in written AmE and AusE texts, NotCs are more frequent than OpeCs with be. Table 14. Total number of tokens found in the corpora according to type of operator 60 As regards the number of NotCs with the operator will, my results differ from those obtained by Berglund (2000: 30). In my study 111 examples of the form won’t have been identified, while in Berglund’s analysis the total number of occurrences of won’t amounts to 108. BE HAVE WILL WOULD TOTAL UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs LOB 830 111 136 331 8 155 109 5 11160 178 3 110 2,087 FLOB 722 194 175 229 9 194 119 6 97 129 - 128 2,002 BROWN 739 89 174 246 2 160 105 2 105 168 2 129 1,921 FROWN 690 257 258 189 2 221 99 2 151 113 1 141 2,124 ACE 487 105 120 173 2 80 73 1 75 90 1 74 1,281 WWC 731 217 168 247 5 174 115 3 114 124 2 133 2,033 LLC 248 760 635 46 40 579 26 4 153 33 1 282 2,807 CSPAE 835 1,654 341 302 11 418 135 - 195 157 2 271 4,321 WSC 252 1,905 881 32 16 985 21 - 424 28 1 631 5,176 TOTAL 5,534 5,292 2,888 1,795 95 2,966 802 23 1,425 1,020 13 1,899 23,752 III. ANALYSIS OF THE CORPUS 127 However, the global figures presented so far reveal very little about the actual status of UncNs, OpeCs and NotCs as true alternatives, since, as mentioned in Section II.3.2.2. above, not all the instances recorded allow the interchangeability of the three variants of negation. In a number of knockout contexts, the use of at least one of the three variant forms is prevented. In such cases, therefore, the distribution of UncNs, OpeCs and NotCs is not merely a question of choice between interchangeable elements. The KO contexts found in my corpus are analysed in the following section. 3.2. Knockout Contexts As mentioned in Section II.3.2.2. above, while UncNs are always possible, the use of the two types of contraction is sometimes blocked. In the corpus I have recorded 4,685 occurrences (19.73% of the total count) which do not freely allow the variation between the three alternatives under study. As in the second chapter (cf. Sections II.3.2.2.1. and II.3.2.2.2. above), blockings in the corpus have been divided into two main types: on the one hand, those instances where OpeC is not possible (cf. Section 3.2.1. below) and, on the other, those KO contexts which do not allow the use of NotCs (cf. Section 3.2.2.). III. ANALYSIS OF THE CORPUS 230 (267) Shrugging, Pettigrew said, “I wasn’t trying to be a hero. There’s a reward.” “Yup. I heered about that reward. What Mrs. Atkinson tells me is you surely earned it.” “When can you hold this inquest?” “Not ’til mornin’. Wish we had a telegraph here. The sheriff’s gone down to New Mexico Territory to eyeball a horse thief they arrested down there, and I wish I could telegraph ’im to get back up here. But we ain't and I cain't, so I’ll do the best I know how.” (FROWN N08 152-156) The third operator analysed in relation to text-type is will (cf. Table 56 below). Table 56. Negative variants with the will-operator in FROWN Number of words in each category UncNs OpeCs NotCs TOTAL CAT A 88,000 18 (2.04) - 15 (1.7) 33 CAT B 54,000 3 (0.55) - 30 (5.55) 33 CAT C 34,000 1 (0.29) - 2 (0.58) 3 CAT D 34,000 1 (0.29) - 2 (0.58) 3 CAT E 72,000 2 (0.27) - 12 (1.66) 14 CAT F 96,000 10 (1.04) - 5 (0.52) 15 CAT G 150,000 7 (0.46) - 1 (0.06) 8 CAT H 60,000 7 (1.16) - 2 (0.33) 9 CAT J 160,000 18 (1.12) - 1 (0.06) 19 CAT K 58,000 5 (0.86) - 17 (2.93) 22 CAT L 48,000 4 (0.83) - 9 (1.87) 13 CAT M 12,000 3 (2.5) - 2 (4.16) 5 CAT N 58,000 6 (1.03) 2 (0.34) 15 (2.58) 23 CAT P 58,000 2 (0.34) - 18 (3.1) 20 CAT R 18,000 - - 5 (2.77) 5 TOTAL 1,000,000 87 (0.87) 2 (0.02) 136 (1.36) 225 III. ANALYSIS OF THE CORPUS 231 As can be seen in this table, the behaviour of will in the FROWN corpus does not differ substantially from that of the operator have, since contractions (1.38) also predominate over UncNs (0.87). Notice that, as with have, some press categories, (B and C), together with D (Religion) and E (Skills, trades and hobbies), come close to the informal categories in their predominance of contractions over UncNs. As in the case of have, only two examples of OpeCs have been recorded in this corpus, both of them in Adventure and western fiction (Cat N). These are given under (268) and (269) below. (268) “You’ll not get away with this, Tom,” a BS rider said. (FROWN N02 3-4) (269) “We’ll not rest," said Cunningham, obviously quoting something or someone, "until the scourge of Mobius is lifted from the land. (FROWN N20 82-84) Turning now to the operator would, the distribution of the examples found in the FROWN corpus is given in Table 57 below. Once again, contractions predominate over UncNs (NF 0.99 for UncNs vs. 1.25 for contractions), especially in the informal set of text-types. However, this operator differs from have and will in that the number of formal categories which prefer contractions is lower, only Press editorial (Cat B). Moreover, as expected, the would-operator clearly favours the use of NotCs over OpeCs. Example (270) below is the only instance of the latter contracted type which has been recorded in FROWN. (270) I’d not sail around the world but say I did. (FROWN K17 123-124) III. ANALYSIS OF THE CORPUS 232 Table 57. The operator would in FROWN Number of words in each category UncNs OpeCs NotCs TOTAL CAT A 88,000 12 (1.36) - 9 (1.02) 21 CAT B 54,000 3 (0.55) - 7 (1.29) 10 CAT C 34,000 3 (0.88) - 2 (0.58) 5 CAT D 34,000 3 (0.88) - - 3 CAT E 72,000 3 (0.41) - 2 (0.27) 5 CAT F 96,000 8 (0.83) - 6 (0.62) 14 CAT G 150,000 23 (1.53) - 10 (0.66) 33 CAT H 60,000 3 (0.5) - - 3 CAT J 160,000 9 (0.56) - - 9 CAT K 58,000 5 (0.86) 1 (0.17) 17 (2.93) 23 CAT L 48,000 3 (0.62) - 18 (3.75) 21 CAT M 12,000 4 (3.33) - 1 (0.83) 5 CAT N 58,000 9 (1.55) - 13 (2.24) 22 CAT P 58,000 6 (1.03) - 27 (4.65) 33 CAT R 18,000 5 (2.77) - 12 (6.66) 17 TOTAL 1,000,000 99 (0.99) 1 (0.01) 124 (1.24) 224 The preference for contractions with the operators have, will and would in the FROWN corpus just discussed makes this corpus differ both diachronically from its AmE counterpart from the 1960s (cf. Section 3.6. below), as well as dialectally from the BrE corpora, since LOB, FLOB and BROWN favour the use of full forms with all operators. The distribution of the examples found in the FROWN corpus according to the type of subject is given in Table 58 below. Table 58. Distribution of negative forms according to type of subject and operator in FROWN BE HAVE WILL WOULD TOTAL UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs Simple 240 (2.4) 236 (2.36) 76 (0.76) 68 (0.68) 2 (0.02) 156 (1.56) 53 (0.53) 2 (0.02) 101 (1.01) 57 (0.57) 1 (0.01) 89 (0.89) 1,081 (10.81) Pronoun Complex 2 (0.02) - - - - - - - - - - - 2 (0.02) Simple 322 (3.22) 13 (0.13) 82 (0.82) 79 (0.79) - 30 (0.3) 27 (0.27) - 30 (0.3) 34 (0.34) - 31 (0.31) 632 (6.32) NP Complex 15 (0.15) - 4 (0.04) 6 (0.06) - 1 (0.01) 6 (0.06) - 2 (0.02) 4 (0.04) - - 38 (0.38) Compound 2 (0.02) - - 2 (0.02) - - - - - - - - 4 (0.04) Exist. there 5 (0.05) 5 (0.05) 15 (0.15) 1 (0.01) - 5 (0.05) 1 (0.01) - - 1 (0.01) - 3 (0.03) 35 (0. 35) Clause 17 (0.17) - 1 (0.01) 2 (0.02) - - - - 3 (0.03) 3 (0.03) - 1 (0.01) 27 (0.27) Other 1 (0.01) - 1 (0.01) - - - - - - - - - 2 (0.02) III. ANALYSIS OF THE CORPUS 234 As can be seen, the different behaviour of the three variant negative forms with respect to pronominal vs. non-pronominal subjects observed in the corpora already examined also holds true for the FROWN corpus: while NP subjects show a tendency to occur with uncontracted forms, pronominal subjects (by far the most frequent type) tend to correlate with contracted negatives. Once again, of the two contracted variants, the operator be prefers OpeCs with pronouns as subjects, in clear opposition to the other operators, as well as to the other types of subject, with which NotCs predominate. The data from the FROWN corpus also confirm the close relationship between existential there and contracted forms and that between clausal subjects and UncNs, at least for the verbs be, have and would. In this corpus, as in the three other corpora considered above (cf. Sections 3.4.1 to 3.4.3.), the string frequency factor also plays an important role in the distribution of uncontracted and contracted variants, since those sequences which are more frequent are precisely the ones which favour the use of contractions to a greater extent. Thus, for instance, in the sequence he had not (41 examples) (cf. examples under (271) below), the have-operator tends to be contracted (12 examples with the full form had not vs. 29 with contractions), while with the sequence the injustice had not (two occurrences, none contracted, as shown in example (272) below), full forms predominate. (271) a. That he had not expected at all. (FROWN M02 6) b. For some reason, he hadn’t been affected, or at least not yet. (FROWN M06 28-29) III. ANALYSIS OF THE CORPUS 235 (272) Why not make it now as though the injustice had not happened, for all that its occurrence in the past is immutable and undeniable. (FROWN G22 161-163) 3.4.5. The Australian Corpus of English (ACE) The ACE corpus contains 1,076 examples of the variant forms of negation under discussion, distributed as shown in Figure 18 below. Figure 18. UncN/OpeC/NotC distribution in ACE As can be observed in this figure, written AusE is also characterised by a higher proportion of UncNs than of contractions and by a predominance of NotCs over OpeCs. Concerning the frequency of contractions and uncontracted forms in relation to the type of text, consider Table 59 below. As the data in this table reveal, in ACE UncNs predominate over contractions in all formal text-types from category A to category J, except in C, where contracted forms are more numerous. The preponderance of contractions also applies to all informal categories from K to R. In this respect, the data from AusE do not differ substantially from those of BrE and AmE given so far, since the categories which favour the use of full forms are, in general, those which are more formal. Moreover, of the two types of contracted forms, NotCs are preferred to OpeCs in all categories, except Miscellaneous (Cat H), where only one example of contraction (OpeC) occurs, III. ANALYSIS OF THE CORPUS 236 due to the high level of formality of this text-type, and Cat J (Learned and scientific writings), where OpeCs and NotCs are evenly distributed. Table 59. UncNs and contractions according to text-type in ACE Number of words in each category UncNs OpeCs NotCs TOTAL CAT A 88,000 108 (12.27) 11 (1.25) 36 (4.09) 155 CAT B 54,000 89 (16.48) 2 (0.37) 16 (2.96) 107 CAT C 34,000 19 (5.58) 10 (2.94) 16 (4.7) 45 CAT D 34,000 45 (13.23) 1 (0.29) 10 (2.94) 56 CAT E 76,000 50 (6.57) 3 (0.39) 19 (2.5) 72 CAT F 88,000 58 (6.59) 6 (0.68) 15 (1.7) 79 CAT G 154,000 118 (7.66) 19 (1.23) 34 (2.2) 171 CAT H 60,000 78 (13) 1 (0.16) - 79 CAT J 160,000 99 (18) 6 (0.37) 6 (0.37) 111 CAT K 58,000 18 (3.1) 9 (1.55) 18 (3.1) 136 CAT L 30,000 9 (3) 13 (4.33) 33 (11) 45 CAT M 14,000 9 (6.42) 3 (2.14) 8 (5.71) 55 CAT N 16,000 3 (1.87) 4 (2.5) 6 (3.75) 21 CAT P 30,000 8 (2.66) 8 (2.66) 13 (4.33) 29 CAT R 30,000 8 (2.66) 10 (3.33) 21 (7) 39 TOTAL 926,000 719 (7.76) 106 (1.14) 251 (2.71) 1,076 Once the distribution of the three variant negative forms according to texttype has been analysed, it is time to see how they behave depending on the operator. As shown in Table 60 below, in AusE the be-operator also shows a strong preference for UncNs (NF 4.58) over contractions (NF 1.84) in written texts. As was the case with the general data in Table 59 above, from category A to category J, except in C, as well as in category M, UncNs prevail over contractions. Thus, once again, the informal text-type Science fiction (Cat M) is more closely related to the formal genre Learned and scientific writings (Cat J) III. ANALYSIS OF THE CORPUS 237 than to its sister fictional categories K, L, N, P and R, where contracted forms are preferred to UncNs. As regards the use of contractions, OpeCs (17.11%) are more numerous than NotCs (11.58%), both in the general count as well as in most texttypes, the only exceptions being categories B, D and E. Table 60. Examples with the be-operator in ACE Number of words in each category UncNs OpeCs NotCs TOTAL CAT A 88,000 52 (5.9) 11 (1.25) 6 (0.68) 69 CAT B 54,000 49 (9.07) 2 (0.37) 6 (1.11) 57 CAT C 34,000 13 (3.82) 10 (2.94) 9 (2.64) 32 CAT D 34,000 36 (10.58) 1 (0.29) 4 (1.17) 41 CAT E 76,000 36 (4.73) 3 (0.39) 8 (1.05) 47 CAT F 88,000 33 (3.75) 6 (0.68) 5 (0.56) 44 CAT G 154,000 54 (3.5) 19 (1.23) 12 (0.77) 85 CAT H 60,000 51 (8.5) 1 (0.16) - 52 CAT J 160,000 73 (4.56) 6 (0.37) 2 (0.12) 81 CAT K 58,000 10 (1.72) 8 (1.37) 7 (1.2) 25 CAT L 30,000 6 (2) 13 (4.33) 5 (1.66) 24 CAT M 14,000 4 (2.85) 3 (2.14) - 7 CAT N 16,000 2 (1.25) 4 (2.5) - 6 CAT P 30,000 2 (0.66) 5 (1.66) 1 (0.33) 8 CAT R 30,000 4 (1.33) 10 (3.33) 4 (1.33) 18 TOTAL 926,000 425 (4.58) 102 (1.1) 69 (0.74) 596 Considering the individual forms of the operator be (cf. Table 61 below), full forms outnumber contractions with are not (135 instances for UncNs vs. 38 for contractions) and is not (273 vs. 108), while contractions are preferred for the first person singular (17 for UncNs vs. 26 contracting). In this respect, the data for the individual forms of the operator be in the AusE corpus do not differ much from those obtained in both the BrE and AmE corpora (cf. Sections 3.4.1. to 3.4.4. III. ANALYSIS OF THE CORPUS 238 above). However, as far as contractions are concerned, AusE is closer to AmE from the 1990s (cf. Table 45 in Section 3.4.4. above), since OpeCs are preferred to NotCs with the forms am not, for which no instances of NotCs have been recorded, and is not (60 occurrences under the form ’s not vs. 47 instances of NotC under the forms ain’t (three examples) and isn’t (44 instances). Table 61. Individual forms of the be-operator in ACE Am not Are not Is not UncN OpeC NotC UncN OpeC NotC UncN OpeC NotC CAT A - 3 - 19 - 3 33 8 3 CAT B 2 - - 17 - 2 30 2 4 CAT C 1 2 - - 3 3 12 5 6 CAT D 4 - - 9 1 1 23 - 3 CAT E - - - 14 - 5 22 3 3 CAT F 4 1 - 9 - 2 20 5 3 CAT G 1 6 - 15 3 - 38 10 12 CAT H - - - 24 - - 27 1 - CAT J - 1 - 21 1 - 52 4 2 CAT K 1 2 - 1 3 4 8 3 3 CAT L 3 4 - 3 2 - - 7 5 CAT M - 1 - - - - 4 2 - CAT N - 2 - - - - 2 2 - CAT P 1 3 - 1 - - - 2 1 CAT R - 1 - 2 3 2 2 6 2 TOTAL 17 26 - 135 16 22 273 60 47 Another difference between the AusE corpus and the four other corpora described in the preceding sections is the low number of occurrences of the form ain’t recorded. The only three examples found, all of them in Cat L (Mystery and III. ANALYSIS OF THE CORPUS 239 detective fiction) and used by one and the same character, are (273) to (275) below. (273) “It ain’t for the money,” he said slowly. (ACE L08) (274) “No, that ain’t it either. (ACE L08) (275) You know, that Bone, he’s jus’ a ignorant nigger, he ain’t very bright, he’s rode with half-weights. (ACE L08) These three instances are clearly related to informal style: the character is a black man who speaks a substandard variety of English. Another non-standard feature of the speech of this character is the use of multiple negation, as in example, (175) above repeated here for the sake of convenience as (276): (276) You know, that Bone, he’s jus’ a ignorant nigger, he ain’t very bright, he’s rode with half-weights. (ACE L08) As regards the selection of full and contracted forms with be as a lexical verb or as an auxiliary, the data from AusE are closely related to those obtained in the most recent corpora of written BrE and AmE (cf. Sections 3.4.2. and 3.4.4. above), where contractions with lexical verb be outnumber those found with auxiliary be (30.51% vs. 23.13%) and progressive constructions favour the use of fused forms to a greater extent than passives (45.1% vs. 11.46%) (cf. Table 62 below). Table 62. Occurrences with lexical be and auxiliary be in ACE Full forms Contractions Total Lexical be 312 (69.49%) 137 (30.51%) 449 Progressive 28 (54.9%) 23 (45.1%) 51 Auxiliary be 34 (23.13%) Passive 85 (88.54 11 (11.46%) 96 III. ANALYSIS OF THE CORPUS 246 than UncNs. It must be borne in mind, however, that the low number of examples recorded in some categories do not warrant definitive conclusions in this respect. The last operator studied is would (cf. Table 67 below). In the ACE corpus, this operator, like be and have, prefers UncNs (0.86) to contractions (0.62), although the proportional difference between full forms and contractions is not so high as for be or have. Table 67. Examples with the would-operator in ACE Number of words in each category UncNs OpeCs NotCs TOTAL CAT A 88,000 19 (2.15) - 14 (1.59) 33 CAT B 54,000 12 (2.22) - 3 (0.55) 15 CAT C 34,000 - - 2 (0.58) 2 CAT D 34,000 1 (0.29) - 3 (0.88) 4 CAT E 76,000 2 (0.26) - 2 (0.26) 4 CAT F 88,000 10 (1.13) - 2 (0.22) 12 CAT G 154,000 13 (0.84) - 6 (0.38) 19 CAT H 60,000 6 (1) - - 6 CAT J 160,000 6 (0.37) - 2 (0.12) 8 CAT K 58,000 3 (0.51) 1 (0.17) 3 (0.51) 7 CAT L 30,000 - - 7 (2.33) 7 CAT M 14,000 2 (1.42) - 1 (0.71) 3 CAT N 16,000 1 (0.62) - 1 (0.62) 2 CAT P 30,000 3 (1) - 5 (1.66) 8 CAT R 30,000 2 (0.66) - 6 (2) 8 TOTAL 926,000 80 (0.86) 1 (0.01) 57 (0.61) 138 Moreover, the OpeC alternative with would in AusE occurs only once in a fictional category (General fiction, Cat K) (cf. example (286) below). III. ANALYSIS OF THE CORPUS 247 (286) History had been a bad dream and now that they were awake they’d not mention it - even though everyone had the identical dream, even though there were still bodies waiting to be discovered, bombs that would explode as Kurt and his friends clambered amongst the ruins. (ACE K27) As regards individual types of text, would favours the use of UncNs in most formal categories, such as A, B, F, G, H and J, as well as in the fictional category M (Science fiction), which, as mentioned above, tends to come closer to the formal text-types. In the remaining categories NotCs seem to be preferred, except for E (Skills, trades and hobbies) and N (Adventure and western fiction), where the proportion of variants is alike. To sum up, it can be said that the be, have and would operators behave similarly as regards the distribution of full forms and contractions, since the former type is the preferred option with the three verbs. By contrast, with the willoperator the uncontracted and contracted variants are evenly distributed. Concerning the influence of the type of subject on the selection of contracted and uncontracted forms, consider the data in Table 68 below. Table 68. Distribution of examples according to type of subject and operator in ACE BE HAVE WILL WOULD TOTAL UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs Simple 173 (1.87) 95 (1.03) 35 (0.38) 65 (0.7) 2 (0.02) 47 (0.51) 24 (0.26) 1 (0.01) 44 (0.47) 31 (0.33) 1 (0.01) 48 (0.52) 566 (6.11) Pronoun Complex 2 (0.02) - 1 (0.01) - - - - - - 1 (0.01) - - 4 (0.04) Simple 204 (2.2) - 24 (0.26) 72 (0.78) - 12 (0.13) 32 (0.35) - 17 (0.18) 39 (0.42) - 9 (0.1) 409 (4.41) NP Complex 22 (0.24) - 1 (0.01) 9 (0.1) - 1 (0.01) 5 (0.05) - - 1 (0.01) - - 39 (0.42) Compound - - - 1 (0.01) - - - - - - - - 1 (0.01) Exist. there 5 (0.05) 7 (0.07) 4 (0.04) 3 (0.03) - 1 (0.01) 1 (0.01) - 1 (0.01) 3 (0.03) - - 25 (0. 27) Clause 19 (0.2) - 4 (0.04) - - - 2 (0.02) - 2 (0.02) 5 (0.05) - - 32 (0.35) III. ANALYSIS OF THE CORPUS 249 The most significant conclusion to be drawn from this table is that all operators prefer the use of UncNs with NP subjects. With pronominal subjects, be and have prefer UncNs, while will and would favour contractions. When the subject is existential there, in turn, the choice of variants also depends on the operator, since contractions are preferred with be, while UncNs are favoured with have and would. Finally, clausal subjects tend to select uncontracted forms, especially with be, the only operator for which I have recorded a representative number of examples. Moreover, the ACE corpus provides further proof of the relevance of string frequency to the selection of contracted and uncontracted negative variants, since the most frequent sequences are those which contract to a greater extent. Thus, for instance, the potentially contractible string he is not (cf. examples under (287) below) appears in this corpus 15 times, out of which seven are contracted (46.67%). By contrast, the sequence the Committee is not (cf. example (288) below) appears only twice, with no evidence of contraction. In this respect, there is no noticeable difference between AusE and the other dialects studied so far (cf. Sections 3.4.1. to 3.4.4. above). (287) a. “The writer who possesses the creative gift owns something,” we are told, “of which he is not always master.” (ACE J67) b. He's not merely a producer he’s a creator, keyed up in every nerve to bring something living out of a void ... (ACE G18) c. But he isn't the only big sports name in the show. (ACE C16c) (288) The Committee is not in a position to express a view of what are adequate privacy safeguards as that would be better done by III. ANALYSIS OF THE CORPUS 250 experts in that area, but at least the committee felt that the minimum safeguards would be. (ACE J46) Two examples from the ACE corpus merit further comment. These are given as (289) and (290) below. (289) His odd jocularity that the chicken feet are like ET's hands are not taken askance by the Chinese staff, although they may have no idea what he is on about. (ACE A27) (290) The velocity expressions used within equation 6.1 is not obtained from either of equations 6.7 or 6.8. Rather, we use the centre point, &symbol; as the point of reference giving &formula; 6.10 Substituting equations 6.9 and 6.10 into equation 6.1 we obtain the following recurrence relation &formula; 6.11. (ACE J18) In example (289) the subject of are not is His odd jocularity that the chicken feet are like ET's hands. The subject is, therefore, a third person singular, since the head of the NP is the noun jocularity. In view of this, the operator should also be inflected for the singular, that is, is not rather than are not. Therefore, in this example there is lack of agreement between the subject and the operator. The fact that the operator occurs in the plural may be due to the proximity of the plural noun hands. A similar case of lack of concord is found in example (290), where the subject of is not is a third person plural, the velocity expressions used within equation 6.1, whose head is expressions. Thus, the operator should also be inflected for the plural, that is, are not rather than the singular is not. Once again, a likely explanation for this lack of agreement is the influence of the singular noun equation. III. ANALYSIS OF THE CORPUS 251 3. 4. 6. The Wellington Written Corpus (WWC) As seen in Section 3.3. above, the WWC corpus contains 1,733 negative occurrences of the constructions at issue, which corresponds to 9.09% of the total of examples analysed in the present piece of research. These 1,733 examples are distributed in the following way: Figure 19. UncN/OpeC/NotC distribution in WWC As can be seen, once again, UncNs markedly predominate over contracted forms in written NzE, and, as regards contractions, there is a clear preference for NotCs. Concerning text-type, the distribution of these 1,733 examples is given in Table 69 below. Table 69. Tokens according to text-type in WWC Number of words in each category UncNs OpeCs NotCs TOTAL CAT A 88,000 115 (13.06) 24 (2.72) 34 (3.86) 173 CAT B 54,000 105 (19.44) 15 (2.77) 20 (3.7) 140 CAT C 34,000 29 (8.52) 4 (1.17) 13 (3.82) 46 CAT D 34,000 44 (12.94) 1 (0.29) 6 (1.76) 51 CAT E 76,000 68 (8.94) 3 (0.39) 23 (3.02) 94 CAT F 88,000 106 (12.04) 9 (1.02) 27 (3.06) 142 CAT G 154,000 169 (10.97) 26 (1.68) 32 (2.07) 227 CAT H 60,000 85 (14.16) - - 85 CAT J 160,000 200 (12.5) 1 (0.06) 4 (0.25) 205 CAT K 252,000 142 (5.63) 144 (5.71) 284 (11.26) 570 TOTAL 1,000,000 1,063 (10.63) 227 (2.27) 443 (4.43) 1,733 III. ANALYSIS OF THE CORPUS 252 These data evince that the WWC corpus shows an overall preference for UncNs in all categories with the exception of category K (Fiction), where contractions clearly predominate over uncontracted forms (NF 5.63 vs. 16.97). The fact that in category K contractions are far more commonly used than their uncontracted counterparts is undoubtedly related to the degree of formality of the texts included, since K is the most informal category in the corpus and the one most closely related to the spoken language. In highly formal categories, by contrast, contracted forms do not occur at all, as in H (Miscellaneous), or are practically non-existent, as in J (Learned and scientific writings) (NF for UncNs 12.5 vs. NF 0.31 for contractions). Another important feature shown in Table 59 above is that NotCs are preferred to OpeCs in all categories, thus confirming the general trend for contracted types in WWC given above. Nevertheless, the proportional difference between NotCs and OpeCs is higher in the informal text-type (5.55) than in the most formal categories A to J, where it fluctuates between 0.19 in category J and 2.65 in category C. The following paragraphs will be devoted to the distribution of the variants under study in relation to the type of operator used. The first operator analysed is be. As can be seen in Table 70 below, the operator be shows an overall preference for UncNs (NF 6.33 vs. 3.06). However, as expected, not all text-types behave in the same way, since category K (Fiction) contains a larger proportion of contractions (NF 2.46 for UncNs vs. 7.41 for contractions), undoubtedly related, as mentioned above, to the lower degree of formality of the texts included. As regards contractions, in NzE, the be-operator prefers OpeCs to NotCs (NF 2.17 vs. III. ANALYSIS OF THE CORPUS 253 0.89), thus behaving in the same way as in LOB (cf. Section 3.4.1. above), FLOB (cf. Section 3.4.2. above), FROWN (cf. Section 3.4.4. above) and ACE (cf. Section 3.4.5. above). Nevertheless, this statement only holds true for the following categories: A (Press reportage), B (Press editorial), C (Press review), G (Belle lettres, memoirs and biographies) and K (Fiction). Table 70. Occurrences with the be-operator in WWC Number of words in each category UncNs OpeCs NotCs TOTAL CAT A 88,000 51 (5.79) 23 (2.95) 10 (1.13) 84 CAT B 54,000 70 (12.96) 15 (2.77) 3 (0.55) 88 CAT C 34,000 21 (6.17) 4 (1.17) 3 (0.88) 28 CAT D 34,000 33 (9.7) 1 (0.29) 2 (0.58) 36 CAT E 76,000 47 (6.18) 3 (0.39) 3 (0.39) 53 CAT F 88,000 74 (8.4) 9 (1.02) 11 (1.25) 94 CAT G 154,000 89 (5.77) 24 (1.55) 5 (0.32) 118 CAT H 60,000 50 (8.33) - - 50 CAT J 160,000 136 (8.5) 1 (0.06) 2 (0.12) 139 CAT K 252,000 62 (2.46) 137 (5.43) 50 (1.98) 249 TOTAL 1,000,000 633 (6.33) 217 (2.17) 89 (0.89) 939 As far as individual forms of the verb are concerned (cf. Table 71 below), the overall predominance of full forms over contractions is also evident for the patterns are not (206 for UncNs vs. 73 for contracted forms: 48 OpeCs and 25 NotCs) and is not (409 for UncNs vs. 178 (115 OpeCs and 63 NotCs) for contracted forms). For the first person singular, by contrast, contractions are preferred (18 for UncNs vs. 55 for contracted forms). The vast majority of these contractions (54 examples in all) are of the OpeC type, while only one occurrence of NotC, under the form ain’t (cf. example (291) below) has been found. III. ANALYSIS OF THE CORPUS 254 (291) “I don’t know what’s in store for me when I get home and I ain’t too much worried about it either,” said Mr Couch. (WWC A11 051052) Table 71. Individual forms of the be-operator in WWC Am not Are not Is not UncN OpeC NotC UncN OpeC NotC UncN OpeC NotC CAT A 2 2 1 16 6 - 33 15 9 CAT B - 2 - 25 3 1 45 10 2 CAT C - - - 8 1 2 13 3 1 CAT D - - - 17 - - 16 1 2 CAT E - - - 8 - - 39 3 3 CAT F 1 3 - 29 1 4 44 5 7 CAT G 4 5 - 22 7 1 63 12 4 CAT H - - - 18 - - 32 - - CAT J 2 1 - 50 - - 84 - 2 CAT K 9 41 - 13 30 17 40 66 33 TOTAL 18 54 1 206 48 25 409 115 63 Ain’t is also used in WWC as a contraction for is not (one example, given as (292) below) and has not (one example, (293) below), both occurrences recorded in the fictional text-type General fiction (Cat K). In this respect, NzE behaves similarly to ACE, (cf. examples (274-276) in Section 3.4.5. above) or LOB, since in the three corpora the number of instances with ain’t is very low. (292) It ain’t for long. (WWC K28 140) (293) See, Jah say that the big one don't know how to fight, that he ain’t got no killer instinct, and the little one does. (WWC K61 083-084) III. ANALYSIS OF THE CORPUS 255 The distribution of the variants in the WWC corpus, as regards the distinction lexical be vs. auxiliary be is clearly related to that attested in the most recent written corpora (FLOB, FROWN and ACE), since contractions are more numerous with copula be than with auxiliary be (32.9% vs. 23.13%), and, as in all the written corpora analysed in the present piece of research, in progressive sentences (55.91%) than in passives (17.95%), as seen in the following Table: Table 72. Lexical be and auxiliary be in WWC Full forms Contractions Total Lexical be 463 (67.1%) 227 (32.9%) 690 Progressive 41 (44.09%) 52 (55.91%) 93 Auxiliary be 80 (23.13%) Passive 128 (82.05%) 28 (17.95%) 156 As in the other written corpora (cf. Sections 3.4.1. to 3.4.5. above), NotCs are more common with lexical be (30.4% vs. 26.25%), while OpeCs predominate with auxiliary be (69.6% vs. 73.75%) (cf. Table 73 below). Moreover, NotCs are, once again, more closely associated with passive constructions, while OpeCs are more frequent with progressives. Table 73. Contracted forms with lexical be and auxiliary be in WWC OpeCs NotCs Total Lexical be 158 (69.6%) 69 (30.4%) 227 Progressive 42 (80.77%) 10 (19.23%) 52 Auxiliary be 59 (73.75%) 21 (26.25%) Passive 17 (60.71%) 11 (39.29%) 28 Table 74 below shows how the have-operator behaves in this corpus as far as text-type is concerned. The aforementioned preference for uncontracted forms III. ANALYSIS OF THE CORPUS 262 their uncontracted counterparts. In turn, the number of OpeCs is very low, NotCs being, once again, the preferred contracted variant. The only two occurrences of the OpeC type, both of them of the form ’d not, examples (305) and (306) below, are found in the least formal category, namely Fiction (Cat K). (305) Paora waited as the man want to get them and after five minutes he began to think something was wrong and he’d not be able to get the heads. (WWC K02 142-144) (306) Then Mark to say he’d not be home for lunch. (WWC K31 158-159) By way of summary, it can be said that in written NzE full forms clearly predominate over contractions with the operators be, have and will, although the proportional difference between UncNs, on the one hand, and OpeCs and NotC, on the other, is not so highly marked with the latter two operators. In the case of would, in turn, the choice between uncontracted negatives and negative contractions is rather balanced. Besides text-type, the other factor analysed in this corpus is the type of subject. The data obtained for the different operators in WWC in accordance with subject-type are shown in Table 78 below. As can be observed, most tokens are found with simple pronominal subjects, which prefer NotCs with all operators except be. Notice that with this verb uncontracted forms predominate over contractions with subjects of this kind. With the remaining types of subjects, with the exception of existential there-constructions, UncNs are more common than contractions with the four operators examined. Table 78. Distribution of examples according to type of subject and operator in WWC BE HAVE WILL WOULD TOTAL UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs Simple 273 (2.73) 200 (2) 44 (0.44) 100 (1) 5 (0.05) 123 (1.23) 53 (0.53) 3 (0.03) 66 (0.66) 64 (0.64) 2 (0.02) 86 (0.86) 1,019 (10.19) Pronoun Complex - - - 1 (0.01) - - 1 (0.01) - - - - - 2 (0.02) Simple 306 (3.06) 7 (0.07) 30 (0.3) 101 (1.01) - 26 (0.26) 41 (0.41) - 21 (0.21) 39 (0.39) - 21 (0.21) 592 (5.92) NP Complex 22 (0.22) - - 8 (0.08) - - 7 (0.07) - 1 (0.01) 2 (0.02) - 2 (0.02) 42 (0.42) Compound 3 (0.03) - - - - - - - - 1 (0.01) - - 4 (0.04) Exist. there 13 (0.13) 10 (0.1) 13 (0.13) 1 (0.01) - 2 (0.02) 2 (0.02) - 5 (0.05) 2 (0.02) - - 48 (0.48) Clause 15 (0.15) - 4 (0.04) 2 (0.02) - - 3 (0.03) - - 2 (0.02) - - 26 (0.26) III. ANALYSIS OF THE CORPUS 264 Thus, once again, the strong connection between the degree of complexity of the subject and the distribution of contracted and uncontracted variants, mentioned by scholars such as Quirk et al. (1985: 123), Krug (1998: 289) or Biber et al. (1999: 1129f), among others (cf. Section 3.4.1. above) is confirmed here. In this connection, the frequency of sequences also plays an important role, since, as in the five other written corpora, contractions turn out to be more common with the most frequent strings. This is proved by the comparison of the two potentially contractible sequences he is not (cf. examples under (307)) and the government is not (cf. (308)). I have recorded 27 occurrences of the former sequence, out of which 14 (51.85%) are of the contracted type, and two instances with the string the government is not, none of them contracted. Thus, once again, the data in the WWC corpus seem to confirm Krug’s statement that the more frequent a given sequence is, the higher is its contraction ratio. (307) a. He is not a judge or a lawyer. (WWC B17 235) b. He’s not a flamboyant man or a dynamic speaker, but his selfcomposure is striking. (WWC B19 117-118) c. Big Dave's magic dipstick ought to be worth a bob or two, and since he isn’t likely to need it, perhaps Roger the Dodger has been negotiating to sell it to Bodgie Bill, to take back to Oz and keep his bolshie anti-nuke left wing amused. (WWC B19 250-253) (308) “The Government is not supporting a wild and irrational Maori cause. (WWC B14 213-214) III. ANALYSIS OF THE CORPUS 265 3.4.7. Summary The comparison of the data provided in Sections 3.4.1. to 3.4.6. above shows that, generally speaking, uncontracted forms are preferred to a greater extent than contractions in all the written corpora. As regards contractions, NotCs predominate over OpeCs in all the written dialects analysed in this study. As far as text-types are concerned, it must be noted that, while in the most formal categories uncontracted negatives are the preferred option, contractions turn out to be the predominant choice in the less formal categories, due to their proximity to spoken style. Concerning the distribution of the negative variant forms with individual operators, UncNs are the most common option with almost all verbs and in all corpora, with the exception of FROWN, which prefers contractions with all operators except be. As regards the variation between the two contracted types, there is a clear preponderance of NotCs over OpeCs with all verbs except be, thus confirming the exceptional behaviour of this operator described in the literature on the topic. However, this does not hold true in the case of written AmE texts from the 1960s, where NotCs are used more frequently than OpeCs with all operators including be. Finally, as regards the distribution of negative forms according to type of subject, contractions, mainly NotCs, are more frequent with pronominal subjects and with existential there, while UncNs constitute the predominant alternative with more complex subject types, such as NPs or clauses. III. ANALYSIS OF THE CORPUS 266 3.5. Analysis of the Spoken Corpora This section is concerned with the variation between the three alternative negative forms under analysis in spoken texts of BrE, AmE and NzE. For this purpose, the LLC (Section 3.5.1. below), the CSPAE (Section 3.5.2.) and the WSC (Section 3.5.3.) will be analysed in detail, taking into consideration variables such as texttype, operators and type of subject. 3.5.1. The London-Lund Corpus (LLC) As seen in Section 3.3. above, the LLC contains 1,742 examples which allow the interchangeability between the three negative variants under consideration. These 1,742 examples are distributed in the following way: 149 (NF per 10,000 words 2.98) for UncNs, 688 (NF 13.76) for OpeCs and 905 (NF 18.1) for NotCs, as illustrated in Figure 20 below. Figure 20. UncN/OpeC/NotC distribution in LLC The most immediate conclusion to be drawn from this figure is that the LLC shows an overall preference for contractions (91.44%, NF per 10,000 words 31.86) at the expense of uncontracted forms (8.56%, NF per 10,000 words 2.98). Such a preference contrasts sharply with the predominance of uncontracted negatives witnessed in the written corpora examined in Section 3.4. above. This corroborates the statement found in the literature that, in contemporary English, III. ANALYSIS OF THE CORPUS 267 contractions are more commonly used in speech than in writing (cf. Fries 1940: 8; Forsheden 1983: 36; Biber 1988: 243 and, Section II.3.1.1. above). As regards contractions, NotCs are the predominant contracted type. The distribution of the different negative forms found in LLC according to text-type is given in Table 79 below. Table 79. Full forms and contractions according to text-type in LLC Number of words in each category UncNs OpeCs NotCs TOTAL S.1 65,000 9 (1.38) 131 (20.15) 181 (27.85) 321 S.2 73,000 20 (2.73) 133 (18.21) 184 (25.2) 337 S.3 37,000 8 (2.16) 54 (14.59) 99 (26.75) 161 S.4 35,000 4 (1.14) 74 (21.14) 83 (23.71) 161 S.5 65,000 39 (6) 86 (13.23) 104 (16) 229 S.6 45,000 6 (1.33) 54 (12) 66 (14.66) 126 S.7 15,000 3 (2) 29 (19.33) 43 (28.66) 75 S.8 20,000 7 (3.5) 42 (21) 48 (24) 97 S.9 25,000 6 (2.4) 25 (10) 41 (16.4) 72 S.10 55,000 9 (1.63) 31 (5.63) 25 (4.54) 65 S.11 30,000 20 (6.67) 13 (4.33) 19 (6.33) 52 S.12 35,000 18 (5.14) 16 (4.57) 12 (3.42) 46 TOTAL 500,000 149 (2.98) 688 (13.76) 905 (18.1) 1,742 As the data in this table show, contractions are preferred to uncontracted forms in all text-types, even in the most formal ones. However, the proportion of contracted and full forms differs considerably among categories. Thus, the less formal text-types, such as S.1 (Conversation between equals) (NF per 10,000 words 1.38 for UncNs vs. NF 48 for contractions) or S.4 (Conversation between intimates and equals) (NF 1.14 for UncNs vs. 44.85 for contractions) are those III. ANALYSIS OF THE CORPUS 268 where the ratio of contractions is the highest. By contrast, in a very formal type of text, such as S.12 (Prepared but unscripted oration), the proportional difference between contractions and full forms is not so highly marked (NF 5.14 for UncNs vs. 7.99 for contractions). Concerning the selection of contractions, it should be noted that NotCs are favoured over OpeCs in most text-types, with the exception of S.10 (Spontaneous commentary) and S.12 (Prepared but unscripted oration), the latter being the most formal text-type in LLC. The data in this corpus therefore suggest that, in spoken BrE, examples such as (310) and especially (311) are more common than that in (309). (309) you have not produced a scrap of evidence. (LLC S5 3 44 7780 1 1 l 11) (310) you’ve not read many nineteenth century novels or present twentieth century novels D H Lawrence say. (LLC S3 5b 6711780 1 3-5b 6711780 1 1) (311) so you’ve never I mean you haven’t been in the habit of interviewing many candidates. (LLC S2 6 50 4640 1 1 A 12-6 51 4660 1 1 A 11) Once the general data from the LLC have been examined, it is time to consider the results obtained for the different operators which allow the three alternative ways of negation under analysis. First of all, I shall concentrate my attention on the be-operator. Table 80 below gives clear evidence of how be behaves in spoken BrE in different text-types. As can be seen, the operator be shows an overall preference for contractions (NF 16.86 vs. 2.1), which also holds true for all individual text-types. Thus, in spoken BrE, the operator be favours the use of contractions not only in less formal categories, such as S.5 (Conversation), but also in the most formal ones, such as S.12 (Prepared but unscripted oration). III. ANALYSIS OF THE CORPUS 269 Table 80. Examples with the be-operator in LLC Number of words in each category UncNs OpeCs NotCs TOTAL S.1 65,000 6 (0.92) 126 (19.38) 29 (4.46) 161 S.2 73,000 14 (1.91) 123 (16.84) 42 (5.75) 179 S.3 37,000 6 (1.62) 49 (13.24) 25 (6.75) 80 S.4 35,000 3 (0.85) 70 (20) 18 (5.14) 91 S.5 65,000 31 (4.76) 84 (12.92) 25 (3.84) 140 S.6 45,000 6 (1.33) 51 (11.33) 22 (4.88) 79 S.7 15,000 3 (2) 27 (18) 5 (3.33) 35 S.8 20,000 4 (2) 40 (20) 9 (4.5) 53 S.9 25,000 4 (1.6) 24 (9.6) 5 (2) 33 S.10 55,000 6 (1.09) 29 (5.27) 7 (1.27) 42 S.11 30,000 10 (3.33) 13 (4.33) 3 (1) 26 S.12 35,000 12 (3.42) 16 (4.57) 1 (0.28) 29 TOTAL 500,000 105 (2.1) 652 (13.04) 191 (3.82) 948 As was the case with the general data (cf. Table 79 above), such a preference is more noticeable in the informal categories than in the formal ones. Thus, for example, in category S.4 (Conversation between intimates and equals), a very informal text-type, the ratio of contractions over full forms is as high as 24.29 (NF per 10,000 words), while in a very formal category such as S.12 (Prepared but unscripted oration), the proportional difference between the two variants is considerably lower (NF 1.43). Perhaps the most interesting feature of the operator be in LLC concerns the variation between the two kinds of contractions. Thus, contrary to the general data presented in Table 79 above, be selects OpeCs more frequently than NotCs in all categories. This is especially evident in a text-type such as Surreptitious telephone III. ANALYSIS OF THE CORPUS 270 conversations between business associates (S.8) (NF 20 for OpeCs vs. 4.5 for NotCs). Such a predominance of OpeCs over NotCs in LLC confirms, once again, the statements by scholars such as Dillard (1980), Freeborn (1986), Hughes & Trudgill (1996) or Anderwald (2002), among others (cf. Sections II.3.2.3.1. and II.4.1. above), on the tendency of the operator be to select OpeC. As far as individual forms are concerned (cf. Table 81 below), the data from the LLC differ from those provided in the study of the written corpora developed in Section 3.4. above. Table 81. Individual forms of the be-operator in LLC Am not Are not Is not UncN OpeC NotC UncN OpeC NotC UncN OpeC NotC S.1 - 36 - 1 23 11 5 67 18 S.2 - 33 1 2 22 4 12 68 37 S.3 1 12 - 3 10 7 2 27 18 S.4 - 14 - - 24 4 3 32 14 S.5 3 33 - 10 16 9 18 35 16 S.6 - 20 - 2 13 1 4 18 21 S.7 - 10 - 1 7 - 2 10 5 S.8 - 14 - 1 11 1 3 15 8 S.9 - 9 - 2 5 1 2 10 4 S.10 - 3 - 2 9 2 4 17 5 S.11 - 5 - 6 3 - 4 5 3 S.12 2 8 - 4 2 - 6 6 1 TOTAL 6 197 1 34 145 40 65 310 150 Here contracted forms are by far the predominant choice (six instances for the UncN of am not vs. 198 for contractions (197 OpeCs and one NotC), 34 UncNs for are not vs. 185 for contractions (145 OpeCs and 40 NotCs) and 65 for is not III. ANALYSIS OF THE CORPUS 271 vs. 460 for contractions (310 OpeCs and 150 NotCs)). Besides, the sequence is not, which is the most frequent in the corpus (525 occurrences), is also the one which favours the use of contracted forms to a greater extent (87.62%), confirming therefore Krug’s (1998: 294) string frequency factor (cf. Section II.3.1.5. above). Moreover, as the data show, OpeC is the predominant type for the three individual grammatical forms, being especially frequent for the first person singular. For this grammatical person, only one example of NotC has been recorded in Conversation between equals (S.2) (cf. example (312) below). (312) and flirting with the conductor and one of them said to the other you’re a dirty whore and she said no I ain’t I`m a clean whore. (LLC S2 5a 50 4380 1 1 C 11 1-5a 51 4420 1 1 C 11 1) Contrary to my expectations, this is the only occurrence of the form ain’t, both for the operators be and have, I have recorded in LLC. In this connection, spoken BrE as represented in the LLC corpus differs from the data obtained for written BrE, where four instances were recorded in LOB (cf. Section 3.4.1. above) and nine in FLOB (cf. Section 3.4.2. above). As regards the distribution of contractions and full forms with the operator be depending on whether it functions as a lexical verb or as an auxiliary for passive and progressive constructions, the data from spoken BrE do not differ too much from those of written BrE from the 1990s. Here again, as seen in Table 82 below, contractions are somewhat more frequent with be as a lexical verb than as an auxiliary (89.08% vs. 88.35% respectively), thus corroborating Philips & Reynolds’ (1987) statement that contractions of be are preferable with main verb III. ANALYSIS OF THE CORPUS 278 cases of UncNs (NF 0.3) and 198 examples of contractions (NF 3.96), all of them NotCs. Therefore, as was the case with the have and will-operators, there is a clearly marked preference for NotCs with would. As Table 87 evinces, the proportional difference between contractions and uncontracted forms in those categories where both types are present, S.1 and S.2 (Conversations between equals), S.3 (Conversations between disparates), S.5 (Conversations), S.11 (Spontaneous oration) and S.12 (Prepared but unscripted oration), is much lower in the most formal text-types, such as S.12 (0.28), than in less formal categories, such as S.1 (6.92). Table 87. Instances with the would-operator in LLC Number of words in each category UncNs OpeCs NotCs TOTAL S.1 65,000 1 (0.15) - 46 (7.07) 47 S.2 73,000 4 (0.54) - 46 (6.3) 50 S.3 37,000 2 (0.54) - 16 (4.32) 18 S.4 35,000 - - 17 (4.85) 17 S.5 65,000 2 (0.3) - 30 (4.61) 32 S.6 45,000 - - 10 (2.22) 10 S.7 15,000 - - 4 (2.66) 4 S.8 20,000 - - 5 (2.5) 5 S.9 25,000 - - 11 (4.4) 11 S.10 55,000 - - 3 (0.54) 3 S.11 30,000 4 (1.33) - 7 (2.33) 11 S.12 35,000 2 (0.57) - 3 (0.85) 5 TOTAL 500,000 15 (0.3) - 198 (3.96) 213 In summary, in spoken BrE the four operators examined in this piece of research favour the use of contractions over full forms in all kinds of texts. However, the proportional difference between the former and the latter type of III. ANALYSIS OF THE CORPUS 279 negation is considerably higher with the be-operator than with the three other verbs. As far as contractions are concerned, NotCs are clearly preferred with all operators except be, thus corroborating what scholars such as Quirk et al. (1985: 1596f), Biber et al. (1999: 166, 1129ff) and Kortmann (2003: 70f), among others, affirm about the predominance of NotCs with all operators with the exception of be (cf. Section II.3.2. above). As in the case of the written corpora (cf. Section 2.4. above), another variable analysed is the type of subject. The data obtained for the different operators in LLC in relation to the type of subject are provided in Table 88 below. As can be seen, most tokens are found with simple pronominal subjects, which prefer NotCs with all operators except be, for which the most frequent choice is that of OpeC. With the remaining subject-types, except for complex NPs with the be-operator, NotC is also the most frequent variant. Therefore, the tendency found in all the written corpora (cf. Section 3.4. above) for more complex subjects to favour the use of full forms and for simple subjects to prefer contractions does not hold true in spoken BrE as represented in LLC. Table 88. Distribution of examples according to type of subject and operator in LLC BE HAVE WILL WOULD TOTAL UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs Simple 58 (1.16) 625 (12.5) 104 (2.08) 17 (0.34) 32 (0.64) 396 (7.92) 3 (0.06) 4 (0.08) 92 (1.84) 14 (0.28) - 180 (3.6) 1,525 (30.5) Pronoun Complex - - - - - - 1 (0.02) - - - - - 1 (0.02) Simple 38 (0.76) 11 (0.22) 53 (1.06) 4 (0.08) - 14 (0.28) 4 (0.08) - 5 (0.1) 1 (0.02) - 14 (0.28) 144 (2.88) NP Complex 4 (0.08) - 2 (0.04) - - - - - 1 (0.02) - 2 (0.04) 9 (0.18) Compound 1 (0.02) - - - - - - - - - - - 1 (0.02) Exist. there 4 (0.08) 16 (0.32) 32 (0.64) - - 3 (0.06) - - 5 (0.1) - - 1 (0.02) 61 (1.22) Clause - - - - - - - - - - - 1 (0.02) 1 (0.02) III. ANALYSIS OF THE CORPUS 281 As to the potential effect of string frequency (cf. Krug (1998: 294), I have selected, as in the other corpora analysed so far, two potentially contractible sequences, one with a pronominal subject (he is not) and one with a NP subject which appears at least twice in the corpus (dhobi is not)76 (cf. examples (320) and (321) below, respectively). A total of 37 examples have been recorded with the sequence he is not, out of which 36 are contractions, and only two instances with Dhobi is not, none of them contracted. Although the low number of potentially contractible forms with NPs do not allow me to draw definite conclusions in this respect, the data seem to indicate that more frequent sequences are those which favour the use of contractions to a greater extent. (320) a. He is not a very clever boy. (LLC S1 114b 9 9460 1 1 A 11 2) b. He’s not a hasty man. (LLC S1 1 39 5980 1 1 A 11) c. well he isn’t anything. (LLC S1 13 20 1620 1 1 B 11) (321) but dhobi is not quite so common. (LLC S2 14 25 2080 1 1 A 11) 3.5.2. The Corpus of Spoken Professional American English (CSPAE) The analysis o f the CSPAE has yielded 3,792 examples of the three variants of negation under analysis. These 3,792 examples are distributed as follows: 1,176 cases of UncNs (NF 11.76), 1,667 of OpeCs (NF 16.67) and 949 of NotCs (NF 9.49), as illustrated in Figure 21 below. 76 According to the OED (s. v. dhobi 1), dhobi means ‘a native washerman in India.’ III. ANALYSIS OF THE CORPUS 282 Figure 21. UncN/OpeC/NotC distribution in CSPAE The most immediate conclusion to be drawn from this figure is that, as was the case with the BrE spoken corpus (cf. Section 3.5.1. above), in spoken AmE contractions are far more frequent than full forms. However, such fused forms are somewhat less common in the AmE corpus than in the BrE one (NF 11.76 for UncNs and 26.16 for contractions in CSPAE vs. 2.98 for UncNs and 31.86 for contractions in LLC). This seems to contradict the statements by Biber (1987: 11f), Tottie (1991: 12) or Castillo-González (2003: 678ff), among many others, that AmE favours the use of contractions to a greater extent than BrE (cf. Section II.3.2.2. above). It must be noted, however, that the lower frequency of contracted forms in my AmE data in comparison with those from the LLC can be related to the higher degree of formality of the texts included in CSPAE, which, as seen in Section III.1.2., comprises spoken material from university meetings and White House press briefings. Another significant feature found in the spoken AmE corpus concerns the distribution of contracted types, since OpeCs predominate over NotCs, in contrast to the general tendency in favour of the latter variant detected in all the corpora analysed so far. Such a preference for OpeCs in CSPAE may be related to the large number of examples with the operator be. As seen in the following III. ANALYSIS OF THE CORPUS 283 paragraphs, examples with the be-operator (2,477 in all) represent 65.32% of the total number of instances recorded in this corpus (cf. Table 89 below). As concerns text-type, the distribution of the 3,792 examples found in CSPAE is given in Table 89 below. Table 89. Negative examples according to text-type in CSPAE Number of words in each category UncNs OpeCs NotCs TOTAL Math Com 6/97 110,000 79 (7.18) 185 (16.81) 115 (10.45) 379 Read Com 6A/97 100,000 100 (10) 213 (21.3) 109 (10.9) 422 Read Com 6B/97 140,000 150 (10.71) 296 (21.14) 192 (13.71) 638 North Carol 95 51,000 51 (10) 54 (10.58) 38 (7.45) 143 North Carol 96 45,000 41 (9.11) 45 (10) 27 (6) 113 North Carol 97 66,000 89 (13.48) 68 (10.3) 47 (7.12) 204 WH 95 100,000 170 (17) 91 (9.1) 57 (5.7) 318 WH 97A 180,000 213 (1.83) 310 (17.22) 181 (10.05) 704 WH 97B 208,000 283 (13.6) 405 (19.47) 183 (8.79) 871 TOTAL 1,000,000 1,176 (11.76) 1,667 (16.67) 949 (9.49) 3,792 The data in this table evince that contractions are the preferred option in most text-types, with the exception of WH 95, where UncNs outnumber contractions (NF 17 vs. 14.8 respectively). Another important feature shown in Table 89 above is that OpeCs are preferred to NotCs in all text-types, thus confirming that the AmE data in CSPAE behave differently from those obtained for spoken BrE (cf. Section 3.5.1 above). Thus, while in spoken AmE UncNs (NF 11.76) and OpeCs (NF 16.67) are more frequent than in LLC (NF 2.98 and 13.76, respectively), NotCs are more numerous in the BrE corpus than in the AmE one (NF 9.49 in CSPAE vs. 18.1 in LLC). As mentioned above, the different degree of formality of III. ANALYSIS OF THE CORPUS 284 the spoken AmE texts and their BrE counterparts may go a long way towards explaining such a wide discrepancy between the two sets of data. The following paragraphs will be devoted to the distribution of the variants under study in relation to the type of operator used, in order to check whether contractions, mainly OpeCs, are also the preferred option with each verb individually. The first operator examined is be. Table 90. Instances with the be-operator in CSPAE Number of words in each category UncNs OpeCs NotCs TOTAL Math Com 6/97 110,000 56 (5.09) 185 (16.81) 28 (2.54) 269 Read Com 6A/97 100,000 73 (7.3) 213 (21.3) 33 (3.3) 319 Read Com 6B/97 140,000 111 (7.92) 292 (20.85) 52 (3.71) 455 North Carol 95 51,000 35 (6.86) 54 (10.58) 9 (1.76) 98 North Carol 96 45,000 21 (4.66) 45 (10) 6 (1.33) 72 North Carol 97 66,000 49 (7.42) 68 (10.3) 11 (1.66) 128 WH 95 100,000 82 (8.2) 89 (8.9) 10 (1) 181 WH 97A 180,000 89 (4.94) 305 (16.94) 21 (1.16) 415 WH 97B 208,000 118 (5.67) 403 (19.37) 19 (0.91) 540 TOTAL 1,000,000 634 (6.34) 1,654 (16.54) 189 (1.89) 2,477 As shown in Table 90, the statements made above concerning the predominance of contractions over uncontracted forms and about the preference for OpeCs at the expense of NotCs hold true in all text-types with the be-operator. The data from the individual forms of the be-operator (cf. Table 91 below) reveal that contracted forms are, without exception, the most frequent negative variants. Nevertheless, the proportional difference between full forms and III. ANALYSIS OF THE CORPUS 285 contractions is particularly marked for the first person singular (21 occurrences of UncNs vs. 586 of contractions, all of them OpeCs). The forms is not (376 examples of UncNs vs. 790 of contractions) and are not (237 of UncNs vs. 467 of contractions) show a similar proportion of uncontracted vs. contracted variants. Furthermore, the predominance of OpeCs over NotCs also holds true for the three individual forms. In this respect, spoken AmE does not differ from its BrE counterpart. Table 91. Individual forms of the be-operator in CSPAE Am not Are not Is not UncN OpeC NotC UncN OpeC NotC UncN OpeC NotC Math Com 6/97 - 44 - 25 46 12 31 95 16 Read Com 6A/97 - 49 - 30 62 22 43 102 11 Read Com 6B/97 - 87 - 43 59 27 68 146 25 North Carol 95 - 13 - 15 12 5 20 29 4 North Carol 96 2 18 - 10 16 5 9 11 1 North Carol 97 2 19 - 20 21 5 27 28 6 WH 95 4 22 - 21 25 5 57 42 5 WH 97A 7 148 - 34 59 12 48 98 9 WH 97B 6 186 - 39 65 9 73 152 10 TOTAL 21 586 - 237 365 102 376 703 87 As in the case of the spoken BrE corpus (cf. Section 3.5.1. above), in spoken AmE, the form ain’t (both for the be and have operators) is practically nonexistent. In this respect, the CSPAE contrasts with the written AmE corpora BROWN and FROWN, which contain a relatively high number of instances of such a form (34 and 44 examples, respectively). The only example recorded in this corpus of the form ain’t, which corresponds to are not, is the following: III. ANALYSIS OF THE CORPUS 286 (322) MCCURRY: [...]There’s no money left in that program. There was $166 million in the bill to put money in for that -- VOICE: Can you finish talking about – MCCURRY: We can get some more – that’s probably easier to do, without reading the whole thing. Can we see if Larry can do that? He said we can probably scrub this and put it out. Okay. VOICE: So what you’re saying, in effect, is that the Republicans – MCCURRY: The Department of Interior -- remember out at Yosemite, when they were trying to repair all the stuff that was damaged out at Yosemite, rebuilding some of the facilities within the park. […] VOICE: So you’re saying that Republicans were more interested in going on vacation than in -- MCCURRY No, I’m just saying that they ain’t doing a good job of running the place. That’s what I’m saying. (CSPAE WH97B) This example occurs in a dialogue between Mike McCurry, a White House Press secretary (cf. Barlow 2000: 18), and a reporter (‘VOICE’). The use of ain’t in this example on the part of the former speaker could be related to his annoyance, at being interrupted all the time by the reporter, so that he changes from formal to informal style as the conversation proceeds, as shown in example (322) above. Spoken AmE as represented in CSPAE shows the tendency found in LOB and BROWN (cf. Sections 3.4.1. and 3.4.3. above) for contractions to occur more frequently with the operator be functioning as an auxiliary (77.13%) rather than as a lexical verb (72.87%) (cf. Table 92 below). Thus, despite comprising data from III. ANALYSIS OF THE CORPUS 287 the 1990s, the CSPAE is closer in this respect to the oldest corpora than to the most recent ones. Moreover, as expected, fused forms are more common with progressive be (81.87%) than with passive be (61.54%) Table 92. Lexical be and auxiliary be in CSPAE Full forms Contractions Total Lexical be 430 (27.13%) 1,155 (72.87%) 1,585 Progressive 124 (18.13%) 560 (81.87%) 684 Auxiliary be 688 (77.13%) Passive 80 (38.46%) 128 (61.54%) 208 In turn, the most significant characteristic of the selection of the two contracted variants with the operator be in CSPAE is that both OpeCs and NotCs are evenly distributed with lexical be and with auxiliary be, as shown in Table 93 below. It is also noticeable that, as in spoken BrE, in spoken AmE NotCs are more numerous with the operator be functioning as an auxiliary for the passive (17.19%) than as an auxiliary for the progressive (8.57%). Table 93. OpeCs and NotCs with lexical be and auxiliary be in CSPAE OpeCs NotCs Total Lexical be 1,036 (89.7%) 119 (10.3%) 1,155 Progressive 512 (91.43%) 48 (8.57%) 560 Auxiliary be 618 (89.83%) 70 (10.17%) Passive 106 (82.81%) 22 (17.19%) 128 Let us concentrate now on the have-operator as far as text-type is concerned (cf. Table 94 below). The preference for negative contractions witnessed above for be is also evident with the have-operator, though it is not so strongly marked as in the case of the verb be. Table 98. Distribution of examples according to type of subject and operator in CSPAE BE HAVE WILL WOULD TOTAL UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs Simple 309 (3.09) 1,611 (16.11) 88 (0.88) 203 (2.03) 11 (0.11) 327 (3.27) 78 (0.78) - 130 (1.3) 97 (0.97) 2 (0.02) 200 (2) 3,056 (30.56) Pronoun Complex - - - 1 (0.01) - - 1 (0.01) - - - - - 2 (0.02) Simple 238 (2.38) 6 (0.06) 74 (0.74) 64 (0.64) - 34 (0.34) 39 (0.39) - 30 (0.3) 35 (0.35) - 11 (0.11) 531 (5.31) NP Complex 30 (0.3) - 4 (0.04) 5 (0.05) - 1 (0.01) 3 (0.03) - 1 (0.01) 3 (0.03) - 3 (0.03) 50 (0.5) Compound 4 (0.04) - - - - - 1 (0.01) - - - - - 5 (0.05) Exist. there 28 (0.28) 37 (0.37) 21 (0.21) 4 (0.04) - 11 (0.11) 3 (0.03) - 6 (0.06) 3 (0.03) - 5 (0.05) 118 (1.18) Clause 25 (0.25) - 21 (0.21) 1 (0.01) - - 1 (0.01) - 1 (0.01) - - - 30 (0.3) III. ANALYSIS OF THE CORPUS 295 As can be observed, most tokens are found with simple pronominal subjects, which prefer contractions, especially NotCs, with all operators except be, with which the most common choice is the OpeC variant. With the remaining subjecttypes, with the exception of existential there-constructions, UncNs are more common than contractions. Such a pattern of distribution could be related to the different degree of complexity of the elements functioning as subjects, since, as already mentioned, more complex subjects tend to correlate with a higher frequency of occurrence of uncontracted forms (cf., among others, Quirk et al. 1985: 123, Krug 1998: 289 or Biber et al. 1999: 1129f). In turn, clauses where the subject is the existential there, contractions, especially NotCs, are the predominant type with all operators, thus testifying to the tendency found from OE to PDE for this construction to favour the use of fused forms (cf. Section II.2. above). As with the other corpora examined so far (cf. Sections 2.4. to 2.5.1. above), in order to analyse the potential influence of string frequency and its relation to subject-type in spoken AmE, I have selected two potentially contractible sequences, namely he is not (cf. examples under (329) below) and the president is not (cf. example (330) below), with a pronominal and a nominal subject respectively. A total of 47 instances with sequence he is not have been recorded, out of which 35 show one of the two contracted variants. By contrast, only 13 occurrences of the string the president is not have been found, but none of them is contracted. As these data show, once again, sequences which are more common (here those involving pronominal subjects) favour the use of contractions III. ANALYSIS OF THE CORPUS 296 to a greater extent than those which do not occur so frequently (in my case, those with a NP subject). (329) a. He is not satisfied and he is not in any sense or any fashion going to indicate that we are not committed to meeting our obligations. (CSPAE WH95) b. I don't like to take Mike McCurry’s name in vain when he’s not here to defend himself, but I think Mike would love for them to go camping. (CSPAE WH95) (330) And the President is not going to go there with an instant solution. (CSPAE WH95) 3.5.3. The Wellington Spoken Corpus (WSC) This section is concerned with the analysis of the variation between the three alternative forms of negation in spoken NzE. As mentioned in Section 3.3. above, the WSC contains 3,755 examples of the variants at issue. These 3,755 examples are distributed as follows: 210 for UncNs (NF 2.1), 1,837 for OpeCs (NF 18.37) and 1,708 for NotCs (NF 17.08), as shown in Figure 22 below. Figure 22. UncN/OpeC/NotC distribution in WSC The most significant conclusion to be drawn from this figure is that, in spoken NzE texts, contractions (NF 35.45) are clearly preferred to uncontracted forms (NF 2.1). Moreover, as regards contractions, OpeCs (NF 18.37) predominate over NotCs (NF 17.08). Thus, in general terms, the results from the analysis of the data III. ANALYSIS OF THE CORPUS 297 from spoken NzE are closer to those obtained for spoken AmE (cf. Section 3.5.2. above) than to those of spoken BrE (cf. Section 3.5.1.). However, this statement should be verified taking into account other factors, such as type of text, operator and type of subject. This is to what the following paragraphs will be devoted. As concerns text-type, the distribution of the 3,755 examples recorded in WSC is given in Table 99 below. Table 99. Negative examples according to text-type in the WSC Number of words in each category UncNs OpeCs NotCs TOTAL DGB 80,000 31 (3.87) 185 (23.12) 129 (16.12) 345 DGI 80,000 32 (4) 96 (12) 105 (13.12) 233 DGU 20,000 22 (11) 18 (9) 14 (7) 54 DGZ 100,000 28 (2.8) 220 (22) 185 (18.5) 433 DPC 500,000 43 (0.86) 1,060 (21.2) 984 (19.68) 2,087 DPF 70,000 9 (1.28) 128 (18.28) 134 (19.14) 271 DPH 20,000 3 (1.5) 22 (11) 20 (10) 45 DPP 30,000 1 (0.33) 27 (9) 39 (13) 67 MSN 20,000 4 (2) 2 (1) 23 (11.5) 29 MST 4,000 9 (22.5) 2 (5) 5 (12.5) 16 MSW 28,000 - - - - MUC 12,000 5 (4.16) 27 (22.5) 31 (25.83) 63 MUJ 24,000 4 (1.66) 4 (1.66) - 8 MUL 10,000 16 (16) 32 (32) 24 (24) 72 MUS 2,000 3 (15) 14 (70) 15 (75) 32 TOTAL 1,000,000 210 (2.1) 1,837 (18.37) 1,708 (17.08) 3,755 As the data in this table evince, spoken NzE reveals an overall preference for contractions in all text-types, with the exception of Broadcast monologue (MST), where uncontracted forms predominate (NF 22.5 for full forms vs. 17.5 for III. ANALYSIS OF THE CORPUS 298 contractions), and Judge’s summation (MUJ), where only UncNs and OpeCs have been identified and are alike in number (NF 1.66). In sum, in spoken NzE texts, even very formal categories, such as Parliamentary debate (DGU), favour the use of negative contractions (NF 11 for UncNs vs. 16 for contracted forms). It must be noted, however, that in this category the difference between full and contracted forms is not so conspicuous as in the most informal text-types, where the frequency of contractions is very high. Consider in this respect the NF per 10,000 words in MUS (Teacher monologue) (15 for UncNs vs. 145 for the two contracted variants), MUC (Sport commentary) (4.16 for uncontracted negatives vs. 48.33 for contracted ones) or DPC (Conversation) (0.86 vs. 40.88). As regards the selection of one of the two types of contractions, the data in Table 99 evince that in seven categories (DGI (Broadcast interview), DPF (Telephone conversation), DPP (Social dialect interview), MSN (Broadcast news), MST (Broadcast monologue), MUC (Sports commentary) and MUS (Teacher monologue), NotCs predominate over OpeCs. By contrast, in the remaining seven categories where occurrences of the negative variants under consideration have been recorded, OpeCs are preferred. In what follows each operator will be analysed on its own. Table 100 below gives the distribution of occurrence of the three variants under analysis with the operator be. The most significant conclusion to be gained from this table is that the be-operator shows an overall tendency for negative contractions (NF 19.89 vs. UncNs 1.57). Such a preference is the strongest among the three spoken corpora (16.86 vs. 2.1 for UncNs in LLC, and 18.43 vs. 6.34 for UncNs in III. ANALYSIS OF THE CORPUS 299 CSPAE) (cf. Tables 80 and 90 in Sections 3.5.1. and 3.5.2. above, respectively). This tendency is corroborated by the analysis of individual text-types, since contractions predominate over uncontracted forms in all categories, except MUJ (Judge’s summation), where UncNs and contractions (OpeCs) are evenly distributed (NF 1.66). Such a predominance is, nevertheless, especially noticeable in informal categories, such as DPC (Conversation) (35 instances of UncNs, NF 0.7 vs. 1,127 occurrences of contractions, NF 22.54) or MUS (Teacher monologue) (one example of UncN, NF 5 vs. 16 cases of contractions, NF 80), and DPP (Social dialect interview), where no full forms have been recorded. Table 100. The operator be in WSC Number of words in each category UncNs OpeCs NotCs TOTAL DGB 80,000 25 (3.12) 182 (22.75) 25 (3.12) 232 DGI 80,000 25 (3.12) 96 (12) 18 (2.25) 139 DGU 20,000 13 (6.5) 18 (9) 2 (1) 33 DGZ 100,000 23 (2.3) 214 (21.4) 14 (1.4) 251 DPC 500,000 35 (0.7) 1,055 (21.1) 72 (1.44) 1,162 DPF 70,000 8 (1.14) 128 (18.28) 15 (2.14) 151 DPH 20,000 3 (1.5) 22 (11) 1 (0.5) 26 DPP 30,000 - 27 (9) 4 (1.33) 31 MSN 20,000 2 (1) 2 (1) 4 (2) 8 MST 4,000 3 (7.5) 2 (5) 4 (10) 9 MSW 28,000 - - - - MUC 12,000 4 (3.33) 27 (22.5) 3 (2.5) 34 MUJ 24,000 4 (1.66) 4 (1.66) - 8 MUL 10,000 11 (11) 32 (32) 2 (2) 45 MUS 2,000 1 (5) 14 (70) 2 (10) 17 TOTAL 1,000,000 157 (1.57) 1,823 (18.23) 166 (1.66) 2,146 III. ANALYSIS OF THE CORPUS 300 The analysis of individual text-types also confirms the general preference of be for OpeC, which is more frequent than NotC in most categories, with the exception of MSN (Broadcast news) and MST (Broadcast monologue), both of which contain a very low proportion of relevant negative forms. Notice also that 1,823 out of a total of 1,837 OpeCs in the whole corpus correspond to the operator be. As was the case with the CSPAE, this may explain the general preference of spoken NzE for OpeCs mentioned above when discussing the general data (cf. Table 99 above). The predominance of OpeCs is also evident with individual forms of the verb. Thus, the three potentially contractible sequences of the be-operator analysed in this piece of research (am not, are not and is not) favour the use of OpeCs, not only over the variant contracted type, but also over full forms (four instances of UncNs vs. 404 for contractions with am not, 61 vs. 431 with are not and 92 vs. 1,154 with is not), as Table 101 below clearly shows. In this respect spoken NzE does not differ from spoken BrE and spoken AmE (cf. Sections 3.5.1. and 3.5.2. above). In WSC, two instances have been recorded of the form ain’t corresponding to the operator be (is not), 78 both of them in Radio talkback (DGB). These are the following: (331) it’s not he ain’t heavy he’s my brother (WSC#DGB005:0295:HS) (332) i also understand that er that you know there are people who are never going to be able to make ends meet and it’s easy for someone like me i have an income and I’m comfortable to sit here and pontificate on these sorts of subjects but glory be if the economy’s 78Two other instances with such a form correspond to the operator have, as seen below. III. ANALYSIS OF THE CORPUS 301 not working then it ain’t going to work for anybody (WSC#DGB056:0880:HG) Table 101. Individual forms of the be-operator in WSC Am not Are not Is not UncN OpeC NotC UncN OpeC NotC UncN OpeC NotC DGB 2 43 - 12 48 10 11 91 15 DGI - 13 - 9 26 3 16 57 15 DGU - 6 - 5 5 - 8 7 2 DGZ - 51 - 9 47 5 14 116 9 DPC 2 231 - 14 192 28 19 632 44 DPF - 34 - 3 24 4 5 70 11 DPH - 10 - - - - 3 12 1 DPP - 7 - - 8 2 - 12 2 MSN - - - 1 1 - 1 1 4 MST - - - 1 1 1 2 1 3 MSW - - - - - - - - - MUC - 1 - 2 6 - 2 20 3 MUJ - - - 2 2 - 2 2 - MUL - 5 - 2 13 - 9 14 2 MUS - 3 - 1 4 1 - 7 1 TOTAL 4 404 - 61 377 54 92 1,042 112 Another feature of the be-operator in spoken NzE which is common to its BrE counterpart (LLC) is the higher proportion of contracted forms of be as a lexical verb than as an auxiliary for both passive and progressive constructions (93.23% for main verb be vs. 91.31% for auxiliary be) (Cf. Table 102 below). In this respect my data confirm the assertions by Philips & Reynolds (1987) or Westergren (1998), contra Quirk et al.’s (1985: 123) statement. III. ANALYSIS OF THE CORPUS 302 Table 102. Lexical be and auxiliary be in WSC Full forms Contractions Total Lexical be 104 (6.77%) 1,432 (93.23%) 1,536 Progressive 27 (5.87%) 433 (94.13%) 460 Auxiliary be 557 (91.31%) Passive 26 (17.33%) 124 (82.67%) 150 As regards the distribution of OpeCs and NotCs with auxiliary and lexical verb be, no significant differences are detected between the WSC and its spoken companions LLC and CSPAE. As illustrated in Table 103 below, NotCs are more common with lexical be than with auxiliary be, while OpeCs are somewhat more frequent with auxiliary be. However, what is most noticeable in this corpus is that, unlike the written corpora, in WSC, NotCs seem to be preferred with passive than with progressive sentences, in clear contrast with the tendency found by Biber et al. (1999: 1129) in their data. Table 103. OpeCs and NotCs with lexical be and auxiliary be in WSC OpeCs NotCs Total Lexical be 1,302 (90.92%) 130 (9.08%) 1,432 Progressive 410 (94.69%) 23 (5.31%) 433 Auxiliary be 521 (93.54%) 36 (6.46%) Passive 111 (89.52%) 13 (10.48%) 124 The distribution of examples in WSC with the have-operator is shown in Table 104 below. Broadly speaking, this operator behaves in the same way as the previous one, since negative contractions (NF 7.58) are markedly preferred to full forms (NF 0.22). Moreover, contractions are the only alternative variant in DPF (Telephone conversations), DPH (Oral history interviews) and MUS (Teacher III. ANALYSIS OF THE CORPUS 303 monologue). However, in contrast to the operator be, NotCs are more commonly used than OpeCs in all text-types, thus confirming my expectations. As a matter of fact, OpeCs with have are found in only 13 examples in the whole corpus (NF 0.13). All this leads me to conclude that, in spoken NzE, the selection of one of the three variants under consideration with the operator have is not necessarily related to the degree of formality of the texts, since contractions, mainly NotCs, are favoured not only in informal texts, such as Conversation (DPC), but also in more formal categories, such as Parliamentary debate (DGU). Table 104. Instances with the have-operator in WSC Number of words in each category UncNs OpeCs NotCs TOTAL DGB 80,000 2 (0.25) 3 (0.37) 46 (5.75) 51 DGI 80,000 4 (0.5) - 48 (6) 52 DGU 20,000 3 (1.5) - 6 (3) 9 DGZ 100,000 3 (0.3) 5 (0.5) 96 (9.6) 104 DPC 500,000 2 (0.04) 5 (0.1) 422 (8.44) 429 DPF 70,000 - - 61 (8.71) 61 DPH 20,000 - - 6 (3) 6 DPP 30,000 1 (0.33) - 16 (5.33) 17 MSN 20,000 1 (0.5) - 5 (2.5) 6 MST 4,000 3 (7.5) - 1 (2.5) 4 MSW 28,000 - - - - MUC 12,000 1 (0.83) - 17 (14.16) 18 MUJ 24,000 - - - - MUL 10,000 2 (2) - 12 (12) 14 MUS 2,000 - - 9 (45) 9 TOTAL 1,000,000 22 (0.22) 13 (0.13) 745 (7.45) 780 III. ANALYSIS OF THE CORPUS 310 (Teacher monologue) two examples have been recorded, one with an UncN and another showing a NotC (cf. examples (349) and (350) below). Table 107. Occurrences with the would-operator in WSC Number of words in each category UncNs OpeCs NotCs TOTAL DGB 80,000 4 (0.5) - 40 (5) 44 DGI 80,000 2 (0.25) - 26 (3.25) 28 DGU 20,000 1 (0.5) - 5 (2.5) 6 DGZ 100,000 1 (0.1) 1 (0.1) 34 (3.4) 36 DPC 500,000 4 (0.08) - 292 (5.84) 296 DPF 70,000 1 (0.14) - 22 (3.14) 23 DPH 20,000 - - 11 (5.5) 11 DPP 30,000 - - 12 (4) 12 MSN 20,000 - - 2 (1) 2 MST 4,000 1 (2.5) - - 1 MSW 28,000 - - - - MUC 12,000 - - 5 (4.16) 5 MUJ 24,000 - - - - MUL 10,000 3 (3) - 4 (4) 7 MUS 2,000 1 (5) - 1 (5) 2 TOTAL 1,000,000 18 (0.18) 1 (0.01) 454 (4.54) 473 Moreover, OpeCs with would are practically non-existent in WSC (only one example in Transactions and meetings (DGZ); cf. example (351) below). Also noteworthy is the high number of contractions found in Conversation (DPC) (292 examples), a feature also shared by the have and will operators (cf. Tables 105 and 106 above, respectively). (348) you may associate her with the english literary context of d h lawrence john middleton murray virginia woolf and lady ottoline morrell and you would not be wrong (WSC#MST043:0050:KC) III. ANALYSIS OF THE CORPUS 311 (349) most of you in this class would not and you will need to work well would not do as well as you will do when you’ve got when you work steadily throughout the year (WSC#MUS003:0285:TT)80 (350) you’re SO purified and sensitive that i wouldn’t WANT any contact with with this man’s body catherine (WSC#MUS001:0285:FG) (351) and they buy it in obviously for um for their own purposes for making waistcoats for suits and stuff like that and there there’d not there wouldn’t i think be a place that you could buy it on a retail basis (WSC#DGZ111:0085:SF) To summarise so far, in spoken NzE contracted forms are preferred to a greater extent than full forms with all operators, although the frequency of contractions is higher with be than with the remaining verbs. Moreover, all operators except be favour the use of NotCs to OpeCs, thus confirming what has been mentioned in the literature by scholars such as Dillard (1980), Freeborn (1986), Hughes & Trudgill (1996) and Anderwald (2002), among others (cf. Sections II.3.2.3.1. and II.4.1. above). The last factor considered in the analysis of this corpus is the type of subject. Consider in this respect the data in Table 108 below. Concerning the selection of contractions and full forms, no significant differences can be detected as regards subject-types, since contractions are preferred to UncNs in all cases, except with complex NP subjects with the operator be. However, the low number of examples recorded for this subject-type do not allow to draw definite conclusions. What is significant here is the distribution of OpeCs and NotCs with 80 The second occurrence of would not recorded in this example has not been included in the total count, since OpeC is not allowed in such a context. It is therefore included in the group of KO contexts for OpeCs (cf. Section 3.2.1. above). III. ANALYSIS OF THE CORPUS 312 the operator be. Although there is an overall predominance of OpeC over NotC with this verb, the latter contracted type is more frequent with simple NP subjects (53 examples of OpeCs vs. 64 of NotCs). By contrast, with simple pronominal subjects OpeCs are preferred (1,706 occurrences of OpeCs vs. 56 of NotCs). Finally, as regards string frequency, those potentially contractible sequences which are more numerous in the corpus are also the ones which favour the use of contractions to a greater extent. Thus, as in the preceding corpora, in spoken NzE fairly frequent strings, such as he is not (cf. examples under (352) below), which appears in WSC 78 times, exhibit a high degree of contraction, 76 examples (97.44%). In contrast, more complex sequences, such as the price is not (cf. example (353) below), occur less commonly (only three times) and always show a full form. The WSC data, thus, serve to confirm the relevance of string frequency to the variation between contracted and uncontracted forms (cf. Section II.3.1.5. above and Krug 1998). (352) a. he is not beautiful he's not beautiful but he's HANDSOME. (WSC#DPC206:1655:JM) b. he he really sounded as as if he was unbalanced and er i i think he’s not suited to the job. (WSC#DGB023:0180:Z1) (353) right well if you buy something and the price is not set or discussed in advance then the guarantee is that you need only pay a reasonable price (WSC#DGI158:0235:AW) Table 108. Distribution of examples according to type of subject and operator in WSC BE HAVE WILL WOULD TOTAL UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs Simple 71 (0.71) 1,706 (17.06) 56 (0.56) 8 (0.08) 11 (0.11) 684 (6.84) 8 (0.08) - 292 (2.92) 15 (0.15) - 404 (4.04) 3,255 (32.55) Pronoun Complex 1 (0.01) - - - - - - - - - - - 1 (0.01) Simple 71 (0.71) 53 (0.53) 64 (0.64) 13 (0.13) - 54 (0.54) 4 (0.04) - 43 (0.43) 3 (0.03) - 38 (0.38) 343 (3.43) NP Complex 5 (0.05) - 1 (0.01) - - 2 (0.02) 1 (0.01) - 3 (0.03) - - 2 (0.02) 14 (0.14) Compound - - 2 (0.02) 1 (0.01) - 1 (0.01) - - - - - - 4 (0.04) Exist. there 3 (0.03) 64 (0.64) 35 (0.35) - 2 (0.02) 3 (0.03) - - 4 (0.04) - 1 (0.01) 9 (0.09) 121 (1.21) Clause 6 (0.06) - 8 (0.08) - - 1 (0.01) - - 1 (0.01) - - 1 (0.01) 17 (0.17) III. ANALYSIS OF THE CORPUS 314 3.5.4. Summary The data from the three spoken corpora analysed in the preceding sections show that contracted forms are preferred to a greater extent than the uncontracted variant in the spoken medium. As far as contractions are concerned, NotCs predominate over OpeCs only in BrE, while OpeCs are the prevailing contracted type both in AmE and in NzE. This is probably so due to the large number of examples with the verb be which have been recorded in the CSPAE and the WSC corpora. As regards text-types, in the spoken texts of the three dialects under consideration, contractions are the predominant type of negation, not only in the less formal categories, but also in the most formal ones. Concerning the behaviour of individual operators, contractions turn out to be the most frequent way of negation with all verbs and in all dialects. As regards the two types of fused forms, spoken BrE, AmE and NzE show an overall preference for NotCs to OpeCs with all operators except be. Finally, as regards the distribution of negative forms according to the type of subject, it must be noted that contractions, mainly NotCs, are the most frequent variant with all kinds of subject, with the exception of NPs with all operators in CSPAE, with which UncNs outnumber contractions. III. ANALYSIS OF THE CORPUS 315 3.6. Comparisons across Corpora This section is devoted to the comparison of the results obtained from the individual analyses of the nine corpora under study presented in Sections 3.4. and 3.5. above. These comparisons allow me to establish: (a) diachronic differences and similarities from the 1960s to the 1990s between the two written BrE corpora, LOB and FLOB (Section 3.6.1.1. below) and between the two written AmE corpora, BROWN and FROWN (Section 3.6.1.2.); (b) dialectal differences and similarities between all the written corpora, on the one hand (Section 3.6.2.1.), and all the spoken corpora (Section 3.6.2.2.), on the other; and, finally, (c) differences and similarities regarding medium, i.e. written vs. spoken texts from different dialects dating from the same time span (Section 3.6.3.), such as LLC vs. LOB and FLOB (Section 3.6.3.1.), CSPAE vs. FROWN (Section 3.6.3.2.), and WSC vs. WWC (Section 3.6.3.3.). 3.6.1. Changes over Time: from the 1960s to the 1990s It is generally acknowledged that all living languages are constantly subject to change over time. The variation between contracted and uncontracted negative variants should, therefore, not be an exception in this respect. The sections which follow will be concerned with the differences and similarities in written BrE (Section 3.6.1.1.) and in AmE (Section 3.6.1.2.) in the course of the last decades of the twentieth century. III. ANALYSIS OF THE CORPUS 316 3.6.1.1. Diachronic Comparison between the Written British English Corpora: LOB vs. FLOB As seen in Sections 3.4.1. and 3.4.2. above, the LOB corpus contains a larger number of examples of negative constructions with the four operators which allow the three alternatives to negate than the FLOB (1,790 vs. 1,728). However, the latter corpus shows more negative contractions, 682 (NF 6.82) in FLOB vs. 519 (NF 5.19) in LOB (cf. Tables 19 and 29 above). Thus, we witness here a considerable increase in the use of contractions on the part of BrE speakers in the course of the second half of the twentieth century. Both OpeCs and NotCs exhibit a moderate advance, from 1.23 to 2.08 in the case of OpeCs and from 3.96 to 4.74 in the case of NotCs. As regards individual text-types, the tendency just described applies to most text-categories. The most drastic decline of uncontracted forms from the 1960s to the 1990s is found in some fictional categories, namely category K (NF 12.41 in LOB vs. 4.13 in FLOB), category M (NF 9.16 in LOB vs. 5 in FLOB) and category R (NF 12.22 vs. 3.33). By contrast, in Learned and scientific writings (Cat J), Mystery and detective fiction (Cat L) and Romance and love story (Cat P), there is a parallel advance of UncNs in the texts from the 1990s than in their counterparts from the 1960s, as seen in Table 109 below: Table 109. UncNs in categories J, L and P in LOB and FLOB LOB FLOB Cat J NF 12.43 NF 13.87 Cat L NF 5.62 NF 6.04 Cat P NF 7.75 NF 9.82 III. ANALYSIS OF THE CORPUS 317 The only exceptions to the trend correspond to the following groups: (a) Religion (Cat D), where the number of contractions (both OpeCs and NotCs) is considerably lower in the 1990s than in the 1960s (OpeCs NF 1.76 and NotCs 2.35 in LOB vs. OpeCs NF 0.29 and NotCs 0.58 in FLOB). (b) Popular lore (Cat F) and Adventure and western fiction (Cat N), where NotCs are less numerous in FLOB than in LOB (NotCs NF 2.95 and 11.03, respectively in LOB vs. NF 2.15 in F and 9.48 in FLOB). (c) In category P (Romance and love story), there is a decrease of NotCs in the most recent corpus (NF 17.24 in LOB vs. 12.06 in FLOB). (d) The only text-type which resists the introduction of contracted forms even in the 1990s is Miscellaneous (Cat H) (no instances in either corpus), due to obvious reasons of formality. Furthermore, concerning the two contracted types, the categories showing a more significant increase of the OpeC variant are General Fiction (Cat K), Romance and love story (Cat P), Mystery and detective fiction (Cat L), Skills, trades and hobbies (Cat E) and Learned and scientific writings (Cat J), as seen in the following Table: III. ANALYSIS OF THE CORPUS 318 Table 110. OpeCs in categories K, P, L, E and J in LOB and FLOB LOB FLOB NF 1.72 Cat K NF 7.06 NF 5.51 Cat P NF 8.1 NF 4.37 Cat L NF 6.66 - Cat E NF 1.18 - Cat J NF 0.18 As regards NotCs, all text-types except D, F, N and P show a larger number of occurrences in the 1990s than in the 1960s. Clear instances of the increase of NotCs in the course of time can be found both in formal text-types, such as categories B (Press editorial) or C (Press Review), and in more informal categories, such as K (General fiction), L (Mystery and detective fiction) and R (Humour), as seen in Table 111 below, though the rise of such forms is much more conspicuous in the latter type of text. Table 111. NotCs in categories B, C, K, L and R in LOB and FLOB LOB FLOB NF 0.92 Cat B NF 4.44 NF 2.05 Cat C NF 4.11 NF 1 Cat K NF 15 NF 14.37 Cat L NF 21.45 NF 3.38 Cat R NF 10.55 The second feature relevant to the comparison between the LOB and FLOB corpora is the distribution of examples according to operator and text-type. In relation to the be-operator (cf. Tables 20 and 30 above), the number of contractions is also higher in FLOB than in LOB (NF 1.74 in LOB vs. 3.03 in FLOB). In the 1990s there are less UncNs than in the 1960s in all text-types, III. ANALYSIS OF THE CORPUS 319 except in categories J, L, N and P. In the latter three (fictional) categories, however, we also witness a moderate increase of contracted forms. The decrease in number of UncNs from the 1960s to the 1990s is very conspicuous in an informal category such as Humour (Cat R) (NF 7.22 in LOB vs. 1.66 in FLOB). Moreover, the advance of the two contracted variants with the operator be is somewhat more noticeable for OpeCs (NF 1.08 vs. 1.94) than for NotCs (NF 0.66 vs. 1.09). This is especially obvious in some text-types, such as General fiction (Cat K), where OpeCs increase from 1.72 to 6.72 in contrast to NotCs, which show a more modest advance (NF 2.24 vs. 3.27). Regarding individual forms of the be-operator, some differences and similarities have been detected between written BrE from the 1960s and its 1990s counterpart (cf. Tables 21 and 31 above). Firstly, in both corpora the selection of full forms and contractions with the three sequences under consideration (am not, are not, and is not) is similar, since the potentially contractible strings are not and is not prefer full forms while the contracted variant is more frequent with the form am not, as illustrated in Figure 23 below. The main difference between the two corpora is that, as far as contractions are concerned, the preference for OpeCs holds true for the three sequences in the FLOB corpus, while in LOB the string is not favours the use of NotCs rather than OpeCs (45 OpeCs vs. 58 NotCs). Also noticeable is the increase in the use of the form ain’t, both referring to the be and have operators, from the 1960s to the 1990s (four examples in LOB vs. nine in FLOB). Nevertheless, the low number of instances recorded with such a form in the two corpora do not allow me to draw definite conclusions in this respect. III. ANALYSIS OF THE CORPUS 326 of the verb help in the LOB, FLOB, BROWN and FROWN corpora. Such a decrease of UncNs and the parallel increase of contractions are noticeable both in formal (e.g. categories B, C or E) and in informal (e.g. K, M, P or R) text-types. In all of them, except in P and in R, UncNs predominate over contractions in the BROWN corpus, while in FROWN contractions are more common than uncontracted forms. The general increase in the use of contractions in AmE in the latter part of the twentieth century is also witnessed in very formal texts, such as Miscellaneous (Cat H), which shows no evidence of contractions in the 1960s, while in the 1990s the frequency of contracted forms amounts to 1.99. As regards the choice between the two kinds of contractions, it must be noted that both OpeCs and NotCs increase from the 1960s to the 1990s, NotC being the predominant type in both decades in all kinds of texts, except in Religion (Cat D) in BROWN, where OpeCs and NotCs are alike in number, and Learned and scientific writings (Cat J) in FROWN, where OpeCs are somewhat more common than NotCs. Concerning individual operators, significant results have been obtained, especially for be. First of all, as can be seen in Tables 40 and 50 above, this operator also shows fewer UncNs (NF 6.08 vs. 6.04) and more contractions (NF 2.02 vs. 4.32) in the 1990s than in the 1960s. Moreover, while in the BROWN corpus NotCs are the preferred contracted variant, in FROWN OpeCs (NF 2.54) outnumber NotCs (NF 1.78). The difference between the two AmE corpora is shown graphically in Figures (25a-b) below. III. ANALYSIS OF THE CORPUS 327 Figure 25. Contractions with be in BROWN and FROWN a. BROWN b. FROWN However, as mentioned in footnote 18 above, factors usually operate in combination. Therefore, operators need to be also examined in the light of the types of texts in which they appear. A close look at the data reveals that, while in BROWN the most formal text-types (Cat A to Cat J) favour the use of full forms with the operator be, some formal categories in FROWN, namely Press reportage (Cat A) (NF 4.77 for UncNs vs. 7.72 for contractions), Press review (Cat C) (NF 4.41 vs. 4.7) and Skills, trades and hobbies (Cat E) (NF 5.13 vs. 5.96), already show a higher ratio of contractions. As regards the variation between the two contracted types in relation to genre, while in the BROWN corpus NotCs are the predominant choice in most kind of texts (categories A, B, E, F, G, J, K and N), in the FROWN corpus NotCs tend to be less numerous than OpeCs. In this respect, therefore, written AmE shows a stronger tendency to select OpeCs rather than NotCs from the 1960s to the 1990s, thus confirming what has been claimed by scholars such as Dillard (1980), Freeborn (1986), Hughes & Trudgill (1996) and Anderwald (2002), among others (cf. Section II.4.1. above). As in the case of the comparisons between the two written BrE corpora in Section 3.6.1.1. above, the analysis of individual forms of the be-operator reveals significant differences (cf. Tables 41 and 51 above). While in both corpora III. ANALYSIS OF THE CORPUS 328 uncontracted forms are more frequent than contractions with the sequences are not and is not, the latter variant is preferred with the first person singular am not (cf. Figure 26 below). Figure 26. Full forms and contractions with individual forms of the be-operator in BROWN and FROWN 0 50 100 150 200 250 300 350 400 450 am not are not is not am not are not is not BROWN FROWN full forms contracted forms As regards the choice of the contracted variants with individual forms, there are also some similarities and differences between the two corpora. Thus, for instance, the sequence are not favours the selection of NotCs in both BROWN and FROWN, while, by contrast, the form am not, as expected, prefers OpeCs, since there is no standard NotC for the first person singular (cf. Sections II.1., II.3.2.3.1. and II.4.1. above). However, BROWN and FROWN differ in their selection of OpeCs or NotCs with the sequence is not: in the 1960s NotCs clearly outnumber OpeCs, whereas in the 1990s OpeCs have become the predominant contracted form. III. ANALYSIS OF THE CORPUS 329 As in the BrE data discussed above (cf. Section 3.6.1.1.), AmE also exhibits an increase in the use of the form ain’t, both for the operators have and be (34 instances in BROWN and 44 in FROWN). It is also noticeable that contractions are used to a greater extent with the verb be when it functions as an auxiliary for passive or progressive constructions in AmE from the 1960s than when it is used as a copula (28.02% vs. 23.88%, respectively), while in the 1990s contractions are favoured with lexical be (42.07% vs. 40.66%, respectively). Another difference detected between the two written AmE corpora concerns the distribution of the two fused forms. OpeCs predominate with auxiliary be and NotCs with be as a copula in BROWN. By contrast, in FROWN OpeC is the preferred contracted choice irrespective of whether the verb is a copula or an auxiliary. It is also worth noting that in the last decades of the twentieth century passive be favours the use of NotCs, while progressive be prefers OpeCs. The three other operators (have, will and would) show a very similar diachronic development in AmE texts. In the three cases I find less UncNs and more NotCs in the 1990s than in the 1960s, while OpeCs are practically nonexistent with any of the three operators in both BROWN and FROWN, and almost restricted to fictional categories. However, individual text-types differ from operator to operator. Thus, for instance, with have the FROWN corpus prefers contractions to a greater extent than full forms in most text-types, both formal categories, such as A, C or E, and informal ones, such as K to P. This contrasts III. ANALYSIS OF THE CORPUS 330 with the state of affairs found in the 1960s, when contractions predominate only in some fictional categories (L, N, P and R) (cf. Tables 44 and 54 above). As far as individual forms of the have-operator are concerned (cf. Tables 45 and 55 above and Figure 27 below), the written AmE data from the 1990s differ from those of the 1960s in the selection of the negative variants at issue: in the 1960s full forms prevail over contracted ones with the sequences has not and had not, whereas in the 1990s uncontracted forms are only favoured with the string has not (39 instances with UncNs vs. 25 of contractions (NotCs)). Figure 27. Full forms and contractions with individual forms of the have-operator in BROWN and FROWN 0 20 40 60 80 100 120 have not has not had not have not has not had not BROWN FROWN full forms contracted forms As in the case of the be-operator examined above, the dichotomy lexical have vs. auxiliary have also plays an important role. Here the former favours the use of contractions to a greater extent both in BROWN and in FROWN. This also applies to both OpeCs and NotCs in written AmE from the 1990s. By contrast, in the 1960s, fused forms with lexical have are exclusively of the NotC type. III. ANALYSIS OF THE CORPUS 331 Another important feature of the written AmE material analysed in this piece of research is the low number of occurrences of the have got construction, 12 examples in BROWN and only nine in FROWN. Such a low number of instances is related to the fact that, as Quirk et al. (1985: 131f) mention (cf. Section II.4.2. above), in AmE the lexical verb have tends to be negated with the do-operator instead of resorting to the have not got construction. This is borne out by the data in the two corpora. Moreover, as mentioned in Section 3.4.4. above, the use of do-support with the lexical verb have has increased in written AmE in the latter part of the twentieth century (68.83% in BROWN vs. 85.19% in FROWN). Regarding will, there have been a decrease of full forms and a parallel advance of contracted variants from the 1960s to the 1990s, as the data in Tables 46 and 56 clearly show (NF 0.95 for UncNs and 0.94 for contractions in BROWN vs. 0.87 and 1.38, respectively, in FROWN). This is especially conspicuous in some text-categories, such as Press editorial (Cat B) (NF 2.77 for UncNs and 1.48 for NotCs in BROWN vs. 0.55 and 5.55, respectively, in FROWN). In the case of would (cf. Tables 47 and 57 above), full forms are more common than contractions in the BROWN corpus (NF 1.45 for UncNs vs. 1.2 for contracted forms), while in FROWN the balance is reversed in favour of contractions (NF 0.99 vs. 1.25). Thus, from a diachronic point of view, in written AmE, we witness a moderate rise in the number of contractions and a parallel decrease of full forms with the operator would. This is clearly reflected in the III. ANALYSIS OF THE CORPUS 332 informal category K (General fiction) (NF 2.93 for UncNs and 1.54 for contractions in BROWN vs. 0.86 and 3.1, respectively, in FROWN). Concerning the patterns of distribution of the three negative variants depending on the nature of the subject both in the 1960s and in the 1990s, some conclusions can be drawn from the comparison of the data in Tables 48 and 58 above. In the BROWN corpus, contracted forms of all operators except be are preferred to UncNs with simple pronominal subjects; in the FROWN corpus, on the other hand, contractions are favoured over UncNs, irrespective of the operator being used. However, the proportional difference between full forms and contractions with simple pronominal subjects is higher in FROWN than in the BROWN corpus. Furthermore, in both corpora, with the exception of will in FROWN, more complex subjects, such as NPs, tend to occur with uncontracted forms. However, when the subject is existential there, the tendency is to use contractions with the be and would operators in the BROWN corpus and with be, have and would in the FROWN corpus. The strong preference for contractions with simple pronominal subjects can be related to the notion of string frequency discussed by Krug (1998: 294) (cf. Section II.3.1.5. above). Sequences such as he is not or he had not, which are very frequently used, are those with show a higher ratio of contracted variants. By contrast, strings such as man is not or the injustice had not, which are not so common, do not favour the use of contractions. Thus, once again, the most III. ANALYSIS OF THE CORPUS 333 frequent sequences are those which favour the use of contractions to a greater extent. 3.6.2. The Dialect Factor: British English, American English, Australian English and New Zealand English 3.6.2.1. Dialect Distinctions in Written English: a Comparison between LOB, FLOB, BROWN, FROWN, ACE and WWC This section is concerned with the study, from a dialectal point of view, of the differences and similarities, between all the written corpora under consideration. For this purpose, I will compare the data of the LOB corpus with those of FLOB, BROWN, FROWN, ACE and WWC. It should be noted that, although the two latter corpora date from the 1980s, I have decided to compare them not only with the BrE and AmE corpora from the 1990s, but also with those from the 1960s in order to check whether the AusE and NzE texts are more closely related to the BrE and/or AmE data from the 1960s, or, by contrast, they are more similar to the most recent BrE and/or AmE texts. As seen in Tables 19, 29, 39, 49, 59 and 69 above, uncontracted negative forms are the predominant option in all dialects. This is not at all surprising, given that these corpora comprise exclusively written texts. The data also show the existence of a greater fondness for contracted forms in AmE, mainly in the 1990s, than in the remaining dialects examined (BrE, AusE and NzE). A similar opposition between BrE and AmE, with the latter variety being more advanced III. ANALYSIS OF THE CORPUS 334 than the former, has also been proved to hold for other areas of grammar. Thus, for example, Mair (2002), in his study of the proportional use of the gerund after the verb begin in LOB, FLOB, BROWN and FROWN, finds a “persistent contrast between British and American usage” (2002: 116f), AmE making use of the construction to a greater extent than BrE. A comparison of the individual analyses of the six written corpora provided in Section 3.4. above allows me to organise these four dialects along a continuum in the way displayed in Figure 28 below. Figure 28. Degree of contraction in written texts according to dialect Another similarity between the six dialects under consideration is their preference to select NotCs over OpeCs, as shown in Figure 29 below. Figure 29. Distribution of negative occurrences in the six written corpora 0 200 400 600 800 1000 1200 1400 LOB FLOB BROWN FROWN ACE WWC UncNs OpeCs NotCs Nevertheless, the proportion between uncontracted negatives and contractions varies from dialect to dialect. Thus, while BrE from the 1960s (LOB) shows the highest difference between the two variants, AmE from the 1990s III. ANALYSIS OF THE CORPUS 335 (FROWN) presents the lowest, the remaining corpora occupying an intermediate position. As far as contractions are concerned, both AusE and NzE are more closely related to BrE than to AmE, since the difference between the two contracted forms in ACE and WWC is not so conspicuous as in the BROWN and the FROWN corpora. The comparison of individual categories in the six written corpora reveals the following: 81 (a) In AmE from the 1960s, full forms are less numerous (NF 12.71 in LOB vs. 10.54 in BROWN) and contractions are more frequent (NF 5.19 in LOB vs. 5.6 in BROWN) than in its BrE counterpart (cf. Tables 19 and 39 above). However, as far as individual categories are concerned, this statement only holds true for the formal categories: Press review (Cat C), Skills, trades and hobbies (Cat E) and Belle letters, memoirs and biographies (Cat G). This allows me to claim that in AmE from the 1960s, contractions are used to a greater extent than in BrE texts from the same decade. Nevertheless, contrary to my expectations, in the most informal text-type, Fiction (Cat K), both full forms and contractions are less frequent in BROWN than in the LOB corpus (NF 8.21 for UncNs and 17.57 for contractions in LOB vs. 6.9 and 16.42, respectively in BROWN). 81 For this section, it was deemed advisable to group categories K to R in LOB, FLOB, BROWN, FROWN and ACE under category K, named fiction, in order to compare the data from these corpora with those from the WWC. This grouping has been maintained when dealing with individual operators and text-types. The general data corresponding to category K in each corpora after grouping the aforementioned categories are the following: LOB: UncNs 207 (8.21), OpeCs 104 (4.12) and NotCs 339 (13.45). FLOB: UncNs 151 (5.99), OpeCs 165 (6.54) and NotCs 346 (13.73). BROWN: UncNs 174 (6.9), OpeCs 72 (2.85) and NotCs 342 (13.57). FROWN: UncNs 161 (6.38), OpeCs 140 (5.55) and NotCs 365 (14.48). ACE: UncNs 55 (2.18), OpeCs 46 (1.82) and NotCs 99 (3.92). III. ANALYSIS OF THE CORPUS 342 operator. As far as fused forms are concerned, OpeCs in LOB, BROWN, ACE and WWC are more common with have as the auxiliary for the perfect, whereas in FLOB and FROWN, both types of fused forms predominate with lexical have. In this respect both written AusE and NzE are closer to BrE and AmE from the 1960s. Another important difference found between the six written corpora concerns the negation of the operator have followed by got. In AusE the combination is practically non-existent (1.88% of the relevant cases), whereas the number of instances is higher in BrE (4.11% in LOB and 5.26% in FLOB), NzE (4.59%) and AmE (3.43% in BROWN and 2.55% in FROWN). Although examples of this kind are too few to draw definite conclusions, the data suggest that in BrE there is an increasing tendency to use this construction, while in AmE its use seems to be decreasing. This can be related to the fact that, as mentioned by scholars such as Quirk et al. (1985: 131f), the trend in AmE is to negate the lexical verb have with the do-operator instead of using the have not got construction (cf. Section II.4.2. above). Concerning the use of the auxiliary do, the analysis of the different corpora yields interesting results. All corpora except LOB prefer negation with do not to that with not alone. However, the proportion of the former construction is higher in AmE (68.83% in BROWN and 85.19% in FROWN), in AusE (87.21% in ACE) and NzE (77.32% in WWC) than in BrE from the 1990s (63.11%). III. ANALYSIS OF THE CORPUS 343 However, the most significant difference between the six corpora analysed in this section concerns the operator will (cf. Tables 26, 36, 46, 56, 66 and 76 above). As shown in Figure 32 below, three corpora show a slightly higher proportion of contractions than of UncNs, namely LOB (NF 1 for UncNs vs. 1.01 for contractions), FROWN (NF 0.87 vs. 1.38) and ACE (0.69 vs. 0.7), while the remaining three, FLOB (NF 1.14 vs. 0.94), BROWN (NF 0.95 vs. 0.94) and WWC (NF 1.07 vs. 0.96), show the reverse situation. Thus, referring to the will-operator, the AusE corpus is clearly related to BrE from the 1960s and AmE from the 1990s, while the NzE corpus is, once again, closer to BrE from the 1990s and AmE from the 1960s. Figure 32. Negative occurrences with the will-operator in the six written corpora 0 20 40 60 80 100 120 140 LOB FLOB BROWN FROWN ACE WWC UncNs OpeCs NotCs The preferred contracted form for will is that of NotCs in all dialects and in all texts. Individual categories may, nevertheless, behave differently in the six corpora. Thus, for example, in the three corpora where contractions outnumber UncNs, the latter prevail in the most formal categories, from Cat A to Cat J, with the exception of Cat F (Popular lore) in LOB, in C (Press review) and D (Religion) and E (Skills, trades and hobbies) in FROWN and in A (Press reportage), C (Press review) and E (Skills, trades and hobbies) in ACE. In III. ANALYSIS OF THE CORPUS 344 contrast, in the three other corpora contractions predominate in category K (General fiction) in FLOB, in categories C (Press review), E (Skills, trades and hobbies) and K (General fiction) in BROWN and in C (Press review) and K (General fiction) in WWC. Therefore, I can conclude that all dialects favour the use of contractions at the expense of full forms in the most informal categories, i.e. those somewhat related to the spoken medium (Cat K), while the most formal text-types, such as Learned and scientific writings (Cat J), prefer the uncontracted negative variant. With the operator would, the selection of full forms and contractions varies depending on the dialect and sometimes also on the date of composition of the texts. In BrE (both LOB and FLOB), AmE from the 1960s (BROWN) and AusE (ACE), the preferred variant is that of UncN, while in FROWN and WWC contractions are favoured (cf. Tables 27, 37, 47, 57, 67 and 77 above and Figure 33 below). Thus, while AusE is closer to BrE (both LOB and FLOB) and AmE from the 1960s, NzE is related to AmE from the 1990s. Figure 33. Negative examples with the would-operator in the six written corpora 0 20 40 60 80 100 120 140 160 LOB FLOB BROWN FROWN ACE WWC UncNs OpeCs NotCs III. ANALYSIS OF THE CORPUS 345 As with the other operators already analysed, the selection between the three alternatives of negation should, nevertheless, be studied taking into account text-types, since in all dialects the fondness for contractions is more obvious in Fiction (Cat K), while in formal categories there is a clear predominance of UncNs over contractions. A comparison of the data in Tables 28, 38, 48, 58, 68 and 78 above reveals the patterns of distribution of contracted and uncontracted forms depending on the nature of the subject of the clause in the six corpora under analysis. Thus, for example, in LOB, simple pronominal subjects prefer UncNs with all operators except will (41 occurrences of UncNs vs. 82 occurrences of contracted forms). In FLOB, BROWN and WWC, by contrast, subjects of this kind prefer contractions with all operators except be (260 occurrences of UncNs vs. 231 occurrences of contracted forms in FLOB, 266 vs. 148 in BROWN and 273 vs. 244 in WWC). In ACE only be (173 vs. 130) and have (65 vs. 49) favour UncNs, while in FROWN contractions are preferred with the four operators under consideration. The state of affairs just described clearly contrasts with that obtained for NP subjects. Here all corpora exhibit a similar behaviour, since in all dialects all operators, with the exception of will in FROWN (27 occurrences with UncNs vs. 30 for contracted forms), tend to select UncNs, though the proportion of full forms is higher with the operator be than with the other verbs. Besides, as far as contractions are concerned, NotCs are more frequent than OpeCs with subjects of this kind with all operators, be included. No significant differences between III. ANALYSIS OF THE CORPUS 346 dialects can be detected either for the remaining types of subjects for which a sufficient amount of data has been gathered, with the exception of existentialthere subjects. In such a context, contracted forms outnumber uncontracted negatives with the operators be, will and would in LOB, with be and would in BROWN, with be, have and would operators in FROWN and with be, have and will in WWC. Nevertheless, in FLOB and ACE this holds true only with the beoperator, so that, in this respect, AusE is closer to BrE from the 1990s. Therefore, the statement found in the literature that contractions are favoured in existential clauses (cf. Section II.2. above) is perhaps too general and should be revised taking into account several other factors, such as dialect, operator or subject-type, among others. Finally, another feature shared by the six written corpora is that the most frequent sequences are those which favour the use of the contracted variants. As stated in Sections 3.4.1. to 3.4.6. above, contractions are more likely to occur in a sequence like he is not, which is rather commonly used, than in other less frequent strings, such as the government is not. Thus, Krug’s string frequency factor is confirmed by my written English data for all dialects examined. 3.6.2.2. Dialect Distinctions in Spoken English: a Comparison between LLC, CSPAE and WSC In this section I will be concerned with the comparison between the three spoken corpora, LLC, CSPAE and WSC, which will allow me to establish differences and similarities between BrE, AmE and NzE from a dialectal point of view. Judging III. ANALYSIS OF THE CORPUS 347 from the data in Tables 79, 89 and 99 above, I can conclude that contractions are, by far, the preferred negative variant in the spoken language of the three dialects under analysis, as Figure 34 below clearly demonstrates. This constitutes a clear difference between spoken English and the written corpora discussed in Section 3.6.2.1. above, which showed a preference for the uncontracted form in all dialects (cf. Figure 29 above). As can be seen, the proportional difference between contractions and uncontracted negatives is very strongly marked for BrE (NF 2.98 for UncNs vs. 31.86 for contractions) and NzE (NF 2.1 vs. 35.45), while it is not so conspicuous for AmE (NF 11.76 vs. 26.16). Such a difference may be due to the greater formality of the texts included in the AmE corpus in comparison to those of the LLC and WSC (cf. Section III.1.2. above). Figure 34. Distribution of negative occurrences in the three spoken corpora 0 300 600 900 1200 1500 1800 LLC CSPAE WSC UncNs OpeCs NotCs Thus, unlike the written corpora, where AmE favoured the use on contractions to a greater extent than the other dialects (cf. Figure 28 in Section 3.6.2.1. above), among the spoken corpora, AmE is precisely the dialect which shows a lower proportion of contractions. The analysis of the three individual corpora yields the continuum given in Figure 35 below: III. ANALYSIS OF THE CORPUS 348 Figure 35. Degree of contraction in spoken texts according to dialect Another important difference between the three spoken corpora concerns the distribution of the two types of contracted variants. Thus, while in BrE NotCs predominate over OpeCs (NF 13.76 for OpeCs vs. 18.1 for NotCs), in both AmE (NF 16.67 vs. 9.49) and NzE (NF 18.37 vs. 17.08), OpeCs are favoured to a greater extent than NotCs. This may be explained by considering the high ratio of tokens recorded with the operator be in the CSPAE and WSC corpora. As seen in Tables 80, 90 and 100 above, the three dialects behave similarly with this verb, since contractions represent by far the preferred negative variant and, among contracted types, OpeCs are vastly more frequent than NotCs (cf. Figure 36 below). However, the proportion of OpeC forms is higher in the NzE corpus (NF 18.23 for OpeCs vs. 1.66 for NotCs) than in the AmE (NF 16.54 vs. 1.89) or the BrE corpora (NF 13.04 vs. 3.82). This way, the spoken corpora corroborate the information provided in the literature on the topic (cf. Section II.4.1. above). Figure 36. The be-operator in the three spoken corpora 0 300 600 900 1200 1500 1800 LLC CSPAE WSC UncNs OpeCs NotCs III. ANALYSIS OF THE CORPUS 349 When dealing with the behaviour of the individual forms of the operator be which allow the three alternatives to negate, no difference is detected between the three dialects either. Once again, contractions, mainly OpeCs, are preferred to a greater extent than UncNs with the sequences am not, are not and is not, ’s not being the most common string in the three spoken corpora (cf. Tables 81, 91 and 101 above). The use of contractions with the lexical verb be vs. the auxiliary verb be is also a clear dialectal marker between the three spoken dialects. In BrE and NzE contractions are more frequent with lexical be, while in AmE such fused forms are used to a greater extent with the auxiliary be (cf. Tables 82, 92 and 102 above). Nevertheless, when dealing with the variation between OpeCs and NotCs, no noticeable distinction is detected between the three dialects: the NotC type predominates with lexical be (cf. Tables 83, 93 and 103 above), while OpeCs are preferred with auxiliary be. Besides, the three dialects share a stronger preference for NotC in passive constructions than in progressives. As can be seen in Figure 37 below, the data for the verb have differ from those of the previous operator in that the former shows a preference for NotCs rather than OpeCs or UncNs in all three corpora (cf. Tables 84, 94 and 104). Nevertheless, the proportion between full forms and contractions is not so conspicuous in CSPAE (NF 2.78 for UncNs vs. 3.84 for contracted forms) as in the two other corpora (NF 0.42 for UncNs vs. 8.9 for contractions in LLC; and III. ANALYSIS OF THE CORPUS 350 0.22 vs. 7.58 in WSC), probably because of the higher degree of formality of the texts included in the CSPAE corpus. Figure 37. Occurrences with the have-operator in the three spoken corpora 0 100 200 300 400 500 600 700 800 LLC CSPAE WSC UncNs OpeCs NotCs Another difference between the operator have and the be-operator concerns the distribution of the variants with individual forms of these two verbs. While spoken BrE and NzE favour the use of contractions with the sequences have not, has not and had not, in spoken AmE only have not and had not prefer contractions to uncontracted negatives, and both variants are evenly distributed for has not. The data for the use of contracted forms with the operator have when it functions as a lexical verb or as an auxiliary for the perfect are rather similar in LLC, CSPAE and WSC, contractions being more frequent with lexical have, in contrast with the statements made by Sinclair (1990: 453) or Biber et al. (1999: 1129), among others. However, some dialectal differences are detected regarding the use of OpeCs and NotCs, the data in spoken NzE differing from those of BrE and AmE. In the latter dialects, NotCs are more frequently used with lexical have, while OpeCs predominate with auxiliary have. By contrast, in WSC both OpeCs III. ANALYSIS OF THE CORPUS 351 and NotCs are more common with lexical have than with have as perfective auxiliary. Moreover, the combination have got is also a clear dialect marker in spoken English: while in AmE its frequency of occurrence is very low (only two instances, NF per 10,000 words 0.02), in BrE (98 examples, NF 1.96) and, especially, in NzE (158 instances, NF 1.58), it is fairly commonly used in speech. This suggests that, as mentioned by Quirk et al. (1985: 131f), in AmE the operator have functioning as a lexical verb tends to be negated with the do-operator (cf. Section III.4.2. above). This is borne out by the data in my corpus, since in CSPAE 99.08% of the relevant instances of lexical have show negation with the auxiliary do. Such a preference is also witnessed in spoken NzE (WSC), as in the case of the written corpora (cf. Section 3.6.2.1. above), although the frequency of negation with do is lower in WSC than in its written counterpart (WWC). By contrast, in spoken BrE the balance is reversed in favour of negation with not alone (61.05%). Thus, the LLC corpus can be said to occupy an intermediate position between the most conservative written BrE texts, those in the LOB corpus, where do-support is used in only 20% of the cases with lexical have, and the most advanced written material from the FLOB corpus, where the proportion of instances with do is considerably higher (63.11%). The preference for contractions with the operator have also holds true for will (cf. Tables 86, 96 and 106 above and Figure 38 below). However, in the CSPAE corpus, once again, the distribution between the two variants is more even III. ANALYSIS OF THE CORPUS 358 Figure 42. Full forms and contractions with individual forms of the be-operator in LOB, FLOB and LLC 0 50 100 150 200 250 300 350 have not has not had not have not has not had not have not has not had not LOB FLOB LLC full forms contracted forms The choice of contractions in relation to the distinction lexical verb vs. auxiliary verb is also relevant to the have-operator. Generally speaking, contractions are used to a greater extent with lexical have in both written and spoken BrE texts. However, the written and the spoken media differs as regards the distribution of OpeCs and NotCs: while in LOB and in LLC, NotCs are more frequent with lexical have and OpeCs with auxiliary have, in FLOB both OpeCs and NotCs are more common with have as an lexical verb. Another interesting aspect found in the comparison of the BrE corpora is the higher proportion of examples recorded with the form got in spoken texts (NF 1.96 in LLC) than in writing (NF 0.18 in LOB and 0.2 in FLOB). In this connection, it is also worth noting that FLOB favours the pattern of negation with do-support (63.11%), whereas both LOB (20%) and LLC (38.95%) prefer negation without the auxiliary do. III. ANALYSIS OF THE CORPUS 359 The remaining operators, will (cf. Tables 26, 36 and 86 above) and would (cf. Tables 27, 37 and 87 above), show similar patters of distribution in the spoken and the written media: in both cases, written texts prefer UncNs (NF 1.07 for UncNs vs. 0.97 for contractions with will and 1.35 vs. 1.02 with would), while contractions are favoured in the spoken language (NF 0.16 vs. 2.14 and 0.3 vs. 3.96 respectively). Moreover, the proportion of NotCs is higher than that of OpeCs in the three corpora with will and would. In order to compare the behaviour of written and spoken texts in accordance to the type of subject, I have grouped the data corresponding to the LOB and FLOB corpora (cf. Tables 28 and 38 above). The results are given in Table 112 below. The comparison of the data in this table with those of Table 88 above for the LLC corpus suggests that, in spoken BrE, contractions are preferred to UncNs with all kinds of subjects with the exception of complex NPs with the be-operator. By contrast, in written BrE, contractions are only favoured with simple pronominal subjects with the operators will and would, and in existential there-constructions with be, will and would. Thus, the general tendency to associate the use of contractions with those subject-types which are less complex, such as pronouns, and the use of full forms with more complex kinds of subject, such as NPs, it is not always followed, since the variation between alternative negative forms depends on several factors: operator, medium, type of subject and so on. Table 112. Distribution of examples according to type of subject and operator in the written BrE corpora (LOB/FLOB) BE HAVE WILL WOULD TOTAL UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs UncNs OpeCs NotCs Simple 623 (3.12) 285 (1.43) 89 (0.45) 274 (1.37) 13 (0.07) 250 (1.25) 101 (0.51) 11 (0.06) 144 (0.72) 138 (0.69) 3 (0.02) 164 (0.82) 2,095 (10.48) Pronoun Complex 3 (0.02) - 1 (0.01) - - 2 (0.01) 1 (0.01) - - - - - 7 (0.04) Simple 585 (2.93) 11 (0.06) 54 (0.27) 188 (0.94) 1 (0.01) 51 (0.26) 88 (0.44) - 34 (0.17) 109 (0.55) - 32 (0.16) 1,153 (5.77) NP Complex 54 (0.27) - 2 (0.01) 17 (0.09) - 3 (0.02) 17 (0.09) - 1 (0.01) 13 (0.07) - 2 (0.01) 109 (0.55) Compound 1 (0.01) - - 1 (0.01) - 1 (0.01) - - - 1 (0.01) - - 4 (0.02) Exist. there 18 (0.09) 5 (0.03) 22 (0.11) 8 (0.04) 1 (0.01) 1 (0.01) 1 (0.01) - 3 (0.02) 4 (0.02) - 2 (0.01) 65 (0.33) Clause 55 (0.28) 1 (0.01) 7 (0.04) 3 (0.02) - 2 (0.01) 6 (0.03) - 2 (0.01) 5 (0.03) - 1 (0.01) 82 (0.41) Other 2 (0.01) - - 1 (0.01) - - - - - - - - 3 (0.02) IV. GENERAL SUMMARY AND CONCLUSIONS 361 Finally, concerning the string frequency factor (Krug 1998: 294), those sequences which are very commonly found both in written and in spoken BrE texts, especially those with simple pronouns such as he is not, are also the ones which favour the use of contractions to a greater extent. 3.6.3.2. FROWN vs. CSPAE Let us proceed now to the analysis of the behaviour of contracted and uncontracted forms in written and spoken AmE. For this purpose, the data from the two AmE corpora from the 1990s, namely the spoken corpus CSPAE and the written corpus FROWN, will be discussed in detail in the following paragraphs.84 A comparison of the general data from both corpora (cf. Tables 49 and 89 above) yields the following results. As in the case of BrE (cf. Section 3.6.3.1. above), contractions, as expected, clearly predominate over UncNs in spoken AmE (NF 11.76 for UncNs vs. 26.16 for contractions), while in the written language the balance is reversed in favour of uncontracted forms (NF 9.48 for UncNs vs. 8.89 for contractions) (cf. Figure 43 below). However, the less informal categories in FROWN (from Cat K to R) are closer to the spoken texts in that fused forms are the preferred variant. This confirms the general association that contractions are mainly found in spoken material and “in written registers with a large admixture of spoken style, such as fiction writings” (Biber et al. 1999: 1129). 84 I have decided to compare those written and spoken AmE texts which date from the same period. Thus, the BROWN corpus is not included in the analysis since, as mentioned in Section 1.1. above, it comprises only texts from the 1960s. IV. GENERAL SUMMARY AND CONCLUSIONS 362 Figure 43. Negative forms in the AmE corpora a. FROWN b. CSPAE As far as contractions are concerned, NotCs outnumber OpeCs (NF 2.59 for OpeCs vs. 6.3 for NotCs) in the FROWN corpus, while the CSPAE shows a clear preference for OpeCs (NF 16.67 vs. 9.49, respectively) (cf. Figures 44a-b below). Such a predominance of OpeCs in the spoken AmE texts may be related to the high number of instances with be-operator recorded in CSPAE, which amount to 65.32% of the total of relevant instances in the whole corpus (cf. Section 3.5.2. above). Figure 44. OpeCs and NotCs in the AmE corpora a. FROWN b. CSPAE As shown in Tables 50 and 90 above, the be-operator also favours the use of UncNs over contractions in FROWN (NF 6.04 for UncNs vs. 4.32 for contractions), in contrast to the CSPAE (NF 6.34 vs. 18.43). Besides, in both corpora OpeCs predominate over NotCs (NF 2.54 for OpeCs vs. 1.78 for NotCs in FROWN and 16.54 vs. 1.89 in CSPAE), although the proportion of the two fused forms is more prominent in spoken AmE. In view of this, my AmE data IV. GENERAL SUMMARY AND CONCLUSIONS 363 corroborate the assertions by Dillard (1980), Freeborn (1986), Hughes & Trudgill (1996) and Anderwald (2002), among many others (cf. Section II.4.1. above), that the verb be favours the use of OpeCs rather than NotCs. Concerning the individual forms of be, as shown in Tables 51 and 91 above, spoken texts favour the use of contractions to a greater extent than full forms with the am not, are not and is not sequences. By contrast, in written AmE (FROWN) contractions are only preferred with the first person singular am not (cf. Figure 45 below). Figure 45. Full forms and contractions with individual forms of the be-operator in FROWN and CSPAE 0 100 200 300 400 500 600 700 800 am not are not is not am not are not is not FROWN CSPAE full forms contracted forms The data also suggest that the selection of OpeCs or NotCs is conditioned not only by medium, but also by the individual forms involved. Thus, in CSPAE the proportion of OpeCs outnumbers that of NotCs for the three grammatical forms under consideration; in FROWN, on the other hand, the proportion is higher only IV. GENERAL SUMMARY AND CONCLUSIONS 364 for is not and am not, as expected given the lack of a standard NotC for the first person singular). The number of occurrences of the form ain’t recorded in my spoken AmE material is very low, only one example in White House 97B (WH 97B), probably due to the high degree of formality of the texts included in CSPAE (cf. Section 1.2. above). However, in the written corpus, the number of instances with such a form is considerably higher (44 examples), by far the highest among the nine corpora used in the present study. Besides, it is recorded not only in the most informal categories (from Cat K to R, with the exception of Cat M), but also in formal text-types (cf. Section 3.4.4. above), such all the Press categories (from Cat A to C) and Skills, trades and hobbies (Cat E). One noteworthy difference between written and spoken AmE texts is the higher frequency of the contracted variants with the lexical verb be in FROWN and with be as an auxiliary in CSPAE, as shown in Tables 52 and 92 above. Thus, spoken AmE texts corroborate Quirk et al.’s (1985: 123) assertion about “the tendency for the contracted form to be more common when functioning as an auxiliary than as main verb,” while written texts confirm the predominance of fused forms with lexical be mentioned by Philips & Reynolds (1987). As far as OpeCs and NotCs are concerned, the two corpora also behave differently: in the written medium lexical be is more commonly found with OpeCs, while NotCs are preferred with auxiliary be. On the other hand, in spoken AmE texts, the two contracted forms are evenly distributed with both copula be and auxiliary be. IV. GENERAL SUMMARY AND CONCLUSIONS 365 Finally, in both spoken and written texts NotCs are more common with passive be and OpeCs with progressive VPs. The operators have (cf. Tables 54 and 94 above), will (cf. Tables 56 and 96) and would (cf. Tables 57 and 97) behave somewhat differently from be, since both written and spoken texts favour the use of contractions rather than of UncNs (cf. Table 113 below), and NotC is the selected type of fused forms in both corpora. Nevertheless, as already noted, the ratio of both full forms and contractions is higher in the spoken corpus with all operators than in the written one. Table 113. Full forms and contractions with the operators have, will and would in FROWN and CSPAE FROWN CSPAE Full forms NF 1.58 NF 2.78 Have Contractions NF 1.94 NF 3.84 Full forms NF 0.87 NF 1.26 Will Contractions NF 1.38 NF 1.68 Full forms NF 0.99 NF 1.38 Would Contractions NF 1.25 NF 2.21 As regards the individual forms of the have-operator (cf. Tables 55 and 95 above), the data from FROWN differ from those of CSPAE. In the former corpus have not and had not favour the use of contractions; in the latter, by contrast, the preference for contractions is noticeable only for have not (169 instances of UncNs vs. 272 of contractions), while with has not and had not the variants are IV. GENERAL SUMMARY AND CONCLUSIONS 366 evenly distributed (84 UncNs vs. 84 contracted forms, and 25 vs. 28, respectively), as seen in Figure 46 below. Figure 46. Full forms and contractions with individual forms of the have-operator in FROWN and CSPAE 0 50 100 150 200 250 300 have not has not had not have not has not had not FROWN CSPAE full forms contracted forms The two corpora also coincide in their preference for contractions with lexical have. However, noticeable differences are found between spoken and written AmE concerning the variation between the two contracted forms under consideration in relation to the function of have, whether main verb or auxiliary. In FROWN, both OpeCs and NotCs are more frequent with lexical have (contra Sinclair (1990: 453) or Biber et al. (1999: 1129)), while in CSPAE OpeCs are preferred with have as auxiliary for the perfect and NotCs are more common with lexical have. As to type of subject and operator (see Tables 58 and 98 above), the two AmE corpora show similar results. In both FROWN and CSPAE, contractions are preferred with simple pronominal subjects and with all operators. The same IV. GENERAL SUMMARY AND CONCLUSIONS 367 tendency applies to clauses with existential there as subject with all operators but will in FROWN, where no contraction has been recorded. By contrast, the UncN is preferred with more complex kinds of subjects, such as NPs or clauses, with all operators except will in the written corpus. Therefore, the idea that simple subjects favour the use of contractions while more complex subjects prefer full forms is corroborated by my AmE data, in both speech and writing. 3.6.3.3. WWC vs. WSC The last comparison established here is the one between the written NzE corpus (WWC) and its spoken counterpart (WSC) in order to search for the main similarities and differences existing between them. As expected, contractions are far more frequent than UncNs in the spoken corpus, while in the written language UncNs predominate over negative contractions, as can be seen in Figure 47 below (cf. also Tables 69 and 99 above). Figure 47. Distribution of negative contractions and uncontracted negatives in WWC and WSC a. WWC b. WSC The two corpora also show differences as regards the use of contractions. Thus, while in the written corpus there is a predominance of NotCs (NF 4.43) over OpeCs (NF 2.27), the latter variant is the preferred contracted type in the spoken corpus (NF 18.37 OpeCs vs. 17.08 NotCs), as illustrated in Figure 48 below. IV. GENERAL SUMMARY AND CONCLUSIONS 374 3.6.4. Summary From a diachronic point of view, the analysis of LOB and FLOB, on the one hand, and BROWN and FROWN, on the other, has revealed the existence of a considerable increase in the use of contractions from the 1960s to the 1990s, both in BrE and in AmE, although full forms are still the preferred negative variant in both dialects. Another important characteristic is that, while NotCs are more common than OpeCs with be in AmE from the 1990s, the latter contracted variant predominates in texts from the 1960s. Therefore, the data from the BROWN corpus do not reflect the expected preponderance of OpeCs reported by Dillard (1980), Freeborn (1986), Hughes & Trudgill (1996) and Anderwald (2002), among many others. From a dialectal point of view, the evidence from written English suggests that, in the four dialects under consideration, BrE, AmE, AusE and NzE, uncontracted negatives are the predominant choice. There is, however, a greater fondness for contractions in AmE than in the three other dialects, NotC being the preferred negative contracted form. In the case of spoken English, contractions are more common than full forms in the three dialects analysed in the present piece of research, BrE, AmE and NzE. Nevertheless, the proportion of fused forms is lower in AmE than in the other varieties due to the formality of the texts of the CSPAE corpus. Regarding contractions, both AmE and NzE prefer OpeCs to NotCs, while the latter contracted type is the predominant form in BrE texts. IV. GENERAL SUMMARY AND CONCLUSIONS 375 As regards medium, the tendency is to favour the uncontracted negative variant in written texts, whereas in speech contractions are preferred to a greater extent. The degree of formality of the texts also plays a role here, since UncNs predominate in written formal texts, while in those texts more closely related to the spoken style, contractions seem to be the favourite option. 3.7. Comparisons with Related Studies The comparisons across the different corpora used in this study presented in Section 3.6. above will be complemented with a comparative account of the results obtained from my analysis and those provided in earlier research on the topic concerning the selection of the three alternative ways of negation in different varieties of English. As mentioned in Section II.3.1.1. above, Biber (1987) studied the use of contracted and full forms in some categories of the LOB and the BROWN corpora, in order to establish comparisons between BrE and AmE. He concludes that AmE favours the use of contractions to a greater extent than BrE even in the most formal text-types, fused forms in the LOB corpus being virtually non-existent. Similar conclusions are reached by Yaeger-Dror, Hall-Lew & Deckert (2002: 81ff): in their data AmE accepts contraction of not in the written language to a greater extent than BrE does. Moreover, in Biber’s (1987) material, contractions are, in general, more common in the informal categories than in the most formal ones. This is also confirmed in Yaeger-Dror, Hall-Lew & Deckert’s (2002: 81ff) analysis, where full forms are favoured in what they call ‘informative registers’, IV. GENERAL SUMMARY AND CONCLUSIONS 376 such as News, Tutorials or Written descriptive texts, while ‘interactive registers,’ such as Conversation or Written dialogue, prefer the contracted variant (cf. Section II.3.1.1.). In my study, the dichotomy informal vs. formal registers also holds in the four written dialects, since contractions outnumber full forms in the most informal categories, that is, the fictional text-types. Kjellmer (1998), in turn, also studied the distribution of the three variants of negation under consideration in the LOB corpus (cf. Sections II.3.1.1. and II.3.2.3.1. above), concluding that, in written BrE texts, uncontracted negatives are more numerous than contracted variants, and that NotCs predominate over OpeCs with all operators but be. He also mentions that, as in Biber’s (1987) study, the most formal categories (Cat A to J) are those in which contractions are less commonly used. However, he also adds that, in an informal text-type such as Science fiction (Cat M), the number of contracted forms is lower than in the other fictional categories, and in this respect, the data recorded here is closer to those obtained in the highly formal category (Learned and scientific writings (Cat J)). In principle, my results for the LOB corpus should not differ much from those obtained by Kjellmer, and this is, in fact, the case, full forms predominating in the most formal kind of texts (from A to J, and in M (Science fiction)) and contractions being more common in the informal categories (from K to R). However, it must be noted that, as regards LOB, Kjellmer’s data do not entirely coincide with mine, since he found only 988 negative occurrences of the variants at issue with simple pronominal and existential there-constructions as subjects (cf. Figure 1 above and Table 20 in Kjellmer 1998: 180), while, as shown in Table 28 IV. GENERAL SUMMARY AND CONCLUSIONS 377 above, I have recorded a total of 1,077 examples with pronominal subjects and 36 instances in clauses of the existential type (KO contexts excluded). As stated in Sections II.3.2.3.2. and II.4.1. above, Hiller (1987) also analysed the distribution of OpeCs and NotCs in a collection of spoken BrE texts and in BrE and AmE written texts, reaching the conclusion that, in both dialects, there is an overall predominance of NotCs over OpeCs with all operators except be. Similar results are obtained in the present study for both BrE and AmE. However, as already mentioned, this statement does not hold true for the AmE corpus from the 1960s (the BROWN corpus), where NotCs predominate over OpeCs with all operators, be included. In what follows, I will compare the use of contractions and full forms in Biber et al.’s (1999) data for BrE (cf. Sections II.3.1.1. and II.3.2.3.1. above) and those obtained in my corpus for the same dialect (LOB, FLOB and LLC) (cf. Sections 3.4.1, 3.4.2. and 3.5.1 above).85 As already seen (cf. section II.3.1.1.), these scholars selected for their analysis four different textual categories, namely News, Academic Prose, Fiction and Conversation, all of them from BrE. Their study, thus, includes not only written texts, but also spoken material. They conclude that contractions are mainly used in spoken texts or in those written texts which are similar to the spoken style, while in formal texts, such as Academic prose, UncNs seem to be preferable (1999: 166, 1129). They also affirm that, in 85 Although most texts analysed by Biber et al. (1999: 26) date from the 1980s, I have decided to compare their results with those obtained for all my BrE texts, since my spoken BrE corpus contains material from the 1960s to the 1990s (cf. Table 10 above). In this respect, the comparison with Biber et al.’s data will be made without taking into account the date of compilation of the texts. IV. GENERAL SUMMARY AND CONCLUSIONS 378 written texts, both NotCs and OpeCs are mainly found in Fiction (1999: 166, 1129). In order to contrast my data with Biber et al.’s, some correspondences have been established between their textual categories and some of the text-types of my written and spoken corpora. The equivalences are the following: Biber et al. (1999) BrE corpora in the present study (LOB, FLOB, LLC) News = All press categories (A-C) in LOB and FLOB Academic prose = Learned and scientific writings (J) in LOB and FLOB Fiction = General fiction (K), Mystery and detective fiction (L), Science fiction (M), Adventure and western fiction (N), Romance and love story (P) in LOB and FLOB Conversation = Conversation between equals (S.1 and S.2), Conversation between disparates (S.3), Conversations between intimates and equals (S.4), Conversations (S.5), Non-surreptitious conversations between disparates (S.6), Surreptitious telephone conversations between personal friends (S.7), Surreptitious telephone conversations between business associates (S.8) and Surreptitious telephone conversations between disparates (S.9) in LLC The distribution of full forms and contractions in my corpus in these four general categories is shown in Table 114 below. The data in Table 114 reveal that my corpus differs from Biber et al.’s, since here not only does Academic prose favour the use of UncNs over contractions, but also News. For the remaining categories, my results are similar to Biber et al.’s, since contractions are specially used in spoken texts, that is, in Conversation, and also in the written medium in those text-types closer to the spoken language, as is the case with Fiction. IV. GENERAL SUMMARY AND CONCLUSIONS 379 Table 114. UncN/OpeC/NotC variants in the BrE corpora according to four general textual categories Number of words in each category UncNs OpeCs NotCs TOTAL News 352,000 466 (13.24) 29 (0.82) 101 (2.87) 596 Academic prose 320,000 421 (13.16) 3 (0.09) 7 (0.22) 431 Fiction 468,000 330 (7.05) 260 (5.56) 626 (13.38) 1,216 Conversation 380,000 102 (2.68) 628 (16.53) 849 (22.34) 1,579 TOTAL 1,556,000 1,319 (13.19) 920 (9.2) 1,583 (15.83) 3,822 Other features considered by Biber et al. (1999) are type of clause and type of subject (cf. Section II.3.2.3.4. above). They analyse only the occurrences of is and are plus not in their contracted forms and argue that NotCs with be are mainly found in clause initial position, while OpeCs with this operator are mostly found in clause final position and with pronominal subjects. Taking into account that clause initial position has been considered here a KO context for OpeC (cf. Sections II.3.2.2.1. and III.3.2.1 above) and that I have analysed all forms of the present of the operator be, including the first person singular, no comparison is possible with Biber et al.’s results in this respect. IV. GENERAL SUMMARY AND CONCLUSIONS 381 I IV V. . G GE EN NE ER RA AL L S SU UM MM MA AR RY Y A AN ND D C CO ON NC CL LU US SI IO ON NS S The present study has been concerned with the distribution of negative contractions and uncontracted negatives in contemporary written and spoken English. My point of departure has been the distinction between two types of negation following the criteria given by Klima (1964), Quirk et al. (1985) and Huddleston (1995): on the one hand, clausal negation, by means of which the whole clause is made negative, and, on the other, subclausal negation, in which only one part of the clause is negated. Such a distinction proves to be especially relevant for my purposes, since my study has been devoted only to negation at clause level, i.e. those cases in which the negator not modifies the whole clause and is placed after the first operator. Clausal negation with not can be expressed in PDE in three different ways: (a) uncontracted negatives or full forms, (b) operator contractions and (c) not-contractions. In English, there are 16 operators which can be negated by means of post-verbal not: be, have, do, will, would, shall, should, can, could, must, may, might, dare, need, ought to and used to, but not all of them allow the three different alternative ways of negation mentioned above. In fact, only be (in the present tense), have (both in the present and in the past), will and would allow UncNs, OpeCs and NotCs. Therefore, my study has been restricted to those operators which allow interchangeability between these three different patterns of negation (cf. Section II.1.). IV. GENERAL SUMMARY AND CONCLUSIONS 382 Although negative contracted forms have existed in English from the time of the earliest written records, negative contractions have undergone considerable diachronic changes. Therefore, I have considered it appropriate to provide an outline of the most important developments in the system of negation and in negative contractions from the early stages of the English language up to PDE. These changes are delineated in Section II.2. After this brief account of the relevant historical background to the topic of clausal negation and negative contractions, the selection of the three variants at issue is analysed in more detail. In doing so, I have considered a number of factors which have been discussed in the literature on the topic as conditioning the choice of one form over the others. Among these determinants of variation, we find the following: (a) Type of text (cf. Section II.3.1.1.). Following scholars such as Fries (1940: 8), Forsheden (1983: 36), Quirk et al. (1985: 123ff) or Biber (1988: 243), among many others, contractions have traditionally been associated with informal text-types and with conversation due to fast and easy production. Thus, for example, Biber (1987) mentions that contractions are practically non-existent in formal categories, such as Official documents and Academic prose, whereas the frequency of negative contractions is higher in fictional texts. The dichotomy formal vs. informal texts is also confirmed by Kjellmer (1998), Biber et al. (1999), Huddleston & Pullum (2002: 91; 800) or Yaeger-Dror, Hall-Lew & Deckert (2002: 81ff). IV. GENERAL SUMMARY AND CONCLUSIONS 383 (b) Geographical variation (cf. Section II.3.1.2.). The choice of contractions varies from dialect to dialect. Thus, for instance, in AmE contractions are used to a greater extent than in BrE, even in written and formal text-types (cf. Biber 1987: 11f; Tottie 1991: 12; Yaeger-Dror, Hall-Lew & Deckert 2002: 80ff). The use of the form ain’t for the operators be and have is also a dialectal marker, since its use is more recurrent in AmE than in BrE texts. (c) Social stratification (cf. Section II.3.1.3.). Following Freeborn (1986) or Jørgensen (1979), among others, contractions such as ain’t are considered as non-standard and illiterate and are mainly associated with the speech of lower social classes. (d) Gender and age (cf. Section II.3.1.4.). Contractions are more commonly used among the youngest and, in general, men tend to use contractions more frequently than women (cf. Coates 1986: 119f). (e) Structural factors, such as type of clause, or type of subject (cf. Section II.3.1.5.). According to Kjellmer (1998: 175ff), Westergren (1998: 135ff), Biber et al. (1999: 1131) or Yaeger-Dror, Hall-Lew & Deckert (2002: 85), contractions are more common in interrogatives, especially in question tags, and in imperative clauses than in declaratives. Moreover, contractions tend to be associated with auxiliary verbs rather than with lexical verbs for Quirk et al. (1985: 123) or Kjellmer (1998: 164), whereas scholars such as Philips & Reynolds (1987) affirm that with the operator