MINDING THE GAPS:
INFLECTIONAL DEFECTIVENESS IN A PARADIGMATIC THEORY
DISSERTATION
Presented in Partial Fulfillment of the Requirements
for the Degree Doctor of Philosophy
in the Graduate School of The Ohio State University
By
Andrea D. Sims, M.A.
The Ohio State University 2006
Dissertation Committee:
Professor Brian Joseph, Adviser Approved by
Professor Daniel Collins
Professor Mary Beckman ______Adviser Linguistics Graduate Program
ABSTRACT
A central question within morphological theory is whether an adequate description of inflection necessitates connections between and among inflectionally related forms, i.e. paradigmatic structure. Researchers have recently used periphrasis and paradigmatic gaps to demonstrate the descriptive capabilities of paradigms, but in doing so have expanded the domain of inflectional structure into the traditional realm of syntax. This expanded structure is justified only if an adequate description of gaps not only allows for but also requires reference to paradigmatic structure.
In this dissertation I argue that crucial evidence that paradigmatic structure underpins gaps is to be found in speakers’ reactions to inflectional defectiveness. I show through a series of experiments and distributional statistics that gaps in the genitive plural of Modern Greek nouns and the first person singular non-past of Russian verbs arose from speakers’ insecurity over competition between paradigmatic patterns of inflection.
The appearance of the gaps can thus be adequately explained only with reference to the inflectional paradigm. I formalize this approach using a Word and Paradigm model incorporating multidimensional inheritance hierarchies.
At the same time, historical causation is not to be confused with synchronic structure. The distributional patterns of the Modern Greek and Russian gaps resemble those which previous researchers have used to posit that gaps are optimal failures –
ii
synchronically epiphenomenal to productive word formation processes. However, a detailed analysis of speakers’ reactions to the Greek and Russian data shows gaps and productive inflectional forms to pattern differently. I interpret this to mean that the Greek and Russian gaps have become disassociated from their original causative factors, leaving the former as idiosyncratic facts of their respective languages. This conclusion throws previous gaps-as-epiphenomena accounts into doubt.
This dissertation makes a substantive contribution on three levels. First, it adds a new type of evidence to the body of research on paradigmatic gaps by exploring speakers’ resolution strategies, beliefs about language structure, and how those beliefs shape defective inflection. Second, it suggests that paradigmatic predictability is a significant force in morphological systems, in ways that are not typically acknowledged even by paradigm-based models. Finally, while the historical development of gaps may be well motivated, the synchronic reality can be very different. This supports viewing language as a series of small-scale, overlapping generalizations.
iii
ACKNOWLEDGMENTS
It is not possible to thank all of the people who deserve to be thanked here. There are simply too many people, who have helped to improve this project in too many ways. But
I would like to acknowledge a number of people whose influence I have particularly felt, and for whose guidance, friendship, and help I have been particularly appreciative.
First, I would like to thank my dissertation committee – Brian Joseph, Mary
Beckman, and Daniel Collins. For providing unique and sometimes unexpected perspectives. For challenging me. For keeping me focused on the theory, but rooted in the data at all times. And perhaps most of all for believing that I could finish this dissertation on time, when I had started to lose faith. And to Brian in particular for shepherding me through my years at Ohio State, and for the constant mantra that language is about the speakers. Without that, this dissertation would have lost its way.
I would also like to thank Charles Gribble, Rich Janda, and Keith Johnson, who were gracious enough to serve on one or several of my pre-dissertation committees. Each helped to shape my thinking and my research skills at critical stages.
I am lucky to have been surrounded by a varied, talented, and supportive group of graduate students at OSU. I owe something to every grad student that I have known these past seven years, but to a few I owe more. To Robin Dodsworth, Jeff Mielke, and
Wes Collins. As my friends and immediate predecessors, they showed me that it could
iv
be done, cheered me to the end, and set a high standard. To Jeff for those early, epic conversations in Oxley 225. To Tom Stewart, who with unbridled enthusiasm was the first person to set me on the morphological path. To Hope Dawson, for reminding me that historical linguistics is not dead. To Grant McGuire, for fixing my Russian experiment in a moment of despair and for much needed distraction, especially during the
Tour de France. To Vanessa Metcalf, for watching over my life while I wandered
Croatia. To Elizabeth Smith and Craig Hilts, who made sure I didn’t go over the edge during the final three months. To all 18 of my officemates, past and present, for all manner of things big and small.
The Russian data in this dissertation was collected by a three person team that consisting of Maria Alley, Bryan Brookes, and me. Without their ideas and their efforts, the project would have been much the poorer. Not to mention that Masha and Bryan probably deserve medals for putting up with me and my controlling, nitpicking tendencies.
A variety of people provided technical support for this work. Georgios
Tserdanelis helped me to understand Greek noun usage and worked with me to develop questionnaire materials. Anton Rytting alerted me several times to resources which kept
Greek data collection from becoming too burdensome. Shari Speer let me invade her lab meetings and ask basic methodological questions. Anastasia Smirnova and Larissa
Bondarchuk translated parts of the Russian experiment on short notice. Jim Harmon
v
went beyond the call of duty to find me cables, microphones, and all manner of other equipment.
I conducted part of my research at the Aristotle University of Thessaloniki, in
Greece. I received a very warm welcome there, and am deeply grateful to everyone from the Linguistics Section and the French Section. I would particularly like to thank Loukas
Tsitsipis and Savas Tsohatzidis, both of whom went out of their way to ensure that my data collection was efficient and successful.
And finally, there are three people who I owe so much that I hardly know where to begin. Thank you to my parents, for training me from birth to write my dissertation and supporting me at every step along the path. They were my first models of success in academia, and are still the best ones.
Thank you to Jason. For being my technical consultant, my sounding board, my personal chef, a vanquisher of bad days, my best friend. For single-handedly moving us to new apartments not once but twice in the final three months of dissertation writing.
For believing in me. For being my life.
This dissertation was supported by a Presidential Fellowship (2005-2006) and an Alumni
Grant for Graduate Research and Scholarship (2006), both awarded to me by The Ohio
State University. This research was conducted under Ohio State University IRB protocol
2004B0214.
vi
VITA
June 30, 1977 ...... Born – Exeter, New Hampshire
1999 ...... A.B. in Anthropology, The University of Chicago
2001 ...... M.A. in Linguistics, The Ohio State University
2003 ...... M.A. in Slavic linguistics, The Ohio State University
1999-2000 ...... Graduate Fellow, The Ohio State University
2000-2005 ...... Graduate Teaching and Research Associate, The Ohio State University
2001-2002 ...... Foreign Language and Area Studies Fellow, The Ohio State University
2003-2004 ...... J. William Fulbright Fellow, U.S. Student Program
2005-2006 ...... Presidential Fellow, The Ohio State University
vii PUBLICATIONS
Research Publications
1. Sims, Andrea. 2003. “Resolving gender conflict in dialectal Croatian: Two predictions of frequency.” Daniel E. Collins and Andrea D. Sims, eds. The Ohio State University working papers in Slavic studies, v. 2: Slavic and Balkan linguistics in honor of the 40th anniversary of the Department of Slavic & East European Languages and Literatures. Columbus, OH: Department of Slavic and East European Languages and Literatures, 145-167.
2. Sims, Andrea. 2005. “Declension hopping in dialectal Croatian: Two predictions of frequency.” Geert Booij and Jaap van Marle, eds. Yearbook of Morphology 2005. Amsterdam: Springer, 201-225.
3. Sims, Andrea. 2006. “On avoidance strategies (and their potential implications for dialect loss).” Christian Voss and Dieter Stern, eds. Marginal Linguistic Identities (=Eurolinguistische Arbeiten 2). Wiesbaden, Germany: Harrassowitz Verlag, 245-260.
Edited Volumes
1. Collins, Daniel E. and Andrea D. Sims, eds. 2003. The Ohio State University working papers in Slavic studies, vol. 2: Slavic and Balkan linguistics in honor of the 40th anniversary of the Department of Slavic & East European Languages and Literatures. Columbus, OH: Department of Slavic and East European Languages and Literatures.
2. Sims, Andrea D. and Miriam Whiting, eds. 2004. The Ohio State University working papers in Slavic studies, vol. 3: Proceedings of the First Graduate Colloquium on Slavic Linguistics. Columbus, OH: Department of Slavic and East European Languages and Literatures.
3. Ivanova, Tanya, Andrea D. Sims and Miriam Whiting, eds. 2006. Ohio Slavic papers, vol. 6: Proceedings of the Second Graduate Colloquium on Slavic Linguistics. Columbus, OH: Department of Slavic and East European Languages and Literatures.
FIELDS OF STUDY
Major Field: Linguistics
viii
TABLE OF CONTENTS
Page
Abstract ...... ii Acknowledgments ...... iv Vita ...... vii List of Tables ...... xiii List of Figures ...... xviii Glossary ...... xx
Chapters
1. What is (and is not) a paradigmatic gap? ...... 1 1.1. The term paradigmatic gap, broadly defined ...... 1 1.1.1. Sidebar: On the differences between inflection and derivation ...... 9 1.2. The term paradigmatic gap as used in this dissertation ...... 13 1.3. Examples of paradigmatic gaps ...... 16 1.4. Paradigmatic gaps as probabilistic phenomena ...... 25 1.5. The debate over the value of marginal phenomena ...... 29 1.5.1. The core/periphery distinction in generative syntax ...... 30 1.5.2. The (lack of a) core/periphery distinction in morphological theory ...... 32 1.6. Paradigmatic gaps as marginal phenomena ...... 35 1.7. Summary ...... 39
2. An overview of the paradigm in a Word and Paradigm model ...... 42 2.1. An overview of the paradigm ...... 42 2.2. The paradigm in traditional Word and Paradigm models ...... 43 2.3. The (lack of a) paradigm in most generative, morphemic theories ...... 46 2.3.1. Classic generative grammar ...... 46 2.3.2. Modern generative, morphemic theories ...... 47 2.4. The rebirth of Word and Paradigm models ...... 49 2.4.1. Similarities to traditional Word and Paradigm models ...... 50 2.4.2. Differences from traditional Word and Paradigm models ...... 53 2.4.3. The structure of modern Word and Paradigm models ...... 56 2.5. Remaining questions: Paradigmatic gaps as evidence for the paradigm ...... 73
ix
3. The Modern Greek genitive plural: How gaps are the epiphenomenal result of paradigmatic structure ...... 78 3.1. Introduction ...... 78 3.1.1. Grammar competition models: predictions and assumptions ...... 80 3.2. Overview of genitive plural stress ...... 85 3.3. The distribution of genitive plural gaps ...... 93 3.4. Avoidance strategies in the genitive plural ...... 97 3.4.1. Predictions ...... 101 3.4.2. Methodology ...... 102 3.4.2.1. Target words ...... 103 3.4.2.2. Questionnaire composition ...... 104 3.4.2.3. Questionnaire administration and subjects ...... 106 3.4.3. Results ...... 109 3.5. Confidence ratings and stress variability in the genitive plural ...... 113 3.5.1. Predictions ...... 113 3.5.2. Methodology ...... 114 3.5.2.1. Target words ...... 114 3.5.2.2. Questionnaire Composition ...... 116 3.5.2.3. Subjects ...... 117 3.5.2.4. Three minor methodological snags ...... 118 3.5.3. Results ...... 122 3.6. Greek gaps in a paradigmatic competition model ...... 128 3.6.1. Setting the stage: multidimensional inheritance hierarchies ...... 129 3.6.1.1. The stress inheritance hierarchy ...... 132 3.6.1.1.1. A comparison to Revithiadou (1988) ...... 136 3.6.1.2. The combined inheritance hierarchy for plurals ...... 138 3.6.2. Conflicting stress inheritance and the paradigmatic dimension ...... 142 3.6.3. Conclusions ...... 147 3.7. Summary ...... 148
4. The Modern Greek genitive plural revisited: How gaps are not the epiphenomenal result of paradigmatic structure ...... 150 4.1. Introduction ...... 150 4.2. Confidence ratings and stress variability in the genitive plural, revisited ...... 151 4.3. Reconciling opposing conclusions through principles of language change ..... 160 4.4. Sometimes defectiveness is just defectiveness ...... 168 4.5. Language as a non-optimizing system ...... 174 4.6. Summary ...... 176
x
5. Russian first person singular non-past verbs: the aftermath of synchronically motivated paradigmatic gaps ...... 178 5.1. Introduction ...... 178 5.2. Overview of morphophonological alternations in verbal inflection ...... 179 5.3. The distribution of paradigmatic gaps in the first person singular ...... 185 5.4. An experiment on speaker confidence and unexpected non-alternation ...... 189 5.4.1. Predictions ...... 191 5.4.2. Methodology ...... 194 5.4.2.1. Experiment structure ...... 194 5.4.2.2. Stimulus composition ...... 198 5.4.2.3. Types of data collected ...... 201 5.4.2.4. Subjects ...... 202 5.4.3. Results ...... 202 5.4.4. Paradigmatic connections, and the divergence of defective and non- defective dental stems ...... 219 5.5.The development of the standard language as a historical source of paradigmatic competition ...... 223 5.6. Summary ...... 230
6. Concluding thoughts ...... 232
Appendices
Appendix A. A survey of paradigmatic gaps ...... 235 A.1. Arabic ...... 235 A.2. Chaha ...... 238 A.3. English ...... 240 A.4. Finnish ...... 246 A.5. French ...... 248 A.6. Hebrew ...... 251 A.7. Hungarian ...... 252 A.8. Icelandic ...... 255 A.9. Mandarin Chinese ...... 260 A.10. Norwegian ...... 261 A.11. Russian ...... 265 A.12. Spanish ...... 268 A.13. Swedish ...... 271 A.14. Tiene ...... 276 A.15. Turkish ...... 278
xi
Appendix B. Genitive plural gaps in Modern Greek ...... 279 B.1. Feminine nouns with nominative singular –α and plural –ες ...... 279 B.2. Feminine nouns with nominative singular –η and plural –ες ...... 308 B.3. Masculine nouns with nominative singular –ας and plural –ες ...... 312 B.4. Neuter nouns with nominative singular –ι ...... 315 B.5. Neuter nouns with nominative singular –ος ...... 316 B.6. Masculine nouns with nominative singular –ης and plural –ες ...... 318 B.7. Neuter nouns with nominative singular –ο ...... 319 B.8. Masculine nouns with nominative singular –ας and nominative plural –αδες. 321 B.9. Neuter nouns with nominative singular –µα ...... 322
Appendix C. Survey instrument: Greek periphrasis ...... 324
Appendix D. Survey instrument: Greek genitive plural gaps ...... 337
Appendix E. Russian first person singular non-past gaps ...... 351
Appendix F. Survey instructions and stimuli: Russian verbal gaps ...... 355
List of References ...... 369
xii
LIST OF TABLES
Table Page
Table 1: A criterial representation of the various uses of the term “paradigmatic gap ..... 2
Table 2: Examples of Russian noun declension ...... 22
Table 3: Russian nominal paradigm showing end stress in all oblique plural forms except the genitive plural ...... 23
Table 4: Russian nominal paradigm gap in the genitive plural ...... 24
Table 5: Some frequency information about the English verb go ...... 26
Table 6: Some frequency information about the English verb undergo ...... 26
Table 7: Some frequency information about the English verb forego ...... 27
Table 8: Two BCS verb paradigms demonstrating implicational relationships within paradigms ...... 43
Table 9: Leveling in Yiddish verbal paradigms (Albright 2002:2) ...... 48
Table 10: Four Bosnian/Croatian/Serbian inflectional paradigms ...... 58
Table 11: Paradigm of BCS word dijete ‘child’ ...... 63
Table 12: Two BCS nominal paradigms, repeated ...... 64
Table 13: Present tense indicative gaps in Spanish ...... 81
Table 14: Morphophonological alternations in the present indicative ...... 82
Table 15: Greek noun inflection classes with genitive plural stress consistently columnar (type 1) ...... 87
xiii
Table 16: Greek noun inflection classes with genitive plural stress consistently fixed relative to the end of the word (type 2) ...... 89
Table 17: Greek noun inflection classes with variability in genitive plural stress placement (type 3) ...... 91
Table 18: Distribution of nouns with genitive plural gaps vs. all nouns, according to stress type ...... 94
Table 19: Distribution of type 3 nouns with genitive plural gaps vs. all type 3 nouns, according to whether the (expected form could have a stress shift ...... 96
Table 20: Genitive vs. prepositional phrase according to frequency ...... 109
Table 21: Genitive vs. prepositional phrase according to stress type ...... 109
Table 22: Examples of the inflection class from which genitive plural stimuli were drawn ...... 114
Table 23: Greek inflectional patterns ...... 131
Table 24: Example of stress fixed on the final syllable in the Greek genitive plural .... 133
Table 25: Example of stress fixed on penultimate syllable in the Greek genitive plural ...... 134
Table 26: Example of fixed stress on final syllable in the Greek genitive singular and genitive plural ...... 134
Table 27: Example of fixed stress on penultimate syllable in the Greek genitive singular and genitive plural ...... 135
Table 28: Example of fixed stress on the penultimate syllable in the Greek genitive singular, genitive plural, and accusative plural ...... 135
Table 29: Selected future forms for the Sanskrit verb ‘give’ – synthetic forms in 1st person ...... 161
Table 30: Selected future forms for the Sanskrit verb ‘give’ – “periphrastic” forms in the 1st person ...... 161
xiv
Table 31: Imperfective paradigm of Latin verb laudo ‘I praise’ – no periphrasis ...... 171
Table 32: Perfective paradigm of Latin verb laudo ‘I praise’ – periphrasis ...... 172
Table 33: Semantic/morphological mismatch in Russian verbal inflection (красть / украсть ‘to steal’) ...... 180
Table 34: Example of the 2nd conjugation non-past inflectional pattern of Russian ..... 181
Table 35: Stem-final palatalization in the 2nd conjugation first person singular non-past of Russian ...... 182
Table 36: Non-alternation in the 2nd conjugation non-past of Russian ...... 183
Table 37: Paradigmatic gap in the first person singular non-past ...... 185
Table 38: Sample of responses to nonce items in a Russian sentence completion task. 211
Table 39: A sample of responses for defective stems in a Russian sentence completion task ...... 212
Table 40: Average and median degrees of intersubject agreement in response to a forced production sentence completion task, according to stimulus type ...... 216
Table 41: Subjects’ average and median confidence in their responses to a forced production sentence completion task, according to stimulus type ...... 216
Table 42: Example of stem reanalysis in Russian (morphological doublets) ...... 217
Table 43: Example of leveling of morphophonological alternation in Russian ...... 218
Table 44: Distribution of gaps relative to all stems with same final C(C) ...... 222
Table 46: Present tense indicative paradigm of be ...... 244
Table 47: More present tense indicative paradigm of be ...... 244
Table 48: Simple tenses of the French verb frire ‘to fry’, with numerous paradigmatic gaps (Kendris 1996) ...... 250
xv
Table 49: Russian nominal paradigm showing end stress in all oblique plural forms except the genitive plural ...... 267
Table 50: Russian nominal paradigm, gap in the genitive plural ...... 267
Table 51: Present tense indicative gaps in Spanish ...... 269
Table 52: Morphophonological alternations in the present indicative ...... 270
Table 53: Examples of Modern Greek feminine nouns with nominative singular –α and plural -ες ...... 280
Table 54: Genitive plural gaps among feminine nouns with nominative singular –α and plural -ες ...... 281
Table 55: Examples of Modern Greek feminine nouns with nominative singular –η and plural -ες ...... 309
Table 56: Genitive plural gaps among feminine nouns with nominative singular –η and plural -ες ...... 309
Table 57: Examples of Modern Greek masculine nouns with nominative singular –ας and plural -ες ...... 313
Table 58: Genitive plural gaps among masculine nouns with nominative singular –ας and plural -ες ...... 314
Table 59: Examples of Modern Geek neuter nouns with nominative singular -ι ...... 315
Table 60: Genitive plural gaps among neuter nouns with nominative singular -ι ...... 316
Table 61: Examples of Modern Greek neuter nouns with nominative singular -ος ...... 317
Table 62: Genitive plural gaps among neuter nouns with nominative singular -ος ...... 317
Table 63: Examples of Modern Greek masculine nouns with nominative singular –ης and plural -ες ...... 318
Table 64: Genitive plural gaps among masculine nouns with nominative singular –ης and plural -ες ...... 319
xvi
Table 65: Examples of Modern Greek neuter nouns with nominative singular -o ...... 320
Table 66: Genitive plural gaps among neuter nouns with nominative singular -o ...... 320
Table 67: Examples of Modern Greek masculine nouns with nominative singular –ας and plural –αδες ...... 321
Table 68: Genitive plural gaps among imparisyllabic masculine nouns with nominative singular –ας and plural –αδες ...... 322
Table 69: Examples of Modern Greek neuter nouns with nominative singular -µα ...... 323
Table 70: Genitive plural gaps among neuter nouns with nominative singular –µα ..... 323
xvii
LIST OF FIGURES
Figure Page
Figure 1: A Venn diagram representation of the various uses of the term “paradigmatic gap” ...... 3
Figure 2: The three paradigms of the BCS noun prozor ‘window’ ...... 61
Figure 3: Syncretism in the BCS noun prozor ‘window’ ...... 62
Figure 4: Three paradigms of the heteroclite BCS nouns dijete ‘child’ ...... 66
Figure 5: Schematization of information sharing in a default inheritance network ...... 67
Figure 6: Another schematization of information sharing in a default inheritance network ...... 68
Figure 7: Syncretism in BCS nominal paradigms as a default inheritance information- sharing network ...... 70
Figure 8: Interaction of frequency and stress predictability as conditioning factors for avoidance of the synthetic genitive plural form ...... 111
Figure 9: Variance in the confidence ratings reported by Greek speakers, relative to the average for words sharing the same inflectional properties ...... 122
Figure 10: Greater variance among genitive plural forms than among nominative singular forms ...... 124
Figure 11: Effect of interspeaker agreement on speaker confidence ...... 126
Figure 12: Effect of shifting stress on speaker confidence ...... 127
Figure 13: A default inheritance hierarchy for Greek nominal stress ...... 133
Figure 14: A default inheritance hierarchy for segmental information of plural Greek nouns ...... 140
xviii
Figure 15: A multidimensional default inheritance hierarchy for plural inflection of Greek nouns ...... 141
Figure 16: An inheritance hierarchy for the classes in which 98% of Greek genitive plural gaps occur ...... 143
Figure 17: A similar inheritance hierarchy, but without a significant number of gaps . 145
Figure 18: Effect of interspeaker agreement on production confidence, differences between gaps and productive inflection ...... 153
Figure 19: Effect of stress variability on production confidence, differences between gaps and productive inflection ...... 155
Figure 20: Lack of effect of lexeme familiarity on confidence for genitive plural forms ...... 157
Figure 21: Relationship between sentence completion reaction time and speaker confidence for Russian ...... 204
Figure 22: Slightly faster responses for familiar words than for non-familiar words ... 206
Figure 23: Lexeme familiarity vs. speakers’ confidence in their word form productions ...... 207
Figure 24: Relationship between lexeme familiarity and production confidence ...... 208
Figure 25: Sample of responses to non-defective, non-nonce items in a Russian sentence completion task ...... 210
Figure 26: Interspeaker agreement in a cloze procedure task vs. speaker confidence, for dental items ...... 214
Figure 27: Interspeaker agreement in a cloze procedure task vs. speaker confidence for non-dental stems ...... 215
xix
GLOSSARY
Word and Paradigm (WP) models of inflection assume a separation between inflectional form and inflectional meaning. In some areas of linguistics it is common to refer to the former as lexemes and the latter as lemmas. However, this is not the terminology typically used within WP models. Moreover, a cursory search of the linguistics literature reveals that the terms lexeme and lemma have no consistent definition or usage across the entire field of linguistics. The following are among the attested definitions of the terms lexeme and lemma.
In psycholinguistics, the term lemma is used for either (a) the morphosyntactic properties expressed by a given word form, but not the word form itself or the lexical meaning, (b) all possible morphosyntactic property sets that can be combined with a given lexical meaning, but not the lexical meaning itself or any word forms (i.e., lemma = an inflectional paradigm if we define it as the combinatory possibilities of inflectional properties in the language, (e.g. Spencer 2004), or (c) the combination of a lexical meaning and all possible morphosyntactic property sets that combine with that lexical meaning, but not any word forms. The term lexeme is used to refer to the phonological instantiation of any of these definitions.
In corpus linguistics, the term lemma is used to refer to the set of all inflected forms of a lexical item. The term lexeme is synonymous with lexical meaning (as opposed to inflectional meaning).
xx
In lexicology, the term lexeme is used to refer to all inflected forms of a lexical item, thus lexeme for the meaning of GO is the set go, going, goes, went, etc. The term lemma is used to refer to the citation form of a lexical item (e.g. go), also known as headwords, and thus lemmas are subsets of lexemes.
Finally, in WP morphology, the term lexeme is either (a) synonymous with lexical meaning – “the paradigm of a lexeme L is the inventory of syntactic atoms which may instantiate L in phrase structure” (Stump 2001a:148), or (b) the set of possible combinations of lexical meaning and inflectional properties (i.e., the entire paradigm).
Lexeme is not used in reference to phonological form – word form fills this role. The term lemma is not used at all.
Thus, a lemma in any of the psycholinguistic senses basically corresponds to lexeme in WP morphology, and herein lies the largest terminological problem for this dissertation, which is framed within a WP model, but which in some respects overlaps with psycholinguistic concepts and methodology. While I find the psycholinguistic distinction to be the more useful, I have ultimately chosen a usage most closely in line with that used in other WP work. I thus use the following terms:
• lexeme: a lexical meaning, apart from the form which instantiates it, or the
morphosyntactic properties which can combine with it,
• lexeme paradigm cell: the combination of a lexeme and a set of morphosyntactic
properties that can combine with that lexeme into a well-formed syntactic atom,
• lexeme paradigm: the set of all lexeme paradigm cells which have a common
lexeme,
xxi
• stem: a form belonging to inflection class I, to which inflectional processes apply
in order to generate an inflected word form, apart from the meaning that it
instantiates,
• stem paradigm cell: the combination of a stem and a set of morphosyntactic
properties that can combine with that stem into a well-formed morphological
atom,
• stem paradigm: The set of stem paradigm cells which have a common stem
• word: The phonological instantiation of a stem paradigm cell that is linked to a
lexeme paradigm cell. This is the unit which is inserted into syntactic structure,
• word paradigm: The phonological instantiation of the set of stem paradigm cells
that are linked to a single lexeme paradigm,
• lemma: no usage
While this may seem like more distinctions than is necessary for the lemma/lexeme distinction, each term represents a separate theoretical concept within many modern
Word and Paradigm theories.
xxii
CHAPTER 1
WHAT IS (AND IS NOT) A PARADIGMATIC GAP?
This dissertation explores paradigmatic gaps. However, this statement is insufficient to delineate the range of data considered in the following pages, because the term paradigmatic gap has been used to describe virtually any situation in which an observed morphological or morphologically-related structure is in some sense unexpected. Since this dissertation is actually concerned with only the most narrow subset of those phenomena which have been described as paradigmatic gaps, I begin here with an overview of those phenomena which are – and are not – the present topic of study.
1.1. The term paradigmatic gap, broadly defined
The various uses of the term paradigmatic gap (and the frequent shorthand form gap) can
(almost) be categorized as being in hyponymic/hypernymic relationships to each other, based on the number of definitional criteria employed. This categorization is represented in Table 1 and Figure 1. I will exemplify each of these categories, working from the broadest definition to the most specific.
1
A B C D E F unexpected form/structure? X X X X X X language internal expectation? X X X X X defectiveness? X X X reflected in morphological system? X X X reflected in inflectional system? X
Table 1: A criterial representation of the various uses of the term "paradigmatic gap"
In the most broad sense, the term paradigmatic gap has been used to refer to words or structures which may be considered “missing” from a cross-linguistic perspective, such as when one language has a word that has no direct semantic parallel in another language. This is represented by set A. A famous example comes from Russian color terms. Russian has two terms – sinij ‘light blue’ and goluboj ‘dark blue’ – where
English has a general term (blue) in addition to a variety of more specific ones. Notably,
Russian does not have a single term which encompasses both sinij and goluboj.
Assuming that two cultures have the same percepts, we might in some sense expect that both of their languages would, contrary to fact, have a word to express that concept. In this sense, a “gap” in the lexical inventory of Russian is identifiable by cross-linguistic comparison. Geckeler (1974) calls these interlingual gaps, and notes that they are the most easily identified type of “deficiency” due to their close relationship to thought and meaning. At the same time, they do not stem from or bear on the structure of language in particularly interesting ways and so are ignored here.
2
A
B
C “paradigmatic gaps” as the term is used in E this dissertation F
D
defectiveness morphology
Key A: A “missing” word or grammatical structure, from any (including cross-linguistic) perspective. B: An unexpected (but grammatical) structure, based on expectations formed based on the internal structure of the language. C: An unexpected (but grammatical) word form; a situation in which the morphological structure of a language leads to an expectation of an overt exponent of a morphological category, but a “null morpheme” (i.e. no unique, segmentable form) expresses the category instead. D: Syntactic or semantic defectiveness, often resulting from a mismatch between the grammatical properties required by the syntax, and those expressed by the morphology. E: Derivational defectiveness – for any word X, the non-existence of any word Y which stands in a given derivational relationship to X, despite the language normally having pairs of words representing the same derivational relationship. F: Inflectional defectiveness – for a given lexeme, the non-existence of any synthetic form expressing a given set of inflectional properties, despite that set normally being expressed on lexemes of the relevant type.
Figure 1: A Venn diagram representation of the various uses of the term "paradigmatic gap"
3
Parallel examples can be found in cross-linguistically suspicious absences in grammatical inventories. Larson reports that “To my knowledge, no language has yet been reported to contain the equivalent of a ‘definite adverb.’ Thus definite adverbs represent an interesting ‘gap’ in the paradigm of determiner-adverb correspondences. I wish to suggest that Fon and Haitian definiteness markers fill the gap just referred to, and complete the expected paradigm. Specifically, I propose that whereas Fon and Haitian nominal definiteness markers are definite determiners, clausal definiteness markers represent the missing paradigm element: definite adverbs” (Larson 2003:manuscript p. 5, emphasis mine). Larson motivates the existence of a definite adverb “gap” in the majority of the world’s language by virtue of its existence in Haitian and Fon. It is unlikely that any of the languages presenting the so-called gap could be considered to have a gap based on system-internal criteria. Instead, the combination of definiteness and adverbials is simply a hypothetical combination of grammatical properties which could be realized in any language, but which for most languages is not.
Another use of the term paradigmatic gap, represented by set B above, is more constrained in the sense that the expectation but non-attestation of some structure is drawn from language-internal facts. For example, according to Phillips (1996), Lema and Rivero (1991) “...claim that VP-topicalization and perfect inversion [in Old Spanish] differ in that topicalization across negation is possible, but perfect inversion is not attested with negation. Here, unfortunately, it is again difficult to evaluate the status of this paradigmatic gap, given the absence of any information about how many tokens we might expect to find in the corpus” (Phillips 1996: footnote 158, emphasis mine). The details of the analysis are not important here, but what is what is noteworthy is the use of
4
the term paradigmatic gap to describe the non-attestation of a syntactic structure in Old
Spanish (perfect inversion in a negated clause) which we might reasonably expect to find based on the combinability of negation and structures related to perfect inversion.
This same use, in which the resulting form is different than that which might be expected on language-internal grounds, is most common at the morphological level. This is represented by set C in the above Venn diagram, and may be exemplified by
Wunderlich’s discussion of Yimas (language family:1 Sepik-Ramu, Nor-Pondo, Pondo).
Yimas “...exhibits several sets of pronominal affixes (agreement morphemes) that are attached to the verb. Foley (1991) lists these sets in terms of A(ctor), O(bject or theme),
S(ubject of intransitive verbs) and D(ative)... All pronominal affixes are prefixed to the verb, except elements of D[ative], which are suffixed after aspect/tense...” (Wunderlich
2001: manuscript p. 10). The relevant issue is that some prefixes (e.g. negation marker) block the realization of others (e.g. nominative markers). The data gets quite complicated, but I present a simple example in (1), in which the second person singular nominative marker “disappears” under negation.
(1) Nominative agreement affixes in Yimas (Wunderlich 2001: manuscript p. 25)
a. ma-a-tpul 2SG.NOM-1SG.ACC-hit ‘You hit me.’
b. ta-a-tpul NEG-1SG.ACC-hit ‘You didn’t hit me.’
1 According to Ethnologue (www.ethnologue.com). 5
Wunderlich describes examples like (b) as having paradigmatic gaps, by which he means the lack of an overt, segmentable form expressing a set of morphosyntactic properties, despite those properties being expressed affixally under a variety of conditions, as in (a).
However, since the grammatical properties are still carried by the entire word, i.e., the structure in (b) is grammatical and conveys a second person singular subject despite the lack of unique surface exponent, it is clear that from a purely descriptive perspective, this is not an example of defectiveness and not a paradigmatic gap in the strictest sense of the word (see Section 1.2).
The term paradigmatic gap is most commonly used in reference to defectiveness
(in the sense that the relevant linguistic structure crashes). Examples external to the morphology are represented as set D in Table 1 and Figure 1. Hetzron (1975:860) provides a particularly clear example from Russian.
[Russian represents an example] ...of morphosyntactic conflict resulting in a gap on the surface. At least five nouns in Russian require the preposition na ‘on’ for their locative use (stationary ‘at’ with the Prepositional case, ‘(going) into’ with the Accusative), and s with the Genitive for their ablative use (‘from’); these are počta ‘post office, fabrika ‘factory’, zavod ‘industrial plant’, stancia ‘station, vokzal ‘(large) railway station’.... On the other hand, Russian has verbs that govern the preposition v + Accusative, such as vojti ‘enter’ and verbs that govern the preposition iz such as vyjti ‘go out, leave’. When the speaker wants to use these verbs with one of the five nouns above, the conflict between the agreement rules (the noun governs na or s, while the verb requires v or iz) results in the ungrammatical utterances [and repair strategies in (2)].
6
(2) Morphosyntactic conflicts resulting in defectiveness in Russian
a. *Ja vošёl v / na počtu. I went-in in on postoffice-ACC.SG ‘I entered the postoffice.’
b. Ja vošёl v zdanije počty. I went-in in building-ACC.SG postoffice-GEN.SG ‘I entered the post office building.’
c. *Ja vyšёl iz / s počty. I left from from postoffice-GEN.SG ‘I left the post office.’
d. Ja vyšёl iz zdanija počty. I left from building-GEN.SG postoffice-GEN.SG ‘I left the post office.’
To Hetzron’s credit, he refers to the examples in (a) and (c) above simply as gaps, and not as paradigmatic gaps per se. However, the more specific terminology has been used for the same type of example by Kiparsky (2001), among others. Certainly, it is easy to understand why the term paradigmatic gap has been applied to this type of situation; the defectiveness is at the level of the syntax (neither vojti nor počta nor the relevant preposition is morphologically defective – it is only the particular combination in (2) which is at issue), but closely tied to how the morphology and the syntax interact, under the lexicalist assumption that the syntactic requirements of these verbs and nouns are at least partially represented as lexical properties.
In the final stop on this tour of the various uses of paradigmatic gap, we can see that the term has been especially used to cover any sort of morphological defectiveness, represented by set E. Examples of derivational non-productivity are easily found in
7
English and other languages. For example, Berg (1998) notes that nouns like output and adjectives like overflowing stand in a derivational relationship to the verb to output and to overflow (whether this relationship is view as an active process of generation or as a static correspondence), and in fact “zero derivation” between nouns and verbs is widespread in English (e.g. a jump, to jump). However, such a relationship cannot be posited for a number of nouns and adjectives which are seemingly similar to output and overflowing. He presents the following data (Berg 1998:245):
(3) Nouns and adjectives with “missing” verbal counterparts
a. However high the income of an average physician may be, the good they do to society is priceless. b. We don’t drink wine or beer. This is part of our religious upbringing. c. She was looking very downcast because she had just lost five pounds on the horses. d. The ongoing debate about expropriation has created an atmosphere of insecurity in this country.
In particular, the issue is that “A derivational analysis of the nouns income [a], and upbringing [b] and the adjectives downcast [c] and ongoing [d] is halted by the fact that there are no such verbs as to income, to upbring, to downcast and to ongo in the living language...” (Berg 1998:246), although certainly verbs with the same parts in different orders sometimes do exist (e.g. to bring up). If we assume that upbringing is not synchronically related to the verb to bring up (and Berg’s analysis provides support for this), the lack of a verb *to upbring can be considered a derivational gap (notably not in my terminology a paradigmatic gap!) in English.
8
1.1.1. Sidebar: On the differences between inflection and derivation
It can be tricky to separate defectiveness such as *to upbring from the paradigmatic gaps that are the topic of this dissertation because such a distinction relies on a separation of inflection and derivation which is not universally incorporated by morphological theories.
The following are among the cited criteria for distinguishing inflection from derivation.
(4) Previously proposed characteristics distinguishing derivation from inflection
a. Derivation adds meaning and changes word-class membership as a result;
inflection does not.
b. Inflection only and always encodes a closed class of functions (e.g. case,
number, gender, person...); derivation is everything else.
c. Inflection encodes configurational (received) properties (e.g. case,
number, person); derivation encodes inherent properties (e.g. gender,
inflectional class).
d. Derivation is more difficult to process than inflection due to conceptual
complexity, semantic transparency, phonological transparency, pseudo-
affixation, affixal homonymy, etc. (Schreuder and Baayen 1995:146).
e. Inflection is fully productive; derivation is not.
f. Derivation appears inside inflection.
g. Derivation originates historically in compounds; inflection originates
historically in other types of material, especially functional categories
(Hall 1992).
9
h. Inflection is only suffixal; derivation is both prefixal and suffixal (in
English).
i. Inflection tends to be subject to speech errors whereby suffixes appear
outside of the phonological word (e.g. tell-us-ing instead of telling us);
derivation tends to not be subject to these kinds of errors (Stemberger and
MacWhinney 1986).
j. “...if an agreement rule causes item X to agree with item Y in property P,
then P is an inflectional property for both X and Y” (Anderson 1982:588).
These types of evidence have led many, if not most researchers who posit a morphological component separate from the phonology and the syntax to also assume a principled distinction between inflection and derivation, both descriptively and theoretically (e.g. Anderson 1982, 1992, Aronoff 1994, Baker 1990, Booij 1996,
Hoeksema 1985, Sells 2004, Steele 1995, Stump 2001b, Zwicky 1990).
At the same time, these criteria are clearly too strong; counterexamples to many of these points are well known. For example, /re-/ as in redo, rewrite, rework is considered derivational and adds meaning in the intended sense, but does not change word-class membership. Second, in Slavic languages diminutives may change the gender of the noun to which they attach, but in other languages this is not true (Anderson
1982:586). Thus, in comparison with other criteria, gender seems to be sometimes derivational, sometimes inflection, countering the claim that inflection only and always reflects a closed class of functions. Third, defective paradigms are an example of the non-productivity of apparent inflection, while nominalization in English is an example of
10
the productivity of apparent derivation. Finally, ablauting (e.g. in the past tense of
English verbs such as run, swim, drink, etc.) and other phenomena are notorious problems for the idea that inflection appears outside of derivation. Many more examples could be added to the list.
Counterexamples such as these have led some to believe that the difference between inflection and derivation is not categorical, but rather a cline or scale along which lexical items may vary. Several researchers have therefore posited a unified set of principles according to which “inflection” and “derivation” are posited to operate (e.g.
Bochner 1993, Bybee 1985, Sadock 1991, Schreuder and Baayen 1995).
I maintain a principled distinction between inflection and derivation, as defined by Anderson (1982), given in (j) above.2 His definition of inflection is qualitatively different from the preceding ones and is a formalization of his well-known claim that
“inflectional morphology is what is relevant to the syntax” (1982:587). It overlaps significantly with some of the other criteria listed above but cannot be subsumed by them for two reasons. First, the definition is “strictly theory-internal” (587) since agreement is a theoretical concept. Second, it is also strictly language-internal. Inflection cannot be defined cross-linguistically since agreement patterns are not identical from one language to another. Gender is one of the areas in which there is disagreement as to whether the feature is inflection or derivation. Anderson’s proposal would claim that gender is an inflectional property if and only if a particular language uses it as the basis for sentential
2 Anderson did not completely pioneer this concept of inflection; it was clearly nascent in Sapir (1921) and his distinction between derivational concepts, concrete relational concepts, and pure relational concepts. Derivational concepts correspond to the Anderson’s definition of derivation, while both concrete relational concepts and pure relational concepts correspond to inflection. These basically correspond to what is now typically called inherent inflection (=concrete relational concepts) and contextual inflection (=pure relational concepts) (Booij 1996). 11
agreement. There is no necessity, and perhaps not even the expectation, that inflection will reflect the same properties from one language to another. In short, Anderson distinguishes inflection from derivation based upon the behavior/distribution of particular linguistic features on a language-internal basis, thereby removing much of the traditional difficulty of defining these concepts.
Thus, in this dissertation I assume that derivation and inflection operate according to different principles, and in contrast to most of the literature on paradigmatic gaps (esp. that in the framework of Optimality Theory), I exclude derivational gaps from the present topic of study. Theories which do not make the same theoretical assumption treat examples such as *to income and *to upbring as paradigmatic gaps.
Making a distinction between derivation and inflection entails assigning examples of morphological defectiveness to one category or the other – derivational gap, or paradigmatic gap. Most cases are clear-cut according to any criteria, but practical problems can occasionally arise. For example, English has a well-known reduplicative phenomenon expressing rejection or devaluing of the significance of the base word. The reduplicated part is formed by replacing a word-initial syllable onset by
If the noun is vowel-initial, [Sm-] is prefixed to the word. This reduplication is possible for any noun except one which already begins with [Sm-]; examples of this type are defective.
12
(5) English
e. Oedipus-schmeodipus holiday-schmoliday Pedro-schmedro food-schmood
f. *Schmidt-Schmidt *schmuck-schmuck *schmooze-schmooze
On the one hand these examples seem to be inflectional because the reduplication does not change the lexical category of the base word. On the other hand, the examples do not express any obviously inflectional category (certainly, there is no agreement), instead seeming to fall into the range of expressivity and pragmatics, loosely defined. By
Anderson’s definition of inflection, these examples represent derivational gaps.
1.2. The term paradigmatic gap as used in this dissertation
We may thus define paradigmatic gaps, as the term is used in subsequent chapters of this dissertation, as follows.
(6) Definition of paradigmatic gaps: For a lexeme L belonging to lexical class C, a paradigmatic gap exists if there is no synthetic form F(Lc) expressing a set of inflectional properties I, when the language normally has an F(xc) expressing I. Any otherwise well-formed syntactic structure into which a hypothetical F(Lc) is placed is ungrammatical.
There are several notable aspects of this definition.
13
First, I define paradigmatic gaps only with reference to a particular lexical class.
In many languages nouns inflect for gender but verbs do not. Logically, for these languages we do not want to say that all verbs have paradigmatic gaps for the inflectional property of gender. Only properties which are normally expressed for a particular lexical class can “go missing”. Thus, the expectation against which the definition is set is tightly constrained by the (prevalent) structures of the language.3
3 Note that this aspect of the definition is not as clear as it might seem. If we set our definition of gaps as the abnormal failure of inflection, we potentially run into a problem when the absence of a form F(LC) is the norm. Consider the example of the verb to be in English. It is the only verb which has a form for counterfactuals which is distinct from the indicative past tense form.
(i) If I were you, I would study harder. Last night I dreamed that I was you.
(ii) If I had a million dollars, I would sleep till noon. Last night I dreamed that I had a million dollars.
According to our definition of paradigmatic gaps (and as Zwicky 2004 argues), had does not have a paradigmatic gap in the counterfactual because there is an inflectional form expressing that property. It is immaterial that the form happens to be distinct from the past tense form for the verb to be, but not distinct for the verb to have. This example is thus parallel to the Yimas example presented in (1) above. However, imagine a language quite similar to English, except that the sentence If I had a million dollars, I would sleep until noon were ungrammatical for the sole reason that no form of to have existed which expressed the counterfactual. And imagine that this were true for every other word of the language except for to be. Would we want to argue that every verb except to be has a paradigmatic gap? This hypothetical example would be problematic because the proposed gaps would be the prevalent structure in the language, and the existence of a counterfactual form for to be would be the exception. On the one hand, if we think of paradigmatic gaps as the unexpected failure of inflection, it seems ludicrous to say that a failure of inflection in all but one verb is unexpected. Instead, it seems that the success of inflection in that one verb is the unexpected result. We would not want to call this an example of widespread paradigmatic gaps. However, two entrenched theoretical concepts would argue the opposite. First, paradigms are typically defined by the combinatory possibilities of inflectional properties (see CHAPTER 2 for discussion), meaning that the expression of an inflectional property set for just one verb is sufficient to posit an inflectional cell in the paradigms of all verbs (Zwicky (2004) is an explicit proponent of this formulation). Second, following the principle of a phonology-free syntax, we would not want to require the syntax to be sensitive to the existence or non-existence of an inflectional form. Thus, in this hypothetical language similar to English, so long as to be expresses some inflectional property set, the theory forces us into claiming that every other verb has a paradigmatic gap. Which solution would be preferable – abandoning common sense or abandoning fundamental theoretical principles with the consequences that would have for our understanding of language – is unclear. Thankfully, I do not know of any real-language examples where such a situation arises, and so this implication of the definition of paradigmatic gaps is more a thought experiment than a flaw. However, if such a distribution as described here were to appear (and it is certainly possible, considering what we know 14
Second, I define paradigmatic gaps as the non-existence of a synthetic form. This implies that where inflectional property sets are realized via periphrastic forms, these periphrases do not occupy cells in the relevant inflectional paradigms. Periphrasis is the use of a multi-word phrase to express a set of morphosyntactic properties, despite that set normally having one or more synthetic realizations in the language (e.g. more beautiful, instead of *beautifuler). Traditional grammars have treated periphrastic expressions as part of the morphological system, as the above layout of the Latin paradigm implies. On a descriptive level, this is tenable, but on a theoretical level the issue of whether periphrases are part of inflectional paradigms, and thus a product of the morphological component rather than the syntactic one, is controversial (Ackerman and Stump 2004,
Börjars et al. 1997, Embick 2000, Kiparsky 2005, Sadler and Spencer 2000, Spencer
1999, Vincent 1987), and has wide-ranging implications for the morpho-syntax interface.
For the purposes of defining the phenomena under discussion, I treat periphrases as the other half of paradigmatic gaps. The difference between paradigmatic gaps and periphrastic expressions is as much a matter of tradition and perspective as one with an empirical basis, because speakers who are faced with phenomena described as paradigmatic gaps will often use periphrastic constructions as a way to circumlocute around the gap and express the intended meaning. (Note, however, that this description of paradigmatic gaps should not be confused with a theoretical account of either paradigmatic gaps or periphrases.)
about inflectional loss), it would have significant implications not only for our definition of paradigmatic gaps, but also for the definition of the paradigm, and the syntax-morphology interface. 15
Third, the most crucial criterion for paradigmatic gaps is that they represent a complete failure of inflection. Any synthetic form proposed to express the inflectional properties of a cell with a paradigmatic gap will cause an utterance to crash. Substitute synthetic forms (as seen for Yimas in (1) above) may represent an unexpected inflectional form, but paradigmatic gaps represent true defectiveness. Simply put, paradigmatic gaps are the existence of “nothing” where we expect “something” in the inflectional structure of a language.
1.3. Examples of paradigmatic gaps
While paradigmatic gaps have long been noted in the grammatical traditions of various languages, they have received relatively little theoretical attention in linguistics (see below for discussion of the reasons behind this oversight). In fact, this is the first monograph on the topic that I am aware of. Correspondingly, relatively few examples of gaps have been the topic of theoretical scrutiny (see APPENDIX A for a survey). I repeat here only a few of the prominent cases, by way of demonstrating the limits and most frequent characteristics of paradigmatic gaps.
Paradigmatic gaps may be grouped, very roughly, into three categories – those which are semantically oriented, those which are related to phonotactic constraints, and those which have no apparent motivation from outside (or, often, inside) of the morphological component. This last group is what Halle (1973) had in mind when he coined the term “accidental gap”. Examples need not belong exclusively to one category
– in some cases multiple issues seem to conspire.
16
By far the most famous and most numerous paradigmatic gaps are semantic in origin, especially pluralia tantum and singularia tantum nouns. Karlsson (2000:648-9) provides a short survey of the cross-linguistic common semantic characteristics of pluralia tantum nouns:
(7) Cross-linguistically common semantic characteristics of pluralia tantum nouns
a. “Nouns denoting human and divine beings: Lat. majores ‘ancestors’, Quirites ‘bourgeoisie’... b. “Nouns denoting meetings, celebrations, or contests: Lat. comitia ‘election meeting’, nuptiae ‘wedding’... c. “Nouns denoting parts of the (human) body: Lat. tonsillae, glandulae ‘tonsils’, intestina, vicera, exta ‘intestines’... d. “Nouns denoting conditions of illness: Lat. tormina ‘stomach ache’, Eng. measles... e. “Nouns denoting clothes: Lat. exuciae ‘dress’, crepidae ‘sandals’... f. “Nouns denoting gifts, money, or wealth: Lat. divitiae ‘wealth’; Eng. amends... g. “Nouns denoting materials and utensils: Lat. impedimenta ‘baggage’, armamenta ‘[naval] fortifications’... h. “Nouns denoting tools and (more or less) technical devices: Lat. scalae ‘ladder’, cancelli clathri ‘bars [in prison]’... i. “Nouns denoting acts or results of communication: Lat. acta ‘official newspaper’, monumenta ‘annals’... j. “Nouns denoting human customs or actions: Lat. caeremoniae ‘ceremonies’, apinae, gerrae, ineptiae ‘hoax, practical joke’... k. “Place names: Lat. Athenae, Syracusae...” (Karlsson 2000:648-9).
Like the example of Russian sinij ‘light blue’ and goluboj ‘dark blue’ versus English blue, pluralia tantum nouns, singularia tantum nouns, and other instances of semantically- driven gaps are perhaps interesting for lexical semantics, or for theories of cognition and perception. Furthermore, our theories of language should be able to account for these examples and other (more quirky) kinds of paradigmatic gaps equally well. However,
17
since these sorts of gaps do not obviously interact in significant ways with the phonological, morphological, or syntactic structures of languages, they are the least interesting from the perspective of a study (like this one) of the light which paradigmatic gaps shed on morphological structure.
Somewhat more interesting are gaps which are at least in part phonotactically conditioned. These are gaps in which some expected (surface) string of sounds resulting from inflectional processes (e.g. concatenation of morphemes) would violate the phonotactics of the language. A clear example, from Swedish possessive constructions, is again provided by Karlsson (2000:648). “...[I]n Swedish the ordinary genitive ending is -s. However, -s cannot be suffixed to stems ending in sibilants. Nouns such as hus
‘house’, svans ‘tail’, Max thus have no indefinite genitive singular forms like *hus-s,
*Max-s (but they do have definite genitive singulars such as bil-en-s, Max-en-s).” 4
Karlsson considers these examples to be “borderline case[s]” because he does not accept phonological blocking as a legitimate cause of defectiveness, reserving that term for semantically driven gaps and gaps governed internally to the morphology. However, if we accept that paradigmatic gaps may arise from phonotactic restrictions but have
4 Note that Swedish possessive constructions are in many respects similar to English constructions. However, whereas the affixation of the possessive marker to a sibilant in Swedish results in a paradigmatic gap, in English speakers tend towards haplology rather than a gap. Thus, the hypothetically possible word Joneses’, as in the Joneses’ house – one belonging to many members of the Jones family, in its full, expected form would be [d°ZownzIzIz], but for most native (American?) English speakers, either [d°Zownz] or [d°ZownzIz] is preferable. The fact that similar conditions result in different resolutions in Swedish and English highlights the fact that paradigmatic gaps are, in some sense, never “forced” by the structure of the language. Where they are a response to a problematic point in the grammar, they are invariably just one among several resolution strategies that speakers could choose. 18
significant consequences for morphological structure, this case is clearly within the purview of the term paradigmatic gap as it has been defined here.5
Finally, gaps in which the expected form would be neither phonotactically nor semantically problematic have been least frequently uncovered, but have garnered some of the most intense attention. For example, the verb to forego is commonly cited as having a paradigmatic gap in the past tense (e.g. Frampton 2001). Should the form be forewent, on analogy to go-went? Or should it be foregoed, since most words have -ed in the past tense? For many native speakers of English, neither form sounds natural.
Consider the following examples, taken from a Google web search (April 2006, emphasis mine).6
(8) Speaker insecurity in using the past tense of forego
a. I for-what? I want the past tense of “forgo”. I forwent? I forgone? I forleft? Anyway, I avoided reading the bio of Zelda Fitzgerald.
b. I got it installed, but I foregoed (forewent?) the IIS install for now.
c. What's the 1st person past tense of "to forego?" As in, "I fore**** that opportunity last week." Forgoed? Forwent? Or something else? I'm boggled by this verb.
d. Because of the fussiness of the decals, I forgoed....forewent....err.... I didn't put on any of the small stencil decals.
5 Some examples of phonotactically-related gaps are not as clear as this case, however. In many instances it seems that originally phonological processes have become morphologized and picked up “eccentricities” along the way. This is typical of synchronic morphological processes, and the historical path of morphologization (Janda 1982, Zwicky 1990). See APPENDIX A for examples. 6 At the time of writing, Google (www.google.com) was the most popular web search engine. In April 2006 a search on google.com produced 123 examples in which the same person wrote multiple past tense forms of the verb to forego. 19
If the past tense of this verb is necessary, speakers often avoid the issue by finding a circumlocution (see a and d in (8)). The (consistent) use of a circumlocution rather than an inflected word form (whether it be foregoed or forewent) results in a paradigmatic gap in the past tense of this verb.
A clearer English example, but one which has been less discussed in the literature is the verb beware. The verb beware has only one form, variously considered to be either a base form or the infinitive. Fodor (1972:531) argues that “The real generalization about beware is that it can occur wherever uninflected be followed by an adjective can occur, e.g. in imperatives, infinitival complements, following modals.” Beware has paradigmatic gaps elsewhere.
(9) Sample constructions allowing beware
g. Beware of the dog. h. Do beware of the dog. i. I will beware of the dog. j. I will have to beware of the dog. k. You must beware of the dog.
(10) Sample constructions NOT allowing beware – paradigmatic gaps
l. *John’s bewaring of the dog was unnecessary. m. *John bewares of the dog. n. *John bewared / bewore of the dog. o. *John doesn’t beware of the dog. p. *John has bewared / beworn of many dogs in his lifetime. q. *John is bewaring of your dog. r. *Bewaring of the dog, John circled the yard.
These gaps in the paradigm of beware seem to have no motivation.
20
These are nearly isolated examples in English, but in languages with elaborate inflectional systems, the possibility of gaps is more significant. There are a handful of imperative gaps in Hungarian, dozens of gaps in the first person singular present of
Spanish verbs and in definite Swedish adjectives, perhaps a hundred gaps in the first person singular non-past in Russian (Halle 1973), and thousands of gaps in the genitive plural of Modern Greek nouns, for example. Here I present an example from the genitive plural of Russian nouns. See CHAPTER 3 and APPENDIX B for Modern Greek,
APPENDIX E for Russian verbs, and APPENDIX A for Hungarian, Spanish, Swedish, and a variety of other examples.
Russian has a number of paradigmatic gaps in the genitive plural of nouns. The distribution of these gaps may be delineated based on inflection class (or, alternatively, the expected form of the genitive plural), stem shape, and the paradigmatic stress pattern, but the motivation for their existence is unclear.7
For most nouns (i.e. excluding i-stem nouns), the genitive plural marker takes one of three forms: It can be realized by the suffix /-ov/, by the suffix /-ej/, or with a “zero ending”, meaning that the genitive plural has the same form as the noun stem. Example paradigms are given below.8
7 But see Pertsova (2005) for a theory based on the idea of lexical conservatism. 8 These paradigms do not represent vowel reduction or some other aspects of pronunciation. 21
чин ‘rank’ конь ‘horse’ singular plural singular plural nominative čín čin-ý kónj kónj-i accusative čín čin-ý konj-á konj-éj genitive čín-a čin-óv konj-á konj-éj dative čín-u čin-ám konj-ú konj-ám prepositional čín-e čin-áx konj-é konj-áx instrumental čín-om čin-ámi konj-óm konj-ámi
болото ‘swamp’ ущелье ‘gorge’ nominative bolót-o bolót-a ušjélj-e ušjélj-a accusative bolót-o bolót-a ušjélj-e ušjélj-a genitive bolót-a bolót ušjélj-a ušjélj-ij dative bolót-u bolót-am ušjélj-u ušjélj-am prepositional bolót-e bolót-ax ušjélj-e ušjélj-ax instrumental bolót-om bolót-ami ušjélj-em ušjélj-ami
Table 2: Examples of Russian noun declension
Jakobson (1984:109) was the first to formulate the rule of thumb which describes the distribution of the genitive plural suffixes in the majority of cases: If the nominative singular is marked with a suffix (e.g. knig-a ‘book’, bolot-o ‘swamp’), the genitive plural is not and will be the same as the nominal stem. If the nominative singular form is the same as the stem (e.g. konj ‘horse’, čin ‘rank’), the genitive plural has either the suffix -ov or the suffix -ej, depending on palatalization of the stem-final consonant (-ov with non- palatalized stems, -ej with secondarily palatalized stems). Nouns like ušjélje ‘gorge’ are among the few significant exceptions to this rule. As can be seen from the above example, nouns in this declension class which have palatalized stems have the genitive
22
plural suffix /-ej/ (realized here as [-ij]), despite also having an overt marker in the nominative singular.
Notably, when the genitive plural is not marked by an overt suffix, but the word is stressed on the ending in the plural oblique cases (i.e. in the genitive, dative, instrumental, locative/prepositional), stress retracts onto the stem.
место ‘place’ singular plural nominative mje!st-o mjest-a! accusative mje!st-o mjest-a! genitive mje!st-a mje!st dative mje!st-u mjest-a!m prepositional mje!st-e mjest-a!x instrumental mje!st-om mjest-a!mi
Table 3: Russian nominal paradigm showing end stress in all oblique plural forms except the genitive plural
“Interestingly, it is a subset of precisely this group of nouns that have gaps in the genitive plural...” (Pertsova 2005:16-17). Pertsova argues that the crucial generalization is that the words with genitive plural gaps have stress on the first syllable of the suffix throughout the paradigm (i.e. no stem stress at all). Compare the stress and genitive plural of mesto above to mečta below.
23
мечта ‘dream’ singular plural nominative mječt-a! mječt-y! accusative mječt-u! mječt-y! genitive mječt-y! -- dative mječt-e! mječt-a!m locative mječt-e! mječt-a!x instrumental mječt-o!j mječt-a!mi
Table 4: Russian nominal paradigm - gap in the genitive plural
(11) Examples of genitive plural gaps in Russian
nominative singular genitive plural gloss fata *fat ‘veil’ yula *yul ‘weasel’ mečta *mečt ‘dream’ mzda *mzd / *mozd / *mzod / etc. ‘bribe (archaic)’ dno *dn / *dan / *don ‘bottom’
Two of these nouns (mzda, dno) lack stem vowels, making phonological illicitness a likely factor for the lack of a genitive plural form. However, the majority of the expected but non-attested forms (N = appx. 20) would be pronounceable and would not violate the phonotactics of Russian (e.g. *fat, *yul). Yet there are also no obvious morphological or semantic motivations for the gaps.
Ultimately, the point here is that even the narrow use of the paradigmatic gap that I employ in this dissertation covers wide-ranging phenomena. Some gaps seem to be synchronically motivated – semantically, phonologically, morphologically, or by some mixture of factors. Others have no obvious logic behind them. Paradigmatic gaps affect
24
nouns, verbs, adjectives, and it seems any other lexical category which realizes inflectional categories. Some gaps are idiosyncratic within their languages, while others are widespread. In short, there is a wealth of data to explore.
1.4. Paradigmatic gaps as probabilistic phenomena
In defining paradigmatic gaps ((6) above), a key criterion was that there be no synthetic form to express a particular inflectional property set. Unfortunately, this aspect of the definition is not entirely tenable. Speakers are wont to innovate, variation is rampant, and with a sufficiently large corpus it is possible to find attested synthetic forms for virtually all, if not all, purported examples of paradigmatic gaps. However, I argue that this sort of evidence need not vitiate either the given definition or the significance of defective inflection. I propose here a “save” – paradigmatic gaps as probabilistic phenomena.
Consider the following data on the frequency of forms of the verbs go, undergo, and forego (alternative spelling forgo). As noted in Section 1.3, forego has been cited as having a paradigmatic gap in the simple past tense.
25
source go goes went goed gone going British 83,770 14,536 45,872 1 18,455 62,663 National 37.2% 6.5% 20.4% 0% 8.2% 27.8% Corpus 43.7% 20.4% 36%
Google 6.32 bill 657,000,000 414,000,000 1,860,000 302,000,000 1.1 billion 71.9% 7.5% 4.7% 0.02% 3.4% 12.5% 79.4% 4.7% 15.9%
Table 5: Some frequency information about the English verb go9
source undergo undergoes underwent undergoed undergone undergoing British 613 123 550 0 570 582 National 25.1% 5% 22.5% 0% 23.4% 23.9% Corpus 30.1% 22.5% 47.3%
Google 55,000,000 13,300,000 20,700,000 395 31,900,000 50,900,000 32% 7.7% 12% 0.0002% 18.6% 29.6% 39.7% 12% 48.2%
Table 6: Some frequency information about the English verb undergo
9 A word of caution about interpreting the data in Table 5 through Table 7... First, the information in these tables is not directly parallel, despite all having the formative
source for(e)go for(e)goes for(e)went for(e)goed for(e)gone for(e)going British 197 4 0 0 26 13 National 82.1% 1.7% 0% 0% 10.8% 5.4% Corpus 83.8% 0% 16.2%
Google 12,500,000 690,000 147,500 950 4,710,000 41,620,000 21% 1.2% 0.2% 0.002% 7.9% 69.8% 22.2% 0.2% 77.7%
Table 7: Some frequency information about the English verb forego
Taken cautiously, these tables allow us some insight into the question of to what degree a paradigmatic gap is the complete non-existence of a relevant inflected form. In the
British National Corpus, both go and undergo have productive past tense forms, went and underwent, representing slightly more than 20% of the total lemma frequency. At the same time, the corpus does not contain a single example of forgoed, foregoed, forwent, or forewent. Considering that the British National Corpus (BNC) contains one hundred million words (balanced, 90% text, 10% speech), this is a notable absence. Based on the
BNC, the definition of paradigmatic gaps as given above is sustainable.
However, with a sufficiently large sample (one hundred million words not apparently being enough) it is possible top find numerous examples of both for(e)goed and for(e)went. Using Google, I located approximately 147,500 examples of for(e)went, and 950 examples of for(e)goed.
I argue that the number of attested examples of a particular inflected form is largely irrelevant to the question of whether there is a gap. Instead, it is the relative
27
frequency which is important.10 If we take the results of the Google searches at face value, the past tense forms of for(e)go constitute only 0.2% of total attestations. This is presumably well below the expected type frequency. Thus, we should perhaps think of paradigmatic gaps as instances in which the number of attested forms is significantly lower than would statistically be expected by the combination of lexeme frequency and type frequency.11 In short, we cannot consider gaps as absolute phenomena, but we may identify them probabilistically while staying in the spirit of the definition in (6).
Albright (2003) comes to a similar conclusion. He argues that gaps are gradient phenomena, based on evidence that Spanish speakers’ confidence ratings in their own production of present tense Spanish verbs (which he assumes to correlate with the presence or absence of a gap – low confidence means a gap) approached a normal distribution, rather than falling into two sharply defined groups depending on whether the relevant form prescriptively had a paradigmatic gap or was productive.
Furthermore, paradigmatic gaps trigger at least two peculiar and even contradictory reactions from speakers which signal that gaps are complicated phenomena unto themselves, and not simply examples of infrequent inflection. First, speakers can often produce the expected form, but then often reject it without offering another form as an alternative (Albright 2003, Hetzron 1975, Orgun and Sprouse 1999). Why should they reject the relevant forms, but not do so with other infrequent words? Second, as noted above, speakers often display signs of being unsure about their own productions. They
10 There is good reason to believe that the relative frequency of different forms of the same lemma is psychologically real for speakers, and not just a convenient tool of linguists. Hay (2003) demonstrates the effects of relative frequency of inflectionally-related form on lexical access, semantic drift, and other areas of language. Hay does not specifically address the issue of defining paradigmatic gaps, but we can extrapolate the possibility that speakers themselves use relative frequency to identify the location of gaps. 11 See the GLOSSARY for discussion of use of the terms lexeme and lemma. 28
may report insecurity directly (Albright 2003), or indirectly by “self-correcting”, as in (8) above. From these two phenomena alone, we can conclude that paradigmatic gaps represent a quirky idiosyncrasy lurking at the fringes of the inflectional system, but nonetheless a phenomenon which cannot be reduced to a random side-effect of frequency distributions.
1.5. The debate over the value of marginal phenomena
There is a long-standing debate in linguistics over the value of data that is “marginal” to the overall linguistic system. Paradigmatic gaps, which are among the most idiosyncratic and language-specific phenomena imaginable, are naturally an important object of this debate.
A distinction between “central” and “marginal” linguistic traits, broadly defined, has been employed in many areas of linguistics. To take just two examples, Ito and
Mester (1995) argue that the Japanese lexicon is structured according to a core/periphery distinction, and exemplar models of speech perception and acquisition posit different effects depending upon whether a given example falls in the center of an exemplar cloud or on its periphery (Hume and Johnson 2001, Pierrehumbert 2001). The list could easily go on. However, the most formalized distinction between central and marginal phenomena comes from generative syntax, where central phenomena are often defined as those which are cross-linguistically universal, while marginality is equated with language-specificity. It is in this sense of a core/periphery distinction that the value of language-specific phenomena has been most acutely questioned.
29
1.5.1. The core/periphery distinction in generative syntax
Proponents of generative grammar have often shown little interest in non-general phenomena. As Culicover and Jackendoff (1999:544) succinctly explain,
For more than thirty years it has been widely accepted within generative grammar that the only syntactic phenomena truly worth studying are those that are maximally general – those that require minimal language-specific detail in their formulation. The motivation for this view is that these general phenomena are likely to reflect something fundamental and universal about the human capacity for language. By idealizing away from language-specific idiosyncrasies and by formulating maximally general rules, constraints, or relations on structures, generative grammar is seen to provide some preliminary insights into the nature of the language faculty.
If we accept that biological encoding has the capacity to explain similarities between languages (i.e. accept the hypothesis of Universal Grammar (UG)), language-specific phenomena are relatively uninteresting because they do not reveal UG. Chomsky goes so far as to describe accounts of language-specific phenomena as “taxonomic” because he argues that the only true explanation lies in the universal properties of the language faculty (Chomsky 2000b:14). Generative research has thus long maintained a loose distinction between universal and language-particular phenomena which has greatly influenced the focus of study.
Developments within the Principles and Parameters framework and Minimalism have left even less room for study of non-universal phenomena. Chomsky (1981) went one step further than previous generative work by crystallizing the distinction between
“core” and “peripheral” phenomena, turning them into fixed theoretical concepts and hypothesizing that they operate according to different principles. The core consists of
30
parameters with limited settings which restrict the degree to which languages may vary and broad principles for operations on structures (e.g. movement). Together the principles and parameters are argued to capture implicational relationships between seemingly distinct syntactic constructions (e.g. between general word order and order of modifiers relative to their heads). The peripheral domain is a combination of lexical specification and rules to account for “performance factors”, “historical residue”, etc.
Under this assumption, Chomsky (2000a:112) formulates the role of non-general phenomena (what he calls ‘apparent imperfections’, reflecting his view that languages are self-optimizing systems) as follows.
Apparent imperfections come in several varieties. Given some apparent property P of language, the following outcomes are possible: a. P is real, and an imperfection. b. P is not real, contrary to what had been supposed. c. P is real, but not an imperfection; it is part of a ‘best way’ (perhaps not unique) to meet design specifications. The most likely – hence least interesting – outcome is [a]. Conclusion [b] is less likely, therefore more interesting; efforts within the Minimalist Program have sought to establish it in particular cases. Conclusion [c] is the most interesting possibility.
In short, language-particular phenomena that can be reduced to more general principles, or shown to be epiphenomenal at a theoretical level, may be brought within the fold of the language faculty. The goal of linguistic theory, according to Chomsky, should be to explore to what extent such explanations are possible. Other language-particular phenomena are among the “least interesting” topics for study, and amount to semi- random miscellany.
31
Within generative syntax, the core/periphery distinction and even the more general premise that maximally general phenomena will be maximally fruitful topics of study have not been universally accepted. There is not space here to go into the details of the arguments, but syntactic accounts represent varying degrees of opposition to
Chomsky’s original core/periphery formulation. Much work essentially accepts
Chomsky’s formulation in full (e.g. Avgustinova 1994, Iwakura 1981, Koster 1987,
Tracy 1986). Others largely accept the distinction but question its structure/function (e.g.
Hirschbühler and Rivero 1981). Yet others have been more critical of the assessment that studying peripheral phenomena will not lead to significant insight or explanation of linguistic structure. Notably, Baker (1991:388-389) claims that “...if core grammar is defined as the realm of principles and parameters, then a peripheral component must also be assumed to exist, a component in which are found rules of grammar that are basic in their own right rather than epiphenomenal” to principles of a more fundamental nature.
Culicover and Jackendoff (1999) come to similar conclusions. Finally, some researchers simply reject the core/periphery hypothesis outright (e.g. Hudson 1985, McCawley
1988).
Yet, in spite of this debate, the majority of syntactic research has continued to skew towards study of maximally productive phenomena.
1.5.2. The (lack of a) core/periphery distinction in morphological theory
Within morphological theory, the degree to which marginal phenomena have been devalued is proportional to the degree to which any particular theory applies syntactic principles below the level of the word. Word Syntax (Lieber 1992), Distributed
32
Morphology (Halle and Marantz 1994), and Williams’ theory of morphological headedness (Williams 1981) have each, to some extent, assumed that morphological and syntactic principles are fundamentally the same (headedness, movement, etc.). These theories have also been almost exclusively concerned with productive morphological phenomena. Theories which assume a separate morphological component operating according to its own independent principles have been less under the sway of arguments about the significance of core and peripheral phenomena. This is particularly true of research within Word and Paradigm models (Aronoff 1994, Blevins to appear 2006,
Bochner 1993, Matthews 1972, Robins 2001[1959], Stump 2001b, Zwicky 1985).
Hypothetically, however, we could posit a morphological component operating according to principles distinct from those of syntax which nonetheless distinguishes universal and language-specific morphological properties. In evaluating a related idea,
Anderson phrases the issue as follows:
...the alacrity with which the syntactician strives to know immediately, for any language under discussion, whether it is SVO, SOV, or VSO... comes from the fact that at least since the work of Greenberg (1966[1963]), we have felt confident that knowing the relative order of the main constituents of a simple transitive main clause will allow us to predict a whole host of other things [e.g. the relative order of modifiers]... When we have achieved such a result, we are entitled to say we have (at least part of) a substantial typology, where the range of language types is then defined by the range of available parameter settings... It makes excellent sense, however, to ask just how likely it is that a situation comparable to that hypothesized in syntax will in fact turn out to obtain in morphology. How plausible is it, that is, that the system of Word Formation Rules in a language will turn out to be globally parameterized in the way syntactic structure seems to be[?]... When one turns to the details of how words are composed of phonological material, an immediate impression is that the number of substantial correlations within a language that could give rise to a productive parameterization of the sort sketched above is very limited. (Anderson 1992:323- 324)
33
The claims that (a) there is a morphological component and (b) morphology does not have any equivalent to the implicational relationships holding at the level of syntax together lead to a simple conclusion: there is no a priori reason to believe that morphological systems are organized according to the universality of phenomena. The simplest hypothesis is that all inflection operates according to the same principles, and barring evidence to the contrary, we should not introduce unnecessary mechanisms; the distinction between core and periphery thus has no apparent function in morphological theories holding to these claims.12
Proponents of Word and Paradigm models assume (albeit implicitly) that all inflection operates according to a single set of principles. This places both central/productive and marginal/idiosyncratic morphological phenomena on equal theoretical footing. Not surprisingly then, a number of researchers in these frameworks have recently taken up study of form-meaning mismatches and other (semi-) idiosyncratic phenomena (see Section 2.1) (Hippisley et al. 2004, Sadler and Spencer
2000, Spencer 2006, Stump 2006, Zwicky 1985). Also not surprisingly, proponents of
WP models have not actively engaged in a defense of the value of marginal phenomena.
12 Anderson himself seems to hold conflicting views on this point. On the one hand, he argues that morphology is epiphenomenal to syntax and phonology (p. 324 and elsewhere). We might therefore expect him to posit morphological structures parallel to those in either syntax or phonology. On the other hand, he argues that morphology operates according to its own principles, distinct from those of either syntax or phonology (also p. 324 and elsewhere). 34
Inasmuch as there is justification for the claim of a morphological component, the potential value of language-specific data is theoretically fundamental.13
1.6. Paradigmatic gaps as marginal phenomena
Especially considering the different stances often held by morphologists and syntacticians on the theoretical relevance of language-specific phenomena, it is perhaps surprising that paradigmatic gaps have received little attention until recently, and that even fewer studies have considered them interesting phenomena in their own right.
Paradigmatic gaps have historically been treated as curious, but isolated phenomena which do not reveal linguistic structure; there are relatively few studies from the 1970’s and 1980’s that mention paradigmatic gaps in any significant detail (Eliasson
1975, Fodor 1972, Halle 1973, Hetzron 1975, Iverson 1981, Maslov 1964, Soboleva
1979, Strunk 1977, Woolf 1980, Zwicky 1989), and only four are concerned with the implications of paradigmatic gaps for synchronic linguistic theory (Halle 1973, Hetzron
1975, Iverson 1981, Zwicky 1989). This correlates, perhaps not coincidentally, with a period during which theories treating morphology according to syntactic principles were dominant.
Recently, there has been a surge of interest in paradigmatic gaps (Albright 2003,
Baerman and Corbett 2006, Baronian 2005, Fanselow and Féry 2002, Frampton 2001,
Hansson 1999, Hudson 2000, Johansson 1999, McCarthy and Wolf 2005, Morin 1995,
Orgun and Sprouse 1999, Pertsova 2005, Prince and Smolensky 2002[1993], Raffelsiefen
13 I believe that justification for a separate morphological component exists, but there is not space here to support this statement. See Zwicky (1990) on synchronic characteristics which are unique to morphological systems. See Janda (1984) for a similar argument from a historical perspective. Finally, see van Marle (2000) for an argument that paradigmatic and syntagmatic relations are of different importance to morphological and syntactic systems. 35
2004, Rice 2005b, Törkenczy 2002, Vincent 2005, Wunderlich 1999). While none of this work is explicitly formulated within a Chomskyan framework, much of it has followed the principle of subsuming paradigmatic gaps to other phenomena. For example, the majority of recent work on paradigmatic gaps has utilized Optimality Theory, a framework which takes competition between general grammatical principles (constraints) as a foundational assumption (Fanselow and Féry 2002, McCarthy and Wolf 2005,
Pertsova 2005, Prince and Smolensky 2002[1993], Raffelsiefen 1996, Rice 2005b,
Törkenczy 2002, Wunderlich 1999). Ultimately, while there has been a notable increase recently in the study of paradigmatic gaps, most of these studies have focused on removing at a theoretical level the apparent idiosyncrasy that gaps present. They essentially follow option (c) in Chomsky’s explication of the role of peripheral phenomena: “P is real, but not an imperfection; it is part of a ‘best way’ (perhaps not unique) to meet design specifications” (Chomsky 2000a:112).
Virtually no research has been conducted on the whether paradigmatic gaps are ever phenomena unto themselves theoretically, and the ways in which they might reflect and/or interact with more productive aspects of inflection. Put differently, there has been little work on defective inflection which challenges the assumption that gaps are isolated and not integrated into linguistic structure. This places the study of paradigmatic gaps behind the recent study of other language-specific phenomena, such as deponency (e.g.
Spencer 2006), periphrasis (e.g. Ackerman and Webelhuth 1998, Ackerman and Stump
2004, Börjars et al. 1997, Kiparsky 2005, Sadler and Spencer 2000, Vincent 1987), suppletion (Aski 1995, Fertig 1998, Hippisley et al. 2004, Juge 1999) and heteroclisis
(Stump 2006). These studies argue that the inflectional structures that we posit should be
36
driven by peripheral phenomena because structures that can handle idiosyncratic phenomena are likely to be able to handle productive phenomena, but not vice versa.
They have thus made a point about the value and nature of marginal inflection. Again, probably not coincidentally, this new-found focus on marginal morphological phenomena has correlated with the rise of modern WP theories (for all intents and purposes beginning with Matthews (1972), although with roots in traditional models, Hockett (1954), and
Robins (2001[1959])).
Baerman and Corbett (2006), Halle (1973), Joseph (1997), Orgun and Sprouse
(1999), and Vincent (2005) are among the rare studies which contend that gaps are not always epiphenomenal to other grammatical principles. For example, Baerman and
Corbett (2006) argue that Russian verbal gaps must be treated as theoretically “real”, and therefore promote a theory which allow form-meaning mismatches. The primary goal of
Joseph (1997) is to argue that marginal phenomena, including paradigmatic gaps, exhibit patterns that are different from those of more productive phenomena, but which must equally be accounted for by linguistic theory. Finally, Orgun and Sprouse (1999), working within Optimality Theory (OT), argue that paradigmatic gaps cannot be handled by the general constraints typically considered to inhabit EVAL, the component within which output forms are selected. Instead, they argue for an additional component
CONTROL, which contains language-specific constraints and serves as an absolute filter on grammatical output. These accounts operate within widely different theoretical assumptions, but they have one thing in common in that they all posit that paradigmatic gaps reflect general inflectional structure, and that our linguistic theories should be able to account for the full range of language data.
37
In line with the position of Joseph (1997), Baerman and Corbett (2006), etc., I maintain that there are several potential reasons to investigate marginal phenomena such as paradigmatic gaps. The most important ones have to do with potential interactions between paradigmatic gaps and more productive inflectional phenomena. If we can show that paradigmatic gaps directly result from the inflectional structure of the relevant language, this would give clues to both the expansiveness and the nature of that structure.
In the following chapters I assess the validity of paradigmatic gaps and other superficially defective phenomena as evidence for one particular structure that has been posited – inflectional paradigms – and test the limits of inflectional morphological structure more generally. We identify paradigmatic gaps by the system of oppositions presented by the inflectional categories of a given language.14 However, we might wonder whether there are deeper relationships between paradigmatic gaps and inflectional structure (in this case the inflectional paradigm). In particular, in this dissertation I explore one simple question, with four subquestions:
Is paradigmatic structure necessary to adequately explain the appearance and/or distribution of paradigmatic gaps?
o Do paradigmatic gaps arise from regular inflection? If so, how?
o Does paradigmatic change begin from points of paradigmatic weakness?15
14 Vincent (2005) clearly had this idea in mind when he said that paradigmatic gaps are fundamentally underpinned by the concept of the paradigm. 15 The term “paradigmatic weakness” needs some explanation. If we define paradigmatically weak areas as those which are likely to be subject to change, then to say that paradigmatic change begins from weak points in the paradigm is circular. I instead use this term to capture the idea that speakers may not be completely confidence in their usage of all areas of inflectional morphology. It is easy to imagine a variety of reasons that speakers might report low confidence in their use of the their native language – variation 38
o Do paradigmatic gaps and regular inflection pattern similarly/differently in
any significant ways, especially in language change?
o Is the paradigm, as currently posited in morphological theory, sufficient to
account for paradigmatic gaps?
My most general conclusion is that paradigmatic gaps cannot be understood without the concept of the paradigm. Gaps thus stand as evidence for paradigmatically-based theories of inflection. At the same time, gaps call into question many of the central tenets of those theories, such as the reliance on morphological blocking and the common theoretical definition of the paradigm as the combinatory possibilities of inflectional properties.
1.7. Summary
In this chapter, I have explored uses of the term paradigmatic gap. While the phrase has been used to describe wide-ranging data, I have outlined that particular use which defines the topic of the subsequent chapters. A paradigmatic gap for the present purposes is defined as the non-existence of a synthetic form expressing a set of inflectional properties for a particular lemma, despite the language normally having a synthetic form expressing that same set for comparable lexemes. In short, it is the failure of inflectional morphology.
Throughout the following pages I discuss examples of paradigmatic gaps, details of their distribution and relation to inflectional structure, hypotheses regarding their
which carries prescriptivist valuation, low frequency, indeterminate inflectional models, etc. Those areas about which speakers are insecure, for whatever reason, are “weak” points in the grammar. 39
causes, and theories for their account, but in this chapter I have attempted to provide a quick overview of the primary types of paradigmatic gaps – those which are semantically-oriented (e.g. pluralia tantum), phonotactically-oriented (e.g. Swedish possessive markers), and morphologically- or otherwise-oriented (e.g. Russian genitive plural gaps, English forego and beware). This taxonomy was intended for clarificatory purposes, and should not be taken as a serious classification of the entire range and scope of paradigmatic gaps. (A survey is available in APPENDIX A.) Still, these few examples provided here demonstrate that paradigmatic gaps, even in the most narrow sense of the term, have diverse characteristics.
Finally, the skeptical reader might observe that paradigmatic gaps rarely (if ever) represent completely failed inflection. After all, even the verb forego – a famous past tense gap in English – produces thousands of examples, if a sufficiently wide sample of the language is considered. Thus, we might wonder whether gaps are simply a case of infrequent inflection, and not a phenomenon unto themselves. I argue, however, that such a perspective would not be useful; paradigmatic gaps show some unique and often peculiar characteristics. I thus propose a slight modification of the definition of paradigmatic gaps – a change from thinking of them in terms of absolute non-attestation to defining them in probabilistic terms. A paradigmatic gap is the overwhelming failure of inflection, as evidenced by a frequency that is significantly lower that what would be expected based on the combination of lexeme and type frequency.
And with this foundation, I now turn to the more detailed issues at hand. In
CHAPTER 2 I outline my theoretical framework. CHAPTERS 3 through 5 present the meat of the original data – explorations of how paradigmatic gaps in Modern Greek,
40
Russian, and Croatian are shaped by paradigmatic structure, and mediated by speakers’ strategies.
41
CHAPTER 2
AN OVERVIEW OF THE PARADIGM IN A WORD AND PARADIGM MODEL
2.1. An overview of the paradigm
This dissertation hangs crucially on the paradigm, a term which has been subject to many definitions, and a concept which has been incorporated (or not) into morphological theory in a variety of ways. Whether an adequate description of one inflectional form necessitates reference to other inflectionally related forms, the essence of paradigmatic structure, is currently a central question in morphological theory. Two camps – Word and Paradigm (WP) models and a subset of Optimality Theory known as Optimal
Paradigms Theory – argue that there are systematic co-occurrence restrictions and other paradigmatic effects which demonstrate that inflected forms are not atomistic, but rather are integrated into a larger inflectional structure (e.g. Carstairs 1983, papers in Downing et al. 2005, Joseph 2005, Matthews 1972, Stump 2001b, Wunderlich 1995, Zwicky
1985). WP models in particular place connections between inflectional forms at the center of the theory by positing paradigms as theoretical constructs and the locus of inflectional structure. Opponents of the paradigmatic view maintain that paradigms are superfluous and that so-called paradigmatic phenomena may be explained through other theoretical tools (e.g. Bobaljik 2001, Noyer 1998, Raffelsiefen 2005).
42
2.2. The paradigm in traditional Word and Paradigm models
We may think of inflectional structure in the pre-generative world as being fundamentally organized according to two main components – words and paradigms.
Bloomfield’s (1933:223) definition of paradigms is representative: “...English nouns occur, for the most part in parallel sets of two: a singular noun (hat) and a plural noun derived from the former (hats). Given one of these, the speaker is usually capable of producing the other. Each such set of forms is called a paradigmatic set or paradigm.”
This definition shows that Bloomfield (among others) conceptualized the paradigm as an exemplar-based model in which concrete forms represent the abstract patterns of the inflectional class, and the predictability of that pattern from exemplars was central to understanding inflectional structure and inflectional change.16
Not all words are equally good predictors of inflectional patterns. For example, a hypothetical Bosnian/Croatian/Serbian (BCS) verb form mode could be either a third person singular present tense form, belonging to the pattern for ići ‘to go’ (3.SG.PRES ide), or be a third person plural present tense form, belonging to the pattern for raditi ‘to work’ (3.PL.PRES rade). Mode is thus a poor predictor of inflection class membership.
SINGULAR PLURAL SINGULAR PLURAL 1ST PERSON idem idemo radim radimo 2ND PERSON ideš idete radiš radite 3RD PERSON ide idu radi rade
Table 8: Two BCS verb paradigms demonstrating implicational relationships within paradigms
16 It is unclear whether the inflection class was considered to exist apart from particular exemplars; a non- mentalist view of language made this largely a moot point. 43
By contrast, in BCS all of the first and second person forms are excellent predictors; if we know whether the hypothetical verb is modim or modem in the first person singular form, we are able to absolutely determine all of the other word forms because the form uniquely signals the class, and the class uniquely signals the form (i.e. this is a biunique implicational relationship). Forms which share this sort of relationship with inflection classes are traditionally called principal parts. We can thus think of the (pre-generative) paradigm as the instantiation of a set of implicational relationships between individual word forms on one level and between words and the more abstract notion of inflectional class on another. These relationships are not always biunique, but biunique relationships are more important to paradigmatic structure than other types of relationships.
These relationships serve more purposes than simply delineating inflectional types. For example, they offer an explanation for analogy (12).
(12) A typical example of four-part linguistic analogy
dog : dogs cow : X ______X = cows
Historically, cow had the plural form kine. Speakers arrive at cows because analogy, as a process of logic applied by speakers to data, is fundamentally abductive and then deductive in its reasoning (Andersen 1973, 1978). Abductive reasoning works as
44
follows: C is an observable fact. A, if true, would explain C. There is no better explanation for C than A. Conclusion: A is probably correct.
(13) Analogy as an abductive process
/-z/ is an observed plural for nouns. (rule) cow is a noun. (observation) ______The plural of cow is probably cows. (case)
In terms of the pre-generative conception of the paradigm, cows is the result of abduction from a word (observation) and an implicational relationship (rule) to an inflectional paradigm (case). Abduction may lead to the “correct” result (as when the plural of dog is concluded to be dogs), or it may lead to an innovative one, as when the plural of cow is concluded to be cows. In both cases, the process is analogical, but we only define the latter case as language change (and often only attach the term “analogy”, incorrectly, to these cases). This is possible because abductions leak; some word forms are bad predictors, i.e. they implicate more than one inflectional pattern. Fundamentally, analogy thus served as a crucial piece of evidence for the paradigm in traditional conceptions because it relied on inflectional connections for its reasoning.17
17 Moreover, the prevalence of analogy among inflectionally related forms but its relative paucity among derivationally related forms indicates that different connections are made across the two types of relations, and supports the traditional organization according to inflectional but not derivational patterns. 45
2.3. The (lack of a) paradigm in most generative, morphemic theories
2.3.1. Classic generative grammar
Classic generative grammar treated analogy as a deductive rather than abductive logic process. In deductive reasoning a person starts with a rule, applies it to a particular case, and produces a result which is in line with the rule; the conclusion is of no greater generality than the input conditions. In the classic generative approach, analogy is rule generalization of a rule to a new underlying form, or constriction of a rule so that it no longer applies to an underlying form. In other words, analogy is no different than other word formation processes, except that from a diachronic perspective we can see that the scope of the generalization has changed over time.
The reconceptualization of analogy as a relationship between underlying and surface forms seems to often produce the same result as the pre-generative approach, but with a different kind of explanatory force. In the pre-generative paradigmatic conception, the non-biunique nature (“leaking”) of some implicational relationships between words and paradigms provides the explanatory force for analogy. However, in the (early) generative view, the force behind analogy lay not in the implicational ambiguity of the surface form, but in the claim that grammatical systems value simplicity.18 Analogy represented either the removal of rules of allophonic variation, or extension of rules to broader conditioning environments.
18 See Thomason (1976) for an argument that language change is not driven by grammar simplification. 46
The issue to which I should like to address myself here is analogical change. Traditionally visualized as the extension of surface patterns (in terms of proportional schemata) it has more recently been given another interpretation as the elimination of arbitrary complexity in the linguistic system... At the back of this lies a new view about the nature of the concrete process of analogical change and how it relates to the acquisition and use of language. A proportional view of analogy fits naturally into a theory of language acquisition based on substitution- in-frames techniques and equivalent ‘taxonomic’ devices. The idea that analogy is simplification of grammar jibes better with the idea that language acquisition is based on a general rule schematism in conjunction with an evaluation measure which selects the simplest grammar... (Kiparsky 1978:78)
This approach eliminates the concept of the paradigm because surface forms are the product of generalizations, not the source of them, and so there is no (obvious) reason to connect surface forms theoretically. In short, in the classic generative approach, there were no longer direct relations between surface forms. Inflectionally related forms were connected by virtue of having the same underlying form, and this connection is used to account for so-called paradigmatic effects without any theoretical equivalent to the paradigm.
2.3.2. Modern generative, morphemic theories
Modern morphological theories are split with regard to the need for paradigmatic structure, but most generative, morpheme-based approaches follow the classic generative approach.19 To take a single example, input-output relations are still central to most versions of Optimality Theory (OT) for describing phenomena like paradigm uniformity
(Kiparsky 2000, Wedel 2004).20
19 The work of Carstairs(-McCarthy) is the exception to the rule (Carstairs-McCarthy 1994, Carstairs 1983). 20 The exception is the recent work of McCarthy (2005), and work that has followed his lead (Downing 2005, Lloret to appear 2004, Rice 2005b). McCarthy argues that entire inflectional paradigms are 47
Paradigm uniformity has been defined in both historical and synchronic terms. In language change, paradigm uniformity is the end-state result of paradigmatic leveling,
“...the complete or partial elimination of morphophonemic alternations within paradigms” (Hock 1991:168).
grObn ‘to dig’ SINGULAR PLURAL 1ST PERSON grOb grObn 2ND PERSON *grebst → grObst grObt 3RD PERSON *grebt → grObt grObn
Table 9: Leveling in Yiddish verbal paradigms (Albright 2002:2)
Synchronically, paradigm uniformity is the lack of morphophonemic alternations within paradigms where those alternations would be expected based on purely phonological criteria. These definitions are clearly related, although the purely synchronic definition is somewhat narrower.
The example above is most obviously accounted for via some generalization that all inflectionally-related forms must share the relevant aspect of form. However, those who consider paradigms to be too unconstrained (e.g. Bobaljik 2001) might seek to produce the same result as an epiphenomenon. In a basic OT formulation, it is possible to capture paradigm uniformity effects by using correspondence constraints.
Correspondence constraints force some relation between two forms. One useful and now long-standing constraint is IDENT(I,O), which forces inputs and candidate output candidate outputs. This allows McCarthy to account for paradigm uniformity as an output-output effect – a constraint states that all inflectionally-related forms of the same lemma must be identical. 48
forms to be segmentally identical. In OT, the relative ranking of constraints is more important than the absolute number of violations that a form incurs. Thus, if IDENT outranks phonotactic constraints, morphological concatenation leads to phonotactic violations when the juxtaposed sounds are not generally licit in the language.21 The hypothesized ranking of constraints dictates that identity between the input and the output is more important than repair of the phonotactic constraint. Importantly, because all inflected forms of a word share the same underlying root, and have the same constraint rankings, the result would be paradigm uniformity. A relationship between inflected forms is an illusion.
There is not space here to explore all of the ways that modern generative, morphemic approaches have accounted for superficially paradigmatic phenomena without recourse to paradigmatic structure.22 The important point is that for these theories the paradigm is a convenient heuristic – a descriptive tool of linguists but not a theoretical construct.
2.4. The rebirth of Word and Paradigm models
In recent years, some frameworks of morphological study have reincorporated connections among inflectionally related forms. Hockett (1954) and Robins
(2001[1959]) in many respects were the inspiration for this trend, but Matthews (1972) was, for all significant purposes, the spark which has led to the current rebirth of Word
21 It is not possible for IDENT to force phonotactic violations except at concatenation boundaries due to the postulation of Richness of the Base (Smolensky 1996). 22 For the most part, theories which treat (inflectional) morphology according to the same principles as syntax (see SECTION 1.5.2) have not taken the relevant phenomena as a topic of study. However, the interested reader is directed to work within Distributed Morphology which uses underspecification of morphosyntactic representations (impoverishment) in order to describe syncretism (Bobaljik 2001, Halle 1997, Noyer 1998). 49
and Paradigm accounts. The return to a claim that inflectional structure is fundamentally organized according to paradigms was driven largely by inflectional phenomena which cannot be generated without reference to multiple inflectional forms
(and/or inflectional property sets), or for which generating a form without such references misses significant generalizations.
The relevant data for modern WP models has been various. There is good reason to counter the classic generative view of analogy as a deductive process and return to a more surface-oriented, abductive view (see Joseph to appear 2006), which in turn supports a WP model. However, the key data for modern WP studies has largely followed a different track in which the connections needed between inflectionally related words is not at the level of surface forms, but at the level of morphosyntactic properties and stems. It has included paradigm economy23 (Carstairs 1983), parasitic formations24
(Matthews 1972, Morin 1990) and especially form-meaning mismatches such as syncretism (see Baerman et al. 2005 for an overview). Syncretism, essentially homophony in an inflectional paradigm, is detailed in Section 2.4.3.
2.4.1. Similarities to traditional Word and Paradigm models
In many respects, modern WP models operate under the same fundamental assumptions as more traditional ones. As Blevins (to appear 2006) notes, the core features of all Word and Paradigm theories include:
23 Paradigm economy is a term to reflect that inflected forms of a word are not randomly drawn from the set of possible forms, i.e. that inflectional classes exist. For example, a language with three accusative singular suffixes (A, B, C) and three accusative plural suffixes (D, E, F) might be expected to present nine combinations, but as Carstairs (1983) notes, this is overwhelmingly not the case. All of the lexemes with accusative singular suffix A may have accusative plural suffix D, and vice versa, with no instantiations of the remaining combinations: (A, E), (A, F), (B, D), or (C, D). 24 Parasitic formations are inflected words which have as their stems other inflected forms. 50
• The postulation that the word is the most basic level at which form is connected to
(inflectional) meaning,
• The claim that inflected word forms are determined, at least in part, by
connections to other inflectional paradigmatic cells,
• Competition for syntagmatic slots,
• Panini’s Principle/morphological blocking/Elsewhere Condition/disjunctive
ordering, and
• What Blevins calls “paradigmatic deduction”, the principle that speakers can
interpret inflectional properties despite the absence of an overt marker, or
sometimes because of the absence of an overt marker, because inflectionally
related forms derive their meaning from the paradigmatic oppositions which they
enter into.25
Taking these in turn, the first point means that Word and Paradigm models are non- morphemic26 (see Section 2.4.2 for a qualification of this claim). While there may be correlations between inflectional properties and subparts of words, those subparts do not inherently carry the relevant inflectional properties. Instead, the properties are inferred from the presence (or absence) of a marker in relation to the presence/absence of markers on related forms. Related to this is the tenet in modern WP models that inflectional
25 More straightforwardly, the Bosnian/Croatian/Serbian word muž ‘husband’ is interpreted as nominative singular, despite not having an affix marking it as such, because it is not anything else – not nominative plural (muževi), not accusative or genitive singular (muža), not accusative plural (muževe) or genitive plural (muževa), or any of the other members of the paradigm. It is interpretable as nominative singular in opposition to these forms. 26 Or alternatively, “morphemes” are definable/identifiable as relations between and among forms – inflectional properties are only definable via the contrastive relations among members of the paradigm. 51
morphology is realized via operations on stems, rather than fundamentally concatenative.
The generation of an inflected word form may include concatenation but need not – lenition, deletion, ablauting, and other processes are equally available. Stump (2001b) coins the term inferential-realizational to describe these fundamental assumptions of WP models.
The nature of the connections between inflectionally related forms is the topic of the following section.
Syntagmatic competition can manifest in two ways. First, the realization of one set of inflectional properties may preclude the overt realization of other sets. If we return to Wunderlich’s Yimas example ((1) above, given again below as (14)), we can see how this principle might work.
(14) Nominative agreement affixes in Yimas (Wunderlich 2001: manuscript p. 25)
a. ma-a-tpul 2SG.NOM-1SG.ACC-hit ‘You hit me.’
b. ta-a-tpul NEG-1SG.ACC-hit ‘You didn’t hit me.’
Wunderlich described the example in (b) as having a paradigmatic gap, but in Word and
Paradigm terms, this is syntagmatic competition (not a gap!) because the realization of the negation marker ta- precludes the realization of the second person singular nominative affix ma-. The rules realizing these two suffixes occupy the same slot, thus
52
allowing only one to be realized.27 The second sense of syntagmatic competition is the feature that one realization of an inflectional property set precludes other realizations of the same set in the same slot (although it is possible to have multiple realizations of the same set if they occupy different slots, as in circumfixation).28
Panini’s Principle, which goes by several names, states that more specific examples block more general examples. This is the actual mechanism guiding syntagmatic competition between various realizations of the same inflectional property set.29
Finally, for examples of paradigmatic deduction, see the Russian examples in
Section 1.3, where virtually every paradigm has a case/number combination – either the nominative singular or the genitive plural – which is realized by a form identical to the stem (a “zero marker”) rather than an overt affix. Also see the Yimas example in (14), where the second person singular nominative value is interpreted under negation, despite the lack of an overt form.
2.4.2. Differences from traditional Word and Paradigm models
Modern WP models also, in many respects, differ from traditional WP models and differ from each other. In this section I outline four differences between traditional WP models
27 See Anderson (1992) for an argument that morphological theory needs slots. 28 Slots are often represented as positions relative to the stem, e.g. first prefixal position, perhaps because the most common (or at least most canonical) realization of morphological categories is via affixation. However, this is a shorthand in WP models, and not a necessary interpretation. Realistically, slots are more abstract, as evidenced by cases where inflectional realization does not involve concatenation. See Stewart (2004) for details. 29 See Stump (2001b) for an argument that Panini’s principle can account for all cases of syntagmatic competition, not just competition between realizations of the same properties. If Stump is correct, this relieves the need for ad hoc ordering of rules, as assumed by Anderson (1992), among others. 53
and Paradigm Function Morphology (PFM, Stump 2001b), as representative of a currently popular type of WP model.
First, the earlier definition of the paradigm as implicational relations drawn from inflectional forms has largely been replaced by a definition based on inflectional properties. Spencer’s work represents a typical approach within modern paradigm-based theories: “I follow contemporary morphologists in appealing to a more abstract notion of paradigm, one which is at one level of abstraction removed from the list of forms. In this sense, a paradigm is a definition of the set of morphological contrasts that a given class of lexemes can make” (Spencer 2004:72). Details of this formulation are given in
Section 2.4.3.
Second, the approach is generative rather than interpretive. As Blevins points out, (to appear 2006:6-7) “The intuition behind this type of analysis is that the lexicon should be largely ‘redundancy free’, with predictable patterns expressed independently – and symbolically – by means of general combinatoric devices...”30 Word forms are created by applying processes to stems. Thus, as in morphemic accounts, the word form is in a basic sense the product, not the source of generalizations.
Third, to call PFM and similar theories Word and Paradigm models is a misnomer, in a sense, because the fundamental structures for the theory are not paradigms and words, but rather, paradigms and stems. For example, A-morphous morphology (Anderson 1992), Aronoff’s theory (Aronoff 1994), and Paradigm Function
Morphology (Stump 2001b) all assume that the forms which occupy paradigmatic cells at the abstract level indicated by Spencer are stems and inflectional property sets. While
30 Reminder to myself to get permission to quote, because this is a pre-publication version. 54
modern WP models reject many of the assumptions of other generative approaches, “...a constructivist bias remains implicit in the composition of the lexicon, which, in many modern WP models, consists largely of roots and stems. These sub-word elements are morphemes in all but name” (Blevins to appear 2006:5). Thus, it might be better said that WP models are partially morphemic; only inflection (not derivation) is clearly non- morphemic. Blevins describes modern WP models as theoretical hybrids, and the term is apt. Inflectional meaning is carried by the entire word rather than subparts of the word, which makes models such as PFM like other WP approaches. However, these word forms are generated rather than interpreted, unlike in traditional WP models, which makes it impossible to avoid morphemic assumptions for stems.
Fourth, traditional and modern WP models have opposite assumptions about which is basic – the inflectional class, or the instantiations of it. In traditional WP models, the inflection class is derived from particular instantiations. In other words, it is the exemplars which are basic, and the classes are secondary. In modern WP models, the classes are basic. Stems are marked for membership in a particular inflectional class, and word forms are generated based on that classification.
Of course, as alluded to above, not all current Word and Paradigm theories share these characteristics. There are several theories which are rooted in whole word forms with patterns abstracted away from those forms, and in some respects these theories are thus more similar to traditional WP accounts. They vary quite a bit in the details, but include Janda and Joseph’s theory of morphological constellations (Janda and Joseph
1999), Bybee’s connectionist-based model (Bybee 1985), Lexical Relatedness
Morphology (Bochner 1993), the Theory of Connected Word Constructions (Baronian
55
2005), and a variety of exemplar models (Pierrehumbert 2001, Skousen et al. 2002).31
This list is not complete, but sufficient to demonstrate that modern WP theories in fact fall into two robust categories: those that are truly word-based, and those which are stem- based. Both types are typically conflated under the rubric of Word and Paradigm
Morphology.
2.4.3. The structure of modern Word and Paradigm models
A variety of evidence has played a role in the reintroduction of paradigms to morphological theory and guided the structures which have been posited. Form-meaning mismatches such as syncretism, a cross-linguistically common phenomenon in which two sets of inflectional properties share the same exponent, have been chief among these (see
Baerman et al. 2005 for an overview). In this section I outline broadly why form- meaning mismatches are evidence for modern paradigmatic theories, as a way of demonstrating in somewhat more detail the structure which I am assuming, and the role that paradigmatic gaps can play in testing that structure.
I tread a path between stem-based and word-based models, assuming that both inflected words and stems are stored in the lexicon. Paradigm Function Morphology and other stem-based theories do not explicitly incorporate word paradigms or any equivalent into the structure of the lexicon; this implies that word paradigms are epiphenomenal to paradigmatic connections, and in particular cannot serve as the environment for paradigmatic effects. I specifically incorporate word paradigms here as a way to capture the idea that some paradigmatic effects are driven by surface phenomena (e.g. analogy).
31 Construction morphology (Booij 2002) also makes many of the same assumptions about the primacy of exemplars, but without much of the explicitly/tacitly paradigmatic approach of the other theories. 56
However, I will otherwise assume that the Paradigm Function Morphology theory of inflectional structure applies virtually in its entirety, and that the word-based assumptions are added on, not fundamentally changing of the functioning of the inflectional system, with one basic exception. Incorporating inflected words into the lexicon requires me to reject the assumption of stem-based models that the theory should remove redundancy.
This makes the theory still largely generative, but I assume that once word forms are generated, they are stored and accessed under certain conditions. I do not have much to say about those conditions, but the interested reader is referred to the literature on dual route models within psycholinguistics, which use frequency effects to argue that words are both fully stored and composable from parts; the fastest route wins, as determined by relative activation levels and other factors (Baayen 2006, Baayen and Martín To appear,
Hay 2001, 2003).
Paradigm Function Morphology and closely related stem-based theories
(Ackerman and Stump 2004, Hippisley et al. 2004, Sadler and Spencer 2000, Spencer
2004, Stump 1991, 2001a, 2001b) provide a structure for describing syncretism.
Consider the following noun paradigms from Bosnian/Croatian/Serbian (BCS).
57
o-stem masculine pattern (‘window’) a-stem pattern (‘woman’) SINGULAR PLURAL SINGULAR PLURAL NOM prozor prozori NOM žena žene ACC prozor prozore ACC ženu žene GEN prozora prozōrā GEN žene žēnā DAT-LOC prozoru prozorima DAT-LOC ženi ženama INSTR prozorom prozorima INSTR ženom ženama
o-stem neuter pattern (‘sea’) i-stem pattern (‘thing’) SINGULAR PLURAL SINGULAR PLURAL NOM more mora NOM stvar stvari ACC more mora ACC stvar stvari GEN mora mōrā GEN stvari stvari DAT-LOC moru morima DAT-LOC stvari stvarima INSTR morom morima INSTR stvari, stvarima stvarju
Table 10: Four Bosnian/Croatian/Serbian inflectional paradigms
As these examples show, in BCS the form of the dative/locative plural is the same as the form of the instrumental plural across all inflectional classes, regardless of the form itself.32 Inasmuch as this syncretism is not coincidental, it is arguably a distributional fact which morphological theory needs to capture. The paradigm is necessary in order to do so because the dative/locative and the instrumental are not derived one from the other.
32 The distinction between dative and locative in BCS is largely a historical one. In the modern language the “dative” and “locative” forms are always identical, with the minor exception of prosodic differences in a handful of forms by some but not all speakers (Browne 1993). In some respects, the BCS dative/locative inflectional distinction is roughly parallel to the inflection of English be as opposed to other verbs. Be has distinct forms for the first person singular present (am) and the second person singular (are), but it is the only verb to make this distinction (and some other distinctions as well – see FOOTNOTE 3 in CHAPTER 1). It is traditional to represent the first and second person singular as occupying separate paradigmatic cells in English, even for those verbs which do not maintain separate forms. It is likewise part of the BCS grammatical tradition to distinguish between the dative and the locative. The joint term dative-locative is used here to reflect that tradition, while saving space because prosodic differences are not orthographically encoded in BCS. Whether the dative and the locative should (still) be considered separate cases of the language can only be determined theory-internally, and I do not stake a position on that point. 58
Nor is it possible to posit inflectional properties that are uniquely shared by the dative/locative and the instrumental plural, but not the dative/locative and instrumental singular, allowing a single morpheme to represent both in the former case.33 In short, unlike paradigm uniformity and perhaps analogy, the superficially paradigmatic nature of syncretism cannot be treated as the epiphenomenal result of shared steps in the derivational process or a relationship between inputs and outputs. Instead, the data require reference to multiple inflectionally-related forms at some level, because the generalization is that the morphological forms are the same, but the syntactic properties are not.
Stump argues that inflectional structure is defined by two paradigms rather than one because
Paradigms can be seen as participating in the definition of two different gramma- tical domains. On the one hand, paradigms are objects defined by a grammar’s morphological component: the paradigm of a root R is the inventory of cells that can be projected from R... In the syntactic domain, paradigms enter into the definition of phrasal constituents: the paradigm of a lexeme L is the inventory of syntactic atoms which may instantiate L in phrase structure (Stump 2001a:147-8).
I would argue that inflectional structure is in fact instantiated by three paradigms
– one defining syntactic atoms (the lexeme paradigm), one defining morphological atoms
(the stem paradigm), and one defining the combination of syntactic and morphological atoms into a word which is inserted into syntactic structure (the word paradigm).34 The
33 Noyer (1998) attempts a formulation of this sort within Distributed Morphology, although not for this specific data. His account was essentially an update of Jakobsonian feature specifications. See Carstairs- McCarthy (1998) for why this approach is untenable. 34 See the GLOSSARY for definitions of paradigmatic terms. 59
lexeme paradigm, also known as the syntactic paradigm or content paradigm, contains lexemes paired with morphosyntactic property sets. The stem paradigm, also known as the morphological paradigm or form paradigm, contains stems paired with morphosyntactic property sets. The word paradigm, not surprisingly, contains word forms and morphosyntactic property sets.
These three paradigms are linked to each other. Cells in stem paradigms are linked on one end to corresponding words, and on the other end to cells in the lexeme paradigm. Realization rules (collectively known as the paradigm function) specify the mapping between cells of the stem paradigm and inflected words by dictating the inflectional processes that a stem undergoes. Links between stem and lexeme paradigms specify which cells from the stem paradigm will generate words; every cell in the stem paradigm which is linked to a cell in the lexeme paradigm is also linked to one in the word paradigm, assuming the speaker has ever had reason to generate that word. A stem paradigm cell will not be linked to a word if there is no complementary link to the lexeme paradigm, because the lexeme paradigm is responsible for contributing lexical and inflectional properties. A form generated from a stem paradigm cell which does not have a link with a lexeme paradigm cell could not be inserted into syntactic structure because it would be, quite literally, meaningless. There is thus no reason for such a word to exist.
These linkages may connect paradigmatic cells in a variety of ways. In the normal case, the morphosyntactic properties inhabiting a cell of the stem paradigm are
60
identical to those inhabiting the respective cell in the lexeme paradigm and word paradigm. Consider the following paradigms for BCS word prozor ‘window’.35
Word paradigm of prozor
Realization rules
Stem paradigm of prozor-
Paradigm linkage
Lexeme paradigm of PROZOR
Figure 2: The three paradigms of the BCS noun prozor ‘window’
The solid arrows represent the set of connections for the genitive singular. All connections are one-to-one (biunique). The lexeme paradigm cell < PROZOR, {gen sg}>
35 To provide grounding for the terminology introduced in the GLOSSARY, PROZOR is the lexeme, < PROZOR, {gen sg}> is a lexeme paradigm cell, prozor is the stem,
is connected to a single cell in the stem paradigm,
Word paradigm of prozor
Realization rules Stem paradigm of prozor-
Paradigm linkage Lexeme paradigm of PROZOR
Figure 3: Syncretism in the BCS noun prozor 'window'
However, the linkages need not connect paradigms in a one-to-one fashion. For example, in syncretism there is a mismatch between the lexeme and stem paradigms (2:1), and the stem and word paradigms (1:2). In nominal syncretism two lexeme paradigm cells carry
62
distinct values for some morphosyntactic property set, but share a stem paradigm cell; the relevant morphosyntactic property sets are ambiguous due to merger of cells. Taking prozor again as an example, this linkage may be graphically represented as above. One cell of the stem paradigm,
In principle, there is virtually no limit to the kinds of linkages that can be made between lexeme and stem paradigms. For example, a single lexeme paradigm may have paradigmatic linkages to multiple stem paradigms. When these stems belong to different inflection classes, this is classically known as heteroclisis. The Croatian word dijete
‘child’ provides an example.
dijete ‘child’ SINGULAR PLURAL NOM dijete djeca ACC dijete djecu GEN djeteta djece DAT-LOC djetetu djeci INSTR djetetom djecom
Table 11: Paradigm of BCS word dijete 'child'
In the singular, the inflected forms of dijete are o-stem neuter singular in morphological form and neuter singular in syntactic properties. However, in the plural the forms 63
correspond to the a-stem singular pattern, but still carry neuter plural syntactic properties.
Compare the forms of dijete with the o-stem and a-stem inflectional patterns (repeated from Table 10 above) and the agreement patterns in the singular and plural.
o-stem neuter pattern (‘sea’) a-stem pattern (‘woman’) SINGULAR PLURAL SINGULAR PLURAL NOM more mora NOM žena žene ACC more mora ACC ženu žene GEN mora mōrā GEN žene žēnā DAT-LOC moru morima DAT-LOC ženi ženama INSTR morom morima INSTR ženom ženama
Table 12: Two BCS nominal paradigms, repeated
A-stem inflection implies feminine agreement (as in ((15)e) and ((15)f) below); the vast majority of words falling into this inflection class trigger feminine agreement.36 In
((15)b), however, djeca has an a-stem nominative singular form, but the morphosemantics are neuter plural.37
36 See Sims (2005) for a discussion of the implicational relationships connecting inflectional forms and agreement in BCS. Summarized, there are two classes of exceptions to the pattern that a-stem morphology implies feminine agreement. The first comprises dijete and other collectives with a-stem inflectional morphology in the plural. The second are a-stem masculine nouns, such as kolega ‘colleague’ (feminine counterpart: kolegica). Together, these exceptions still make up a small percentage of the total a-stem nouns. 37 Realistically, the situation is more complex. Some agreement targets, such as verbs, routinely take neuter plural agreement, but other targets often appear with feminine singular agreement. The pattern seems to follow the Agreement Hierarchy (Corbett 1991). For a true account of this pattern, we would need to assume that morphosyntactic properties are inherited both from the lexeme paradigm, and from the stem paradigm. This is essentially the approach followed by Wechsler and Zlatić (2000) and Kathol (1999); both studies are framed within Head-driven Phrase Structure Grammar (HPSG), a lexicalist syntactic theory which is largely compatible with the Word and Paradigm approach to morphology. 64
(15) Examples of verbal agreement for the BCS word dijete ‘child’
a. Dijete je došlo. child-O-STEMNEUT.NOM.SG AUX.3.SG come-NEUT.SG.PAST ‘The child arrived.’
b. Djeca su došla. child-A-STEM.NOM.SG AUX.3.PL arrive-NEUT.PL.PAST ‘The children arrived.’
c. More je bilo široko. sea- O-STEMNEUT.NOM.SG AUX.3.SG be- NEUT.SG.PAST wide-NEUT.SG ‘The sea was wide.’
d. Mora su bila široka. sea- O-STEMNEUT.NOM.PL AUX.3.PL be- NEUT.PL.PAST wide-NEUT.PL ‘The seas were wide.’
e. Žena je došla. woman-A-STEM.NOM.SG AUX.3.SG arrive-FEM.SG.PAST ‘The woman arrived.’
f. Žene su došle. woman-A-STEM.NOM.PL AUX.3.PL arrive-FEM.PL.PAST ‘The women arrived.’
In Paradigm Function Morphology terms, the plural cells of the lexeme paradigm are linked to singular a-stem paradigm cells.
65
Word paradigm of dijete
Stem paradigm of dijet- Stem paradigm of djec-
Lexeme paradigm of DIJETE < DIJETE, {nom sg}> < DIJETE, {nom pl}> < DIJETE, {acc sg}> < DIJETE, {acc pl}> < DIJETE, {gen sg}> < DIJETE, {gen pl}> < DIJETE, {dat/loc sg}> < DIJETE, {dat/loc pl}> < DIJETE, {inst sg}> < DIJETE, {inst pl}>
Figure 4: Three paradigms of the heteroclite BCS noun dijete 'child'38
38 Inflection class membership is marked here with subscripts for reasons of space. 1 = o-stem neuter inflectional class. 2 = a-stem inflectional class. 66
Furthermore, we may posit that the lexicon is not composed of stem and lexeme paradigms as isolated entities, but rather is made up of paradigms which are nodes in a static information-sharing network, as formalized in Network Morphology (Brown et al.
1996, Corbett et al. 2001, Hippisley 1999, Hudson 2000). In these networks, inflectional information which is shared across multiple groups (whether they be individual paradigms or larger groupings) is encoded as templates. These are typically represented as a tree structure; mother nodes share information with daughter nodes by default, unless the daughter node has lexically specified information that is more specific than the information on the mother node.
x=A
x=A x=A y=C y=D
x=A x=A x=A x=B x=A y=C y=C y=L y=D y=D z=E z=F z=G q=H q=I
Figure 5: Schematization of information sharing in a default inheritance network
67
The metaphor of a tree structure is somewhat misleading, however, because the relationship between mother and daughter notes is not strictly hierarchical – daughter nodes may inherit from more than one mother node.39
x=A
x=A x=B y=C y=D
x=A x=A x=B x=B x=A y=C y=C y=C y=D y=D z=E z=F z=G q=H q=I
Figure 6: Another schematization of information sharing in a default inheritance network
Furthermore, note that not all templates may be captured through default inheritance because it is not always possible to identify a node which uniquely dominates the affected lower nodes. However, we may hypothesize that paradigm linkage operates within the network in much the same way that linkages connect stem and lexeme paradigms. These linkages run parallel to default inheritance; nodes may specify other nodes as the source from which a particular form is inherited, even if this is contradictory to the path of inheritance.
39 This means that contradictory information might, in theory, be inheritable by a single daughter node from two mother nodes. Hudson (2000) argues that inheritance of contradictory inflectional information leads to paradigmatic gaps. More typically, it is simply assumed that one mother node has precedence in feature sharing (Hippisley 1999). 68
The role of default inheritance networks can be made more concrete by considering how this system allows for BCS syncretism. As noted earlier, in
Bosnian/Croatian/Serbian the dative/locative plural is the same as the instrumental plural for all inflection types (see Table 10 above). The structure of the stem paradigm and its connections to the lexeme paradigm, as given in Figure 3, are sufficient to account for syncretism in any single inflectional class, but as given cannot account for the pattern across all inflectional classes. A default inheritance hierarchy allows us to capture this larger generalization. The key information to capture is that the identity of form holds regardless of the individual form itself; it must therefore be part of the stem paradigm hierarchy, but also must be at a level higher than the level of the inflection class, so that all classes are affected. The syncretism may be represented graphically as follows.
69
Figure 7: Syncretism in BCS nominal paradigms as a default-inheritance information-sharing network40
The information in bold is introduced at that hierarchical node; the information in plain face is inherited from higher nodes. The lowest nodes are stem paradigms. The higher nodes may be thought of as paradigms at greater levels of abstraction
(metageneralizations in the terminology of Stump (2001b) or templates in the terminology of Aski (1995) and Hippisley et al. (2004)).
Three aspects of these accounts of syncretism and heteroclisis, as examples of the wider phenomenon of form-meaning mismatches, are significant for inflectional structure generally.
40 Note that this tree is highly simplified for the purpose of demonstration. Fully developed, the final nodes would be entire stem paradigms, not individual cells, and far more information would need to be represented on higher nodes to capture the entire pattern of inflection in BCS. 70
First, form and meaning are separate and distinct concepts (Separation
Hypothesis, Beard 1995). They are not bundled together into lexical units, i.e. there are no inflectional morphemes, and while form and inflectional meaning may stand in a one- to-one relationship, they need not do so.
Second, stem paradigms and lexeme paradigms differ in type. Both paradigms are defined by the combinatory possibilities of inflectional properties, but may have a different number of cells. If there are five cases and two number values, there will be exactly ten cells in the lexeme paradigm because this represents the number of syntactic atoms. By contrast, stem paradigms may have the same number and distribution of cells, fewer cells (as in syncretism, where two cells are “combined”), or more cells (as in heteroclisis, where twice as many cells exist as are linked to lexeme paradigm cells).
Since stem paradigms and the realizational rules that apply to them are by definition the locus of inflectional form, this mutability of the stem paradigm and its relationship to the lexeme paradigm amounts to different kinds of connections among inflectionally related elements, i.e., paradigmatic structure.
Third, the term “paradigm” applies at many levels. It may be instantiated by concrete words, or by stem paradigms, or by lexeme paradigms, or by templates at various levels. Inasmuch as the levels are all representations of inflectional structure, we might expect paradigmatic structure to apply at any and all points, as we see with BCS syncretism in the dative/locative and instrumental plural, which applies at a level abstracted away from particular inflectional classes.
71
Form-meaning mismatches cannot be adequately described without these aspects of the theory, and so these phenomena have provided ample evidence for paradigmatic structure as a necessary theoretical construct.
We are now in a position to understand the potential value of studying marginal phenomena within the specific framework of a modern Word and Paradigm account like
Paradigm Function Morphology. The most significant point is that the inflectional patterns exemplified in Figures 2 through 4 all operate according to the same mechanisms. The connections which result in the genitive singular word prozora
‘window’ are not accorded any special status in the language, even though this is the canonical case – a single form to express meaning and a single meaning associated with the form. If this sort of one-to-one form-meaning relationship were the only sort of data that we were to take as the topic of study, the structure in Figure 2 would clearly seem too complex; morphemic accounts which bundle form and meaning into lexical units would be both more logical and simpler. We might build a theory which follows the simpler route, and then tries to handle the more marginal phenomena such as syncretism and heteroclisis (and deponency, periphrasis, suppletion, and a variety of other form- meaning mismatches) through some special mechanism. Word and Paradigm theories do exactly the opposite. The structure in Figure 2 is more complicated than is needed to account for biunique form-meaning relationships, but it is exactly what is needed to account for non-biunique relationships. By making the fundamental assumption that all inflection operates according to a single set of principles, Word and Paradigm models build a theory around the marginal phenomena which is still sufficient to account for the more central/frequent/productive aspects of inflection. Differently stated, there is an
72
implicit argument that marginal phenomena uniquely reflect something about core inflectional structure.
2.5. Remaining questions: Paradigmatic gaps as evidence for the paradigm
This focus on marginal phenomena raises the question of what paradigmatic gaps could reveal about inflectional structure. On the one hand, paradigmatic gaps are similar to syncretism, heteroclisis, suppletion, deponency, etc. in that they represent a problem in the form-meaning mapping. On the other hand, paradigmatic gaps, unlike these phenomena, represent language defectiveness. They should thus allow us to test not only the nature of paradigmatic structure, but also its scope.
The scope of the paradigm has not been thoroughly investigated, although there have been attempts to incorporate the arguably syntactic phenomenon of periphrasis into morphological structure (Ackerman and Stump 2004, Börjars et al. 1997, Sadler and
Spencer 2000).41 The fundamental point of these arguments is that periphrasis represents another type of override of a one-to-one form-meaning mapping which paradigm-based theories of morphology are capable of describing. The theoretical argument is then turned around, using the ability of paradigmatic models to describe periphrasis as evidence for this expanded structure.
For an example of how this argumentation works (and does not work), consider the Latin perfective passive.
41 Periphrasis is the use of a multi-word phrase to express a set of morphosyntactic properties, despite that set normally having one or more synthetic realizations in the language. 73
(16) Latin verbs (e.g. laudo ‘I praise’)
a. no periphrasis: IMPERFECTIVE
ACTIVE PASSIVE PRESENT laudat laudatur PAST laudabat laudabatur FUTURE laudabit laudabitur
b. periphrasis: PERFECTIVE
ACTIVE PASSIVE PRESENT laudavit laudatus/a/um est PAST laudaverat laudatus/a/um erat FUTURE laudaverit laudatus/a/um erit
In Latin, the passive is expressed with a single word formation when it is imperfective, as is the perfective when it is active. The combination of perfective and passive, however, entails a periphrastic construction. Within a Paradigm Function Morphology account we may formulate a realization rule which matches a multi-word realization in a word paradigm cell to a cell in a stem paradigm. Ackerman and Stump (2004) and Sadler and
Spencer (2000) follow this approach to periphrasis.
In some respects, this postulation is justified. We know that periphrasis is potentially different from a normal syntactic phrase precisely because the component morphosyntactic features are normally realized synthetically in the language, and because the periphrastic phrases are not fully decomposable semantically (Sadler and Spencer
2000). There are interesting parallels here with idioms and compounds. If we think of wordhood as being a scale from least decomposable to most decomposable, and employ the lexeme/stem distinction discussed previously, we can think of periphrasis, idioms and
74
compounds as all being examples of mismatches between the hierarchies of decomposability – the forms are more decomposable than the semantics. This suggests that Latin periphrasis is structurally intermediate between a typical synthetic phrase and a typical syntactic phrase, and not canonically one or the other. Recent WP accounts have chosen to draw the line so as to include the periphrastic phrases in morphology.
There is also a theory-internal reason for incorporating periphrasis into morphological structure: the paradigm is defined as the combinatory possibilities of morphosyntactic properties, and this principle is assumed to apply to both the lexeme and stem paradigms. However, if periphrasis is considered part of the syntactic component, then either the stem paradigm, or the lexeme paradigm, or perhaps both must be considered to be incomplete in the sense that there is a possible combination of morphosyntactic properties which is not realized at some level. WP theories go out of their way to avoid this possibility; note for example that the stem paradigms in Figure 4 overgenerate so as to complete the paradigm. Expanding the domain of morphology to include periphrasis allows for a unified description of inflectional categories in the way implied by the definition of the paradigm.
Does periphrasis thus stand as evidence for an expanded paradigmatic structure in the same manner that syncretism motivates the core structure? Simply put, no. While both the dissociation of form and meaning and a mismatch between the stem and lexeme paradigms are necessary for a description of syncretism, describing periphrasis requires only the first half of this equation. Like syncretism, periphrastic constructions represent form-meaning mismatches, but the mismatch does not reside in the connections between lexeme and stem paradigms. Instead, the relationship is between the stem paradigm and
75
the word paradigm. In periphrasis, realization rules link a single cell in a stem paradigm to a periphrastic rather than synthetic form in the word paradigm. Realization rules are a feature of all theories assuming a separation of form and meaning (the Separation
Hypothesis), even those which are not paradigmatic (e.g. Distributed Morphology) – the relationship between stem and word paradigms is essentially a relationship between inputs and outputs.
In the end, previous research has shown that periphrasis is intermediary between syntax and morphology, and it need not necessarily be treated as exceptional at a theoretical level. But it has failed to demonstrate that an adequate description of these phenomena necessitates connections between inflectionally related forms. In short, periphrasis has provided no significant evidence to date in favor of paradigmatic theories, as opposed to any theory assuming the Separation Hypothesis. Moreover, without evidence that periphrasis is a paradigmatic phenomenon, the entire logic of providing morphological accounts of these phenomena at the expense of the boundary between morphology and syntax is called into question. Thus, the scope of inflectional paradigms has yet to be significantly probed.
From this discussion it might seem that evidence for the structure and scope of paradigms is unlikely to come from periphrasis or paradigmatic gaps (these often being two sides of the same coin). There is a difference between a theory being able to describe a particular phenomenon, and that description standing as evidence for the theory. The difference lies in constraint. If WP theories such as Paradigm Function Morphology
(Stump 2001b) are capable of describing periphrasis and gaps only because they may
76
describe virtually any relationship between form and meaning, this weakens the relevance of defective inflection as evidence for paradigmatic structure.
I claim that defective inflection does not stand as evidence for the paradigm as a theoretical construct simply because paradigmatic theories are capable of describing them. However, paradigmatic gaps would stand as evidence for the paradigm if the paradigm is necessary for an adequate description of the phenomenon, for instance if the rise or distribution of paradigmatic gaps were governed by paradigmatic structure.
Furthermore, one way to test the extent of word structure is to find its limits – the point at which it breaks down. If we find evidence that paradigmatic gaps (and periphrastic constructions) rely on paradigmatic structure, this is evidence that the scope of paradigmatic structure (and thus inflection generally) should really be expanded to incorporate these phenomena. This, in turn, raises the question of whether paradigmatic structure, as currently posited, is not only necessary but also sufficient to account for gaps.
77
CHAPTER 3
THE MODERN GREEK GENITIVE PLURAL:
HOW GAPS ARE THE EPIPHENOMENAL RESULT OF PARADIGMATIC
STRUCTURE
3.1. Introduction
In this dissertation I focus on two overarching questions: In order to describe gaps, do we ever need to reference paradigmatic structure? In order to describe gaps, do we need to reference the existence of a gap?
In this chapter, I examine the genitive plural in Modern Greek from three different angles, and demonstrate that the ways in which paradigmatic gaps in that inflectional cell are distributed, and the ways in which speakers react to them, cannot be understood without reference to rules of stress placement among Modern Greek nouns, and that those rules of stress placement must be considered to be fundamentally paradigmatic. Modern
Greek thus provides evidence that paradigmatic gaps are dependent upon paradigmatic structure.
Whether morphological theory needs to make explicit provision for gaps has been a central question within morphological theory in the most recent wave of studies of paradigmatic gaps (Albright 2003, Baerman and Corbett 2006, Bresnan 2001, Fanselow and Féry 2002, Frampton 2001, Hudson 2000, Johansson 1999, Joseph and Sims 2006, 78
McCarthy and Wolf 2005, Orgun and Sprouse 1999, Pertsova 2005, Rice 2005b, Vincent
2005). Modern Greek provides some evidence in favor of the predominant view that morphological theory does not need to make specific reference to gaps because they are the epiphenomenal result of conflicting grammatical principles (e.g. Albright 2003,
Bresnan 2001, McCarthy and Wolf 2005, Rice 2005b). As in previous gaps-as- epiphenomena accounts, I present evidence suggesting that gaps are synchronically motivated/caused by conflicting grammatical structures, or more technically, the ideological issues through which speakers interpret grammatical structure.
At a theoretical level, a grammar conflict approach is appealing for at least two reasons. First, it allows the grammar to be maximally general, without the ad hoc, idiosyncratic generalizations about paradigmatic gaps that have long been a thorn for morphological theory. Second, for linguists who believe that language is fundamentally an optimal/optimizing system, treating gaps as epiphenomena appeals because it suggests that even those aspects of language which are superficially the most idiosyncratic exist for the optimization of the whole (McCarthy and Wolf 2005, Rice 2005b).
In CHAPTER 4 I reject the hypothesis that the Greek genitive plural gaps can be treated solely as byproducts of morphological structure, and with it the thesis of language as an optimal system. However, it is first worth exploring the truth which lies in this approach, beginning with an explanation of grammar competition models, and the predictions that they make with regard to paradigmatic gaps.
79
3.1.1. Grammar competition models: predictions and assumptions
Many inflectional models incorporate competition into the grammar in at least one of two interrelated ways. One focuses on competition among morphosyntactic properties, the other on competition among forms. Both types of competition have been argued to cause gaps, but the latter is more relevant here.
The classic case of form-based competition is between periphrastic and synthetic forms (Ackema and Neeleman 2001, Ackerman and Stump 2004, Andrews 1990, Börjars et al. 1997, Bresnan 2001, Embick 2000, Kiparsky 2005, Poser 1992, Sadler and Spencer
2000, Spencer 1999, Vincent 1987, Zwicky 1989), but competing forms are linked to paradigmatic gaps, for example in Spanish (Albright 2003).
Albright (2003) argues that Spanish verbal gaps are caused by speaker uncertainty over the application of morphophonological alternations. In essence, morphophonological variation sets up a situation in which a form with an alternation competes with a form without the alternation. When the competition cannot be resolved in favor of one or the other form, a paradigmatic gap results.
He distinguishes two types of present tense indicative gaps, what he calls anti- stress gaps and anti-egotistic gaps. Anti-stress gaps are demonstrated below on the left; anti-egotistic verbs are on the right.
80
abolir ‘to abolish’ singular plural asir ‘to grasp’ singular plural 1st person -- abolimos 1st person -- asimos 2nd person -- abolís 2nd person ases asís 3rd person -- -- 3rd person ase asen
Table 13: Present tense indicative gaps in Spanish
(17) Sample of verbs with anti-stress gaps
abolir ‘to abolish’ agredir ‘to assault’ aguerir ‘to harden for battle’ arrecierse ‘to stiffen’ aterirse ‘to be numb’ colorir ‘to color’ denegrir ‘to blacken’ descolorir ‘to de-color’ empedernir ‘to harden’ garantir ‘to guarantee’ tra(n)sgredir ‘to transgress’ transhumar ‘to move pastures’
(18) Sample of verbs with anti-egotistic gaps
asir ‘to grasp’ balbucir ‘to stammer’ pacer ‘to graze’
Albright argues that these two patterns of gaps are related to morphophonological alternations.
The gap patterns... are extremely suggestive, since they mirror exactly the distribution of unpredictable morphophonemic alternations. In particular, anti- stress verbs are missing forms where diphthongization and raising occur, while anti-egotistic verbs are missing the form where velar insertion occurs. Furthermore, verbs with gaps generally meet the structural description for alternations: anti-stress verbs mostly have mid vowels, while anti-egotistic verbs have stem-final [s]. In addition, virtually all of the defective verbs belong to class 3 ([i] theme), which is the most susceptible to alternations (Albright 2003:4).
81
Diphthongization and raising apply to mid vowels under stress. Velar insertion applies in the first person singular when there is a stem-final /s/. These alternations are exemplified below; diphthongization and raising in the left tables, velar insertion in the right one.
sentir ‘to feel’ singular plural crecer ‘to grow’ singular plural 1st person s[jé]nto s[e]ntímos 1st person cré[sk]o cre[s]émos 2nd person s[jé]ntes s[e]ntís 2nd person cré[s]es cre[s]éis 3rd person s[jé]nte s[jé]nten 3rd person cré[s]e cré[s]en
pedir ‘to ask’ singular plural 1st person p[í]do p[e]dímos 2nd person p[í]des p[e]dís 3rd person p[í]de p[í]den
Table 14: Morphophonological alternations in the present indicative
Since the alternations do not apply uniformly, especially in the classes where paradigmatic gaps are found, Albright argues that there is a synchronic, causative link between the “unpredictable morphophonemic alternations” and the existence of Spanish verbal gaps, although the gaps only appear in a subset of the eligible lexemes.42 He furthermore shows that when speakers do not agree over whether a word form does or does not have an alternations, they are less confident that the form they produced is
42 A bit of data not mentioned by Albright may also be relevant. Baronian (2005:154) notes that those verbs with gaps in the singular and third person plural also fail to raise the root vowel in the present participle: infinitive past participle 1st p. singular gloss d[o]rmir d[u]rmiendo d[we]rmo ‘to sleep’ s[e]ntir s[i]ntiendo s[je]nto ‘to feel’ ab[o]lir ab[o]liendo -- ‘to abolish’ ag[e]rrir ag[e]rriendo -- ‘to arrange’
82
correct. It thus seems that morphophonological variation causes speakers to be insecure.
Combined with the claim that paradigmatic gaps in Spanish fall into the range in which speakers neither agree nor are confident when forced to produce the gapped form,
Albright concludes that competing word forms cause speakers to be insecure, which in turn causes paradigmatic gaps (presumably through avoidance).
(19) Albright’s grammar competition hypothesis:
grammar competition → speaker insecurity → paradigmatic gap
Albright constructs a grammar in which the rules governing the morphological patterns in
Table 14 compete. A form with an alternation and a form without an alternation go head to head. Competition is present to one degree or another for all word forms meeting the conditions for the alternation, but where the variability in the (expected) word form is sufficient that the two patterns are evenly weighted for a particular lexeme, the grammar cannot choose between the two possibilities. The result is a (probabilistic) failure in the mechanism of inflection, i.e. a gap. In other words, paradigmatic gaps are the extreme end result of speakers’ gradient reactions to grammar competition.43 It is this sense of competition, based in variation between two morphophonological patterns, which is most relevant to the Greek data discussed below.
Crucially, in this view there is no need for morphological theory to incorporate an explicit statement/rule/constraint for the gap; the Spanish verbal gaps are the epiphenomenal result of the competition itself. If the theory can describe the
43 See Baronian (2005) for similar competition among word forms, in which the competition takes place in the lexicon. 83
competition, the existence of the gap comes “for free”. While formalisms vary significantly, this idea that grammatical competition causes gaps is widespread in current morphological theory.
In the following sections I present similar data from the Modern Greek genitive plural. Although descriptive grammars have often noted paradigmatic gaps in this cell, to my knowledge nobody has previously made a systematic study of their distribution or possible causes.
The Modern Greek nominal system has four cases: nominative, accusative, genitive, and vocative, but among these the genitive plural has a special status for at least three reasons. First, there are gaps in the genitive plural, but not in other cells.44 Second, although the synthetic genitive form is possible in a wide variety of uses, a periphrastic prepositional phrase containing an accusative noun phrase is often preferred. Third, the stress of the genitive plural is governed by a separate generalization than is stress in
(most) other inflected forms.
I show that these three facts are connected. My evidence comes from three sources: the distribution of the genitive plural gaps among Greek nominal inflection classes, a forced choice sentence-completion survey of preference for periphrastic or synthetic forms among non-defective nouns, and a forced production survey in which subjects wrote genitive plural forms of words with and without gaps, and rated their confidence in their production and their familiarity with the lexeme.
To jump ahead to the conclusions, the Greek results parallel the results of
Albright’s study in many respects. The genitive plural gaps in Modern Greek are closely
44 Excluding pluralia tantum and singulare tantum, which are (primarily) semantically motivated. 84
connected to the uneven application of a morphophonological alternation, in this case shifting stress. There are two major differences, however. First, the proper generalization about the relationship between genitive plural stress and paradigmatic gaps can only be made with reference to multiple inflectional forms, and that reference defines paradigmatic structure. It thus seems that the gaps in Modern Greek are the byproduct of competing forms, as in Spanish, but with the twist that the competition takes place at the paradigmatic level. Second, based on the notion of paradigmatic competition, I argue that the key factor is predictability of form, not variability of form. To set the stage for the relevant data, I first give an overview of the genitive plural stress system.
3.2. Overview of genitive plural stress
Stress placement in the Modern Greek genitive plural is best described as a series of idiosyncratic generalizations left as residue of other changes from Ancient Greek to the present. Ancient Greek had a pitch accent system. The primary (high) pitch accent was constrained so as to fall within the last three syllables of a prosodic word, or better, the last three moras.45 The genitive plural marker [-ōn] was bimoraic, whereas many other inflectional markers were monomoraic. This difference meant that the genitive plural showed an accent shift relative to other inflected forms in those lexemes for which accent was generally antepenultimate (giving, for example, nsg µ!µ-µ, gpl µµ!-µµ). However, with the loss of distinctive length in post-Classical Greek, stress placement (the language by then having shifted from a pitch accent system to a stress accent system) was reanalyzed as being syllable-based, so that the accent fell within the last three syllables. As a result,
45 Describing accent placement in terms of moras, as opposed to syllables, allows for a better account of what happens with finite verb forms, but it requires the assumption that a non-final long vowel counts as monomoraic. 85
there was no longer a phonological motivation for the difference in the stress placement between the genitive plural and other inflectional markers, since all the endings were equally monosyllabic. The stress shift in the genitive plural thus became an idiosyncratic fact about that inflectional form. Borrowings, stress levellings, the collapse of some inflectional classes, and other historical developments have further diminished the possibility of making motivated generalizations about stress placement in the genitive plural form. And the previously diglossic situation in Modern Greek, in which stress patterns more faithful to the accentual placement of Ancient Greek were given prominence, possibly impeded any widespread leveling of the stress pattern.
Triantafillidis (1998) identifies 69 inflection classes for Modern Greek nouns, not including singleton classes or indeclinable nouns. If we group these classes according to shared segmental material (collapsing differences of gender and stress placement), the number of classes is reduced to 23. These inflectional classes in turn represent three descriptive types, according to the stress pattern of the synthetic genitive plural form relative to the other inflectional forms of the lexeme.46
Noun classes with type 1 stress are those which, for a given word, have stress on the same syllable throughout the inflectional paradigm, including in the genitive plural.
Some noun classes of this type have a different number of syllables in the singular and plural. In these situations type 1 stress is defined as columnar. An example of each of the non-singleton type 1 inflection classes is given below.
46 Neither traditional Greek grammatical descriptions nor theoretical studies of Greek stress use this grouping into three types. However, these three types are meant purely as a descriptive tool, to demonstrate how stress patterns crosscut inflectional classes. In the theoretical discussion in Section 3.6.1 I argue for a somewhat different organization into inflectional classes, based on a multidimensional inheritance hierarchy which separates the singular and plural into separate segmental hierarchies, and stress into a third one. I also address previous theoretical treatments of stress. However, here it is easiest to begin with a more traditional classification, for descriptive purposes. 86
‘greengrocer’ SINGULAR PLURAL ‘coffee’ SINGULAR PLURAL NOM µανάβης µανάβηδες NOM καφές καφέδες manávis manáviđes kafés kaféđes ACC µανάβη µανάβηδες ACC καφέ καφέδες manávi manáviđes kafé kaféđes GEN µανάβη µανάβηδων GEN καφέ καφέδων manávi manáviđon kafé kaféđon
‘countess’ SINGULAR PLURAL ‘grandfather’ SINGULAR PLURAL NOM κόντες κόντηδες NOM παππούς παππούδες kó(n)des kó(n)diđes pappús pappúđes ACC κόντε κόντηδες ACC παππού παππούδες kó(n)de kó(n)diđes pappú pappúđes GEN κόντε κόντηδων GEN παππού παππούδων kó(n)de kó(n)diđon pappú pappúđon
‘mother’ SINGULAR PLURAL ‘fox’ SINGULAR PLURAL NOM µαµά µαµάδες NOM αλεπού αλεπούδες mamá mamáđes alepú alepúđes ACC µαµά µαµάδες ACC αλεπού αλεπούδες mamá mamáđes alepú alepúđes GEN µαµάς µαµάδων GEN αλεπούς αλεπούδων mamás mamáđon alepús alepúđon
‘relative’ SINGULAR PLURAL ‘corporal’ SINGULAR PLURAL NOM συγγενής συγγενείς NOM δεκανέας δεκανείς si(n)genís si(n)genís đekanéas đekanís ACC συγγενή συγγενείς ACC δεκανέα δεκανείς si(n)gení si(n)genís đekanéa đekanís GEN συγγενή συγγενών GEN δεκανέα δεκανέων si(n)gení si(n)genón đekanéa đekanéon
‘circum- navigation’ SINGULAR PLURAL NOM περίπλους περίπλοι períplus perípli ACC περίπλου περίπλους períplu períplus GEN περίπλου περίπλων períplu períplon
Table 15: Greek noun inflection classes with genitive plural stress consistently columnar (type 1)
87
In type 2 nouns, genitive plural stress is fixed relative to the end of the word. For some inflectional classes, genitive plural stress is fixed on the final syllable. For others, it is fixed on the penultimate syllable. The result in both cases is that stress is not necessarily on the same syllable in the genitive plural as in the rest of the inflectional paradigm.
Sometimes the genitive singular is involved in the stress shift as well, although it need not be. For example, with regard to the paradigm represented by αγόρι / aγόri ‘boy’,
Holton et al. (1997:65) notes that “these nouns are always paroxytone... In their genitive singular and plural they move the stress to the final syllable, with the -ι- losing its syllabic value...”
88
‘victory’ SINGULAR PLURAL ‘boy’ SINGULAR PLURAL NOM νίκη νίκες NOM αγόρι αγόρια níki níkes aγóri aγórja ACC νίκη νίκες ACC αγόρι αγόρια níki níkes aγóri aγórja GEN νίκης νικών GEN αγοριού αγοριών níkis nikón aγorjú aγorjón
‘part’ SINGULAR PLURAL ‘sailor’ SINGULAR PLURAL NOM µέρος µέρη NOM ναύτης ναύτες méros méri náftis náftes ACC µέρος µέρη ACC ναύτη ναύτες méros méri náfti náftes GEN µέρους µερών GEN ναύτη ναυτών mérus merón náfti naftón
‘apprentice’ SINGULAR PLURAL ‘(mental) SINGULAR PLURAL power’ NOM κάλφας καλφάδες NOM δύναµη δυνάµεις kálfas kalfáđes đínami đinámis ACC κάλφα καλφάδες ACC δύναµη δυνάµεις kálfa kalfáđes đínami đinámis GEN κάλφα καλφάδων GEN δύναµης δυνάµεων kálfa kalfáđon đínamis đinámeon
‘continent’ SINGULAR PLURAL ‘name’ SINGULAR PLURAL NOM ήπειρος ήπειροι NOM όνοµα ονόµατα ípiros ípiri ónoma onómata ACC ήπειρο ηπείρους ACC όνοµα ονόµατα ípiro ipírus ónoma onómata GEN ηπείρου ηπείρων GEN ονόµατος ονοµάτων ipíru ipíron onómatos onomáton
‘meat’ SINGULAR PLURAL ‘prospects’ SINGULAR PLURAL NOM κρέας κρέατα NOM µέλλον µέλλοντα kréas kréata méllon méllonta ACC κρέας κρέατα ACC µέλλον µέλλοντα kréas kréata méllon méllonta GEN κρέατος κρεάτων GEN µέλλοντος µελλόντων kréatos kreáton méllontos mellónton
Table 16: Greek noun inflection classes with genitive plural stress consistently fixed relative to the end of the word (type 2)
89
Thus, speakers who know the stress pattern that is general to the noun class should be able to predict the stress of the genitive plural (and genitive singular) when presented with a new word of this class.
The third type includes words for which there is variation within the inflection class itself – stress placement is neither columnar nor uniformly fixed relative to the end of the word. In these cases, stress cannot be securely predicted from segmental inflectional markers without historical knowledge. Holton et al. (1997:48-49) provide the following description of one class of this type – nouns in which the nominative singular is marked by stem + /as/.
There are two types, according to the stress of the genitive plural: (i) those that have a genitive plural with stress on the penultimate (mostly nouns deriving from the Ancient Greek 3rd declension); (ii) those that have a genitive plural with stress on the final syllable (nouns deriving from the Ancient Greek 1st declension, but also including some from the 3rd declension and some newer formations). With these exceptions, the stress remains on the same syllable as in the nominative singular.
Two examples of each inflection class of this type are given below, starting with the one described by Holton et al.
90
WITHOUT STRESS SHIFT WITH STRESS SHIFT ‘father’ SINGULAR PLURAL ‘tourist’ SINGULAR PLURAL NOM πατέρας πατέρες NOM τουρίστας τουρίστες patéras patéres turístas turístes ACC πατέρα πατέρες ACC τουρίστα τουρίστες patéra patéres turísta turístes GEN πατέρα πατέρων GEN τουρίστα τουριστών patéra patéron turísta turistón
WITHOUT STRESS SHIFT WITH STRESS SHIFT ‘echo’ SINGULAR PLURAL ‘angel’ SINGULAR PLURAL NOM αντίλαλος αντίλαλοι NOM άγγελος άγγελοι antílalos antílali á(n)gelos á(n)geli ACC αντίλαλο αντίλαλους ACC άγγελο αγγέλους antílalo antílalus á(n)gelo a(n)gélus GEN αντίλαλου αντίλαλων GEN αγγέλου αγγέλων antílalu antílalon a(n)gélu a(n)gélon
WITHOUT STRESS SHIFT WITH STRESS SHIFT ‘mother’ SINGULAR PLURAL ‘hour’ SINGULAR PLURAL NOM µητέρα µητέρες NOM ώρα ώρες mitéra mitéres óra óres ACC µητέρα µητέρες ACC ώρα ώρες mitéra mitéres óra óres GEN µητέρας µητέρων GEN ώρας ωρών mitéras mitéron óras orón
WITHOUT STRESS SHIFT WITH STRESS SHIFT ‘iron’ SINGULAR PLURAL ‘face’ SINGULAR PLURAL NOM σίδερο σίδερα NOM πρόσωπο πρόσωπα síđero síđera prósopo prósopa ACC σίδερο σίδερα ACC πρόσωπο πρόσωπα síđero síđera prósopo prósopa GEN σίδερου σίδερων GEN προσώπου προσώπων síđeru síđeron prosópu prosópon
Table 17: Greek noun inflection classes with variability in genitive plural stress placement (type 3)
91
As has been noted previously (Revithiadou 1998, Touratzidis and Ralli 1992), the data entail that stress placement in the genitive plural is sometimes governed by the same generalization as the stress in other inflected forms, and sometimes by a separate generalization. Any theory which adequately describes the facts would seem to need to come to this conclusion, although the formalism and details of the relevant generalizations may vary. Moreover, the classification above highlights the fact that inflectional patterns for stress placement crosscut inflectional patterns for segmental material. For example, νίκη / níki ‘victory’ (Table 16) and ώρα / óra ‘hour’ (Table 17) are typically considered to belong to different inflection classes because /i/ and /a/ are theme vowels, but a comparison of these lexeme’s respective forms shows a strong similarity. The former is classified as type 2, and the latter as type 3 because only the latter has variability in the stress pattern. The complementary pattern is also easily noticed, in which the stress pattern of two word form paradigms is the same, but the segmental inflectional material differs.
Following Albright’s formulation of a form-based grammar competition model, the variable application of stress shift in the genitive plural provides competing inflectional models. If this competition triggers insecurity in speakers, it should lead to avoidance strategies and paradigmatic gaps. There are three pieces to the puzzle: (1)
Distribution. Is there a correlation between words whose classes have stress variability and paradigmatic gaps? (2) Motivation. Do speakers perceive this variation as representing competing patterns? (3) Mechanism. What strategies do speakers use to resolve competing patterns? In the following sections I present one type of evidence in answer to each of these questions.
92
3.3. The distribution of genitive plural gaps
A search of two major Modern Greek dictionaries, Lexiko tis neas ellinikis glossas and the online version of Lexiko tis koinis neoellinikis47 (Babiniotis 1998, Triantafillidis 1998, respectively), uncovered genitive plural gaps in 2,141 distinct Modern Greek nouns. Of these, 470 are marked as genitive plural gaps in both dictionaries. An additional 212 words are marked in the Triantafillidis dictionary as having genitive plural gaps, while
Babiniotis marks them as having gaps in the entire plural. There are two examples of the reverse, in which the Babiniotis dictionary marks a genitive plural gap, but Triantafillidis marks the same words as having plural gaps. There are also five nouns which the
Triantafillidis dictionary marks as having genitive plural gaps, and Babiniotis marks as having gaps in the entire genitive, and no examples of the reverse. Thus, there are a total of 689 nouns which both dictionaries agree have gaps minimally in the genitive plural.
Of the remaining number, 579 genitive plural gaps are unique to the Babiniotis dictionary, either because the Triantafillidis dictionary did not have an entry for the lexical item, or more commonly because the Triantafillidis dictionary does not mark the noun as having a gap of any type, and 873 gaps are unique to the Triantafillidis dictionary.
These gaps represent a surprisingly large portion of all Greek nouns. The
Triantafillidis dictionary contains approximately 27,600 lexical entries for nouns, and
1,560 of these entries are marked as having gaps in the genitive plural. This means that about 5.6% of all nouns in the language have gaps in the genitive plural, and this number
47 http://www.komvos.edu.gr/dictionaries/dictadv/DictAdvSea.htm 93
excludes gaps that are in the entirety of the genitive, or the entirety of the plural, or for which the genitive plural is marked as being “rare” but not as a gap.48 To provide some sense of comparison, consider that the Russian first person singular verbal gaps (e.g. *(ja) pobežu ‘(I) win’) are one of the most famous cases. Halle (1973) claims that there are about 100 gaps in this cell, and it turns out that these 100 represent only about 60 distinct stems (see APPENDIX E). In other known cases, paradigmatic gaps exist in only a handful of forms (see APPENDIX A). The Modern Greek genitive plural gaps thus seem to be unusual in having such a large sample number, and it is interesting to delve into their distribution.
Of the 1,560 genitive plural gaps cited by Triantafillidis (1998), 88.5% occur among nouns of stress type 3. Excluding indeclinable nouns and singleton classes, only
58% of all Greek nouns belong to the same inflection classes. These distributions are disproportionate (x2=596.2, p<0.001), suggesting that the stress pattern of Modern Greek nouns conditions the appearance of gaps.
Noun classes with Noun classes with Noun classes with non- columnar stress fixed gen.pl stress columnar, non-fixed (type 1) (type 2) gen.pl stress (type 3) % N % N % N nouns with gpl gaps 0% 0 11.5% 180 88.5% 1,380 all nouns 2.9% 790 38.8% 10,588 58.3% 15,883
Table 18: Distribution of nouns with genitive plural gaps vs. all nouns, according to stress type (Triantafillidis 1998)
48 Similar figures are not available for the Babiniotis dictionary, but I expect that the results would be substantially similar. 94
Furthermore, by subdividing the gaps in nouns of stress type 3, it is possible to explore whether the relevant issue is variability in fixed stress placement, or stress shift.
The term “stress shift” implicitly compares multiple inflectional forms, but from the above description it seems possible to describe genitive plural stress without necessarily making reference to multiple inflectional forms (e.g., “stress in the genitive plural is always on the final syllable for nouns of inflection class X”). Paradigmatic (stress shift) and non-paradigmatic (stress fixed) accounts make different predictions regarding the distribution of gaps within an inflection class.
If the genitive plural and other inflected forms are subject to separate generalizations, and do not reference each other, there is no reason to believe that they will interact in a significant way. Thus, we would expect that the distribution of paradigmatic gaps within an inflection class should show no effect for whether a given word has stress on the same syllable in the nominative singular and the genitive plural.
By contrast, if the correct generalization compares stress in the genitive plural to that in other forms, paradigmatic gaps could be distributed according to whether the nominative singular and genitive plural place stress on the same syllable.
The most course-grained but also easiest way to distinguish the words which necessarily have the same stress in both forms from those which could have stress on separate syllables is to divide them according to whether the nouns have final or non-final stress in the nominative singular. Stress in Greek nouns is virtually never on an earlier syllable in the genitive plural than in the other inflected noun forms. Thus, words with
95
final stress necessarily have stress on the final syllable in all inflected forms, including the genitive plural; words with non-final stress potentially have a shift in the genitive plural towards the end of the word.
Genitive plural gaps appear exclusively in nouns with non-final stress in the nominative singular. 82.4% of all nouns in the same classes have non-final stress.
Again, this distribution is disproportionate (x2= 300.4, p<0.001).
final stress non-final stress (always “accidentally” (potentially subject to columnar) gpl stress shift) % N % N nouns with gpl gaps 0% 0 100% 1,380 all nouns 17.6% 2,801 82.4% 13,082
Table 19: Distribution of type 3 nouns with genitive plural gaps vs. all type 3 nouns, according to whether the (expected) form could have a stress shift (Triantafillidis 1998)
The data on genitive plural gaps in Modern Greek demonstrates that generalizations about Greek genitive plural stress should be made at the level of the paradigm. Trying to explain the appearance of genitive plural gaps with reference only to the genitive plural cell fails to capture that gaps are tied to nouns in which stress falls on different syllables in the genitive plural and other inflected forms. If we were to allow reference only to the genitive plural, these nouns do not form any sort of natural class. Thus, the evidence suggests that stress is a fundamentally paradigmatic phenomenon in Greek.
96
This is also a first piece of evidence suggesting that paradigmatic competition causes the gaps because of the close correlation between the distribution of gaps and the distribution of stem classes with variable stress patterns.
3.4. Avoidance strategies in the genitive plural
The distributional evidence is suggestive, but better evidence of synchronic motivation comes from variation between synthetic and periphrastic forms. If gaps are synchronically motivated by the competing stress patterns, I would expect to find the results of this competition among productive genitive plurals as well as among gaps.
After all, the hypothesis is that paradigmatic gaps represent only one extreme end result of the normal functioning of the morphological system and speakers’ reactions to it. The interchangeability of synthetic genitive plural forms and periphrastic equivalents sets up an ideal scenario to test the relationship between the stress pattern and genitive plural forms because only the synthetic form is subject to competition. If the variability of the stress shift causes speaker insecurity and thereby gaps, we might expect to find avoidance strategies for words with stress type 3, but not for words of types 1 and 2.
As alluded to in Section 3.1, the genitive is unique in the inflectional system in part because a periphrastic form can be used in place of the synthetic genitive form.49
49 Variation between a genitive NP and a bare accusative NP also exists in Modern Greek, as the following examples show. However, periphrasis is far more common and is the concern of this paper.
a. µια σειρά πηδηµάτων b. µια σειρά πηδήµατα mia seirá piðimáton mia seirá piðímata a series leaps.GEN.PL a series leaps.ACC.PL ‘a series of leaps’ ‘a series of leaps’
97
(20) a. Το έδωσε της Κατερίνας Το éxose tis Κaterínas It gave the-GEN.SG Catherine-GEN.SG ‘(S)he gave it to Catherine.’
b. Το έδωσε στη Κατερίνα To éxose sti Κaterína It gave to-the-ACC.SG Catherine-ACC.SG ‘(S)he gave it to Catherine.’
(21) a. µια µεγάλη µερίδα του λαού mia megáli meríða tu laú a large portion the-GEN.SG people-GEN.SG ‘a large portion of the people’
b. µια µεγάλη µερίδα από τον λαό mia megáli meríða apó ton laó a large portion from the-ACC.SG people-ACC.SG ‘a large portion of the people’
(22) a. Ο Παύλος είναι συνοµήλικος του ∆ηµήτρη O Paúlos eínai sinomílikos tu Ðímitri the Paul is same-age the-GEN.SG Demetra-GEN.SG ‘Paul is the same age as Dimitri’
b. Ο Παύλος είναι συνοµήλικος µε τη ∆ήµητρα O Paúlos eínai sinomílikos me ti Ðímitra the Paul is same-age with the-ACC.SG Demetra-ACC.SG ‘Paul is the same age as Demetra’
Based on grammatical descriptions of genitive usage and the discussion in the previous sections, several factors may be relevant to this variation (Holton et al. 1997, Jannaris
1987, Joseph and Philippaki-Warburton 1987, Thumb 1964[1912]).
First, genitive – prepositional phrase variation is tied to formality. Prepositions such as από/apó ‘from’ are “...often used as a colloquial alternative to the genitive, particularly the possessive genitive, the genitive of type, the genitive of content, or the
98
partitive genitive...” (Holton et al. 1997:383). The register difference is closely connected to the diglossic linguistic situation that Greek maintained until the 1970’s. The high style, katharevousa, represented an archaized version of the modern language. The low style, dhimotiki, represented everyday language. While dhimotiki was raised to the status of the official language in 1976, features of katharevousa have seeped into formal registers of Greek. Where the genitive and a prepositional phrase are functional equivalents in the modern language, synthetic genitive use is often associated with katharevousa and formal registers, periphrasis with dhimotiki and colloquial registers.
Second, different semantic functions are amenable to periphrasis to different degrees. As seen in (20) above, either a prepositional phrase or a genitive noun phrase is possible to express the indirect object relation. Similar situations exist for partitives (21), comparison (22), and a host of other functions. But Greek speakers typically prefer the
(b) form in (20), with the periphrasis, to the (a) form, but prefer the (a) form in (21), with the synthetic genitive form, to the (b) equivalent.
Furthermore, at the extremes only one form or the other may be possible.
Grammars and Greek informants claim that prepositional phrases cannot substitute for animate possessive genitives (23), genitives governed by verbs,50 any genitive weak personal pronoun, etc. Likewise, synthetic genitive forms cannot substitute for prepositional phrases in the locative function.
50 Brian Joseph (p.c.) points out that verbs which subcategorize for genitive tend to be katharevousa forms, and that a prepositional phrase would therefore produce a register clash. 99
(23) a. το σπίτι της Λουκίας to spíti tis Lukías the house the-GEN.SG Lucy-GEN.SG ‘Lucy’s house’
b. *το σπίτι από την Λουκία *to spíti apό tin Lukía the house from the-ACC.SG Lucy-ACC.SG *‘Lucy’s house’
These are only a few examples of the complex relationship between synthetic and periphrastic forms that are created by semantic issues. The basic pattern seems to be that the more prototypical the connection between a form and a particular function, the less likely it is to allow both the periphrastic and synthetic forms. Thus, animate possessors cannot be marked periphrastically. Directional and locational functions cannot be marked synthetically. Most functions, however, lie in the murky middle ground – inanimate possessors, relations between containers and contents, partitives, purposives, etc.
Third, based on the variability in stress placement across and within inflection classes, and the strong correlation between gaps and those word forms which are eligible for a stress shift, I would expect stress to be one of the factors which mediates this middle ground.
To these issues we may add frequency. There are large literatures on frequency effects in morphological processing, lexical storage, and morphological change (Alegre and Gordon 1998, Baayen 1993, Bertram et al. 2000, Butterworth 1983, the papers in
Bybee and Hopper 2001, Caramazza et al. 1988, Hay 2003, Hooper 1976, Phillips 2001,
100
Schreuder and Baayen 1995, and Stemberger and MacWhinney 1986 are just a few of the influential works). It would be impossible here to do justice to previous work in this area, and I have nothing new to suggest about the subtleties of frequency effects.
However, acknowledging that frequency has a pervasive influence, I would expect to find its effects even in crude measures such as corpus counts of lemma frequency.
3.4.1. Predictions
To review, from grammars I have identified four primary factors that may impact genitive/prepositional phrase variation in Modern Greek – register, function, stress, and frequency. Register, function, and frequency are relevant for both the genitive singular and the genitive plural, while stress relates primarily to the genitive plural. Focusing on stress and frequency, I set forth three hypotheses regarding genitive plural usage in
Modern Greek:
• Words with type 3 stress are more likely to show the prepositional phrase than are
words with stress types 1 or 2. (competing inflectional models → genitive plural
avoidance).
• Infrequent lemmas are more likely to take the prepositional phrase than are
frequent lemmas (infrequent → genitive plural avoidance).
• High competition and low frequency will interact to have an additive effect,
increasing the likelihood of genitive plural avoidance.
As outlined above, I also expect the use of genitive vs. prepositional phrase to vary according to the function of the genitive and according to register. However, these issues
101
are set aside for this experiment. Both function and register are controlled for in the following data in order to remove their influence from the analysis.
The logic behind these predictions is that where a morphophonological alternation occurs within a paradigm, there is motivation for avoidance of the alternation. That motivation should increase as the consistency of the alternation decreases. In the case of
Modern Greek, the morphophonological alternation is a stress shift in the genitive plural of some noun classes which does not typically exist in the accusative plural. Since a periphrastic prepositional phrase containing an accusative plural noun phrase is an independently existing alternative to the genitive plural, avoidance of the genitive plural should be promoted under these conditions.
This is a study of motivations. While this study does not measure the causes of
gaps directly, if the same factors which correlate with the distributional skew of gaps
turn out to synchronically motivate avoidance of the synthetic form, and below I show
that they do, it is reasonable to infer that gaps are the end result of this avoidance.
3.4.2. Methodology
I tested these hypotheses through a forced choice task in which subjects completed Greek sentences by choosing among four possible answers: a genitive noun phrase with normative stress, a periphrastic prepositional phrase with normative stress, a genitive noun phrase with a motivated but non-normative stress pattern, and a prepositional phrase with non-normative, unmotivated stress.
102
3.4.2.1. Target words
To compose this task, I mined a sample of Modern Greek nouns from the on-line version of the Λεξικό της Κοινής Νεοελληνικής / Leksikό tis Koinís Neoellinikís (Triantafillidis
1998).51 All inflectional classes were represented among the 2,561 words. I then sorted these words according to the three categories described above: stress type 1 (798 words), stress type 2 (742 words), and stress type 3 (1,021 words).
Words were sorted into two levels of frequency – frequent and infrequent – based on a combination of lemma frequency in the Hellenic National Corpus (HNC), 52 and the judgments of selected words from one adult native Greek speaker from Athens. This native speaker rated the nouns on a 7-point Likert scale according to his familiarity with the words and the frequency with which he uses them. Words which were both within the top 50 most frequent words for that stress type according to the corpus count and which received a rating of 7 (“I use this word frequently”) are considered, for the purpose of this study, “frequent”. Words which had a lemma frequency between 3 and 15 tokens per million according to the corpus count and which received a rating of 5 (“This word is familiar, but I use it only rarely”) are considered, for the purpose of this study,
“infrequent”. Only words meeting both the Likert and HNC criteria were used in the study. The result is a binary opposition between frequent and infrequent words, with words of intermediate frequency discarded.
I further filtered candidate words according to the frequency of the genitive plural form relative to the rest of the word, as represented in the Hellenic National Corpus. I
51 http://kastor.komvos.edu.gr/dictionaries/dictadv/DictAdvTri.htm 52 The Hellenic National Corpus is a 32+ million-word corpus of written Modern Greek which was created by the Institute for Language and Speech Processing in Greece. The frequency produced by the corpus is assumed to be representative of the each noun's frequency in written texts. http://hnc.ilsp.gr/statistics.asp 103
eliminated words in which the genitive plural is disproportionately frequent/infrequent
(less than 2% or more than 25% of lemma frequency) so as to ensure that idiosyncrasies of particular words did not become an undue factor in the analysis and that the genitive plural is an attested form, and thus a possible response.
All selected target words had non-final stress.
3.4.2.2. Questionnaire composition
With the help of a native Greek speaker, I created thirty-six examples by factorial design, permuting three factors: 6 levels of genitive function x 3 levels of stress type x 2 levels of frequency = 36. Each example consisted of a sentential frame with the target noun phrase removed. Each sentential frame was different, and aimed to create realistic sentences of the Greek language, such as might be found in a newspaper or other general-audience literature. In each case, the frame could be completed grammatically by either a synthetic genitive plural noun phrase, or a prepositional phrase. An additional eighteen examples were fillers. These eighteen mostly represent functions that in the history of the language were realized with the synthetic genitive, but that use is obsolete in Modern Greek, with a prepositional phrase being the only “natural” choice.
The six functions were abstract object, content, partitive, possession, prepositional, and purpose. An example of each function is given below. I chose these functions because they represent a wide variety of uses of the genitive, and because they occur robustly both in the genitive and in the periphrastic form. See APPENDIX C for all of the questionnaire materials.
104
(24) abstract object a. η λύση των προβληµάτων του i lísi ton provlimáton tu the solution the-GEN.PL problems-GEN.PL his ‘the solution to his problems’ b. η λύση στα προβλήµατά του i lísi sta provlimáta tu the solution to-the-ACC.PL problem-ACC.PL his ‘the solution to his problems’
(25) content a. µια παρέα εννέα γυναικών mia paréa ennéa γinekón a group nine women-GEN.PL ‘a group of nine women’ b. µια παρέα µε εννέα γυναίκες mia paréa me ennéa γinékes a group with nine women-ACC.PL ‘a group with nine women in it’
(26) partitive a. µεγάλος αριθµός των µηχανών meγálos ariTmós ton mixanón large number the-GEN.PL machines-GEN.PL ‘a large number of the machines’ b. µεγάλος αριθµός από τις µηχανές meγálos ariTmós apó tis mixanés large number from the-ACC.PL machines-ACC.PL ‘a large number of the machines’
105
(27) inanimate possession a. τα κλαριά των δέντρων ta klarjá ton ðéndron the branches the-GEN.PL trees-GEN.PL ‘the trees’ branches’ b. τα κλαριά από τα δέντρα ta klarjá apó ta ðéndra the branches from the-ACC.PL trees-ACC.PL ‘the trees’ branches’
(28) prepositional a. ένα λόγο εναντίον των απεργιών éna lóγo enandíon ton aperγjón a speech against the-GEN.PL strikes-GEN.PL ‘a speech against the strikes’ b. ένα λόγο ενάντια στις απεργίες éna lóγo enándia stis aperγíes a speech against to-the-ACC.PL strikes-ACC.PL ‘a speech against the strikes’
(29) purpose a. ένα ποτήρι του κρασιού éna potíri tu krasjú a glass the-GEN.SG wine-GEN.SG ‘a wineglass’ b. ένα ποτήρι για (το) κρασί éna potíri γja to krasí a glass for (the-ACC.SG) wine-ACC.SG ‘a wineglass’
106
As noted in a previous section, register plays a known role in genitive – prepositional phrase variation. This study does not explore the role of register, but I attempted to compose sentences of the same register in order to control for this problem.
Participants chose the best sentence completion from among four choices, each of which featured the same lexeme. In one choice the target noun was a genitive plural noun phrase with normative stress.53 One choice was a prepositional phrase with the target noun appearing in accusative plural, and with normative stress. A third choice was a genitive plural noun phrase identical to the first choice, except that the target noun had non-normative stress placement. In the case of nouns from type 1 stress paradigms, this shift was always within the final three syllables of the phonological word, as required generally in Greek, but was unmotivated because stress never shifts in these declension classes. For nouns from stress type 2 and stress type 3 paradigms, the choice with the non-normative stress represented stress leveling. A fourth choice was a prepositional phrase with an accusative plural target noun, identical to the second choice except for a non-normative stress shift on the target noun. Since none of the nouns used in this study had stress shift in the accusative plural, this stress shift was unmotivated for all noun classes.
Thus, three choices were viable as possible sentence completions – a normative genitive plural noun phrase, a normative prepositional phrase, and a genitive plural noun phrase exhibiting stress leveling (for types 2 and 3). By forcing speakers to choose between these three possibilities (+ one filler), I was able to test native Greek speakers’ preference for genitive vs. prepositional phrase according to frequency and stress type.
53 Stress is encoded orthographically in the Greek language, and I assume that speakers write stress where they would place it in speech. 107
3.4.2.3. Questionnaire administration and subjects
I conducted this research partially as a web survey, and partially as a pencil-and-paper task. While the mediums differed, the materials were identical in each case. Questions were presented in four different orders and response options were in randomized order.
I recruited participants for the web survey through friends, colleagues and academic mailing lists known to be read by native Greek speakers. Volunteers for the pencil-and-paper task were recruited from among the student body at the Aristotle
University of Thessaloniki, primarily through announcements in introductory linguistics classes and French literature classes.
Thirty-one native Greek speakers completed the survey and qualified to have their answers included in the analysis. 23 were women, 9 were men. The skew towards female subjects reflects the reality that students of the French and Linguistics sections at the Aristotle University of Thessaloniki are overwhelmingly (approximately 90%) female. There were sixteen participants aged 20-29, including all ten participants recruited in Thessaloniki, eight aged 30-39, five aged 40-49 and two aged 50-59. All participants were college educated or in the process of pursuing an undergraduate degree.
Some have lived their entire lives in Greece; others currently live abroad. All of the included subjects reported speaking at least some Greek in their daily lives.
These 31 subjects do not include two female and one male subject who had not lived through adulthood in Greece, and whose answers I therefore discarded.
108
3.4.3. Results
The most obvious result of this study is that there are main effects in the predicted directions for both stress type and lexeme frequency.
As shown in Table 20, participants chose the genitive plural 76.7% of the time for frequent words, and 68.8% of the time for infrequent words. Conversely, participants chose the prepositional phrase 21.7% of the time for frequent words and 26.9% of the time for infrequent words. A chi-squared test shows that frequency is a significant factor to the level of p<0.01 (x2 = 13.208).
gpl NP, periphrasis, gpl NP, periphrasis, non-normative non-normative normative stress normative stress stress stress % N%N%N % N frequent 76.7% 428 21.7% 121 1.1% 6 0.5% 3 infrequent 68.8% 384 26.9% 150 2.0% 11 2.3% 13
Table 20: Genitive vs. prepositional phrase according to frequency
gpl NP, periphrasis, gpl NP, periphrasis, non-normative non-normative normative stress normative stress stress stress % N%N%N % N stress type 1 80.9% 301 17.7% 66 0.5% 2 0.8% 3 stress type 2 73.7% 274 24.7% 92 1.3% 5 0.2% 1 stress type 3 63.7% 237 30.4% 113 2.7% 10 3.2% 12
Table 21: Genitive vs. prepositional phrase according to stress type
109
A more robust correlation appeared between noun stress type and periphrasis, given in
Table 21. For target nouns with type 1 stress, survey participants chose the normative genitive plural answer 80.9% of the time, for type 2 stress words – 73.7% of the time, and for type 3 stress words – 63.7% of the time. An opposite but less robust effect is seen for the prepositional phrase – 17.7% for type 1 words, 24.7% for type 2 words and 30.4% for type 3 words. A chi-squared test indicates that stress predictability is a highly significant factor in preference for the synthetic vs. periphrastic form (p<0.001, x2=38.541).
However, if competing stress patterns promote speakers’ avoidance of the synthetic genitive plural by causing them to be unsure about their production of genitive plural forms, we would not expect the effects for frequency and stress type to be independent of each other. Low frequency and low consistency of stress placement within the inflection class should have an additive effect. As can be seen in the following graph, this is exactly the interaction that I found.54
54 The values on the y-axis should be interpreted as follows: 1.0 = 100% selection of the normative genitive plural form for the given condition, 0 = any combination of the other three choices. 110
0.83 2 0.76 1 0.75 0.78 1 1 2 0.72
2 0.51 choice of prescriptive genitivechoice prescriptive pluralof
1=Frequent 2=Infrequent 0.0 0.2 0.4 0.6 0.8 1.0
type 1 type 2 type 3
stress type
Figure 8. Interaction of frequency and stress predictability as conditioning factors for avoidance of the synthetic genitive plural form
Separating the data in this way, we can see that there is no effect for stress type apart from frequency, or for frequency apart from stress type. However, the combination of low frequency and type 3 stress entail a strong drop in the preference for the synthetic genitive plural form. A simple linear regression model for the infrequent tokens across stress types produced a fit line indicating a marginally significant correlation (R2=0.245,
111
p<0.05). By contrast, the frequent items produced a non-significant fit line (R2=0.001, p=0.8794).
I argue that these effects exist because low frequency and competing stress patterns together lead speakers to be uncertain about the form of the synthetic genitive plural and thus to avoid it in favor of the periphrastic prepositional phrase for which noun stress is the same as in the nominative singular and other forms.
This result should not be surprising. Language ideology can be a powerful influence on both language use and perceptions of language use (see, for example,
Blommaert 1999, Schieffelin et al. 1998). Furthermore, there is some evidence that
Greek speakers can feel insecure about linguistic features associated with katharevousa.
Hawkins (1979) presents the results of a cloze procedure experiment designed to test the degree to which speakers are willing to codeswitch between katharevousa (K) and dhimotiki (D). Among various results, Hawkins concluded that “...speakers would switch to K for the ‘easier’ levels of phonology/orthography and morphology, but use D at the more ‘difficult’ levels of lexis and syntax” (169). While Hawkins does not specifically consider the genitive plural, his work provides some empirical support for the frequent layman’s claim that Greek speakers do (or at least in 1979 did) feel insecurity over forms of katharevousa origin, of which the genitive plural is one, and vary their use accordingly.55
Moreover, the similarities between the distribution of gaps and avoidance of the synthetic forms are striking. In both cases, the inflection classes with stress variability are the sites of lesser viability of the genitive plural form. There thus seems to be a
55 I am grateful to Grace Fielder for bringing this study to my attention. 112
connection between a factor which synchronically motivates speaker behavior and paradigmatic gaps. The only remaining question relates to the mechanism which is purported to link the two – do the competing stress patterns cause speaker insecurity?
3.5. Confidence ratings and stress variability in the genitive plural
In this experiment I demonstrate that stress shifts affect speaker confidence, thus providing motivation for preference of periphrastic constructions over synthetic constructions. This study replicates the basic methodology of Albright’s (2003) study of
Spanish, so as to provide comparable data to that which he uses to argue for a grammar competition model.
3.5.1. Predictions
With regard to the Greek data, a grammar competition model makes at least two predictions: (1) Speakers’ confidence in their productions of genitive plural forms will directly correlate with the degree to which speakers agree on whether a given word has a stress shift (agreement over the form of a word being taken as a measure of grammar competition; higher agreement = lower competition). (2) Speakers’ confidence in their productions of genitive plural forms will be inversely correlated with lexeme familiarity.
If these predictions are upheld, we may conclude that paradigmatic competition is an organizing principle of inflectional structure, that it has a significant impact on language use, and that the Modern Greek genitive plural gaps are epiphenomenal, extreme result of its gradient effects.
113
3.5.2. Methodology
Following Albright, I tested these hypotheses through a questionnaire with three primary tasks: a familiarity judgment task, a forced production task, and a production ratings task.
3.5.2.1. Target words
The questionnaire consisted of 80 words: 40 target nouns and 40 filler nouns. The target nouns represented three conditions: gaps (N=15), nouns in the same inflection class as gaps (N=15), nouns in different inflection classes than gaps (N=10). Nouns in this last group all had type 1 stress and the formative /đ/ in the plural; see Table 15 for examples.
NO GAP, WITH A STRESS SHIFT GAP ‘tavern’ SINGULAR PLURAL ‘girl’ SINGULAR PLURAL NOM ταβέρνα ταβέρνες NOM κοπέλα κοπέλες tavérna tavérnes kopéla kopéles ACC ταβέρνα ταβέρνες ACC κοπέλα κοπέλες tavérna tavérnes kopéla kopéles GEN ταβέρνας ταβερνών GEN κοπέλας * tavérnas tavernón kopéla
NO GAP, NO STRESS SHIFT ‘portion’ SINGULAR PLURAL NOM µερίδα µερίδες meríđa meríđes ACC µερίδα µερίδες meríđa meríđes GEN µερίδας µερίδων meríđa meríđon
Table 22: Examples of the inflection class from which genitive plural stimuli were drawn
114
This group was intended as the control because it has columnar stress throughout. The remaining stimuli all belonged to the class that Triantafillidis identifies as
O25/O25a/O26. This class is exemplified above. All of the selected words had penultimate stress in the non-genitive-plural forms.
I chose this inflection class because it has by far the largest number of gaps (see
APPENDIX B), it is one of the largest nominal inflection classes overall in Greek, and because it is a class with no stress shift in any form other than the genitive plural.
I determined the defective or non-defective status of a word according to the listings in two major dictionaries, the Λεξικό της Κοινής Νεοελληνικής / Leksikό tis
Koinís Neoellinikís (Triantafillidis 1998) and the Λεξικό της Νέας Ελληνικής Γλώσσας /
Leksikó tis Néas Ellinikís Glóssas (Babiniotis 1998). The stimuli with genitive plural gaps (e.g. κοπέλα ‘girl’) are listed as such in both dictionaries. Likewise, the words without gaps are listed as being unexceptional by both dictionaries, and a search of the
Hellenic National Corpus (see footnote 52) confirmed that the genitive plural of these forms is used roughly in the expected proportion, relative to the frequency of other inflected forms. I chose the target nouns to represent a variety of levels of familiarity, as judged by lemma frequency within the Hellenic National Corpus.
The filler nouns were chosen to represent a variety of levels of difficulty for speakers, including productive nouns, indeclinable nouns, nouns strongly associated with katharevousa, nouns with known morphological variation, nouns in singleton inflectional classes, etc.
115
3.5.2.2. Questionnaire Composition
For each of the 80 nouns, subjects performed five tasks – two production tasks, two self- judgments of their productions, and a lexeme familiarity ratings task.
In the main section, a forced production/ratings task, participants filled a sentence blank with the most natural form of a provided word. The questionnaire presented the stimulus words in the nominative singular, which is the standard citation form in Greek.
Each lexeme was included twice in order to elicit two different forms. For target nouns, the sentential frames required the genitive plural and either the nominative singular or the nominative plural. For the filler nouns, the case-number form required by the sentential frames varied, and included all inflected forms except the genitive plural. After writing the necessary form of the word, subjects rated their confidence in their own production using free magnitude estimation.
In free magnitude estimation, subjects are asked to randomly assign a score reflecting some perceptual judgment to the first stimulus, and then double that score for the second stimulus if they are twice as confident in their production, half the score if they are half as confident, etc. Free magnitude estimation thus provides a dynamic scale with the range determined by the participants which can be used statistically as a continuous variable. I asked participants to rate the degree to which they were sure that the form of the word that they wrote was correct.
Since the questionnaire presented each stimulus word twice, the main section of the questionnaire contained 160 sentential frames, and 160 corresponding ratings requests
(80 target; 80 filler).
116
The familiarity judgment task was a pretest in which participants rated their familiarity with and use of Greek nouns on a 6-point Likert scale. I presented the stimulus words in the nominative singular. In the analysis, I removed production responses and production ratings for those lemmas which speakers rated below a 3 (“I am familiar with this word, but I don’t use it”) on the pretest. More than half of subjects rated two target words below this threshold: κορβανά / korvaná ‘coffer’, and βαγενά / vaγená ‘barrel maker’. Both of these belong to the “different inflection class than gaps” group. I ultimately did not include this group in the analysis due to an unforeseen methodological problem (see Section 3.5.2.4). For all other target words, familiarity ratings below 3 were sporadic and rare.
I administered this questionnaire as a pencil-and-paper task.
3.5.2.3. Subjects
I recruited participants for the questionnaire from among the student body at the Aristotle
University of Thessaloniki, primarily through announcements in introductory linguistics classes and French literature classes.
Thirty-eight native Greek speakers completed the survey and qualified to have their answers included in the analysis.
• 34 were women, 4 were men. The skew towards female subjects reflects the
reality that students of the French and Linguistics section at the Aristotle
University of Thessaloniki are overwhelmingly (approximately 90%) female.
Thirty-seven of the subjects were 18-30 years old in 2006; the exception was a 38
year old female. I analyzed her answers for deviance from those of the remainder
117
of the subjects. No notable differences were found, so her answers are included in
the following data.
• All participants were in the process of pursuing an undergraduate degree or
graduate degree.
• All had lived through adulthood in Greece and currently live in Thessaloniki.
Eleven had been raised in Thessaloniki. Among the remaining subjects, 6 grew
up primarily in Athens, 4 in the city of Serres, 2 in Veroia, and 14 in other
villages and towns, primarily in northern Greece. The home town of one subject
was not available.
• Among these thirty-eight subjects, all reported knowing at least one language well
other than Greek, usually English and French, but none reported a second native
language.
These 38 subjects do not include four people who volunteered for the study, but whose answers were discarded. Three were non-native speakers of Greek. One did not provide a complete set of responses to the questionnaire.
3.5.2.4. Three minor methodological snags
In the course of collecting data, three minor methodological issues were raised.
First, as described in Section 3.5.2.1, I intended to use ten words as control items.
According to the Λεξικό της Κοινής Νεοελληνικής / Leksikό tis Koinís Neoellinikís
(Triantafillidis 1998), these words all have the formative /đ/ in the plural and belong to inflection classes of stress type 1, i.e. with columnar stress throughout. My intention was that these would be words that are subject to no significant morphological variation, and
118
which therefore would likely represent the maximal confidence that speakers could have in genitive plural forms. However, it turns out that several of these words vary between two inflection classes. Below I list the nominative singular form that I provided on the questionnaire, and the genitive plural forms that speakers responded with.
(30) Variability among the control nouns in the genitive plural
Nom. sing. Gen. plur. Gloss γιαγιά / γjaγjá γιαγιάδων / γjaγjáđon ‘grandmother’ µαµά / mamá µαµάδων / mamáđon ‘mother’ ταξιτζής / taksitzís ταξιτζήδων / taksitzíđon ‘taxi driver’ ταξιτζίδων / taksitzíđon καναπές / kanapés καναπέδων / kanapéđon ‘sofa’ βαγενάς / vaγenás βαγενάδων / vaγenáđon ‘barrel maker’ βαγενών / vaγenón κορβανάς / korvanás κορβανάδων / korvanáđon ‘coffer’ κορβανών / korvanón αµανές / amanés αµανέδων / amanéđon ‘love song’ καφετζής / kafetzís καφετζήδων / kafetzíđon ‘coffee shop owner’ καφετζίδων / kafetzíđon µουσαµάς / musamás µουσαµάδων / musamáđon ‘oil cloth’ παπουτσής / paputsís παπουτσήδων / paputsíđon ‘shoe maker’ παππουτσήδων / papputsíđon παπουτσίδων / paputsíđon παπουτσάδων / paputsáđon παππουτσών / paputsón παπουτζήδων / paputzíđon
Due to this variation, these nouns are not suitable for the purpose for which I intended them. I do not report on them below. This is not a serious loss, however. For the nouns under prime focus, i.e. those exemplified in Table 22, the more interesting comparisons are between the nominative and genitive plural for the same noun, and between gaps and non-gaps belonging to the same class.
119
Second, among the non-defective words belonging to the class exemplified in
Table 22 on page 114, I chose eight nouns which, according to the Triantafillidis dictionary, have a stress shift in the genitive plural to the final syllable, as in the example for ‘tavern’. I chose seven words reported as having no stress shift, as in the example for
‘portion’. Subjects produced the seven non-shifting words with columnar stress, and as the data in the following section show, were almost entirely consistent in doing so.
Interestingly, however, the supposedly shifting words did not form a coherent group.
Three were judged almost uniformly to be non-shifting, indistinguishable from the other seven, while five were produced with a stress shift to one degree or another. Thus, the ratio of shifting to non-shifting genitive plural forms turned out to be 1:2. See
CHAPTER 4 for some potential implications of this variability within (only) the
“shifting” group.
Third, subjects were supposed to rate the word that they wrote to complete each sentential frame based on how confident they felt that the form of the word was correct.
My logic, inherited from Albright (2003), was that I could rely on speakers’ persistent prescriptivism. If speakers have difficulty deciding what the form of the word should be, but feel that there is a single, “correct” variant of each word, they will be less certain that they have produced the form which is “right”, and will report that as lower confidence, relative to other forms of the same lexeme. This result would indicate the possibility that ideological factors influence their linguistic behavior in this area of the grammar – the sought after link between morphophonological variation on the one side, and avoidance behavior and/or gaps on the other. However, if speakers either do not weigh multiple
120
forms, or do not feel that one form is more correct than another (as is often the case with variation), they should report equivalent confidence ratings, relative to other forms.
Unfortunately, the instructions, presented in Greek, were unclear. My discussions with some of the subjects after completing the questionnaire indicated that as a result, subjects had interpreted the instructions in one of two ways. Many had rated the nouns according to whether they thought that the lexeme sounded like a native Greek word
(high score) or like a borrowing (low score), i.e. “right” in the sense of genuinely Greek.
Some had rated the nouns according to the criteria that I had intended.
I determined subjects’ tactics by comparing ratings for seven words with a long history in the Greek language with ratings for seven obvious borrowings into Greek. The questionnaire presented all fourteen words in sentential frames requiring the nominative singular. If a subject consistently gave lower scores to the obvious borrowings than to the words with a long history in Greek, I assumed that the subject had scored words according to perceived nativeness. The ratings of twenty-seven subjects met these criteria. If a subject gave equivalent scores to both groups (there were no instances of consistently higher scores for the borrowings), then I assumed that the subject scored according to the criteria that I had intended, or at least did not base their scale on perceived nativeness. The ratings of eleven subjects met these criteria. Thus, in the following data, the word form statistics that are reported reflect all thirty-eight subjects, but the confidence ratings associated with those word forms reflect the ratings of only eleven speakers. Where this discrepancy potentially affects the analysis it is noted.
121
3.5.3. Results
The first notable result from the data is that speakers are not equally confident in all word forms which share a case-number combination. 0.6 0.8 1.0 1.2 1.4 Confidence ratio (form: other forms w/ same infl. properties) infl. same w/ forms other (form: Confidenceratio 3.0 3.5 4.0 4.5 5.0 5.5 6.0
Average familiarity rating (6=high)
Figure 9: Variance in the confidence ratings reported by Greek speakers, relative to the average for words sharing the same inflectional properties
122
I present the confidence ratings as a ratio relative to a given person’s average confidence rating for all forms sharing the same inflectional property set. For example, a given accusative singular form is measured against the average for that speaker for all accusatives. A given nominative plural is measured against all nominative plurals. Etc.
These ratios are then averaged across subjects to produce a single point per item. All items with an average familiarity rating above 3 (“I am familiar with this word, but I don’t use it”) are included (N=152). If speakers were equally confident in all nominative singular forms, or all accusative plural forms, etc., we would expect the values to cluster around 1.0 on the y-axis, indicating little deviance from the average for words expressing that inflectional property set. While there are many forms which do cluster around 1.0, there are also many that do not. Thus, subjects were not equally confident in all of the forms that they produced.
Furthermore, it is apparent that the genitive plural accounts for most of the variance. In the graph below, I included only the nominative singular forms (N=40) and the genitive plural forms (N=30), so as to not clutter the graph. The same demonstration could have been made with any of the inflected forms except the accusative plural in place of the nominative singular, however. Only the accusative plural resembles the genitive plural in its degree of variance around the mean.56
56 The accusative plural forms of the lexemes included in the questionnaire had their own variable stress placement, so the overlap between genitive plural and accusative plural is not surprising. 123
square=gen plur circle=nom sing 0.6 0.8 1.0 1.2 1.4
3.0 3.5 4.0 4.5 5.0 5.5 6.0 Average Confidence Ratio (form: other forms w/ same infl. properties) infl. same w/ forms other ConfidenceAverage Ratio (form: Average Familiarity Rating (6=high)
Figure 10: Greater variance among genitive plural forms than among nominative singular forms
124
Note also that contrary to expectations, there is no major effect for familiarity. Single linear regression does indicate that confidence decreases slightly as words become less familiar, but was only marginally significant (p<0.05, df=150). This fit also accounted for virtually none of the variance (R2 = 0.02871).57 I conclude that in this task, familiarity with a lexeme had little to no impact on Greek speakers’ confidence in producing a form of that lexeme.
The question, then, is whether the greater variance in confidence for the genitive plural correlates with agreement (and lack thereof) over the stress pattern. Among the fifteen non-defective words of the type given in Table 22, the answer appears to be that it does. Since speakers were asked to rate words using free magnitude estimation, in which they determined their own scale, I normalized across speakers. In Figure 11, I represent subjects’ confidence in the genitive plural as a ratio relative to their confidence in the nominative singular or nominative plural form of the same lexeme.58 For example, if a subject was equally confident in the genitive plural as in the nominative form, the normalized score is 1; if half as confident, the normalized score is 0.5, etc. In the graph below, these normalized scores are averaged across speakers to produce an item analysis.
57 The closer the value is to 1, the more familiarity predicts confidence. 58 Whether this other form was a nominative singular form or a nominative plural form had no noticeable effect on confidence ratings. A linear fit for the nominative singular confidence ratings had an intercept of 0.8985, a slope of 0.0206, and an R2 value of 0.1951, indicating that almost 20% of the variance in the confidence ratings is accounted for by interspeaker agreement on the form. A linear fit for the nominative plural confidence ratings had an intercept of 0.8795, a slope of 0.0277, and an R2 value of 0.0496. This difference was not significant, nor was the result of a difference of the means test (p=0.557). I thus conclude that speakers were equally confident in the nominative singular as in as the nominative plural, and I have lumped together the nominative values. 125
Average Confidence Ratio (gpl form: nom form) nom ConfidenceAverage Ratio (gpl form: 0.6 0.8 1.0 1.2 1.4
0.5 0.6 0.7 0.8 0.9 1.0
Interspeaker Agreement about Stress Shift (0.5 = 50/50 split)
Figure 11: Effect of interspeaker agreement on speaker confidence
Although there are few points in the low agreement range to judge from, it seems that agreement over the genitive plural stress pattern does correlate with speakers’ reported confidence in producing the genitive plural form of that word (R2 = 0.6908, p<0.01).
Notice especially that when speakers agree completely on the genitive plural form, they report the same level of confidence in this form as they have in the nominative form. 126
The effect is even stronger when actual level of stress shift, rather than speaker agreement is considered (R2 = 0.8054, p<0.001) 0.6 0.8 1.0 1.2 1.4 Average Confidence Ratio (gpl form: nom form) nom form: (gpl Ratio Confidence Average
0.0 0.2 0.4 0.6 0.8 1.0
% NO Stress Shift (0 = all shifting, 1 = all non-shifting)
Figure 12: Effect of shifting stress on speaker confidence
Since this is the same class into which the vast majority of genitive plural gaps fall, these results are what we would expect to find if gaps are caused indirectly by grammar competition and directly by the ideological issues which speakers place on that 127
competition. Forms with high agreement effectively are not subject to the competition.
Forms for which there is low agreement correspond to low speaker confidence. This shows that speakers interpret their genitive plural usage through a prescriptivist filter.
In combination with the data on the distribution of periphrasis presented in
Section 3.4 and the distribution of paradigmatic gaps presented in Section 3.3, this study supports the conclusion that the nominal stress pattern of Modern Greek causes speakers to be insecure, insecurity causes speakers to avoid the genitive plural form, preferring a periphrastic form instead, and avoidance ultimately leads to paradigmatic gaps. In the following section I outline how these results translate to a theory of inflection which can describe the distribution without resorting to ad hoc mechanisms such as a CONTROL component (Orgun and Sprouse 1999) or specification [-Lexical Insertion] (Halle 1973), whose sole purpose is to describe the anomalous distribution of gaps.
3.6. Greek gaps in a paradigmatic competition model
To summarize briefly over the preceding data, a model of the Greek paradigmatic gaps needs to take account of the following facts:
• Paradigmatic gaps in the genitive plural are closely tied to the stress patterns of
noun inflection classes.
o Paradigmatic gaps occur primarily in inflection classes which have
internal stress variation in the genitive plural form (assuming classes are
defined by shared segmental inflectional material, an assumption that I
challenge below).
128
o Within these inflection classes, paradigmatic gaps occur exclusively in
lexemes whose attested inflected forms have non-final stress. The
expected genitive plural form would thus potentially have stress on a
different syllable than the attested forms.
• Some inflection classes are subject to stress competition, while others are not.
Only a subset of the nouns with fixed genitive plural stress are subject to
competing patterns.
• Stress is paradigmatically governed. The need to reference multiple inflected
forms in order to adequately describe the distribution of the gaps shows that
competition is between paradigmatic patterns, not individual word forms as in
Albright’s (2003) Spanish examples.
Considering these points, how stems are organized into inflection classes, and the relationship between stress and inflection class membership are obviously critical to any description of paradigmatic gaps. Thus, I first provide a theoretical account of Greek nominal inflection classes, as motivated by productive word forms, and then move on to show how this structure also accounts for paradigmatic gaps.
3.6.1. Setting the stage: multidimensional inheritance hierarchies
My basic claim is that the classification into inflection classes and stress types given in
Section 3.2. is descriptively convenient, but not theoretically adequate. The traditional classification fails to account for two facts: (1) Groupings based on stress patterns crosscut groupings based on segmental patterns. The latter are typically given priority;
129
stress is described as an internal subdistinction. However, since both serve as markers of inflectional properties, there is no a priori reason to prioritize segmental information. (2)
Singular and plural inflectional patterns seem to operate at least partially independently of each other. Neither is strongly predictive of the other in many cases. This can be seen from the following table. Traditionally, the so-called theme vowels are considered part of the inflectional ending in Greek nouns (e.g., the stem for kálfas is kálf-, rather than kálfa-). However, if we assume that theme vowels are part of the stem, we can define an inflection class in the narrowest sense as a cell in the table, thereby collapsing some of the traditional distinctions, and capturing similarities between words whose only inflectional differences is in this vowel.
130
singular formatives (nom, acc, gen) -ς, --, -- --, --, -ς -oς, -o, -ου --, --, -τα --, --, ου --, --, -ού -ος, -ος, -ους (-s,--,--) (--,--,-s) (-os, -o, -u) (--,--,-ta) (--,--,-u) (--,--,-ú) (-os, -os, -us) ‘-δες, ‘-δες, ‘-δων κάλφας (‘-đes, ‘-đes, ‘đon) kálfas -ηδες, -ηδες, κόντες -ηδων kóntes (-iđes, -iđes, -iđon) -δες,-δες,-δων µανάβης µαµά (-đes, -đes, -đon) manávis mamá
καφές αλεπού kafés alepú
παππούς pappús -ες , -ες, -ων πατέρας µητέρα (-es, -es, -on) patéras mitéra -ες , -ες, ‘-ων φύλακας (-es, -es, ‘-on) fílakas -ες, -ες, -ών τουρίστας ώρα (-es, -es, -ón) turístas óra
ναύτης νίκη náftis níki ‘-εις, ‘-εις, ‘-εων δύναµη (‘-is, ‘-is, ‘-eon) đínami -είς, -είς, -έων δεκανέας (-ís, -ís, -éon) đekanéas
-εις, -εις, -ων συγγενής (-is, -is, -on) si(n)genís -οι, -ους, -ων περίπλους αντίλαλος plural formatives (nom, acc, gen) (nom, formatives plural (-i, -us, -on) períplus a(n)dílalos -οι, ‘-ους,’ –ων άνθρωπος (-i, ‘-us, ‘-on) ánθropos -τα, -τα, ‘-των µέλλον (-ta, -ta, ‘-ton) méllon
όνοµα ónoma
ηρέας iréas -α, -α, ‘-ων πρόσωπο (-a, -a, ‘-on) prósopo -α, -α, -ων σίδερο (-a, -a, -on) síđero -α, -α, -ών αγόρι (-a, -a, -ón) aγóri ‘-η, ‘-η, -ών µέρος (-í, -í, -ón) méros
Table 23: Greek inflectional patterns
131
This format highlights that in some cases there is a unique correspondence between the singular forms and the plural forms, as for µέρος / méros ‘part’. However, in many cases there are two or more classes which share inflectional processes only in the singular or only in the plural. For example, µανάβης / manávis ‘greengrocer’ shares plural formatives with µαµά / mamá, but not singular formatives. It shares only singular formatives with a variety of other classes, for example that represented by περίπλους / períplus ‘circumnavigation’. Thus, in many cases, inflection classes are overlapping entities.
I argue that the disconnect between the singular and plural forms and between the segmental and stress patterns should be represented as a multidimensional inheritance hierarchy in which each dimension operates independently.
3.6.1.1. The stress inheritance hierarchy
One dimension of the hierarchy carries information about stress. The nominal stress system, as relevant to the genitive plural, can be represented as follows.59
59 This hierarchy glosses over some details which would be relevant to a full account of Greek nominal stress, the rare case of δεκανέας / đekanéas ‘corporal’ and similar words in which stress appears to be fixed in the nominative plural and accusative plural, but not in the other forms, including the genitive plural. 132
gpl fixed
gpl = ...σσ⁄# gpl = ...σ⁄σ# [1]
gpl = ...σ⁄σ# gpl = ...σ⁄σ# [2] gpl ∪ gsg
gpl = ...σσ⁄# gpl = ...σ⁄σ# gpl = ...σ⁄σ# gpl ∪ gsg gpl ∪ gsg gsg ∪ gpl ∪ accpl [3] [4] [5]
Figure 13: A default inheritance hierarchy for Greek nominal stress
Node [1] represents inflection classes which have genitive plural stress fixed on the final syllable, but potentially on a different syllable in the other inflected forms, as in the following example.
‘tavern’ SINGULAR PLURAL NOM ταβέρνα ταβέρνες tavérna tavérnes ACC ταβέρνα ταβέρνες tavérna tavérnes GEN ταβέρνας ταβερνών tavérnas tavernón
Table 24: Example of stress fixed on final syllable in the Greek genitive plural
133
Node [2] represents inflection classes which likewise have fixed genitive plural stress, but for these examples it is fixed on the penultimate syllable.
‘friend’ SINGULAR PLURAL NOM φύλακας φύλακες fílakas fílakes ACC φύλακα φύλακες fílaka fílakes GEN φύλακα φυλάκων fílaka filákon
Table 25: Example of stress fixed on penultimate syllable in the Greek genitive plural
Node [3] also represents inflection classes with final stress, but includes the generalization that the genitive singular and genitive plural have stress fixed on the same syllable. Note that within the structure given above, this must be considered an override of default inheritance, because penultimate stress is the expected inheritance at this node.
‘boy’ SINGULAR PLURAL NOM αγόρι αγόρια aγóri aγórja ACC αγόρι αγόρια aγóri aγórja GEN αγοριού αγοριών aγorjú aγorjón
Table 26: Example of fixed stress on final syllable in the Greek genitive singular and genitive plural
134
Node [4] represents the stress pattern in which stress is penultimate in both the genitive singular and genitive plural. The similarities between nodes [3] and [4] is captured by their representation as sisters in the hierarchy.
‘face’ SINGULAR PLURAL NOM πρόσωπο πρόσωπα prósopo prósopa ACC πρόσωπο πρόσωπα prósopo prósopa GEN προσώπου προσώπων prosópu prosópon
Table 27: Example of fixed stress on penultimate syllable in the Greek genitive singular and genitive plural
Finally, node [5] is likewise similar, except that the accusative plural also has fixed stress, on the same syllable as in the genitive singular and genitive plural.
‘man’ SINGULAR PLURAL NOM άνθρωπος άνθρωποι ánθropos ánθropi ACC άνθρωπο ανθρώπους ánθropo anθrópus GEN ανθρώπου ανθρώπων anθrópu anθrópon
Table 28: Example of fixed stress on the penultimate syllable in the Greek genitive singular, genitive plural, and accusative plural
135
This network of stem information captures virtually all cases of fixed stress in the Greek nominal inflection system as well as the implicational relationship between accusative plural fixed stress, genitive singular fixed stress, and genitive plural fixed stress. It is not intended to capture columnar stress directly; I assume that stems are specified for stress placement, and that stem stress appears as a default where no other stress pattern is specified for a given inflectional property set (e.g. genitive plural) and stem class.
3.6.1.1.1. A comparison to Revithiadou (1988)
Revithiadou (1998) provides a different theoretical treatment of Greek nominal stress.
She argues that differences in genitive plural stress placement indicate a difference in the inherent lexical marking of roots. In her approach, roots may be lexically marked for stress, unmarked, or marked as unaccentable. Roots marked as unaccentable trigger stress on the inflectional suffix (word final stress). Suffixes may be unmarked or marked for stress, including a specification that the stress fall on the syllable before the suffix.
Stems which are lexically marked for stress or as unstressable have columnar stress throughout the paradigm, because stem-specified stress trumps stress specified by an inflectional suffix. Suffixal stress marking dominates in the absence of stem stress. If neither the root nor the suffix carries stress, the default realization of the stress pattern appears, meaning antepenultimate stress. Nominative singular, nominative plural, and accusative singular suffixes are always unmarked. Genitive plural is always marked.
There are some similarities, and some differences, between the two treatments.
Like Revithiadou, I assume that specification of stress placement is an inherent feature of
136
stems, and that stress in the genitive plural, accusative plural, and genitive singular is sometimes, but not always, governed by a separate generalization. My account differs, however, in that I assume that all stems carry a specification for stress placement, and that the stress pattern inherited from the stress hierarchy overrides stress which is lexically marked on the stem.
Revithiadou is virtually forced into the opposite conclusion because of assumptions rooted in her item-and-arrangement approach to inflection. First, form and meaning are bundled into morphemes, and inflectional meaning is added to a word by adding the relevant morpheme. Second, the number of inflectional morphemes should be kept to a minimum. The result is that she must assume that a genitive plural morpheme specifies word-final stress and that when the inflectional properties {genitive plural} are added to a stem, the stress pattern is also added. This leaves Revithiadou with the problem of explaining why genitive plural stress is not always word-final. In order to produce columnar stress in some instances, she must posit that root marking for stress is dominant over suffix marking for stress. And in order to explain why some words do have word-final stress in the genitive plural, she must further argue that not all roots are marked for stress.
By contrast, if we take a Word and Paradigm approach to inflection, in which information is not added by morphemes but is instead licensed by processes applying on stems, we can remove the assumption that word final stress is expected for all genitive plurals. There is nothing to be gained by that assumption because inflectional properties pre-exist the forms which realize them, and economy of representation is not typically valued. Changing theoretical assumptions in this way further nullifies the need to assume
137
that stem marking is dominant, and that only some stems are lexically marked for stress.
The result is my argument that all stems are lexically marked for stress, and that in the process of inflection, marking inherited through the stress hierarchy dominates by virtue of having a more specific conditioning environment than does stem stress, which is relevant to all inflected forms.
As far as I can tell, the two theories cover similar empirical ground, with the difference being that my approach captures the generalization that there is an implicational relationship between the accusative plural, the genitive singular, and the genitive plural. By treating inflectional morphemes as isolated lexical entries,
Revithiadou’s theory makes these correspondences seem accidental. Better empirical coverage motivates the postulation that stress patterns and segmental patterns operate independently of each other, defining separate hierarchies.
3.6.1.2. The combined inheritance hierarchy for plurals
Another dimension of the hierarchy carries information about plural segmental processes, as represented in Figure 14.60 By overlaying the stress and segmental dimensions of the
60 In terms of a theoretical treatment, there seem to be two possible approaches to representing the relationship between singular and plural forms. One possibility is to treat singular and plural forms as heteroclite, following Stump (2006). This amounts to treating stems as suppletive; the singular stem would belong to one inflection class, and the plural stem would belong to another. The singular and the plural would each have their own inheritance hierarchies, and be united by being linked to cells in a single lexeme paradigm. Under this approach, the singular stems of µανάβης / manávis and περίπλους / períplus would belong to the same inflection class, but the plural stems would belong to different inflection classes. Another possibility is that one inflection class could be deemed basic, and the source of inheritance for the other classes. In this approach, περίπλους / períplus and µανάβης / manávis do not belong to the same inflection class, even in the singular, but they inherit their singular specifications from the same source, thereby linking them. This approach utilizes a version of rules of referral. There is no clear reason to choose one approach over the other. The former is problematic because it posits suppletive stems for most words which are nonetheless usually homophonous. The second approach is also problematic; it forces one inflection class to be considered primary, without there being any evidence for such a status, as the target of a rule of referral. In the end, I posit the former position because 138
hierarchy, the ways in which stress and segmental information intersect become apparent.
This allows the model to capture that two separate morphological generalizations determine inflectional class, without having to relegate stress information to a subdivision of segmentally-determined inflection classes, losing the similarities and differences in stress patterns across these classes. An inflection class inherits along both paths.61
In a multidimensional inheritance hierarchy, Greek nouns may be represented as follows. The segmental hierarchy includes only divides classes according to plural formation, so as not to clutter the picture, but the stress hierarchy includes information on the singular and the plural, since this hierarchy is not complicated to begin with. Note that only together do the hierarchies carry the information necessary for a full representation of any particular stem’s inflection class.
it simplifies the representation to be able to remove the singular forms, but this is an issue of convenience more than a theoretical claim. 61 This proposal is in the spirit of redundancy rules (Jackendoff 1975). Redundancy rules do not serve a primary role in the generation of word forms, but capture the degree to which any particular inflection class is independent of others. In my proposal, this function of generalizing patterns over disparate classes is carried by the multidimensionality of the inheritance hierarchy. 139
{plural, ...ων}
-ων nom pl ∪ acc pl -οι, -ους, -ων
-Vς, -Vς, -ων -V, -V, -ων -τα, -τα, -των
-ες, -ες, -ων -εις, -εις, -ων -η, -η, -ών -α, -α, -ων
140 -δες -ες -εις -εις -δες -ες -εις -εις -δων -ων -εων -ων
-δες -ηδες -δες -ηδες -δων -ηδων
Figure 14: A default inheritance hierarchy for segmental inflection of plural Greek nouns
genitive plural
...σσ⁄# ...σ⁄σ# plural, gpl = ...ων ...σ⁄σ# ...σ⁄σ#, gpl ∪ gsg
...σ⁄σ# -ων ...σσ⁄# ...σ⁄σ# gsg ∪ gpl nom pl ∪ acc pl gsg ∪ gpl gsg ∪ gpl accpl ∪ gpl
-Vς, -Vς, -ων -V, -V, -ων -τα, -τα, ‘-των -οι, -ους, -ων
-ες, -ες, -ων -εις, -εις, -ων -η, -η, -ών -α, -α, -ων µέλλον- -οι -οι όνοµα- -ους ‘-ους
141 µέρ- ηρέα- -ων ‘-ων
-δες -ες -εις -εις -α -α -α περίπλ- -δες -ες -εις -εις -α -α -α αντίλαλ- άνθρωπ- -δων -ων -εων -ων -ων ‘-ων -ών
σύγγεν- πρόσωπ- -δες ‘-δες -ηδες -ες -ες -ες ‘-εις -είς σίδερ- αγόρ- -δες ‘-δες -ηδες -ες -ες -ες ‘-εις -είς -δων ‘-δων -ηδων -ων ‘-ων -ών ‘-εων -έων
µανάβη- κάλφα- κόντ- φύλακ- δύναµ- καφέ- πατέρ- τουρίστ- δέκαν- παππού- µητέρ- ναύτ- µαµά- ώρ- αλεπού- νίκ- Figure 15: A multidimensional default inheritance hierarchy for plural inflection of Greek nouns
3.6.2. Conflicting stress inheritance and the paradigmatic dimension
The hierarchy in Figure 15 provides all of the information needed to represent paradigmatic gaps resulting from competing inflectional stress patterns.
Ninety-eight percent of the paradigmatic gaps in Modern Greek appear in the inflection classes represented by the shaded cells in Table 23 (repeated as node [1] below). By traditional classifications, the inflection class which includes µητέρα / mitéra
‘mother’ and ώρα / óra ‘hour’ contains 1,841 genitive plural gaps (appx. 17% of all nouns in this class). The class represented by νίκη / níki ‘victory’ contains 192 genitive plural gaps (appx. 12% of all nouns of this class), and the class represented by πατέρας / patéras ‘father’, φύλακας / fílakas ‘friend’, and τουρίστας / turístas ‘tourist’ includes 57 genitive plural gaps (appx. 7% of all nouns in the class). Finally, the class represented by
ναύτης / náftis ‘sailor’ includes 12 gaps (appx. 1% of all nouns of the same class). This area of the lexicon is most worth investigating in detail, and is repeated in Figure 16.
142
genitive plural fixed
[2] gpl = ...σσ⁄# gpl = ...σ⁄σ#
[3] [1] gpl = ...σ⁄σ# gpl = ...σ⁄σ# -ες , -ες, -ων gpl ∪ gsg
[4] [5] [6] [7] [8] -ες -ες -ες -η, -η, -ών -τα, -τα, ‘-των -ες -ες -ες -ων ‘-ων -ών µέρ- µέλλον- όνοµα- φύλακ- ηρέα- πατέρ- τουρίστ- µητέρ- ναύτ- ώρ- νίκ-
Figure 16: Inheritance hierarchy for the classes in which 98% of Greek genitive plural gaps occur
Albright argued that morphophonological variation within inflection classes is the cause of gaps. By contrast, I argue that the key issue is the predictability of inflection class membership, which turns out to be subtly different from variation. The more unique a particular inflectional marker is to a particular inflection class, the better a word form that includes that marker is as a predictor of all other forms (principal part).
Principal parts are thus a stabilizing force within an inflectional paradigm. On the other hand, if an inflectional marker is shared by multiple classes, it is a bad predictor. Bad predictors set the conditions for analogy and other changes by virtue of connecting inflectional paradigms along one or more parameters, increasing the likelihood that 143
speakers will make connections along other parameters as well. Bad predictors entail grammar competition by their very nature.
I hypothesize that paradigmatic gaps are clustered under the nodes labeled [1] in
Figure 16 because within these paradigms there is no form which can predict all other forms of the paradigm. No fewer than three word forms of a given lexeme must be known in order to fully predict the inflected forms of a lexeme for each of the three classes under [1] – the genitive plural, another plural form, and a singular form. This many forms is necessary because of the lack of correspondence between singular segmental patterns, plural segmental patterns, and stress patterns. Within a default inheritance hierarchy, this is represented as the lack of a one-to-one linkage between dimensions. Moreover, the genitive plural by itself is a particularly bad predictor; the paradigms under nodes [5] and [6] overlap in the genitive plural with other paradigms along both the segmental dimension and the stress dimension. This lack of both predictiveness and predictability makes the genitive plural a particularly weak spot in the inflectional paradigm of these words.
The importance of predictability rather than simply variability can be seen by comparing the fragment of the inheritance hierarchy in Figure 16 with another fragment of the grammar which is similar in structure, in Figure 17. In both cases the segmental characteristics of the inflection class are crosscut by three stress patterns. A grammar competition analysis based on variability predicts that these classes would be equally likely to develop gaps as those in Figure 16. However, the inflection classes represented under node [1] in Figure 17 contain only twenty-four gaps out of a total of 5,234 nouns
144
listed in Λεξικό της Κοινής Νεοελληνικής / Leksikό tis Koinís Neoellinikís
(Triantafillidis 1998).
...σ⁄σ#, gpl ∪ gsg
[2] ...σ⁄σ# ...σσ⁄# ...σ⁄σ# gsg ∪ gpl [1] gsg ∪ gpl gsg ∪ gpl accpl ∪ gpl -α, -α, -ων
-α -α -α -α -α -α -ων ‘-ων -ών
σίδερ- αγόρ- πρόσωπ-
Figure 17: A similar inheritance hierarchy, but without a significant number of gaps
The difference between these two areas of the lexicon lies in the degree to which the genitive plural is isolated within the paradigm. In Figure 17, the genitive plural is predictable from the genitive singular. The merger of the genitive singular and the genitive plural along the stress dimension entails that whatever syllable stress is on in the genitive singular, it is also on that syllable in the genitive plural. This means that the relevant stress nodes are better predictors for these classes than stress is for the classes where most gaps are found. Also, recall from Table 23 that for the classes in Figure 17 there is a one-to-one correspondence between the singular and plural segmental patterns.
Outside of the genitive plural, there is no overlap with other paradigms along the
145
segmental dimension. This entails that the genitive singular form is sufficient to predict not only the genitive plural form, but all word forms of a given lexeme. The genitive singular serves as a principal part. Greater predictability means less paradigmatic competition, and may mean less motivation for the appearance of gaps.
One conclusion that we can draw from this is that having type 3 stress is a necessary but not sufficient condition for describing the causes and distribution of gaps.
It is not variability within the paradigm which is specifically at issue, but rather the predictability of a particular variant. In the inflection classes with the highest number of gaps, predictability of the genitive plural stress pattern is low due to overlapping paradigms along all three dimensions of the inheritance hierarchy. In the inflection classes represented in Figure 17 there is an equal amount of variation internal to the inflection class if we define the class according to segmental information, as is traditionally done. However, the correlation between the singular and plural segments, and the paradigmatic connections in the stress hierarchy entail that despite this variation, the genitive plural form is far more predictable. These correspondences mean less paradigmatic competition.
Defining competition at the level of the paradigm, and in terms of predictability rather than variability takes full account of the data. For example, the lack of gaps in words with final stress throughout the paradigm makes sense under this view. The genitive plural form of these words is always predictable based on any other inflected form of the same lexeme, regardless of inflection class membership. And predictability is an inherently paradigmatic concept. Variability is not.
146
3.6.3. Conclusions
Hudson argued that in the case of *amn’t, inheritance of conflicting morphosyntactic property sets causes the gap. Grammar competition need not be as literal as the inheritance of contradictory property sets, however. If we think of language as a series of surface forms from which speakers interpret the underlying grammatical principles and patterns (and how could it be any other way...), the possibility of interpreting multiple patterns from a single inflected word form is in and of itself competition. Hence, I call this paradigmatic competition. A default inheritance hierarchy formalizes this generalization; any two nodes which share a mother node compete along some parameter.
Albright’s (2003) model of Spanish and my model of Greek differ in two crucial respects. First, in order to explain the distribution of the Greek gaps, I make reference to other inflected forms, and consider competition to operate paradigmatically. Albright only considers competition between individual forms. Second, in my view the strength of the competition, and thus the chance that a paradigmatic gap will result, rests in the degree to which the form representing the point of contact between two paradigms is predictable from other paradigmatic forms. Albright assumes that variability is the same as predictability.
Above I present two groups of inflection classes. Within each group, there are three stress patterns in the genitive plural, so we might expect that these groups would be equally likely to develop paradigmatic gaps. The fact that they do not can be attributed to the predictability of stress variants in one case but not the other, if a paradigmatic view is taken. Both have equal levels of variability. The reliance of gaps on paradigmatic structure thus forms part of a larger claim that for inflectional morphology, the paradigm
147
is crucial to being able to make the interpretive leap from the surface form to the underlying patterns.
Note that if paradigmatic competition actively causes paradigmatic gaps in the genitive plural, no other grammatical structure is needed to account for the gaps. As
Albright claimed, if the theory incorporates the degree and nature of overlap between inflectional categories, i.e. the paradigmatic competition, gaps become epiphenomena of the competition itself. In combination with semantic function, frequency, register, language ideology, and other factors known to influence genitive plural usage in Modern
Greek, we can describe the appearance of the paradigmatic gaps in a motivated, generalizable way.
3.7. Summary
In this chapter I have presented a variety of evidence which suggests that paradigmatic gaps in the genitive plural of Modern Greek nouns are synchronically and actively caused by speakers’ reactions to competing stress patterns in that cell. Paradigmatic gaps present a skewed distribution; gaps appear in the (segmentally defined) inflection classes with the greatest variability in the stress pattern, and within those classes, in the forms which are least predictable from other paradigmatic information. Via a forced choice task I demonstrated that speakers prefer periphrastic constructions when the synthetic genitive plural form has a potential stress shift and is low frequency, but prefer are more likely to prefer the synthetic genitive plural form otherwise. This data suggests that lack of stress predictability causes Greek speakers to be insecure about their use of the genitive plural
148
and triggers an avoidance strategy. A forced production and ratings task confirmed that speakers’ confidence levels are closely tied to stress patterns.
I presented an analysis in terms of a default inheritance hierarchy. In contrast to previous accounts of Greek stress and traditional descriptions, I argue that information about the singular and the plural should be treated separately, as should stress, because there is often not a strong correlation along these three dimensions. The stress and segmental hierarchies combine to determine the inflectional class of a particular stem.
Separated in this way, it became apparent that the issue is not stress variation per se, but rather the predictability of a particular stress variant. Within some classes the genitive singular serves as a principal part, from which all other inflected forms and the inflection class generally may be inferred. Within other classes with equal amounts of stress variation this is not true. The genitive plural may not be predicted from nor is predictive of other forms of the same class. It is in the latter group of inflection classes that the overwhelming majority of gaps fall. This also explains the disproportionately low number of gaps among stems with word-final stress. In these cases, the genitive plural form may be predicted from any other form, regardless of inflection class membership.
Ultimately, it seems that paradigmatic competition, in combination with the ideological issues that speakers impose onto grammatical structure, actively causes paradigmatic gaps in the Modern Greek genitive plural, and that gaps may be treated as epiphenomenal to the competition itself.
149
CHAPTER 4
THE MODERN GREEK GENITIVE PLURAL REVISITED:
HOW GAPS ARE NOT THE EPIPHENOMENAL RESULT OF
PARADIGMATIC STRUCTURE
4.1. Introduction
In CHAPTER 3 I explored the connection between stress predictability and genitive plural gaps. A variety of evidence pointed to the same conclusion – gaps are actively caused by speakers’ reactions to paradigmatic competition. In response, I formulated an account which provided for paradigmatic competition, but made no explicit provision for gaps, treating them instead as the end product of speakers’ gradient insecurity over the competition and subsequent avoidance strategies.
In this chapter, I present one additional piece of evidence which suggests that the correct analysis for Greek is actually more complicated than initial indications suggested.
I show that the uncertainty that Greek speakers report for gaps, unlike their ratings for productive genitive plural inflections, is not (entirely) correlated with stress patterns.
From this I argue that gaps are (sometimes) phenomena unto themselves, and not simply epiphenomenal to other structures. The gaps represent defective paradigm classes. This does not mean that the reader can now discard the analysis and conclusions in CHAPTER
150
3, however. Both analyses make sense, and in fact complement each other, if we look at the data from the viewpoint of what happens in morphological change.
4.2. Confidence ratings and stress variability in the genitive plural, revisited
Returning to the forced production task presented in Section 3.5., I made the following predictions for the behavior of productive nouns:
• Speakers’ confidence in their productions of genitive plural forms will directly
correlate with the degree to which speakers agree on whether a given genitive
plural form has a stress shift (agreement over the form of a word being taken as a
measure of grammar competition; higher agreement = lower competition).
• Speakers’ confidence in their productions of genitive plural forms will be
inversely correlated with lexeme familiarity.
While there was a surprising lack of correlation between lexeme familiarity and speakers’ confidence ratings, the first prediction was upheld. This seemed to provide the necessary link between the distribution of the morphophonological alternation on the one side, and preference for periphrastic constructions (i.e. avoidance strategies) and paradigmatic gaps on the other.
If gaps are epiphenomenal to speakers’ reactions to paradigmatic competition, paradigmatic gaps should pattern similarly to the productive words. We can thus add two more predictions:
151
• In forced production, agreement rates for non-gaps will vary, with some words
leading to high agreement levels, and some words to low agreement levels. Gaps,
by contrast, should cluster in the low range.
• Gaps and non-gaps should lead to equivalent confidence ratings.
Surprisingly, these predictions do not hold up.
As I presented in the previous chapter, stress agreement rates for productive genitive plural forms do vary, but when we overlay the data for gaps (Figure 18) it is clear that the agreement rates for gaps cover the same range. If gaps are the synchronic result of paradigmatic competition, why are there a substantial number of words for which speakers largely or completely agree on the form which would be used if such a form existed?
152
0.6 0.8 1.0 1.2 1.4 Average Confidence Ratio (gpl form: nom form) nom ConfidenceAverage Ratio (gpl form: square=gap circle=no gap
0.5 0.6 0.7 0.8 0.9 1.0
Interspeaker Agreement about Stress Shift (0.5 = 50/50 split)
Figure 18: Effect of interspeaker agreement on production confidence, differences between gaps and productive inflection
Even more importantly, subjects were on average less confident in their productions of gaps than in their productions of normally inflected words within the same agreement range. In the lowest agreement range, the gaps and nongaps are interspersed, but in the highest agreement range, they form separate groupings (with one anomaly on
153
each side). The average difference in this range was almost 14%. And the effect of agreement on confidence is much greater for nouns with productive genitive plural forms
(p<0.001) than for nouns with genitive plural gaps (p<0.05).62
The independence of gaps and stress can be seen even more clearly if we measure whether speakers produced a particular word form with a stress shift or not, rather than whether speakers agreed on whether there is a stress shift or not. Among genitive plural forms without a stress shift (right side of graph below), speakers had approximately 25% less confidence in their productions of gaps than of non-gaps relative to nominative forms of the same lexemes. One gap was an outlier in this regard. In a grammar competition model, the difference between gaps and non-gaps is unexpected.
62 productive inflection: confidence = 0.484 * agreement + 0.4885. R2 = 0.6908, p<0.001. gaps: confidence = 0.3105 * agreement + 0.5235. R2 = 0.3507, p<0.05. 154
0.6 0.8 1.0 1.2 1.4 Average Confidence Ratio (gpl form: nom form) nom ConfidenceAverage Ratio (gpl form: square=gap circle=no gap
0.0 0.2 0.4 0.6 0.8 1.0
% NO Stress Shift (0 = all shifting, 1 = all non-shifting)
Figure 19: Effect of stress variability on production confidence, differences between gaps and productive inflection
Moreover, there is no statistical effect for stress placement on confidence ratings for gaps
(R2 = 0.0429, p=0.4584), but a strong effect for lexemes with productive genitive plural forms (R2 = 0.8054, p<0.001).
155
In a multiple linear regression model in which confidence levels are the dependent variable and lexeme familiarity, stress agreement, and whether a word has a gap are the independent variables, both agreement and defectiveness are significant predictors of confidence (p<0.001 for agreement and defectiveness, p=0.637 for familiarity).63 In another model replacing stress agreement with percentage of stress shift, defectiveness has the stronger effect (p<0.001), and stress the weaker one (p<0.01). These facts thus suggest that stress alone cannot explain speakers’ reactions to gaps, and gaps also have their own, independent effect on confidence.
Before fully submitting to this conclusion, however, there are at least two related factors through which we might seek to explain this difference in the behavior of gaps and non-gaps – lexeme familiarity and validity of the classification into defective and non-defective words. As it turns out, neither of these possibilities pans out.
First, this difference in confidence ratings cannot be attributed to lexeme familiarity. In the following graph a score of three represents the rating “I know the meaning of this word well, but don’t use it,” four represents “I have used this word once or twice,” five represents “I use this word sometimes,” and six corresponds to “I use this word frequently.”64
63 Multiple linear regression formula 1: confidence = -0.012*familiarity – 0.11*defectiveness + 0.386*agreement + 0.8932. R2 = 0.7032, p<0.001. Multiple linear regression formula 2: confidence = -0.022*familiarity – 0.123*defectiveness + 0.16895*stress + 0.8932. R2 = 0.5973, p<0.001. 64 A given subject’s confidence in a given genitive plural form is represented as a ratio of that person’s average confidence score across all genitive plural forms. This ratio is then averaged across speakers (N=11) to produce one point per item. If subjects had equal confidence in all genitive plural forms, the points in the following graph would cluster around the horizontal line y=1.
156
square = gap Average Confidence Ratio (gpl form: all gplConfidence forms) Average Ratioform: (gpl circle = no gap 0.6 0.8 1.0 1.2 1.4
3.0 3.5 4.0 4.5 5.0 5.5 6.0
Average Familiarity Rating (6=high)
Figure 20: Lack of effect of lexeme familiarity on confidence for genitive plural forms
As the graph shows, gaps and normal genitive plural forms have comparable levels of familiarity,65 perhaps with the exception of one low value for a normally inflected genitive plural form. If lexeme familiarity were responsible for the discrepancy in confidence ratios between gaps and productive word forms, I would expect speakers to
65 In a two-sample difference of the means test, |t| = 1.04, df = 25.2, p = 0.31. 157
have reported less familiarity with lexemes containing gaps than with lexemes with productive inflections.66
Another possibility is that the purported gaps are not in fact gaps. The partial overlap in confidence ratings between the gaps and regular nouns, and especially the one gap encircled by a ring of non-gaps in Figure 18 and Figure 19 (κουλτούρα / kultúra
‘culture’, average agreement = 0.971, gpl:nom confidence ratio = 0.967) should raise suspicions that the lexemes which dictionaries cite as having gaps are not perceived as gaps by the speakers themselves.
On this point, some fortuitous and colloquial data is informative. In the process of collecting data for this forced production and ratings survey, two informants refused to provide certain forms for certain words. As they explained to me, they were concerned that as a non-native (and very poor) speaker of Greek, I perhaps was not aware that some of the words in my survey could not be used in the way the sentence required. They helpfully left these questions blank so that I could identify the flaws in my questionnaire.
(Apparently the number of gaps was small enough that it did not raise their suspicions that defective words were exactly the point of the questionnaire.) In my discussions with these two people, it seemed that the “flaws” were not problems related to semantics, or due to ungrammatical frames or other obvious methodological problems. These subjects furthermore admitted that the genitive singular and the non-genitive plural forms of the
66 Lexeme familiarity is a crude measure. For example, it is common to consider the frequency not of all inflected forms combined, but rather of a particular inflected form in comparison to other inflected forms of the same lexeme. However, in the case of paradigmatic gaps, that path leads to circularity. If there is a correlation between relative non-use of the genitive plural form and confidence ratings, this tells us nothing more than to say that there is a correlation between gaps and confidence ratings because gaps, by definition, are inflected forms which are used at well below the expected level. 158
relevant words are used.67 The words simply are not used in the form that I was asking for. The fortuity of this failure of the survey is obvious. These informants identified exactly which words, in their opinion, have paradigmatic gaps.
All of the gaps that they identified were in genitive plurals. Subject 11 left sixteen of the forty genitive plural frames blank – three of the ten control items
(παπουτσής / paputsís ‘shoe maker’, βαγενάς / vaγenás ‘barrel maker’, κορβανά / korvaná
‘coffer’), two of the fifteen items in the same classes as gaps but prescriptively normally inflected (χαράδρα / xaráđra ‘ravine’, ταβέρνα / tavérna ‘pub’) and eleven of the fifteen prescriptive gaps (ροτόντα / rotó(n)da ‘rotunda’, µουρµούρα / murmúra ‘grumbling’,
κουλτούρα / kultúra ‘culture’, καµπάνια / ka(m)bánia ‘campaign’, νουβέλα / nuvéla
‘novella’, καρότσα / karótsa ‘coach’, κολόνια / kolónja ‘cologne’, σαµπάνια / sa(m)bánja
‘champagne’, λακκούβα / lakúva ‘pothole’, σακούλα / sakúla ‘purse’, κοπέλα / kopéla
‘girl’) Subject 28 left ten genitive plural frames blank – two of the fifteen items in the same classes as gaps (πιτζάµα / pidzáma ‘pajamas’, καρέκλα / karékla ‘doll’), and eight of the fifteen items which prescriptively have genitive plural gaps (πραµάτεια / pramátja
‘merchandise’, ροτόντα, κουλτούρα, καµπάνια, νουβέλα, καρότσα, κολόνια, σαµπάνια).
Note that both of these Greek speakers specifically cited κουλτούρα as a gap. While the overlap between the prescribed gaps and the words that these two informants rejected is not absolute, there clearly is validity to the status of the prescribed gaps as actual gaps in people’s speech. Of the fifteen, these two subjects listed twelve as problematic forms.
67 βαγενάς / vaγenás and κορβανά / korvaná are possible exceptions. They received the lowest ratings on the familiarity pretest, and were outliers in this respect. Subjects seemed generally unhappy with the words, even apart from the status of the genitive plural. Thus, familiarity might have contributed to one subject’s refusal to make genitive plural forms. 159
We therefore cannot explain away any unexpected behavior as a widespread mismatch between prescribed and actual patterns.
Ultimately, I conclude that there is a correlation between defectiveness and confidence ratings which cannot be explained as the independent result of the representation of stress in the inflectional system, or any other obvious factor.
4.3. Reconciling opposing conclusions through principles of language change
This data presents an apparent conundrum for the model presented in Section 3.6. Based on the bulk of the data, the best conclusion is that speakers actively avoid stress unpredictability, but based on the one piece of data in this chapter, the opposite conclusion suggests itself. In this section I argue that the contradictions are more apparent than real, if we consider the nature of morphological systems, and principles of language change.
At issue are the structures which precipitate change, and the sorts of residue that changes leave in the grammar of a language. I formulate these as two principles, neither of which is particularly controversial.
Principle 1: Reanalysis precedes restructuring. Restructuring reveals reanalysis.
To demonstrate this point, consider the following example of periphrasis from Sanskrit.68
In Sanskrit the first person future may be realized either by a synthetic form (Table 29) or by what has traditionally been called a periphrastic form (Table 30).
68 Periphrasis is the use of a multi-word phrase to express a set of morphosyntactic properties, despite that set normally having one or more synthetic realizations in the language. 160
SINGULAR PLURAL dā-syā-mi dā-syā-mas 1ST P give-FUT-1SG give-FUT-1PL ‘I will give.’ ‘We will give.’
dā-sya-ti dā-sya-nti 3RD P give-FUT-3SG give-FUT-3PL ‘He will give.’ ‘They will give.’
Table 29: Selected future forms for the Sanskrit verb 'give' -- synthetic forms in 1st person
SINGULAR PLURAL dātā asmi dātā smas 1ST P giver.MASC-SG COPULA.1SG giver.MASC-SG COPULA.1PL ‘I will give.’ ‘We will give.’
dātā dātāras 3RD P giver.MASC-SG giver.MASC-PL ‘He will give.’ ‘They will give.’
Table 30: Selected future forms for the Sanskrit verb 'give' -- "periphrastic" forms in 1st person
The forms in Table 29 are unquestionably single words. Note, however, that in Table 30 the first person plural form dātā smas inflects for plurality on only one of the two words.
Plural marking on both parts would be expected based upon the third person plural form dātāras. The generalization is that when the copula is present, there is no number agreement on the form for ‘giver’, but when the copula is absent, there is number agreement on that form. Contrary to its description as a periphrastic construction, dātā 161
smas best meets the criteria of being a single word unit. Distributed exponence is a typical characteristic of morphological phenomena (Zwicky 1990), so if dātā smas is treated as a single word, plural marking on only smas and not also on dātā is to be expected. If we were to treat the first person forms in Table 30 as syntactically generated constructions, we would have difficulty capturing that plural marking applies to only one form, in light of marking on third person forms.
This example is one of many which could be used to demonstrate that change reveals prior reanalysis. At some point in the history of the language, Sanskrit speakers must have reanalyzed the periphrasis as a single word rather than a phrasal construction.
We can assume that speakers induce changes which make morphological structures more canonically morphological, syntactic structures more canonically syntactic, etc.
Logically, then, the reanalysis must have occurred prior to the time at which speakers stopped inflecting both of the component parts for plurality. This in turn opened the constructions up to the loss of inflectional marking, making the structure more typically morphological. However, we only have evidence of such a reanalysis when the subsequent change occurs.
Principle 2: Morphological change tends to leave behind remnants which are no longer motivated by the synchronic structure of the language.
For example, there is a tendency for irregular verbs to have high token frequency. In many cases, these verbs represent previously productive patterns; their irregularity in the modern language is the result of their having been left behind when lower frequency
162
verbs underwent analogical change (Hock 1991). If the number of stems belonging to that inflection type is sufficiently reduced, the pattern ceases to be an active product of the generative system. And again, the fact that high frequency verbs tend to be resistant to analogical change shows us that the structure of those forms differed from the structure of forms with lower frequency in a way which was not obvious until the later change happened.
These principles help to explain the apparent discrepancy in the Greek data. They tell us that structures which appear to be productive in a language, including productively derived defectiveness, may or may not in fact be as they appear. The data presented in this chapter hints that a (partial?) reanalysis of the Greek gaps has taken place. The gaps have gone from being synchronically motivated by the stress pattern to being at least in part treated as idiosyncratic phenomena. This reanalysis is then revealed by speakers’ tendency to level stress, which strands the gaps.
There are three facts which indicate that Greek may currently be undergoing a stress leveling in the genitive plural of nouns. First, the Greek language has a history of stress leveling in adjectives. Adjectives in Modern Greek inflect for case and number, just as nouns do, and additionally inflect for gender. Adjectives and nouns used to have parallel stress shifts towards the end of the word in the genitive plural and sometimes in the genitive singular and accusative plural as well.69 However, stress in adjectives has mostly been leveled in favor of the predominant (non-genitive) pattern (Holton et al.
1997, Jannaris 1987). This loss of distinctive genitive plural stress for adjectives has been completed at least since the nineteenth century (Brian Joseph, p.c.) and contrasts with the
69 There is also a parallelism between segmental forms. 163
pattern for nouns which, for whatever reason, by comparison remained close to the patterns inherited from Ancient Greek.
Second, remember that subjects in the production task did not always mark stress in the nouns according to prescriptive patterns. Of the eight nouns expected to show a stress shift in the genitive plural according the dictionary Λεξικό της Κοινής
Νεοελληνικής / Leksikό tis Koinís Neoellinikís (Triantafillidis 1998), only five did to any significant degree, and only one noun was consistently produced (<90%) with a stress shift. For the remaining three, subjects largely agreed that the lexemes respective genitive plural forms did not have stress shifts. Considering that there were no cases of speakers claiming that prescriptively non-shifting nouns did have a stress shift, the tendency is clearly for stress to be columnar more often than dictionaries describe. This is not definitive proof of an ongoing stress leveling. After all, semi-prescriptive sources like dictionaries do not necessarily accurately reflect usage – current or previous. Still, this pattern should make us wonder whether there is an ongoing leveling of the stress pattern in nouns, especially in light of what has already happened for adjectives.
Finally, this is a likely time for a stress shift from a sociolinguistic standpoint.
Katharevousa, the formerly high diglossic variety of Modern Greek, emphasized forms closer to the pattern of Ancient Greek, and was taught to all schoolchildren. As suggested in CHAPTER 3, this may have had a stabilizing influence on the genitive plural form. All of the subjects in this study, however, were educated (and mostly born) after the end of the diglossic era. While we can never predict language change with any certainty, the raising of the status of the low variant in 1976 makes the generation that my subjects represent a likely one to promulgate a stress leveling. Together, these facts make
164
it possible, perhaps even likely, that some of the genitive plural forms which subjects listed with columnar stress previously had shifting stress.
The key question is how stress leveling could reveal the status of gaps. The problems presented by the data in Figure 18 and Figure 19 are twofold. First, there is a weaker or non-existent correlation between stress agreement/ stress shift and confidence ratings for genitive plural gaps, despite a strong effect in words with productive genitive plural forms. Second, if the genitive plural gaps result from stress unpredictability, forced production should lead to low interspeaker agreement about stress placement. Yet most of the gaps cluster in the high agreement range, and these high-agreement forms are universally non-stress-shifting.
Imagine first that the genitive plural gaps are the synchronic result of shifting stress, and that there is an ongoing leveling of stress. The theory would predict that if stress is leveled, it becomes fully predictable, and as a result the gaps should disappear.
The new data is clearly a problem.
Now imagine that the genitive plural gaps have become frozen entities. At one time they were the synchronic, active result of speakers’ avoidance strategies, but either in part or in whole, the stems with genitive plural gaps have become detached from that motivation. This entails a narrowing of the generalization that speakers are making from one which sweeps through the entire class and probabilistically identifies gaps based on paradigmatic competition and other factors to one which lexically marks a particular stem paradigm cell as defective. If gaps were reanalyzed by speakers in this way, there may be no empirical evidence of the reanalysis at the time it occurs, but I would make the opposite prediction with regard to a subsequent stress leveling. If gaps are their own
165
phenomena, not the byproducts of the paradigmatic stress pattern, I would expect them to be maintained despite a widespread leveling. Stress leveling is like the rise of distributed exponence in Sanskrit (Table 30). It reveals the reanalysis that was there already.
Moreover, I would expect that if Greek speakers are forced to produce a form with a gap, they would analogical derive the most likely form, which would often include non- shifting stress. The second postulation more closely matches all of the available data.
Thus, we seem to have some way to explain both the distribution of the gaps, which coincides so well with the presence of a variable morphophonological alternation, and its deviance from that pattern. Not all words need be equally motivated or unmotivated, and a shift to gaps being lexically-specified information does not negate evidence of avoidance strategies and the distributional pattern. Avoidance strategies and distributional evidence demonstrate that there is necessarily a historical connection (or stated differently, that at a point in history there was a synchronic connection), but only possibly a current connection.
This changes the conclusions regarding a grammar competition model. Albright
(2003) made the following postulation about the relationship between speaker confidence and paradigmatic gaps (in Spanish):
grammar competition → speaker insecurity → paradigmatic gap
This may be a correct path of causation for Spanish, and it seems to be the correct basic hypothesis for Modern Greek for at least some speakers at some point in time. However, if we include the data presented in this chapter, this model will not work for all of the
166
Greek data all of the time. We probably need to allow that different speakers will have different generalizations – for some speakers all gaps may be motivated – but the genitive plural gaps in Modern Greek cannot be completely accounted for through independent principles. Moreover, a model which seeks to explain speakers’ variable confidence in their own speech best accounts for the data if defectiveness is included as an independent variable. This result leads to a reversal of the relationship between gaps and insecurity:
paradigmatic gap → speaker insecurity
In the absence of the effects of an unpredictable morphophonological alternation, it is implausible to assume that speaker insecurity causes gaps because there is no understandable cause of the insecurity itself. This is a point that I take up in greater detail in CHAPTER 5, where I discuss Russian verbal gaps.
Critics might contend that there is no motivation for a language learner to make a series of small, idiosyncratic generalizations when they could posit a single generalization which covers more examples in a motivated way. And in some sense this is true. Yet large generalizations often do fracture into smaller ones over time. For example, Janda (1982) surveys umlauting in German. Umlauting is widespread in the inflectional system, and there are tantalizing similarities, both in terms of conditioning environment and grammatical effect, which led Lieber (1980) to put forth a unified treatment of German umlaut. However, Janda demonstrates that on close inspection, umlauting processes overlap in their conditioning environments and effect but are nonetheless sufficiently different that they cannot be collapsed and still provide
167
empirically adequate coverage. The parallelism between the umlauting rules in German presumably results from a fracturing of a single generalization over time. This supports a view of language as fundamentally a system of small-scale generalizations. It may include large-scale generalizations of the type that linguists prefer to investigate, but
Janda and Joseph (1999) among others hypothesize that these are not the types of rules which primarily drive linguistic systems. In the remainder of this chapter I consider how to incorporate the Greek gaps into a Word and Paradigm model of morphology.
4.4. Sometimes defectiveness is just defectiveness
There are several points in the grammar at which genitive plural gaps, as independent entities, could be captured. As Albright (2003:12) notes for paradigmatic gaps generally, assuming a generative theory of morphology “... the underlying forms of certain words may be defective,... the grammar itself may be indeterminate or uncertain, or ... some external mechanism blocks the output from being pronounced at the surface.”
Indeterminacy of the grammar is approach that Albright advocates for Spanish, and I modify to explain the Modern Greek data in CHAPTER 3. This is also the basic approach of many Optimality Theory accounts which argue that gaps result from the process of derivation via the null output/null parse candidate (Ackema and Neeleman
2000, McCarthy and Wolf 2005, Prince and Smolensky 2002[1993], Raffelsiefen 1996,
Rice 2005b).70 For gaps that are not the active product of the grammatical system, however, this approach is not viable.
70 Theories derive these outputs in a variety of ways, but the unifying factor is that the EVAL component produces an optimal candidate which has meaning but no phonological form. 168
Orgun and Sprouse’s (1999) argument for surface blocking has been equally popular, however. They are followed by Fanselow and Féry (2002), Hansson (1999), and
Pertsova (2005). Orgun and Sprouse argue for a new component which applies within
Optimality Theory after EVAL, called CONTROL. A fundamental starting point for their paper is the observation that for paradigmatic gaps speakers often know what the form would be, but reject it anyways. CONTROL is a theoretical encoding of that idea.
Literally, the form is produced, but blocked from being inserted into syntax. The derivational process proceeds unimpeded in EVAL, and an optimal candidate is chosen according to the typical metrics of Optimality Theory. Winning candidates then move to the CONTROL component, which checks the candidates for overall well-formedness.
Constraints within CONTROL are never violable in the language, so a winning candidate which violates one becomes a gap. These constraints are assumed to be non-universal, and may even make highly specific reference to lexemes, morphemes, etc. Orgun and
Sprouse’s paper is thus an argument specifically against treating paradigmatic gaps as the result of the process of word generation on the grounds that paradigmatic gaps cannot be the optimal result of the derivational process.
A parallel approach could be developed within a WP model by adding a component, after realization rules have applied, which checks the well-formedness of the output. Admittedly, this would be less well suited to a WP theory than it is to OT, which is by nature constraint-based, but even apart from this issue, there are at least three reasons to think that surface constraints are not the best approach. First, there is another way to explain the ability of speakers to produce non-used forms, namely analogy.
Second, it is undesirable to posit a mechanism whose only purpose is to block forms
169
which are defective. Third, treating the Greek gaps instead as a problem of lexical organization capitalizes on paradigms gaps as a kind of form-meaning mismatch through parallelism between gaps and other phenomena such as syncretism and heteroclisis.
The highly structured nature of the lexicon in the theory developed by Stump
(2001a, 2001b, 2006) means that lexical marking of defectiveness could in principle apply in a variety of ways – defectiveness within a lexeme paradigm cell, a stem paradigm cell, the association which connects the two (paradigm linkage), or the nodes from which a stem paradigm cell inherits its inflectional class information. Lexeme paradigms encode syntactic atoms, so barring some interaction with the syntax which causes morphological defectiveness, this would intuitively not seem to be the place to mark a gap. Paradigm linkage controls whether there is a one-to-one relationship between form and meaning. It would be possible to stipulate that there is no paradigm linkage from a lexemic genitive plural cell to stem paradigms for certain inflection classes. However, paradigmatic gaps would be the only obvious failure of paradigm linkage to apply and would require the removal of otherwise desirable default connections. By contrast, marking paradigmatic gaps within the stem paradigm, which I advocate, turns out to partially parallel the structures needed for syncretism and heteroclisis.
Incorporating specifications that particular combinations of stems and an inflectional property set are defective is equivalent to claiming that stem paradigms can be defective. This is implicitly contrary to a common definition of the paradigm.
170
I follow contemporary morphologists in appealing to a more abstract notion of paradigm, one which is at one level of abstraction removed from the list of forms. In this sense, a paradigm is a definition of the set of morphological contrasts that a given class of lexemes can make. (Spencer 2004:72)
This focus on morphosyntactic properties has promoted \the expansion of the domain of morphology at the expense of syntax, and the full explication of paradigms which might otherwise be considered defective. A classic example of this logic comes from treatments of periphrasis.
Börjars, Vincent and Chapman (Börjars et al. 1997) argued that periphrasis should be treated as a product of the morphological system, in part so that every syntactic atom will have a morphological form. Consider the following often-cited example.
‘to praise’ ACTIVE PASSIVE PRESENT laudat laudatur PAST laudabat laudabatur FUTURE laudabit laudabitur
Table 31: Imperfective paradigm of Latin verb laudo 'I praise' -- no periphrasis
171
‘to praise’ ACTIVE PASSIVE PRESENT laudavit laudatur/a/um est PAST laudaverat laudatur/a/um erat FUTURE laudaverit laudatus/a/um erit
Table 32: Perfective paradigm of Latin verb laudo 'I praise' -- periphrasis
In Latin, the passive is expressed with a single word formation when it is imperfective.
Likewise, the perfective is expressed synthetically when it is active. The combination of perfective and passive entails a periphrastic construction, however, and we know that this is a periphrasis, and thus in some sense potentially different than a normal syntactic phrase, precisely because the component morphosyntactic features are (individually) realized synthetically (see Sadler and Spencer (2000) for a fuller example of this argumentation). Inasmuch as we might expect a synthetic realization, recent attention has turned to explaining why this expectation is not met, and the degree to which the periphrasis parallels synthetic structure. For example, Ackerman and Stump (2004:128) posit the following research questions: “...why do some morphosyntactic property sets lack single-word realizations? And... why do single-word realizations exclude the use of synonymous periphrasis?” Ackerman and Stump (2004) and Sadler and Spencer (2000) both argue that Latin and other examples of periphrasis exhibit prototypically morphological behavior and should be treated as morphological phenomena – an override of the one-to-one form-meaning mapping within a Word and Paradigm model.
There are several advantages to treating periphrases such as the Latin example as morphological phenomena. Among them is the ability to posit complete paradigms at 172
both the lexeme and stem paradigm levels. If the Latin periphrasis were treated as a syntactic phenomenon, this would entail that the perfective stem paradigm would be defective. It thus seems that defective stem paradigms are not in keeping with a Word and Paradigm model of morphology.
At the same time, however, recent work on form-meaning mismatches has opened the door for less-than-complete stem paradigms within the theory. For example, heteroclisis is the phenomenon in which a lexeme is realized by at least two stems which belong to different inflection classes. See CHAPTER 1 for details and an example. In
Stump’s (2006) Paradigm Function Morphology account, the cells in a single lexeme paradigm correspond to cells in more than one stem paradigm. This raises the question of whether each of the respective stem paradigms generates the same contrasts as the lexeme paradigm, but is only be partially linked, or whether the stem paradigms only generate cells which are linked to lexeme paradigm cells. From the point of view of a speaker setting up a representation of stem paradigms, there is no motivation to posit a cell if a corresponding word form does not exist in the language. Thus, in examples of heteroclisis a given stem paradigm can have fewer paradigmatic cells than its corresponding lexeme paradigm. This mutability of the stem paradigm cell is also used for syncretism, in which two or more lexeme paradigm cells correspond to a single stem paradigm cell. The result is fewer cells in the stem paradigm than in the lexeme paradigm, even though they express the same morphosyntactic properties.
Together, these phenomena show that lexeme paradigms and stem paradigms operate according to different principles. The number of contrasting forms determines the number of cells in a stem paradigm. The number of contrasting property sets defines
173
the number of cells in a lexeme paradigm. Thus, Spencer’s definition of the paradigm applies only to lexeme paradigms. The same organization of the stem paradigm would be the default, but not necessary. If stem paradigms are defined in this way, Greek genitive plural gaps can be treated as a missing stem paradigm cell, or more specifically, a class of stem paradigms since the gaps are highly localized within the structure of the lexicon.
4.5. Language as a non-optimizing system
Central to the discussion in this chapter is the question of whether language change is ever the only motivation for a linguistic structure. Perhaps surprisingly, this is a controversial question. Janda (1984:87-88) nicely summarizes the issue:
The generative-grammatical reaction to [a historical explanation of language structure] has always been that... one can explain linguistic phenomena by positing synchronic grammars with particular structures in different speakers’ heads – grammars whose rules predict, and therefore at least partially explain, numerous facts of language. And the explanation for why these grammars are the way they are is that they result from the interaction of children’s innate capacities for language acquisition with the synchronic data heard in their environment.... Thus, we now believe that at least synchronic explanation is possible, in linguistics. But, since speakers(’ grammars) rarely possess knowledge of the history of their languages, many linguists have gone to the opposite extreme... and denied the possibility of any historical explanation in our field.
He goes on to argue that while synchronic explanations are often valid, sometimes a phenomenon can be explained only in historical terms.71 He focuses on the rarity of
71 It is important here to not confuse explanation with description. A synchronic grammar of a language must be able to describe any facts of the language, regardless of their origins. This does not entail, however, that those facts must necessarily be predicted, explained, caused, or motivated by the other facts in the synchronic structure of the language. It is furthermore important to not confuse historical explanation with the recapitulation of historical changes by a synchronic grammar. At best historical recapitulation in synchronic grammar leads to poor synchronic description. At worst it is an invalid type of 174
morphological metathesis, meaning cases in which the process of metathesis serves to express a grammatical relation. He concludes that
...the best account for why morphological metathesis rules are rare is the diachronic one that the possible sources potentially reanalyzable as such processes are themselves rare; this historical account is free of the unwarranted assumptions inherent in the competing synchronically based claim that morphological metathesis is rare because it is “marked”. (Janda 1984:87)
The point is that in considering the likelihood of a particular linguistic phenomenon synchronically, we must consider the likelihood that processes of language change could lead to such a scenario. In this sense, language change can be explanatory of synchronic linguistic structure. Anderson (1988) makes essentially the same point for a variety of other morphological structures.
Expected but non-attested forms have long fascinated morphologists because they seem to defy the productivity of language; paradigmatic gaps are particularly anomalous.
However, we must be wary of the explicit or implicit assumption that linguistic structures should have language-internal, synchronic explanations. Theories which universally treat gaps as byproducts of grammar competition seek to reconcile the superficially anomalous
(i.e. non-optimal) nature of paradigmatic gaps with this belief by positing gaps to be
“optimal failures” of the inflectional system.
For example, Rice (2005b) argues that gaps in Norwegian result from syllable structure problems. He utilizes a sub-theory of Optimality Theory called Optimal explanation, since speakers do not have historical knowledge of their languages. Historical explanation, as Janda argues, seeks to show how principles of language change shape the range of phenomena which exist across languages, and makes no assumptions that speakers have knowledge of the change. 175
Paradigms Theory (OPT, McCarthy 2005) to frame his analysis. On OPT, entire inflectional paradigms serve as candidates. If candidates are tagged for morphosyntactic information, there are constraints such that all possible morphosyntactic properties must be expressed somewhere within the paradigm, and the entire paradigm is simultaneously evaluated, paradigmatic gaps will appear only where a syllable structure violation or other phonological problem is higher ranked than the need to realize the morphosyntactic value. “This will allow us to check the preference for an incomplete paradigm, and very specifically limit the situations in which a gap is preferred” (Rice 2005b:11). If Rice intends these limitations on the appearance of gaps as universals, this is tantamount to saying that gaps are in fact part of an optimal state for the grammar.
While it may often be possible to provide an account of paradigmatic gaps which is motivated from the synchronic system of the language, the corollary that language exists in or moves towards an optimal state must be rejected. Indications are that the
Greek gaps ride the border between synchronic and historical-only explanation. In the former case paradigmatic structure indirectly causes gaps. In the latter it only describes them. In neither case is this an “optimal” result.
4.6. Summary
In this chapter I argued that while gaps may arise from grammar competition, Greek presents evidence of the genitive plural gaps having been at least in part reanalyzed as idiosyncratic facts of the language. I drew parallels to other phenomena in which historical changes revealed prior reanalyses of the data and argued, albeit tentatively at this stage, that the tendency to level stress is revealing gaps as having undergone a similar
176
reanalysis. In the absence of evidence that paradigmatic gaps in Greek have idiosyncratic properties, the account in CHAPTER 3 is the simplest and most explanatory. But in light of the evidence in this chapter, it is clear that it is not the entire picture. And the entire picture entails a rejection of the claim that gaps are universally epiphenomena, not requiring specific encoding within the morphological component. Finally, I suggested that gaps without synchronic motivation are best captured as defective stem paradigm classes.
177
CHAPTER 5
RUSSIAN FIRST PERSON SINGULAR NON-PAST VERBS:
THE AFTERMATH OF SYNCHRONICALLY MOTIVATED
PARADIGMATIC GAPS
5.1. Introduction
In the preceding chapters I argued that genitive plural paradigmatic gaps in Greek arose from an inability to predict the genitive plural form of some nouns based on other forms of the paradigm. I also argued, however, that the structure underpinning that unpredictability – a stress shift in the genitive plural form – is unstable, and the loss of that original motivation for the gaps reveals their status as idiosyncratic phenomena which must be explicitly accommodated by morphological theory.
In this chapter, I explore gaps in the aftermath of the loss of the motivating structure, and the potential that such cases have to reveal the inner workings of more productive inflection. I look in particular at a famous case of paradigmatic gaps – those in the first person singular non-past of Russian verbs. These gaps have mystified researchers for two reasons. First, the distribution of the gaps is closely tied to the distribution of a morphophonological alternation, a pattern which should be familiar by now from Greek and Spanish among other languages, but in the standard language the alternation applies uniformly within its conditioning environment. This seems to remove any possibility for an explanation via a grammar competition model. Second, when 178
speakers are forced to produce gaps, they sometimes produce forms according to patterns which are not attested in productive verbs. I argue that both of these facts are easily understood if we accept the idea of historical explanation, as outlined in CHAPTER 4, and allow our theory of inflectional morphology to look beyond the first person singular to other parts of the inflectional paradigm.
Ultimately, I have two goals. First, I seek to show that variation is not a sufficient criterion for the spread of gaps, and potentially not even a necessary one. Second, I demonstrate that paradigmatic gaps in Russian reveal paradigmatic connections that are not evident from productive forms alone, and show that these connections can operate at a greater level of abstraction than was evidenced in Modern Greek.
The data at the heart of this chapter comes from a forced production experiment parallel to the one I describe in CHAPTER 3 for Greek. Historical information, dictionary entries, and other information support and flesh out the conclusions drawn from the experiment. As with the Greek, however, my analysis cannot be understood without a global view of the inflectional facts of Russian verbs, so I begin there.
5.2. Overview of morphophonological alternations in verbal inflection
Russian has a set of inflectional forms which morphologically realizes the non-past tense, but semantically either the present tense (for imperfective verbs) or future tense (for perfective verbs). A different inflectional pattern is used for the future tense of imperfective verbs.72 This mismatch is illustrated in Table 33.
72 There are no forms at all for present tense perfective verbs because the meaning expressed via perfective forms, at its most basic a one-time completive sense, is incompatible with the meanings expressed via present tense forms, at the most basic signifying ongoing or generic action. 179
IMPERFECTIVE SINGULAR PLURAL краду / крадём / 1ST PERSON kradu kradjom крадёшь / крадёте / 2ND PERSON semantically kradjoš' kradjote present tense крадёт / крадут / 3RD PERSON kradjot kradut
IMPERFECTIVE SINGULAR PLURAL morphologically буду красть / будем красть / future tense 1ST PERSON budu krast’ budem krast’ будешь красть / будете красть / 2ND PERSON budeš krast’ budete krast’ non-past tense RD будет красть / будут красть semantically morphologically 3 PERSON ’ ’ budet krast budut krast future tense
PERFECTIVE SINGULAR PLURAL украду / украдём / 1ST PERSON ukradu ukradjom украдёшь / украдёте / 2ND PERSON ukradjoš’ ukradjote украдёт / украдут / 3RD PERSON ukradjot ukradut
Table 33: Semantic/morphological mismatch in Russian verbal inflection (красть/ украсть ‘to steal’)
This mismatch is relevant for gaps insofar as gaps affect morphologically non-past inflections, regardless of whether the verb is semantically future or present tense.
Verbs are traditionally divided into two primary conjugation classes, 1st and 2nd, according to their non-past inflectional patterns, each of which has a number of subpatterns. The examples above belong to the 1st conjugation class, and the example 180
below, ‘to decide (perfective)’, belongs to the 2nd conjugation class. Russian verbs usually come in aspectual pairs (imperfective/perfective), for instance красть ‘to steal
(imperfective)’ and украсть ‘to steal (perfective)’, but these pairs often do not belong to the same conjugation class.
‘to decide’ SINGULAR PLURAL решу / решим / 1ST P. re[š]u re[š]im решишь / решите / 2ND P. re[š]iš’ re[š]ite решит / решат / 3RD P. re[š]it re[š]at
Table 34: Example of the 2nd conjugation non-past inflectional pattern of Russian
The primary difference between the first and the second conjugation is the theme vowel –
/e/ or /o/ in the first conjugation, and /i/ in the second – and the third person plural realization. However, morphophonological alternations and other differences define a variety of subclasses.
For example, and relevant for the present purposes, the second conjugation class has a palatalizing alternation that affects stems ending in dental sounds (“dental stems” from here on). The sounds [dj] or [zj] at the end of the stem in the non-first-person-
181
singular form correspond to [ž] in the first person singular, [tj] likewise alternates with
[čj] or occasionally [šj], [sj] alternates with [š], and [stj] alternates with [šj].73
‘to see’ SINGULAR PLURAL ‘to carry’ SINGULAR PLURAL вижу / видим / ношу / носим / 1ST P. 1ST P. vi[ž]u vi[dj]im no[š]u no[sj]im видишь / видите / носишь / носите / 2ND P. 2ND P. vi[dj]iš’ vi[dj]ite no[sj]iš’ no[sj]ite видит / видят / носит / носят / 3RD P. 3RD P. vi[dj]it vi[dj]jat no[sj]it no[sj]jat
‘to cry’ SINGULAR PLURAL ‘to lower’ SINGULAR PLURAL плачу / платим / спущу / спустим / 1ST P. 1ST P. pla[čj]u pla[tj]im spu[šj]u spu[stj]im платишь / платите / спустишь / спустите / 2ND P. 2ND P. pla[tj]iš’ pla[tj]ite spu[stj]iš’ spu[stj]ite платит / платят / спустит / спустят / 3RD P. 3RD P. pla[tj]it pla[tj]jat spu[stj]it spu[stj]jat
Table 35: Stem-final palatalization in the 2nd conjugation first person singular non-past of Russian
The alternation in the first person singular non-past applies uniformly in the standard language. There are no words which fail to alternate if they meet the relevant criteria, although there is the minor variation in the form of the alternation for stems ending in /t/.
This contrasts with the relevant facts of Modern Greek and Spanish, as presented in
CHAPTER 3, in which words that were for all relevant purposes identical were realized sometimes with a morphophonological alternation, and sometimes without it.
73 There is also a palatalization process for stem-final labials, as well as various other kinds of alternations in both the first and second conjugations. These other patterns are not relevant for the present purposes, however. 182
In Russian there are also words in the 2nd conjugation class which do not have alternations in the first person singular by virtue of not meeting the relevant conditioning criteria. These include but are not limited to stems ending in palatal sounds ([š], [ž] or
[čj]). Like решить ‘to decide’, they have the same stem shape throughout the non-past paradigm.
‘to scream’ SINGULAR PLURAL кричу / кричим / 1ST P. kri[čj]u kri[čj]im кричишь / кричите / 2ND P. kri[čj]iš’ kri[čj]ite кричит / кричат / 3RD P. kri[čj]it kri[čj]at
Table 36: Non-alternation in the 2nd conjugation non-past of Russian
Thus, only a subset of the second conjugation stems meet the criteria for having a palatalizing alternation in the first person singular non-past.
The dental stems also have an alternation in the past passive participle.
183
(31) Parallelism, and non-parallelism, between alternations in the first person singular and the past passive participle
INFINITIVE 1ST SING PAST PASSIVE PART. GLOSS tj ~ čj оплатить оплачу оплачен ‘to pay’ oplatit’ oplaču oplačen
tj ~ šj сократить сокращу сокращен ‘to shorten’ sokratit’ sokrašču sokraščen
dj ~ ž зарядить заряжу заряжён ‘to load’ zarjadit’ zarjažu zarjažon
dj ~ ždj освободить освобожу освобождён ‘to become osvobodit’ osvobožu osvoboždjon free’
sj ~ š пригласить приглашу приглашён ‘to invite’ priglasit’ priglašu priglašon
zj ~ ž загрузить загружу загружён ‘to load’ zagruzit’ zagružu zagružon
Superficially, the alternation in the first person singular non-past and the past passive participle seem to be connected. Note, however, that there is one alternation in the participial forms which does not exist in the first person singular forms, namely /dj/ ~
/ždj/. The Russian cases thus seem much more like the fractured morphological generalizations that Janda (1982) discusses for German umlauting (see Section 4.5.). I return to this point later, where I show that this alternation creeps into the first person singular form when speakers are forced to produce stems that normally have paradigmatic gaps, and only for these stems.
184
5.3. The distribution of paradigmatic gaps in the first person singular
While Russian differs from Spanish and Modern Greek in that the alternations presented in Table 35 always apply in the standard language when the conditions are met, it also resembles those languages because the distribution of paradigmatic gaps is closely tied to the distribution of the alternation (first discussed by Halle 1973, and in some detail by
Švedova 1982). Russian has approximately 70 gaps in the first person singular non- past,74 and the vast majority affect verbs belong to the 2nd conjugation class with non-past stems ending in /t, d, s, z, st/. For example, we would expect the verb дерзить ‘to be imprudent’ to have a form держу / deržu ‘I am imprudent’, but according to several dictionaries and grammars, there is no first person singular non-past form which speakers consider acceptable.
‘to be imprudent’ SINGULAR PLURAL дерзим / 1ST PERSON * derzim дерзишь / дерзите / 2ND PERSON derziš’ derzite дерзит / дерзят / 3RD PERSON derzit derzjat
Table 37: Paradigmatic gap in the first person singular non-past
74 Halle (1973) claims that there are approximately 100, but I could reach this number only by counting two citation forms sharing the same root as separate entries. 185
The full list of gaps, as culled from nine dictionaries and grammars, is given in
APPENDIX E. Here I provide the list which these sources most commonly cite as having paradigmatic gaps in the first person singular non-past form.
(32) Paradigmatic gaps in the first person singular non-past of Russian verbs
бдеть bdet’ ‘to keep watch’ бузить buzit’ ‘to protest’ галдеть galdet’ ‘to make a hubbub’ дерзить derzit’ ‘to be imprudent’ дудеть dudet’ ‘to play the pipe’ ерундить erundit’ ‘to do stupid or funny things’ затмить zatmit’ ‘to eclipse’ кудесить kudesit’ ‘to do magic’ очутиться očutit’sja ‘to find oneself; to come to be’ победить pobedit’ ‘to win’ рысить rysit’ ‘to trot’ соседить sosedit’ ‘to be a neighbor’ убедить ubedit’ ‘to persuade’ умилосердить umiloserdit’ ‘to take pity on’ чудесить čudesit’ ‘to do magic’ чудить čudit’ ‘to behave in a weird way’ шкодить škodit’ ‘to misbehave’
Many of these words, especially победить / pobedit’ ‘to win’ and убедить / ubedit’ ‘to persuade’, are commonly used lexemes. Their stems all belong to the second conjugation class. All except затмить / zatmit’ ‘to eclipse’ have a stem ending in /t,d,s,z,st/, and would thus be expected to have a palatalization in the (missing) first person singular form. Very few can be connected to semantic problems, and even fewer to homophony avoidance or other popular explanations.
Variability in the application of an alternation was the basis for Albright’s (2003) claims of grammar competition in Spanish. And unpredictability was the basis for my
186
claim of paradigmatic competition in Modern Greek. However, as Baerman and Corbett
(2006) argue, the fact that all productive verbs of the relevant type have palatalization in the first person singular entails that the same sort of direct competition cannot exist in
Russian. In fact, there is no obvious synchronic motivation for the gaps at all. There are thus three key questions in connection with the Russian data:
First, what are the actual facts of language use with regard to the palatalizing alternation among dental stems? There is anecdotal evidence, cited specifically by
Baerman and Corbett, that when forced to produce first person singular non-past forms for stems usually having paradigmatic gaps in this cell Russian speakers sometimes forms without an alternation, for example derzju ‘I am imprudent’ or pobedju ‘I will win’, instead of deržu, or pobežu, as would be expected based upon the first person singular of non-defective verbs. I call this “unexpected non-alternation”. Does unexpected alternation represent a broader trend, evident not only in forced production of a form where there is ordinarily a gap, but also in regular, fully productive verbs? In other words, does the standard language provide an accurate picture of language use? If not, there may be a basis for claiming that the gaps are synchronically motivated after all.
Second, what causes/caused the gaps? If no significant synchronic motivation is evident, this becomes a significant question of historical analysis. In particular, did paradigmatic competition exist at the time that the gaps first appeared?
Finally, how do speakers react to the gaps? The Russian verbal gaps have a distribution which allows us to further probe the relationship between word production and speakers’ reflections on that word production. Albright showed a correlation
187
between variability, speaker confidence, and paradigmatic gaps. Based upon this, he posited the following causative link:
variability → lower confidence → gaps
The potential import of the Russian distribution is clear. In the absence of grammatical preconditions such as variability or, as I argue, paradigmatic predictability, any correlation between lowered confidence and gaps most likely indicates the opposite direction of causation because there is no apparent reason for lowered confidence to exist otherwise:
gaps → lower confidence
A glimmer of this pattern was evident in the Modern Greek data, but Russian provides an even better arena to test whether the very fact of a gap itself is sufficient to lower speaker confidence. If so, this draws into question directionality within Albright’s model, and with it the entire logic of a gaps-as-epiphenomena account.
To jump ahead to the conclusions, I will show based on the results of a production experiment that the first person singular non-past form of productive dental stems are to some degree paradigmatically unpredictable, but that this cannot be the entire explanation. Speakers’ reactions to the gaps largely do not parallel their reactions to regular inflection, as would be expected within a grammar competition model. I also present evidence, however, that Russian at the end of the 19th century represented the
188
kind of paradigmatic competition combined with sociolinguistic insecurity found in
Modern Greek, suggesting that the gaps are motivated in a historical sense, if not in a synchronic one. It seems that the very fact of a gap does indeed lower speakers’ confidence when forced to produce the expected but non-attested form.
I further argue that the unique way that speakers approach gaps reveals paradigmatic connections between the first person singular non-past and the past passive participle which are not evident based on data from other types of verbs alone.
Moreover, this phenomenon cannot be understood without reference to the paradigm, so
Russian provides further and different evidence that paradigmatic structure is a necessary part of inflectional theory.
5.4. An experiment on speaker confidence and unexpected non-alternation75
The literature on paradigmatic gaps frequently cites the Russian verbal gaps (Albright
2003, Baerman and Corbett 2006, Baronian 2005, Fanselow and Féry 2002, Halle 1973,
Hetzron 1975, Iverson 1981, McCarthy and Wolf 2005), but in most of the studies the data has been limited to the basic distributional information given above. No studies have systematically compared how speakers react to gaps as compared with regular inflections, such as whether speakers are able to produce the expected but unattested form as easily as they can non-defective forms.76 Nor have they tried to quantify the degree to
75 I am very grateful to Bryan Brookes and especially Maria Alley, who worked with me to develop and execute this experiment. It would have been much poorer without their work and their ideas. 76 As I have mentioned elsewhere, the fact that speakers can often produce the expected form but then reject it anyways has fascinated researchers. This ability suggests a parallelism between gaps and regularly inflected forms, and has driven some approaches to paradigmatic gaps, such as the CONTROL component (Orgun and Sprouse 1999) which. However, very little research has given detailed attention to the depth of this parallelism, and none of it has been done on Russian. This is perhaps surprising, considering that the unexpected non-alternation that is so well attested in Russian runs counter to the idea that speakers produce 189
which unexpected non-alternation or other variability occurs. These are potentially important sources of data, however. Some theories, such as the CONTROL component posit that all forms are generated by the same mechanism, and the gaps are later rejected by a surface filter. This predicts parallel behavior between gaps and regularly inflected forms with regard to unexpected non-alternation and other patterns of inflection.
Grammar competition theories also predict parallel behavior. Moreover, as we have seen, distributional evidence does not always tell the entire story. More detailed information on the relationship between gaps and inflectional structure is available from parallelisms, or lack thereof, in how speakers treat gaps and non-defective inflections.
This study was inspired by the work of Baerman and Corbett (2006), who claim that the morphophonological alternation in the first person singular non-past of Russian verbs (tj ~ čj, dj ~ ž, sj ~ š, stj ~ šj and zj ~ ž) may have originally provided motivation for gaps in that cell, but that this factor was later subsumed by an analogic pattern of spread, with the stem-final consonants being the basis for extending gaps to new words. Their position is enticing in light of the similarities and difference between Russian, Spanish, and Modern Greek, but their evidence is weak. No historical evidence is presented for the analogical extension – only modern distribution. And they present no evidence that the morphophonological alternation is synchronically unrelated to (i.e. does not cause) the gaps. The experiment that I present in this section expands the available data on
Russian gaps by exploring speakers’ intuitions about and reactions to defectiveness. This study has a similar methodology to the one used by Albright for Spanish, and my own work on Modern Greek. gaps and regular inflections by the same mechanism, because non-alternation is not attested among the latter type of stimuli. 190
5.4.1. Predictions
Following Baerman and Corbett, I would expect to find evidence that gaps pattern distinctly from non-defective stems, showing that defectiveness cannot be synchronically motivated by the regular functioning of the linguistic system.
Consider four different types of stems: (a) stems with prescribed gaps in the first person singular non-past, (b) stems which meet all of the criteria for having gaps in this cell, but which prescriptively have normally inflected forms (see Table 35 above; from here on I call these “non-defective dental stems”), (c) words which do not meet the criteria for gaps, but which belong to the same inflection class (see Table 34 and Table
35), and (d) stems belonging to an entirely different inflection class (see Table 33).
These types represent an ordinal scale with four levels, representing increasing difference from defective stems: (a) is maximally close, being defective, and (d) is maximally far, not meeting any of the criteria for having gaps. I make several predictions with regard to the speakers’ reactions these different types of stems in a forced production and confidence ratings task.
• Speaker confidence: Speakers will report lower confidence in forms that
correspond to gaps than in any type of non-defective forms (gap → lower
confidence).
This reverses the prediction of a grammar competition model, and is in line with the results of the Greek production study which showed that those gaps are not as closely tied to the stress alternation as other data would suggest. If this prediction is upheld, and 191
stems with gaps pattern differently from non-defective dental stems, this will demonstrate that speakers are reacting to something other than the morphophonological structure of the stem, and would support the claim that gaps cannot be reduced to that structure. If the prediction is rejected, and especially if the resulting pattern groups gaps and non- defective dental stems to the exclusion of the other two types of stems, this would implicate the stem structure and the morphophonological alternation that applies to it as somehow being causative of the gaps.
More interesting, perhaps, is the issue of variability across speakers when producing different types of stems.
• Interspeaker agreement: In forced production, speakers will consistently produce
non-defective dental stems with the relevant morphophonological alternation.
Stems with gaps may induce more variability, however.
When prompted with a stem with a gap in the first person singular non-past, whether different speakers produce different forms for this cell should depend upon the number of patterns upon which speakers could draw to produce the relevant form. If stems with gaps are lexically specified as having no first person singular non-past form,77 speakers must rely on analogical pattern matching when forced to produce a form. If there is only one relevant pattern, subjects should consistently produce nonce words and gapped words according to that pattern. Presumably, the words would have the alternation that applies uniformly to productive stems with the same structure. However, if at some level there
77 In CHAPTER 4 I formalized this as a missing stem paradigm cell. 192
are multiple patterns available for the first person singular, different subjects may produce different forms. Gaps thus have the potential to reveal the connections that speakers make between different areas of the inflectional system, i.e., paradigmatic structure.
• Confidence and interspeaker agreement: Speakers’ confidence in producing 1sg
forms where there are normally gaps will pattern differently (probably lower) than
their confidence in inflectional patterns which are equally variable, but less
sociolinguistically sensitive.
As I have argued elsewhere, a primary deficit of Albright’s argument of grammar competition stemming from variability is that it far overpredicts the appearance of paradigmatic gaps. His model suggests that wherever there is significant variability between two inflectional patterns, a gap should result. This view is overly mechanistic, failing to consider the role of the speaker within a social context. In CHAPTER 3 I argued that we must narrow that idea along two dimensions – predictability rather than variability, and with consideration of social factors. Gaps result most directly from avoidance strategies, which in turn only make sense if the relevant structure is a social marker or stereotype in the Labovian sense (Labov 1994). If not all variation has negative valuation attached to one or more variants, different types of variation could lead to different reports of confidence levels.
193
• Lexeme familiarity and confidence: Speakers’ confidence in their productions of
regularly inflected 1sg non-past forms will go up as lexeme familiarity goes up.
Familiarity for stems with gaps in the first person singular will not affect
confidence (to the same degree).
This is the pattern, found also in Greek, which I referred to as a ceiling effect.
Familiarity is a measure of the entire lexeme, but confidence is a measure of the first person singular non-past form. This difference might not be important if the relative frequencies of the first person singular non-past form and the other inflected forms of the lexeme were constant across the stimuli. However, the (presumed) extremely low frequency of the gaps is a unifying factor for that class, regardless of the frequency or familiarity of the lexeme over all. This may create a “ceiling” in the sense that it nullifies expected effects for lexeme familiarity, leading gaps to pattern closer to nonce words.
5.4.2. Methodology
The methodology for this experiment parallels the Modern Greek study in that it was composed of four tasks: a background questionnaire, a lexeme familiarity ratings task, a forced production sentence completion task, and a production ratings task.
5.4.2.1. Experiment structure
I used the background questionnaire to collect basic demographic information from each subject (e.g. age, sex, current and previous cities of residence, etc.), and information about language use (e.g. native language, what language is used in the home, what language is used at work, etc.). Based on this information I subsequently removed the 194
data of two subjects who were not eligible for the study. I consider an eligible subject to be someone who was born and raised through adulthood in a Russian speaking country, who spoke Russian in the home as a child, and who still speaks Russian on a regular basis. The background questionnaire was administered as a pencil-and-paper task.
I split the other three tasks into two parts each. Subjects completed the familiarity ratings task for half of the stimuli, followed by the production and production ratings task for those same stimuli, which were interleaved as described below. After a short break, they would then repeat the process for the other half of the stimuli. For these tasks subjects sat at the computer and responded to prompts on the screen. The experiment was administered using the psychology experiment software E-prime.78
In the familiarity ratings pretest subjects saw a Russian word on the screen in red letters, and a series of numbers with labels below the word in blue. They judged how familiar that word was to them, and how often they use that word, based on the six-point
Likert scale represented by the numbers. They were be asked to respond as quickly as possible while still being accurate by pushing the corresponding number on a numeric keypad.79 The scale was: 1 = I don’t know the word. / 2 = I can guess the meaning of the word, but have never heard it. / 3 = I am familiar with the word, but don’t use it. / 4 = I may have used this word a couple times. / 5 = I use this word sometimes. / 6 = I use this
78 See Appendix F for the full experiment materials. 79 I used an external numeric keypad rather than a response box because I felt that it was important to distinguish all six of these points on the scale rather than the five possible with response boxes. The problematic value, in a sense, is “I can guess the meaning of the word, but have never heard it.” In my experience, this is a necessary value for familiarity scales used to judge Slavic languages because rich morphological systems create situations in which a word is easily understood by analysis of component parts, but has no frequency in the language as a whole word. This means that the response times reported in this experiment are not accurate to the millisecond. According to the E-prime User’s Guide, keyboards introduce a response delay averaging 5ms, as compared with response boxes. Considering that typical response times were upwards of 1000 milliseconds for word production and in the hundreds of milliseconds for familiarity judgments, this margin of error is minor. 195
word frequently. The task included two practice sessions, each of which consisted of ten stimuli. During the first, subjects saw the full labels for the scale on the screen at all times, but during the second, the labels were shortened (e.g. “don’t know”, “don’t use”,
“use a lot”) to discourage reading and overthinking. The subsequent non-practice items displayed these shortened labels at all times.
The task was self-timed, and subjects received feedback on their pace after every
20 words. If their mean response time for all stimuli to that point was faster than 2.5 seconds, they received the message “x seconds per response. Good job!”, where x is their average response time. If the mean response time for all stimuli to that point was slower than 2.5 seconds, subjects received the message “x seconds per response. Try to go faster!”.
The main section of the experiment consisted of a cloze procedure forced production task and a self-rating task. First, subjects saw an incomplete sentence in green letters on the screen, with a blank in place of a missing verb. Once the person had read the sentence he/she pushed any button on the keypad, and the missing verb appeared on the screen in the infinitival form, in red letters below the sentence. The subject then said the word aloud in the form needed to complete the sentence. For the target words, the sentential frame always required a first person singular non-past form. For fillers, a different present tense form or the infinitive was required. Subjects were instructed to say their responses as fast as they could, while still being accurate. Subjects had only 3 seconds to respond once the word displayed on the screen, before the experiment moved to a ratings slide.
196
The ratings slide prompted subjects to say a number reflecting how confident they felt that the form of the word that they had just said was correct. The instructions encouraged subjects to choose their own scale, emphasizing that only the relative distance between the numbers was important, and not the absolute values. They were asked to randomly assign a number to the first stimulus, and then double that number for the second stimulus if they felt twice as confident in it as in the first stimulus, half it if they felt half as confident, etc. This is free magnitude estimation (see Section 3.5.2.2.).
Subjects had 2.5 seconds to respond with a number once the ratings slide appeared. The experiment then moved on to the next sentential frame.
Preliminary testing indicated two problems with this ratings methodology. First, it showed a tendency for subjects to automatically assign the lowest score in their scale to nonce words, on the apparent logic that if a word does not exist in Russian it cannot possibly be “correct”. As a result, between the familiarity ratings pretest and the main section I orally asked subjects to judge the “made up words” as if they were real words of
Russian, explaining that words are always coming and going in the language, and
Russian speakers must always make decisions about how to use new words. This seemed to solve the problem for all but two subjects during main testing. I removed all data from these two subjects from the subsequent analysis.
Second, while the experiment instructed subjects to use free magnitude estimation, with perhaps one exception they actually used closer to a percentage scale, with fixed maximum and minimum values (although different maximum values for different speakers), and a fixed number of points along the scale, typically about 8. The measures of confidence were not necessarily erroneous, but are more accurately treated
197
as reflecting percentage scales, rather than magnitude scales. In the following analysis I follow this practice.
The same set of 250 stimuli, all verbs, were used in all three tasks. The experiment presented each word once in each task.
5.4.2.2. Stimulus composition
We selected stimuli from a representative sample of Russian verbs based on a variety of criteria.80
We compiled a list of gaps from nine dictionaries and previous research on gaps, then selected twenty words with first person singular gaps from the list (see APPENDIX
E) based on the number of sources which cite the word as having a gap, lexeme frequency, morphological complexity, stem-final consonant, and pairing. We gathered lexeme frequencies for the gaps from the Russian National Corpus.81 We discarded those with a frequency of less than one token per million, and among the remaining words, chose items which ranged widely in frequency, were maximally morphologically simple
(when possible), which represented a variety of stem-final consonants (4 with /t/, 6 with
80 Maria Alley, Bryan Brookes and I created a representative list of potential stimuli in the following way: First, we identified a large sample of Russian verbs by searching a reverse alphabetized list of Russian words. Verbs are easily identified with a reverse alphabetized list due to infinitival suffixes. The list contained 99,430 words. 26,263 were verbs. 4,463 were verbs with /t, d, s, z, st/ at the end of the infinitival stem (both the first and second conjugation classes included). 2,502 were words with /č, š, ž/ at the end of the infinitival stem. Second, we separated verbs with /t, d, s, z, st/ at the end of the infinitival stem into conjugation classes via an automated search of an online version of Ožegov (1972) (http://starling.rinet.ru/cgi- bin/query.cgi?root=/usr/local/share/starling/morpho&morpho=1&basename=\usr\local\share\starling\morp ho\ozhegov\ozhegov). The search retrieved the first and second person singular forms and any comments about usage for all 4,463 verbs with /t,d,s,z,st/ immediately prior to the infinitival suffix, all 2,502 verbs with /č, š, ž/ in this position, and 10% of the remaining verbs. This allowed us to separate verbs according to inflection class. This process resulted in lists of all gaps, all non-defective dental stems (2nd conj.), all palatal stems (2nd conj.), and a substantial number of first conjugation verbs. 81 http://corpus.leeds.ac.uk/ruscorpora.html and also http://ruscorpora.ru/ 198
/d/, 4 with /s/, 6 with /z/), and which approximately matched another gap in terms of frequency, stem type, and number of syllables. This created 10 pairs of gaps.
Three other types of stimuli were matched to these 10 pairs of gaps: dental stems, palatal stems, and first conjugation stems. Dental stems, as described above, are lexical items which belong to the same conjugation class as the gaps, and which have the same morphophonological structure – /t,d,s,z/ at the end of the non-past stem. Palatal stems also belong to the second conjugation class but have /č,š,ž/ at the end of the stems.
Remember that these have the same stem formation throughout the non-past paradigm, unlike the dental stems which alternate. We matched items in these three stem types to the gaps (one example of each stem type for every pair of gaps) according to lexeme frequency,82 number of syllables in the first person singular non-past form, and morphological complexity. This resulted in 50 stimuli which are real words of the
Russian language.
We additionally composed 30 similar nonce words – 10 dental stems, 10 palatal stems, and 10 first conjugation verbs.
These were thus 80 stimuli representing 10 groupings within which stimuli were paired as closely as possible. For example, the word чудить / čudit’ ‘to behave eccentrically’ is a gap. In the (hypothetical) first person singular non-past form it has 2 syllables, according to the Russian National Corpus it has a lexeme frequency of 2.5 tokens per million words, it contains no prefixes, and the stem ends in [d]. It is therefore matched to another gap чадить / čadit’ ‘to singe’, the regularly inflected dental stem
82 In addition to frequency information mined from the Russian National Corpus, Maria Alley verified that in her opinion as a native Russian speaker, the words which were paired across each of the four stem types were comparable in terms of frequency in her variety of spoken Russian. 199
word прудить / prudit’ ‘to dam up’, the nonce dental stem word бадить / badit’, the regularly inflected palatal stem бренчать / brenčat’ ‘to strum’,83 the nonce palatal stem word кленчать / klenčat’, the regularly inflected 1st conjugation stem word нашить / našit’ ‘to sew onto’, and the nonce 1st conjugation stem word зафить / zafit’.
Finally, we included at additional ten first conjugation words which are morphological “doublets”, meaning that in Russian there is widespread variation in the inflectional patterns that they follow. For example, some people inflect the third person singular of the verb махать / maxat’ ‘to wave (a flag)’ as машет / mašet while others use махает / maxaet. We included these verbs partially in order to distract subjects from the fact that most of the “difficult” stimuli, i.e. the gaps, are confined to a very specific type of word. We were also interested in the degree to which variation in a language, apart from any grammatical factor like paradigmatic competition, causes speakers to be less confident when producing a form.
The experiment included 160 filler verbs and 20 practice items, for a total of 270 test items. See APPENDIX F for the complete list of test items. As described above, each item was presented three times – once in the familiarity ratings pretest, once in the cloze procedure forced production task, and once in the confidence ratings task.
Including consent, debriefing, the background questionnaire, the computer work, and a
83 Slavicists will recognize that бренчать belongs to a relatively infrequent subtype of palatal stems. Most second conjugation stems have infinitival forms with /i/ in the infinitival form, where бренчать has /a/. /a/ is more typical of infinitives for first conjugation stems. In this study we tried, where possible, to avoid stems of this subtype where possible, but the overall limited number of second conjugation palatal stems made this at times impossible without sacrificing the other criteria used for matching. Only three palatal stem items have /a/ in the infinitive – бренчать, кленчать, and брюзжать ‘to be grumpy’. The remaining 17 have /i/. This difference apparently led to a few interpretations of кленчать as a first conjugation stem (1sg кленчаю / klenčaju), but otherwise it had no noticeable effect on the data. 200
few breaks, the entire experiment lasted on average 1 hour and 30 minutes per subject.
Subjects were paid for their participation.
5.4.2.3. Types of data collected
From this experiment we collected five kinds of data.
• Lexeme familiarity ratings, on a scale from 1 (low) to 6 (high).
• Speakers’ reaction times for lexeme familiarity ratings.
• The verb form that subjects say to complete a sentential frame when prompted
with the infinitival form.
• The reaction time for subjects’ verb productions.
• Subjects’ ratings for how confident they were that their verb productions were
correct.
E-prime collected the lexeme familiarity ratings, and both types of reaction times. The verb forms and confidence ratings were spoken aloud by the subjects and recorded using a Shure SM10A microphone and a Marantz compact disk recorder. For the verb productions, verb production reaction times, and confidence ratings we employed backup systems, in case of data loss. First, E-prime sounded a tone at the beginning of each slide on which speakers were prompted to produce verb forms (after having read the sentences). This tone was recorded onto the same track as subjects’ responses, and allowed us to check for erroneous verb production reaction times by measuring the time from the tone to the beginning of the subject’s response using the waveform editor Praat.
The subject did not hear the tone. During the experiment, a phonetically trained researcher or a native Russian speaker also listened to and wrote down subjects’ verb
201
forms and confidence ratings for target items. These were then checked against the recording when the response was in doubt.
5.4.2.4. Subjects
Twenty-two native speakers of Russian participated in this experiment. At the time of the experiment all were residents of Columbus, Ohio. They represented a roughly even mix of men and women. More than half were from Russia; approximately 1/3 were from the Ukraine, but had learned Russian as their first language. One person was from
Moldova. The subjects ranged in age from 18 to late 60’s, with the majority falling into the neighborhood of 25-35 years old. All reported Russian as their native language and the native language of at least one parent. All reported speaking Russian on a regular basis at the time of the experiment, typically with other family members. Some subjects were permanent émigrés to the United States; others had come to the United States for education.
These twenty-two subjects do not include seven whose results were removed.
Two had arrived in the United States as young teenagers, and reported that they did not at the time of the experiment speak Russian on a regular basis. For three subjects technical problems while running the experiment caused significant amounts of data to be lost.
Two did not follow directions in the confidence ratings task, giving uniformly the lowest rating to all nonce words, apparently solely because they were nonce words.
5.4.3. Results
Focusing on speakers’ productions of word forms in the sentence completion task and their confidence in their productions, there are several primary results. First, subjects
202
reported lower confidence in their productions of first person singular non-past stems with gaps than their productions of stems with regular inflection, including doublets.
They reported higher confidence in gaps than in nonce words. Second, there is an effect of lexeme familiarity on speakers’ confidence for all regular inflections, but not for gaps.
Third, subjects produced forms corresponding to gaps with a large degree of variability.
Subjects produced non-defective dental stems with a significantly lesser degree of variability but nonetheless some, and produced first conjugation non-doublets with no variation. The most interesting part, however, is that variation differed between gaps and non-defective dental stems. Both resulted in some examples of unexpected non- alternation, but gaps also resulted in examples of the alternation used by the past passive participle (see (31)).
In the sentence completion task, there is an inverse linear relationship between production time and speaker confidence. Words which required more time for speakers say the necessary form also received lower confidence scores. The following graph gives an item analysis for all verbs for which the sentential frame required the first person singular form. Because subjects used their own scales for the judgment task, I normalized across speakers by calculating the confidence score for each item as a ratio to the maximum score for that speaker. Many subjects explicitly chose 0 as a minimum score and the comments and behavior of others (e.g. by not choosing any negative scores) indicated an implicit use of 0 as the lowest possible value for the remaining subjects as well. The ratio thus converts subjects’ confidence ratings to a percentage of maximal confidence. These ratios are then averaged across subjects.
203
average 1sg production confidence ratios confidence 1sg production average square = gap circle = nonce word triangle = doublet diamond = non-defective, non-doublet 0.0 0.2 0.4 0.6 0.8 1.0
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
average 1sg word production response times (z-scores)
Figure 21: Relationship between sentence completion reaction time and speaker confidence for
Russian
The inverse relationship between response time and confidence ratings was significant according to a simple linear regression model (R2 = 0.6417, p<0.001). This indicates that subjects were by and large awake, paying attention, and trying to complete the experiment. 204
Interestingly, however, these data points were not evenly distributed among the various types of stimuli. Words with productive inflectional forms in the first person singular, whether they are doublets, first conjugation stems, dental stems, or palatal stems were both the fastest to produce and resulted in the highest confidence ratings. Gaps were slightly slower, and speakers had slightly lower confidence in them. Nonce words were slowest and triggered the lowest confidence ratings on average.
To some extent the differences in the response times between stimulus types, in particular between the nonce words and the real Russian words, can be attributed to lexeme familiarity. There is a slight trend that the more familiar the word, the less time subjects was required to produce a word form in the cloze procedure sentence-completion task. In the study, six was the highest Likert rating for lexeme familiarity (=I use this word frequently”). Nonce words, which virtually uniformly received Likert ratings of 1
(=I don’t know this word) or 2 (“I have not encountered the word but can guess the meaning”), were logically the slowest to be produced.
205
word production response times (z-scores) times response word production -2 0 2 4 6
123456
Likert ratings
Figure 22: Slightly faster responses for familiar words than for non-familiar words
Confidence ratings for nonce words also fall close to the expected value based on the trend for these stems. Removing gaps and nonce words from the data set, the best fit for the non-defective stems is y = 0.07103x + 0.49183. This fit predicts a confidence score of 0.563 for lexemes receiving a Likert rating of 1. The average confidence score for the
206
nonce words was 0.467, and the median was 0.5. In this respect, nonce words behave as would be expected based upon the patterns of regularly inflected words.
average 1sg production confidence ratios confidence production 1sg average square = gap circle = nonce word triangle = doublet diamond = non-defective, non-doublet 0.0 0.2 0.4 0.6 0.8 1.0
123456
average Likert ratings
Figure 23: Lexeme familiarity vs. speakers' confidence in their word form productions
However, when we examine the gaps, it is clear that lexeme familiarity cannot account for either the word production response time or, much more importantly, the confidence
207
ratios that speakers produce. The following figure replicates the data from Figure 23, but has removed the nonce words, and collapsed the doublets and other non-defective stems. average 1sg production confidence ratios confidence production 1sg average square = gap circle = no gap 0.0 0.2 0.4 0.6 0.8 1.0
3.0 3.5 4.0 4.5 5.0 5.5 6.0
average Likert ratings
Figure 24: Relationship between lexeme familiarity and production confidence
Subjects reported comparable familiarity with the lexemes containing gaps and with those which have regularly inflected first person singular form, yet the confidence scores that the first person singular received were significantly lower for the gaps than for the 208
regularly inflected forms. Moreover, there is a significant correlation between lexeme familiarity and speakers’ confidence for regularly inflected forms (R2 = 0.4784, p<0.001) with no notable differences between palatal, dental and first conjugation stems, but a non- significant relationship for defective stems (y=0.01364x + 0.63654, R2 = 0.0117, p =
0.6495).
While this lack of relevance of lexeme familiarity only for gaps might at first seem anomalous, it makes sense when we consider that familiarity in this experiment was measured for the entire lexeme, but confidence is measured for the first person singular only. Paradigmatic gaps have lower frequency of use relative to other forms of the same lexeme than do regularly inflected words. Thus, while overall familiarity is comparable, the status of the first person singular relative to the rest of the paradigm differs for gaps and non-defective stems. The overall lower confidence ratings for gaps may be explainable as a consequence of this difference.
The split between gaps and regular inflections is further evident from the actual forms that subjects produced to complete the partial sentences. A subset of the data is reproduced below.
209
S11 S13 S14 S16 S17 вкатить вкач'у вкач'у вкач'у вкач'у вкач'у dental vkatit’ vkaču vkaču vkaču vkaču vkaču stems грузить груж'у груж'у сам груж'у гр'ужу гр'ужу gruzit’ gružu gružu sam gružu gružu gružu божиться бож'усь бож'усь бож'усь бож'ился бож'усь božit’sja božus’ božus’ božus’ božilsja božus’ palatal (error) stems бренчать бренч'у бренч'у бренч'у бр'енч'у бренч'у brenčat' brenču brenču brenču brenču brenču вливаться влив'аюсь воль'юсь влив'аюсь влив'аюсь вл'ился vlivat’sja vlivajus’ vol’jus’ vlivajus’ vlivajus’ vlilsja 1st conj. (error) (error) stems влипнуть вл'ипну вл'ипну вл'ипну вл'ипну вл'ипну vlipnut’ vlipnu vlipnu vlipnu vlipnu vlipnu блистать блищ'у блист'аю блист'аю блист'аю блист'аю blistat’ blišču blišču blistaju blistaju blistaju doublets кропать кропл'ю кропл'ю кроп'аю кропл'ю кропч'у kropat’ kroplju kroplju kropaju kroplju kropču (error)
Figure 25: Sample of responses to non-defective, non-nonce items in a Russian sentence completion task
210
S11 S13 S14 S16 S17 бадить баж'у баж'у бад'ю баж'у баж'у badit’ bažu bažu badju bažu bažu втасить втас...ш'у вташ'у втас'ю вташ'у вташ'у dental vtasit’ vtas...šu vtašu vtasju vtašu vtašu stems оретиться ореч'усь ореч'усь ореч'усь отрет'иться отрет'ишься oretit’sja orečus’ orečus’ orečus’ otretit’sja otretiš’sja (error) (error) намушить намуш'у намуш'у намуш'у намуш'у намуш'у namušit’ namušu namušu namušu namušu namušu palatal завужить завуж'у завуж'у завужать завуж'у завуж'у stems zavužit’ zavužu zavužu zavužat’ zavužu zavužu (error) гзинуть гниз'у гз'ину n/a зг'ину зг'ину gzinut’ gnizu gzinu zginu zginu 1st conj. (error) (error) (error) stems чувнуть чувн'у чувн'у n/a чувн'у чувн'у čuvnut’ čuvnu čuvnu čuvnu čuvnu
Table 38: Sample of responses to nonce items in a Russian sentence completion task
211
S11 S13 S14 S16 S17 победить побед'ю побеж'у побеж'у побеж'у pobedit’ pobedju pobežu pobežu N/A pobežu убедить убед'ю убеж'у убеж'у убед'ю убежд'у ubedit’ ubedju ubežu ubežu ubedju ubeždu чудить чуд'ю чуж'у чуж'у чуд'ю чуж'у čudit’ čudju čužu čužu čudju čužu голосить голос'ю голош'у голош'у голос'ю голош'у golosit’ golosju gološu gološu golosju gološu приютиться приюч'усь приюч'усь пр'июч'усь приют'юсь приюч'усь prijutit’sja prijučus’ prijučus’ prijučus’ prijutjus’ prijučus’ грезить гр'езю греж'у гр'ежу гр'ежу гр'ежу grezit’ grezju grežu grežu grežu grežu ощутить ощут'ю ощущ'у ощущ'у ощу-, ощу- ощущ'у oščutit’ oščutju oščušču oščušču не знаю oščušču ošču-, ošču-, I don’t know
Table 39: A sample of responses for defective stems in a Russian sentence completion task
While subjects sometimes made errors in the cloze procedure task, producing the wrong lexeme, or a different inflected form of the same lexeme than what was intended, excluding doublets there is very little variation among the productive stems apart from these mistakes. Nonce words and gaps exhibit more variation. Also significantly, subjects were generally less likely to agree on the form of the word for dental stems
(gaps, nonce and non-defective stems) than for other stem types.
In the following graphs, I calculated interspeaker agreement as a comparison of all pairs of responses for the same item across subjects. For example, the response for pobedit’ ‘to win’ for subject 13 (pobežu) is compared with the response by subject 11
(pobedju). They are different, which is coded as a 0. The response for subject 13 is also 212
compared with that of subject 14. These responses are the same, and are coded as 1.
Averaging across these individual comparisons thus provides a measure of speakers’ average agreement for a given item. For any given item there are 231 pair-wise comparisons assuming no erroneous responses (22 subjects, and thus 21+20+19...+1 comparisons). Variation in stress placement, vowel quality,84 and other minor differences were ignored. Errors were not included in the analysis. Non-responses, circumlocutions, examples of a second conjugation stem produced with a first conjugation inflection or vice versa, or other responses which are not errors but which could be considered possible avoidance strategies were included in the analysis but I treated them as a single response type because they were so various, and individually infrequent.
84 For example, okan’e was a notable characteristic of one subjects’ responses. Contemporary Standard Russian has a variety of vowel reductions. Okan’e is the failure to reduce the vowel /o/ to either /´/ or /ø/, a dialect feature in southern Russia. 213
Average speaker confidence ratio confidence speaker Average
square = gap (2nd conj. dental stem) circle = nonce 2nd conj. dental stem diamond = real 2nd conj. dental stem 0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Interspeaker agreement for 1sg word production, dental stems only
Figure 26: Interspeaker agreement in a cloze procedure task vs. speaker confidence for dental stems
214
Average speaker confidence ratio confidence speaker Average
circle = nonce palatal/1st conj. stem diamond = real palatal/1st conj. stem triangle = doublet 0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Interspeaker agreement for 1sg word production, non-dental stems only
Figure 27: Interspeaker agreement in a cloze procedure task vs. speaker confidence for non-dental stems
215
DENTAL STEMS PALATAL STEMS 1ST CONJ. STEMS MEAN MEDIAN MEAN MEDIAN MEAN MEDIAN NON-DEFECTIVE 87.0% 95.5% 100% 100% 100% 100%
NONCE 51.1% 47.4% 83.6% 90.5% 77.2% 83.3% GAPS 54.1% 51.6% DOUBLETS 63.7% 52.4%
Table 40: Average and median degrees of intersubject agreement in responses to a forced production sentence completion task, according to stimulus type
DENTAL STEMS PALATAL STEMS 1ST CONJ. STEMS MEAN MEDIAN MEAN MEDIAN MEAN MEDIAN NON-DEFECTIVE 84.1% 100% 87.7% 100% 89.2% 100% NONCE 43.7% 50% 47.6% 50% 48.8% 50% GAPS 70.7% 79.4% DOUBLETS 84.9% 100%
Table 41: Subjects' average and median confidence in their responses to a forced production sentence completion task, according to stimulus type
Comparing the graphs and tables, the most obvious difference lies in the real words.
Subjects produced non-defective, non-nonce dental stems sometimes with the expected alternation, and sometimes without it; for five stems there was 100% agreement but for the remaining five agreement ranged from 52.4% to 90.1%. By contrast, subjects uniformly agreed on the first person singular non-past forms for all twenty non-nonce palatal and first conjugation stems. A linear regression line calculated based upon interspeaker agreement rate and the mean confidence ratio accounts for 85.6% of the variation in the non-defective dental stems (p<0.001) – a surprisingly strong correlation 216
considering the few forms for which agreement was less than 100%. A regression line calculated using the median confidence ratio accounts for 96.5% of the variation
(p<0.001). I conclude that interspeaker agreement is somehow tied to speakers’ confidence.
This raises two questions: First, how should the interspeaker agreement rate be interpreted? Is it a representation of morphological variation, paradigmatic unpredictability, or something else? Second, what is the relationship between speakers’ reactions to these non-defective dental stems and gaps?
The answer to the first question is apparent from comparison of the two graphs above. Variation may be a necessary criterion, but it is not a sufficient one to explain speakers’ confidence ratings because the doublets are equally as variable as the gaps and nonce, but the three types of stimuli inspire different confidence ratios.85 Predictability is a better measure. Consider the differences between dental stems (with or without gaps) and doublets; an example of each is given below.
кудахтать кудахтать kudaxtat’ kudaxtat’ ‘to cackle’ SINGULAR PLURAL ‘to cackle’ SINGULAR PLURAL кудахчу кудахчем кудахтаю кудахтаем 1ST P. 1ST P. kudaxču kudaxčem kudaxtaju kudaxtaem куда'хчешь кудахчете кудахтаешь кудахтаете 2ND P. 2ND P. kudaxčeš’ kudaxčete kudaxtaeš’ kudaxtaete кудахчет кудахчут кудахтает кудахтают 3RD P. 3RD P. kudaxčet kudaxčut kudaxtaet kudaxtajut
Table 42: Example of stem reanalysis in Russian (morphological doublets)
85 A pair of two sample t-tests both produced highly significant results (p<0.001). 217
сузить сузить suzit’ suzit’ ‘to narrow’ SINGULAR PLURAL ‘to narrow’ SINGULAR PLURAL сужу сузим сузю сузим 1ST P. 1ST P. sužu suzim suzju suzim сузишь сузите сузишь сузите 2ND P. 2ND P. suziš’ suzite suziš’ suzite сузит сузят сузит сузят 3RD P. 3RD P. suzit suzjat suzit suzjat
Table 43: Example of leveling of morphophonological alternation in Russian
The variation in the doublets involves two analyses of the non-past stem shape. For
кудахтать / kudaxtat’ ‘to cackle’ some people use the non-past stem kudaxč- while others use kudaxtaj-. The latter is probably the newer stem, being a reanalysis based on the stem for the infinitive and past tense. The important thing to note here is that whichever stem shape formative is used, it is used throughout the non-past tense. By contrast, the two different attested forms of the dental stems differ only in the first person singular non-past – with or without the palatalizing alternation. The other non-past forms are the same.
This difference entails that a given first person singular non-past form of a doublet can be predicted from any other form of the non-past paradigm, and there is generally a high degree of predictability between any given pair of forms. While there may be variation within the community, within each variant paradigm the forms are therefore mutually reinforcing. By contrast, in the dental stems paradigm the form of the first person singular non-past cannot be predicted by nor is it predictive of the other 218
forms. In paradigmatic terms the first person singular non-past of dental stems is isolated, but the same paradigmatic cell within doublets is not. In this experiment the gaps and the doublets showed equal variability, and they are comparable in familiarity to the subjects. The lower confidence that subjects’ exhibited in gaps and non-defective dental stems, relative to their confidence in doublets, it thus best explained in terms of paradigmatic predictability. Nonce words are the least predictable since even the inflection class that they belong to cannot be predicted securely – there are no other paradigmatic forms to judge from.
This relationship between paradigmatic predictability and paradigmatic gaps brings the discussion to the second question raised by the data: What is the relationship between speakers’ reactions to non-defective dental stems and gaps? Based upon the variation in the first person non-past cell of dental stems, the way that this variation is paradigmatically isolated in such a way as to make it unpredictable, and the lack of confidence that speakers have in variable forms, it would be easy to conclude that the paradigmatic gaps in Russian are synchronically caused by speakers’ reactions to paradigmatic unpredictability. In the remainder of this chapter, I argue that paradigmatic unpredictability has a role to play, but that based on finer details of the subjects’ responses and historical evidence, this would be a hasty conclusion.
5.4.4. Paradigmatic connections, and the divergence of defective and non-defective
dental stems
While there may be some overlap in the confidence ratings and agreement rates for gaps and non-defective dental stems, the details show that the forms that speakers produced
219
were not always comparable for these two types of stimuli. Remember that the past passive participle, like the first person singular non-past, has a morphophonological alternation for dental stems. The relevant data are repeated below. The important example here is освободить / osvobodit’, which has different alternations in the first person singular non-past and the past passive participle.
(33) Parallelism, and non-parallelism, between alternations in the first person singular and the past passive participle
INFINITIVE 1ST SING PAST PASSIVE PART. GLOSS tj ~ čj оплатить оплачу оплачен ‘to pay’ oplatit’ oplaču oplačen
tj ~ šj сократить сокращу сокращен ‘to shorten’ sokratit’ sokrašču sokraščen
dj ~ ž зарядить заряжу заряжён ‘to load’ zarjadit’ zarjažu zarjažon
dj ~ ždj освободить освобожу освобождён ‘to become osvobodit’ osvobožu osvoboždjon free’
sj ~ š пригласить приглашу приглашён ‘to invite’ priglasit’ priglašu priglašon
zj ~ ž загрузить загружу загружён ‘to load’ zagruzit’ zagružu zagružon
As noted earlier, it is tempting make a single generalization to cover the alternations in the first person singular non-past and the past passive participle. Both apply to dental stems belonging to the 2nd conjugation class. Yet the difference in some stems ending in
/d/ shows that the generalizations are not exactly the same. This small degree of non- parallelism is significant because subjects repeatedly produced two of the test items,
220
победить / pobedit’ ‘to win’ and убедить / ubedit’ ‘to persuade’, with the stem final cluster /žd/ rather than the expected /ž/. Eight subjects produced either pobe[žd]u or pobe[ždj]u, nine produced pobe[dj]u, and seven produced pobe[ž]u.86 For убедить / ubedit’ six provided the response ube[žd]u, four said ube[dj]u, and eleven said ube[ž]u.
The ways in which speakers react to the gaps thus seem to reflect a generalization that when there is an alternation in the first person singular non-past, there is also an alternation in the past passive participle, and these alternations are usually the same.87
For the other four dental stems ending in /d/, the alternation in the non-past first person singular and the past passive participle would be expected to be the same, /ž/.
If paradigmatic gaps are specifically captured within the lexicon as a missing stem paradigm cell, stems with paradigmatic gaps lack information on the formation of the first person singular form, including “override” information when the alternations in the past passive participle and the first person singular non-past are not identical. The subject must analogically derive a form. Thus, we would expect speakers to produce any form according to any pattern which is represented in the lexicon. These two examples thus crucially show that for Russian speakers there is at some level an active, synchronic link between the first person singular non-past stem paradigm cell, and the past passive participle stem paradigm cell. Forcing speakers to produce forms where there are normally gaps causes speakers to rely on the structure of the lexicon to derive a form, and the alternation /žd/ is unexpected without a paradigmatic view of language. It is also
86 These numbers do not add to the number of subjects, 22, because some subjects provided multiple responses, and other subjects made errors (e.g. future tense or infinitival form instead of non-past tense). 87 The stem shape /...žd-/ also occurs in the imperfective pairs of these verbs: побеждать / pobeždat' and убеждать / ubeždat’, and this is likely a contributing factor favoring many speakers’ choice of /žd/ in the stem of their responses. However, their answers could not be interpreted as actually being the imperfective variant because the imperfective forms belong to the 1st conjugation, with the expected first person singular non-past forms побеждаю / pobeždaju and убеждаю / ubeždaju. 221
unexpected if gaps are treated as epiphenomenal to competition between forms with and without the relevant morphophonological alternation displayed by the non-defective stems. In short, this bit of data suggests that the gaps are idiosyncratic phenomena.
The reader at this point might be wondering whether the connection to the past passive participle is somehow causative of the gaps in the non-past tense, by virtue of creating competition where the alternations in the two cells are different. I think that this would be the wrong conclusion. Stems with /d/ and with /t/ are the locus of variation, so if this connection were the source of paradigmatic unpredictability, I would predict that gaps would disproportionately affect these stems. This prediction is not met. While stems with /d/ do have more gaps than any other stem, /t/ has nearly the fewest, both in raw numbers and proportionately to the total number of stems of the same type.
stem type # w/ gap total #88 % w/ gap /stj/ 3 70 4.3% /zj/ 10 275 3.6% /sj/ 15 447 3.4% /dj/ 19 779 2.4% /tj/ 6 652 0.9%
Table 44: Distribution of gaps relative to all stems with same final C(C)
Moreover, the consistency with which dictionaries mark the same groups of stems as defective suggests, albeit tentatively, that currently the gaps are driven more by
88 I took the numbers in this column from Baerman and Corbett (2006), who in turn took them from Ilola and Mustajoki (1989), who base their calculations on Zalizjak’s Grammatical Dictionary of the Russian Language (Zaliznjak 1977). 222
prescriptions about language use than by paradigmatic structure. I therefore follow
Baerman and Corbett, who argue that “defectiveness was motivated for one class of verbs, was morphologized and spread by analogy to others,” perhaps beginning with the root /bjedj/ (2006:5-6). The dental stems are a natural dimension along which to spread gaps, being a motivated grouping both phonetically and morphologically. In the following section I explore the relevant historical issues.
5.5. The development of the standard language as a historical source of
paradigmatic competition
While Contemporary Standard Russian more or less represents a codification of the speech of social elites, the development of the standard and literary languages was not always based in a spoken norm. Prince Vladimir’s conversion to Christianity in 988-9 on behalf of all Rus’ effectively started the literary tradition of East Slavic peoples and their language, including what would develop into the Russian language. Rus’ received liturgical books and teachers from South Slav regions, and while the language of these books was somewhat different from the local version of Slavic, it is generally accepted that the new linguistic features used for (church) writing were not so different as to constitute an impediment to understanding for Russians (Schenker 1995, Vinogradov
1969, Vlasto 1988).
While this language, (Old) Church Slavonic, is no longer used as a basis for the literary language, its influence is still strongly present in Contemporary Standard
Russian, most commonly in the areas of vocabulary and (morpho-)phonology. The relevant example for the present purposes is the reflexes of Common Slavic *dj and *tj –
223
Russian developed /ž/ and /č/, while (Old) Church Slavonic developed the reflex /žd/ and
/šj/. In Modern Russian, there are numerous pairs of words/stems, one with the native
Russian reflex, and one with the Church Slavonic reflex. These forms are often
stylistically or semantically differentiated.
CHURCH SLAVONIC NATIVE RUSSIAN REFLEXES GLOSS REFLEXES GLOSS сопровождать провожать ‘to accompany’ ‘to see off’ soprovoždat' provožat' возбуждённый разбужённый ‘arouse (p.p.p.)’ ‘rouse (p.p.p.)’ vozbuždjonnyj razbužjonnzj тождество тожество ‘identity’ ‘identity’ toždestvo tožestvo еженощный ночной ‘nightly’ ‘nocturnal’ eženoščnyj nočnoj просвещать просвечивать ‘to pass through (of ‘to illuminate’ prosveščat' prosvečivat' light)’
Table 45: Church Slavonic and native Russian reflexes, pairs of stems in Contemporary Standard
Russian (taken from Vlasto 1988:14)
In general, morphological influence from Church Slavonic (ChSl) was minor, but the
ChSl. reflex became widespread in the past passive participle, in large part because the
entire grammatical category was borrowed into East Slavic. The first person singular
non-past seems to have maintained the native Russian reflex, setting up the partial non-
parallelism between the alternations in the these two paradigmatic cells, as presented in
(31) and again in (33), and a potential site of grammar competition.
224
The Russian language underwent substantial changes in the 13th and 14th century and the South Slavic languages had a renewed influence during this same time on the language of East Slavic church writing (the Second South Slavic Influence), causing a significant rift between spoken and written Russian. As Vlasto (1988:366) notes, “By the beginning of the 17th c. Ch[urch] Sl[avonic] had everywhere – whether in Moscow,
Kiev, or Vilna – to be acquired with much study, and therefore a codification of ChSl. grammar became for the first time in its ESl. history an urgent desideratum.” For the first time in Russian, the written norm was both imbued with high social prestige, and was also inaccessible to the average literate person.
From the 17th to the 19th century, the place of Church Slavonic in the emerging literary language was a source of hot debate, characterized by alternating and overlapping waves of conservative return to an archaic, Church Slavonic-based form, and the incorporation of colloquial features (Vinogradov 1969). Lomonosov’s work was particularly influential, advocating what eventually became the basic outline of the modern standard language. According to Vlasto, Lomonosov’s basic premise (Риторики
(1744) and Российская грамматика (1755)) was that
The ‘middle style’... should use this common stock of words with a judicious admixture of Slavonicisms. The ‘low’ style... will avoid virtually all Slavonicisms. The ‘high’ style will use Slavonicisms liberally, including new words formed on ChSl. models, with one important proviso: all such Slavonicisms must be immediately intelligible in Russian terms. (Vlasto 1988:381-2)
225
This proposal, however, potentially caused more problems than solutions for the average literate Russian at the end of the 19th century. As Vinogradov argues, as a result of
Lomonosov’s influence and other coinciding trends,
The grammatical system of the literary language also underwent great changes in the second half of the [19th] century. These were of two kinds. Grammatical doctrine, which had condemned the ‘folkish’ features of the older grammar... nonetheless still allowed those conversational grammatical forms which did not sharply contradict the orthoepic system of the written language. The drive to establish national standards for grammatical categories thus became stronger. On the other hand, however, the tendency toward bookishness in scholarly and journalistic writings, and the instability of the boundaries between the written and conversational languages, led to a development of new grammatical forms based on the old categories of the literary language. A process of granting equal literary rights to opposing grammatical categories thus came into being (Vinogradov 1969:262).
Thus, in the second half of the 19th century, the linguistic state of things is perhaps best described as chaotic. There was still a strong influence from Church Slavonic alongside a growing acceptance of grammatical features based in colloquial language of one sort or another. This was all combined with an increasing valuation of adherence to a prescriptivist norm which was sometimes unclear due to “unstable” stylistic norms.
Finally, “...writers now came from very different social backgrounds, and the choice of subjects had broadened. The hypertrophy of artificial bookishness coexisted with a democratic broadening of the literary language” (Vinogradov 1969:254). These conditions were ripe for a large class of Russian speakers who felt insecurity over social language norms of which they had little command. Combined with paradigmatic
226
unpredictability, these are exactly the conditions which I would expect to promote the development of gaps.
In this light, it is probably not accidental that one of the characteristics of colloquial speech which was debated in the process of normativization was non- alternation in the first person singular non-past of dental stems. Vinogradov (1969:267) states that by the mid-nineteenth century, in city dialects morphophonological alternations had been leveled, including among dental stems, e.g. гордюсь / gordjus’ instead of горжусь / goržus’ ‘I am proud of’. This is the same pattern that I found in the forced choice experiment described above. Of course, the standard language ultimately did not accept these stem levelings. But at the time that the standard language was codifying, the non-alternating forms were in contention with the alternating forms, and both vied for status as standard language variants. And neither would have been paradigmatically predictable. This would have made the first person singular non-past a weak point in the non-past inflectional paradigm, all the more so because of the almost- but-not-quite parallel alternations in the first person singular non-past and the past passive participle.89
89 There is one more piece of evidence which is too speculative to put too much weight behind, but which is nonetheless intriguing: “In official, scholarly, and publicistic styles it was occasionally necessary to obliterate or blur the individualized, often familiar and concrete, depiction of action which was frequently a property of the simple form of the verb. For this purpose, a periphrastic construction was used, consisting of a more or less abstract verb, which designated the action in general (or which had lost all or nearly all concrete meaning of any sort), and a deverbative noun, which designated the specific content of the action... Sometimes these constructions came from the Church Slavonic tradition (одержать победу [=oderžat' pobedu ‘to score a victory’]), and sometimes they were calques of West European phraseology... Synthetic forms of expression were replaced by analytical ones – on the model of West European languages” (Vinogradov 1969:250-51). If there was already a strong tendency to not use verbs such as pobedit’, this may have opened a wedge to a more general pattern of avoidance, and the subsequent appearance of gaps. 227
While it is always difficult to equate historical attestation with historical development, there is some indication that gaps began appearing in the aftermath of unstable norms in the first person singular non-past. Ushakov’s (1935-1940) dictionary was the first to cite a gap, in the verb победить / pobedit’ ‘to lose’. This is also the lexeme most frequently cited in more recent dictionaries and grammars as having a gap, and thus potentially (as Baerman and Corbett argue) the original locus of gaps and the point from which others spread. The timing of Ushakov’s attestation fits well with the social and grammatical conditions surrounding the first person singular non-past form which I would expect to promote gaps.
This timing is significant for the issue of whether defective stems arose from productive ones, or whether they were always defective. Baronian (2005) notes that many of the defective stems have origins in Church Slavonic, most notable those with the root /-bed/, or are denominal. He argues based on a phonologically-based nesting of lexical entries that these stems could not be clearly assigned to one inflectional pattern or another, and as a result are defective. Baronian’s approach has some similarities to my own in the sense that it is rooted in competition between inflectional patterns, and an inability to resolve it in favor of a single pattern.90 The differences between his approach to the Russian gaps and my own, however, illustrate the more general problems with a mechanistic, deterministic view of defective inflection. Baronian’s conclusion is that
90 I do not delve into the similarities and differences between the two accounts because Baronian’s analysis rests on a misunderstanding of the relationship between letters and sounds in Russian, and consequently fundamentally misrepresents the inflectional class divisions for Russian verbs. Since his underlying description of the Russian morphological system is unsustainable on the facts, there is no point in disputing in detail an analysis of gaps which rests exclusively on that description. I discuss his analysis here at all as a way to talk more generally about mechanistic as opposed to sociolinguistically informed approaches to gaps, his approach representing the former and mine the latter. 228
...it is not surprising that Russian speakers have no inherited 1Sing Present forms for these verbs. In the case of the denominal, a 1Sing form never existed, while in the case of the Slavonic borrowings, they were probably borrowed via their Infinitive forms, and given the modern Russian inflections, which is problematic for the 1Sing Present. (2005:164)
There is no basis for his claim that a first person singular non-past form never existed for the denominal verbs, since this grammatical category is realized for a wide variety of denominal verbs, and the currently defective stems by and large preexisted any indication of defectiveness. There is also no foundation for Baronian’s assumption that the lexemes of Church Slavonic origin were inherited via the infinitival form, considering the pervasive use of Church Slavonic as a written norm for several hundred years. In fact, considering that there is every indication that past participle forms were borrowed directly, it seems improbable that first person singular non-past forms would be somehow unborrowable. However, if Baronian were to allow that forms/stems could be borrowed and be fully productive, only later becoming defective, his explanation would fail because it rests entirely on the hypothesis that paradigmatic gaps are the inability of speakers to categorize stems into the appropriate inflection class. Stated differently, he treats gaps as synchronically motivated epiphenomena mechanistically derived from the structure of the lexicon. If the structure of the lexicon is unproblematic, Baronian’s account, and others that are equally mechanistic, are incapable of explaining the gap.
By contrast, if we allow that gaps are only indirectly the result of grammar competition, with the direct causes being speakers’ insecurity and subsequent avoidance strategies, then we have no need to assume that defective stems have been defective since their introduction to the language. If unpredictability and social stigma align, in my
229
approach we could easily find (and in fact have found) the development of defective stems out of previously productive ones.
The best explanation of the Russian verbal gaps, and I would claim all gaps, is thus best understood relative to the ways that speakers use language structure, and not as a mechanistic result of language structure. Furthermore, Russian provides a classic example of correlation not equating to causation. While there is some indication of paradigmatic competition in modern language which may serve to reinforce the presence of a gap, the best explanation for the causes of the gaps lies in the history of Russian, and not in its present structure. The correlation between a palatalizing alternation and the paradigmatic gaps is, as far as the synchronic structure is concerned, largely a coincidence, reflecting that variants with the palatalizing alternation became the prestigious variant.
5.6. Summary
In this chapter, I have made two general points.
First, paradigmatic gaps are not always motivated by the synchronic structure of the language. Researchers have been interested in the Russian verbal gaps in large part because of the distributional parallels between the palatalizing alternation and the paradigmatic gaps. This has led some previous accounts to try to find some structure of the language which predicts (causes) the gaps. However, from the evidence that I presented here, it seems more likely that paradigmatic gaps have a historical source in paradigmatic competition arising from the influence of Church Slavonic, stem leveling, and the preexisting alternation. Russian speakers treat gaps differently from regularly
230
inflected forms along a number of parameters. This entails that our model of morphology must specifically encode information about the gap, and cannot reduce it to epiphenomenal principles. In short, the Russian data provides much greater force to the conclusions that I drew regarding Modern Greek in CHAPTER 4. The generalization of
/ž/ in the first person singular for all 2nd conjugation stems with /d/ removed paradigmatic unpredictability, making the Russian gaps the seeming mystery that they are.
Second, the gaps reveal paradigmatic connections between the first person singular and the present passive participle cells in the stem paradigm. The appearance of the alternant /žd/ among (only) the items which also have the alternation /žd/ in the past passive participle looks almost like a throwback to a previous state of the language, and this variant is unexpected if we look only at other first person singular forms for a model.
However, if we look farther afield, the appearance of this alternation makes perfect sense.
This strongly supports not only a paradigmatic theory, but in particular a theory in which generalizations exist at successively broader levels. On a local scale the first person singular cell, like the umlauting patterns in German, is governed by a separate generalization than the past passive participle. And on a more broad level they are connected. The Russian gaps thus provide evidence for a very specific version of a Word and Paradigm model.
231
CHAPTER 6
CONCLUDING THOUGHTS
In the preceding chapters I have tried to shed light on the nature of the relationship between paradigmatic gaps and paradigmatic structure. Traditionally, gaps have been treated as curious but isolated phenomena within the larger linguistic system. More recently, there has been a surge of interest in trying to explain gaps in exactly the opposite fashion, as epiphenomenal to general grammatical principles. I argued that neither approach is adequate.
I showed that in both Modern Greek and Russian, there is a strong connection between the structure of the lexicon and the appearance of paradigmatic gaps.
Specifically, those inflected forms which are paradigmatically unpredictable based on lexical organization are prone to developing gaps, if the appropriate sociolinguistic conditions exist. Previous research on grammar competition models as a source of paradigmatic gaps, most notably by Albright (2003), had focused on morphological variation, but in both Modern Greek and in Russian there are examples of morphological variation which failed to develop gaps. This shows that variability may be a necessary criterion for a language to develop defective points in the inflectional system, but it is not a sufficient one. The morphological doublets in Russian are highly variable across speakers, but fully predictable, given any other form of the non-past inflectional
232
paradigm. In Modern Greek two inflectional classes show equal internal variability with regard to stress placement in the genitive plural, but for one class the form of the genitive plural is predictable from the genitive singular while for the other class it is not predictable. The latter class is the locus of 98% of the paradigmatic gaps in the genitive plural. This shows us that predictability entails variation, but is narrower in scope. Since predictability is an inherently paradigmatic concept, the data highlights the ways in which gaps are dependent upon inflectional structure, and supports a model of the lexicon in which stems form a network connecting paradigmatic cells and entire paradigms to each other.
The fact that gaps are dependent upon paradigmatic inflectional structure should not be taken as equivalent to saying that paradigmatic structure causes gaps, however.
An area of paradigmatic unpredictability represents a weak point in an inflectional system, but speakers’ reactions to that weak point – insecurity and subsequent avoidance
– are the direct causes of paradigmatic gaps. This is perhaps my point of greatest departure from previous research on paradigmatic gaps, which has tended towards a mechanistic approach. In straightforward terms, mechanistic approaches overpredict the appearance of gaps because they fail to be able to distinguish between multiple possible resolutions to the same type of weakness in inflectional structure. In response to paradigmatically unpredictable stress in the genitive plural of Modern Greek speakers could have leveled the stress, rather than avoided the paradigmatic cell. Russian speakers could have accepted variation in the first person singular non-past, much as variation in the morphological doublets is largely accepted. The grammatical structure of the language cannot distinguish between these outcomes because they are fundamentally
233
rooted in social structure and speakers’ beliefs about their language. A theory which only attends to the grammatical structure is therefore necessarily inadequate. This dissertation is not a sociolinguistic study, but by taking a more sociolinguistically informed approach than previous research on paradigmatic gaps, the result here has been a better understanding not only of the technical relationship between paradigmatic gaps and inflectional structure, but also of how gaps arise, and why they exist.
There is much work still to be done. I have touched upon only a few examples of gaps, and of only one type. Gaps related to phonotactic restrictions remain a separate issue, as do semantically-driven gaps. As is so often the case, answering one question raises five more. For example, one of the more significant questions arising from this dissertation relates to the expansiveness of paradigmatic predictability as an explanation for language change and synchronic linguistic structure. But as the first book-length study of paradigmatic gaps, my primary goal has been to lay a foundation for future study of the topic. If nothing else, it is clear that while paradigmatic gaps are defective, they are not marginal to the linguistic system, and therefore deserve more theoretical attention than they have traditionally garnered.
234
APPENDIX A
A SURVEY OF PARADIGMATIC GAPS
Three kinds of paradigmatic gaps (in the most narrow sense of the term, as discussed in detail in CHAPTER 1) have been documented by previous studies, arbitrary lexical gaps, phonotactically-induced gaps, and semantically induced gaps. In this Appendix, I catalogue paradigmatic gaps of the first two types, as they have been relevant to the theoretical literature. I exclude semantically driven gaps for two reasons. First, on the practical side, there are simply too many of them. As Karlsson notes (2000:648), “this is the most frequently recurring example of inflectional defectivity.” Languages often have thousands of singularia tantum nouns alone. Second, on the theoretical side, gaps which are the result of purely semantic issues are simply uninteresting for my purposes in this dissertation. While our theory should be able to account for them, semantically caused gaps shine little light on the interaction between defective inflection and inflectional structure. The interested reader is referred to Karlsson (2000) for examples of semantically-related paradigmatic gaps, and to Aristar (1997) for a particularly interesting example of the subtle and wide-ranging ways that semantics can lead to gaps in inflectional structure.
For arbitrary lexical gaps and phonotactically-induced gaps, I try here to make a thorough survey of the literature, presenting that data which has been a significant focus
235
of theoretical discussion. With this statement come three disclaimers. First, to the extent possible, I have attempted to maintain a descriptive approach, avoiding theoretical issues.91 However, the distribution of gaps cannot always be fully separated from the theory used to describe it, so certain biases (my own or those of the original source) may seep through at times. Second, since this collection was culled from other sources, none of it being originally collected by me, the descriptions are as good, or as bad, as those in the original sources. Third, this is only a survey of the theoretical literature. For the descriptions below I have not surveyed grammars, dictionaries or other descriptive work for evidence of paradigmatic gaps. This data is thus almost certainly (and woefully) incomplete, and is best considered an assessment of the basis for previous theoretical work, not a typologically accurate summary of paradigmatic gaps. Considering that the theoretical literature has overwhelmingly revolved around a limited range of examples, I would expect a true typological survey of paradigmatic gaps to uncover a far greater number and variety of gaps than what I present here. Such a survey is, however, beyond the scope of this dissertation.
The following descriptions are alphabetized by language. Language family designations are according to www.ethnologue.com.
A1. Arabic
Language family: Afro-Asiatic, Semitic, Central, South Source: Hetzron (1975)
91 Discussion of the theoretical significance of these gaps may be found in the main body of the dissertation. 236
As is well known, Arabic, like many Semitic languages, has verbal roots consisting of consonantal sequences. Inflectional (and derivational) morphology is realized through templates which specify consonant slots and vowels, onto which the verbal roots are mapped. In some cases, however, the combination of a consonantal sequence and a morphological template may trigger phonological processes which produce phonotactic problems, and ultimately gaps.
In particular, problems in some Modern Arabic dialects result from verbal roots in which the final two radical consonants are the same, what Hetzron calls 1-2-2 roots, e.g. m-d-d ‘stretch’, s-r-r ‘make happy’, š-d-d ‘tighten’. These roots “...undergo contraction when there is a short vowel between the last two radicals. The root š-d-d can be used to illustrate the contractions...”
(2) Contraction in Classical Arabic 1-2-2 verbs (Hetzron 1975:862)
a. /šadada/ → [šadda] ‘he tightened’ b. /yašdudu/ → [yašuddu] ‘he tightens’ c. /ša:didun/ → [ša:ddun] ‘a tightener’ d. /ša:dada/ → [ša:dda] ‘he persecuted’
“Normally, Classical Arabic admits long vowels only in open syllables. When a long vowel would be expected to occur in a closed syllable, it has to undergo vowel- shortening... [But in light of the above data], one has to reformulate the statement about the possible positions for a long vowel: it may occur in an open syllable, but also in a closed syllable when that syllable has been closed by a contraction” (Hetzron 1975:862).
237
Many of the modern dialects have lost word final vowels under certain conditions
(which are unclear from Hetzron’s description). “For 1-2-2 verbs, we would expect forms like *ša:dd for both [c] and [d], with the uncomfortable combination of a long vowel in DOUBLY closed word-final syllable. The modern dialects have reacted to this situation in different ways... In many dialects, no III forms at all may be derived from 1-
2-2 roots – e.g. in eastern Maghrebi dialects... and in Mauritian... [S]ome dialects do exhibit a conspicuous absence of the functionally more important (and normally fully productive) active participle for simple (I form) 1-2-2 verbs... Finally, for several
Maghrebi dialects, in Algeria in particular, the complete absence of an active participle for 1-2-2 roots in the I form has been noted” (Hetzron 1975:862-3).
In short gaps among 1-2-2 verbs in the active participle, and potentially other forms, are related to an (excessive) violation of the phonotactic principle that long vowels do not occur in closed syllables. The violating conditions are created by the inflectional processes of the language, in particular the contraction.
A2. Chaha
Language family: Afro-Asiatic, Semitic, South, Ethiopian, South, Outer, tt-Group, Sebat Bet Gurage
Source: Fanselow and Féry (2002), citing Rose (2000)
In Chaha (spoken in Ethiopia, according to Ethnologue), the length of the template for frequentative verbs leads to gaps for some lexemes. Frequentative verbs (i.e. those that express a repeated action) are normally realized by the template CÈCVC´C, in which the
238
C’s are specified by the verbal root, and the V is [´] (treated theoretically as having been copied from the final syllable, and therefore underspecified in the template). “When the consonantal root has three consonants, the second one is copied to fill the extra C slot...
In the case of diliteral root verbs... there is no grammatical output, because no tri-linking or no double copy of Cs is allowed. As a result, one of the C slots cannot be filled, and there is a failure to form frequentative verbs. This is also true of quadriliterals, but for a different reason. Here, the frequentative is indistinguishable from the regular form.
Again, no repair applies” (Fanselow and Féry 2002:manuscript p. 15).
(3) Paradigmatic gaps among frequentatives in Chaha
Root Regular Gloss Frequentative Gloss sbr sab´r´ ‘break’ sÈb´r´r ‘break in pieces’ mzr mez´r´m ‘count’ mÈz´r´r ‘count again’
nd n´d´d ‘burn’ *nÈd´d´d ‘burn again’ t’m t’´m´m ‘bend’ *tÈm´m´m ‘bend again’
(no examples of ungrammatical quadriliteral frequentatives are provided be
Fanselow and Féry (2002))
In more descriptive terms, a frequentative verb is ungrammatical if in the expected form the final three consonants would be identical, or if the frequentative and the non- frequentative would be homophonous. No other synthetic (i.e. “repaired”) forms are used, so a paradigmatic gap in the verbal paradigm results. The gaps appear specifically in the frequentative form not because of any semantic features of this form, but rather due
239
to the interaction between the length of the inflectional template, the length of the verbal root, and (morpho-?)phonological constraints on surface forms.
A3. English
Language family: Indo-European, Germanic, West
Paradigmatic gaps in English have been most frequently noted in the literature, and a wider variety of instances have been cited. Here I present six examples which have been garnered theoretical attention.
A3.1. Comparatives and Superlatives
Main sources: Aronoff (1976), Zwicky (1989) Other citations: Bresnan (1973), Graziano-King and Cairns (2005), Karlsson (2000), Kiparsky (1982), Stump (1991)
In English, comparatives of adjectives and adverbs are realized by suffixation (X + -er)92 and/or by a periphrastic form (more X). The superlative has the same distribution (X + - est vs. most X). Some lexemes have both synthetic and periphrastic forms, but most have only one of these. Where only periphrastic forms exist, it is possible (and has been common) to view adjectival paradigms as having paradigmatic gaps in the comparative and superlative forms.
While judgments as to the grammaticality of one or another formation of the comparative/superlative tend to be fairly uniform, there have been several overlapping but not entirely congruous generalizations made over the data. Here I primarily follow
92 Whether <-er> attaches to an adjective, an adverb, or both has been the subject of much debate. See Zwicky (1989) for an overview of the theoretical issues. 240
Zwicky (1989), who in turn drew many of his descriptive generalizations from Aronoff
(1976).
Monosyllabic adjectives and adverbs usually have synthetic comparative and superlative forms. Periphrastic forms are usually not grammatical. There are exceptions and some variation in judgments.
(4) Synthetic English comparatives, and an exception to the rule (monosyllabic base)
basic form comparative superlative big bigger / *more big biggest / *most big fast faster / *more fast fastest / *most fast deep deeper / *more deep deepest / *most deep quick quicker / *more quick quickest / *most quick
apt *apter / more apt *aptest / most apt
Disyllabic adjectives and adverbs in –y (and especially –ly) typically allow both synthetic and periphrastic forms (Aronoff 1976), as do disyllabic words ending in -er (Zwicky
1989). But again, exceptions exist, usually in the form of excluding the synthetic forms.
(5) Synthetic and periphrastic English comparatives (disyllabic base in -y, -ly or -er)
basic form comparative superlative lovely lovelier / more lovely loveliest / most lovely silly sillier / more silly silliest / most silly happy happier / more happy happiest / most happy early earlier / more early earliest / most early tender tenderer / more tender tenderest / most tender
quickly *quicklier / more quickly *quickliest / most quickly clever *cleverer / more clever * cleverest / most clever
241
Other adjectives and adverbs usually have only the periphrastic comparative.
(6) Periphrastic English comparatives
basic form comparative superlative flagrant *flagranter / more flagrant *flagrantest / most flagrant correct *correcter / more correct *correctest / most correct active *activer / more active *activest / most active tyrannical *tyrannicaler / more tyrannical *tyrannicalest / most tyrannical
Apt, quickly, clever, flagrant, correct, active, tyrannical, and the numerous other forms which do not have synthetic comparative forms can be considered to have a gap in the inflectional paradigm of English adjectives and adverbs (*apter, *quicklier, *cleverer,
*flangranter, *correcter, *activer, *tyrannicaler).
A3.2. Negated copula
Source: Hudson (2000) Other citations: Bresnan (2001), Dixon (1982), Frampton (2001), Gazdar et al. (1982), Hudson (1977), Langendoen (1970), Sells (1985), Zwicky and Pullum (1983)
Several researchers, among whom Hudson (2000) is probably the most prominent, have argued that there is a paradigmatic gap in the first person negated form of the copula
(*amn’t). The details present a more complicated case than many of the other gaps presented here, however, and in the following discussion I follow Hudson’s analysis relatively loosely.
The English negator not has a contracted form n’t [nt], which can suffix to the copula, among other verbs, assuming that the copula is not itself in a contracted form.
242
The form -n’t is most frequently treated as a clitic resulting directly from phonological reduction of not. However, Zwicky and Pullum (1983) argue convincingly that -n’t actually shares more properties in common with affixes than with clitics. This observation has two consequences: First, as argued by Zwicky and Pullum, -n’t should be treated theoretically as an affix which has only an indirect relationship to not. Second, as relevant to paradigmatic gaps, if -n’t is an inflectional suffix and we believe in a paradigmatic theory of inflection, the morphology must contain separate paradigmatic cells for the positive and negative forms of the copula (and other verbs to which this affix attaches). English inflects for three persons and two numbers (although there are not six distinct verbal forms...), and negated forms are available for five of the resulting inflectional cells in American English and many other varieties.93 The first person singular is the exception. There is a gap in this cell of the negated, non-inverted paradigm.
(7) Negated (non-inverted) English copula
a. She’s not your friend. b. She isn’t your friend. c. I’m not your friend. d. *I amn’t your friend. e. *I aren’t your friend.
93 As Hudson (2000) and Bresnan (2001) note, there are varieties of English which have amn’t or some other form, rather than a gap. 243
positive negated, non-inverted syntax singular plural singular plural 1st person am are 1st person -- aren’t 2nd person are are 2nd person aren’t aren’t 3rd person is are 3rd person isn’t aren’t
Table 46: Present tense indicative paradigm of be
Note, however, that when the subject and be are inverted, such as in tag questions, be does not have a gap – the first person singular cell is filled by aren’t (what Frampton
(2001) calls “anomalous” aren’t).
(8) Negated (inverted) English copula
a. She’s your friend, isn’t she? b. I’m your friend, aren’t / (*amn’t) I? c. I’m your friend, am I not?
negated, inverted syntax singular plural 1st person aren’t aren’t 2nd person aren’t aren’t 3rd person isn’t aren’t
Table 47: More present tense indicative paradigm of be
Thus, the status of *amn’t as a paradigmatic gap rests not only on the (largely descriptive) claim that -n’t is an affix and not a clitic, but also on assumptions about the
244
workings of the syntax. If different inflectional forms of the same lemma may be called depending upon the syntactic structure, or alternatively if the morphology is allowed to create separate paradigms based on syntactic structure (as is descriptively given above with one paradigm for “inverted syntax” and one for “non-inverted syntax”), we must treat *amn’t as a paradigmatic gap. And inasmuch as there is good reason to do so,
*amn’t stands as an example of paradigmatic defectiveness in English.
A3.3. Past Participle
Citations: Baronian (2005), Frampton (2001), Lightfoot (1979), Woolf (1980)
Some English verbs have well-known problematic past participle forms. Stride may be the most famous among these. While it is similar in form to words with fully productive forms throughout the paradigm, such as strive, ride, and hide, many speakers are uncomfortable with the potential past participle form(s) for stride, but not strive, ride, hide, etc. There are no obvious phonological, morphological or syntactic features which cause such a gap.94
(9) An example of a gap in the past participle: stride
infinitive present participle past participle to ride riding ridden to hide hiding hidden to strive striving strived to stride striding *strided, *stridden
94 But see Baronian (2005), who argues that the Theory of Connected Word Constructions is able to predict the gap in stride. 245
A3.4. Beware
Source: Fodor (1972)
The verb beware has only one form, variously considered to be either a “base” form or the infinitive. Fodor (1972:531) argues that “The real generalization about beware is that it can occur wherever uninflected be followed by an adjective can occur, e.g. in imperatives, infinitival complements, following modals.” Beware has paradigmatic gaps elsewhere (e.g. present and past participles, all indicative forms).
(10) Sample constructions allowing beware
a. Beware of the dog. b. Do beware of the dog. c. I will beware of the dog. d. I will have to beware of the dog. e. You must beware of the dog.
(11) Sample constructions NOT allowing beware – paradigmatic gaps
a. *John’s bewaring of the dog was unnecessary. b. *John bewares of the dog. c. *John bewared / bewore of the dog. d. *John doesn’t beware of the dog. e. *John has bewared / beworn of many dogs in his lifetime. f. *John is bewaring of your dog. g. *Bewaring of the dog, John circled the yard.
A.4. Finnish
Language family: Uralic, Finno-Ugric, Finnic Source: Fanselow and Féry (2002)
246
Finnish negation presents a possible case of a paradigmatic gap, but the jury remains out, so to speak, on the following example due to insufficient information about the lexical category of the relevant word and the inflectional structure of Finnish. I include this example provisionally until I am able to uncover the necessary details.
Fanselow and Féry (2002:manuscript p. 13) note the following about Finnish negation: “... [N]on-finite sentences in this language cannot be negated by using the negation word ei [(12)a], because this word must carry subject agreement, but cannot do so in an infinitive. One has to use an infinitival form with a suffix meaning ‘without’ for expressing meanings like ‘I promised not to go out.’ [(12)b]. Since, obviously, ei does not have an infinitive form, we are again confronted with a case of a defective paradigm with consequences for syntax.”
(12) Avoiding gaps in the paradigm of ei ‘not’ (Fanselow and Féry 2002:manuscript p. 13)
a. *Lupasin ei mennä ulos promised.1SG not go out ‘I promised to not go out.’
b. Lupasin olla menemättä ulos promised.1SG be go-INF-WITHOUT out ‘I promised to not go out.’ [lit. ‘I promised to be without going out.’]
The issue here is whether there is a reason to expect
Paradigmatic gaps are defined only with respect to an expectation (e.g. we do not consider the English verb be to have a paradigmatic gap in the dual because English does
247
not have DUAL as an inflectionally-expressed grammatical category). If ei is a verb, we might reasonably expect an infinitival (non-agreeing) form because that is a category which is normally expressed by the verbal inflectional system of Finnish. However, if ei belongs to a different lexical category which does not normally realize a non-agreeing form, there is no clear basis for saying that ei has such an inflectional cell to fill. In short, there would be no basis for saying that (a) results from a defective paradigm, although it would certainly represent defectiveness of one sort or another. Fanselow and Féry do not provide enough information about Finnish to be able to assess this case.
A.5. French
Language family: Indo-European, Italic, Romance, Italo-Western, Western, Gallo- Iberian, Gallo-Romance, Gallo-Rhaetian, Oïl Sources: Baronian (2005), Fanselow and Féry (2002), Morin (1987)
There is some disagreement as to which forms are affected, but the sources seem to agree that some French verbs, most notably clore ‘to shut’ and frire ‘to fry’, have gaps in multiple verbal forms. Fanselow and Féry are the most restricted in their claims: “These verbs have no first and second person plural forms, though the other forms are regular.
The non-existent *nous closons / vous closez ‘we/you shut’ and *nous frions / vous friez
‘we/you fry’ are phonologically well formed, as similar forms nous cuisons / vous cuisez
‘we/you cook’ and nous rions / vous riez ‘we/you laugh’ testify, but are nevertheless absent from the vocabulary of most French speakers” (Fanselow and Féry
2002:manuscript p. 9).
However, Baronian (2005:147) claims that “...the French verb clore ‘bring closure’ has no 1PLUR or 2PLUR present or imperative forms, nor any imperfect or simple
248
past forms. According to Morin (1987), some speakers are also lacking a 3PLUR present form... The verb éclore ‘hatch’ is said to be used only in the 3SING, but this is most likely a semantic restriction due to the fact that animates cannot hatch. The verbs forclore
‘exclude’ and déclore ‘open up’ are said to only have infinitive and past participle forms...”. The verb frire ‘to fry’ follows the same pattern.
And finally, the reference book 501 French verbs (Kendris 1996) gives the following set of forms for frire. Note that all of the gaps cited by Fanselow and Féry,
Baronian, and Morin are included here, and the subjunctive is added. Leaving aside the disagreement over which forms actually show gaps, verbs like clore and frire (and it is unclear whether there are few or many) represent an obvious case of so-called “arbitrary lexical gaps”, since there are no apparent phonological, semantic or syntactic reasons for the gaps. The missing forms should be well formed at all levels.
249
present indicative conditional singular plural singular plural 1st person fris -- 1st person frirais fririons 2nd person fris -- 2nd person frirais fririez 3rd person frit -- 3rd person frirait friraient
imperfect indicative present subjunctive singular plural singular plural 1st person -- -- 1st person -- -- 2nd person -- -- 2nd person -- -- 3rd person -- -- 3rd person -- --
simple past imperfect subjunctive singular plural singular plural 1st person -- -- 1st person -- -- 2nd person -- -- 2nd person -- -- 3rd person -- -- 3rd person -- --
future imperative singular plural singular plural 1st person frirai frirons 1st person fris -- 2nd person friras frirez 2nd person faisons frire -- 3rd person frira friront 3rd person faites frire --
Table 48: Simple tenses of the French verb frire 'to fry', with numerous paradigmatic gaps (Kendris
1996)
Whether they are viewed as the direct and predictable result of morphological structure depends on theoretical persuasion. See Baronian (2005) for an argument that gaps are the expected result of a particular structure of the lexicon. See Fanselow and Féry (2002) for the (more mainstream) view that these gaps are idiosyncratic.
250
A6. Hebrew
Language family: Afro-Asiatic, Semitic, Central, South, Canaanite Source: Rice (2005a), citing Becker (2003)
“In Modern Hebrew, there is a correlation between word size, perceived native status of a word, and stress mobility (Becker 2003). Words with mobile stress are maximally disyllabic and perceived as native... Longer stems have come into the language through borrowing, but these all have fixed stress. There are also shorter words lacking perceived native status, and these have fixed stress as well... Becker (2003) argues that ineffability arises with words which have more than two syllables, but which have some surface phonological properties of the native vocabulary, such that they are perceived as native”
(Rice 2005a:4-5). Specifically, lexemes of this type lack plural forms.
251
(13) Plural formation in Hebrew
a. With mobile stress, mono- or disyllabic (perceived as native)
dód ‘uncle’ dod-ím ‘uncles’ dikdúk ‘grammar’ dikduk-ím ‘grammars’
b. With fixed stress, various syllable lengths (perceived as borrowings)
tút ‘strawberry’ tút-im ‘strawberries’ pílpel ‘bell pepper’ pílpel-im ‘bell peppers’ diktátor ‘dictator’ diktátor-im ‘dictators’ fonológ ‘phonologist’ fonológ-im ‘phonologists’ kópirayter ‘copywriter’ kópirayter-im ‘copywriter’
c. With gaps in the plural (multisyllabic but perceived as native)
duxifát ‘hoopoe’ *duxifát-im / *duxifat-ím /izdaréxet ‘Chinaberry tree’ */izdaréxet-im / */izdarexet-ím
A7. Hungarian
Language family: Uralic, Finno-Ugric, Ugric
Main sources: Hetzron (1975), Törkenczy (2002) Other citations: Dressler and Ladanyi (1998), Fanselow and Féry (2002), Rice (2005a, 2005b)
Hungarian has been taken as a classic case of gaps induced by phonotactic restrictions, although Hetzron (1975) is more circumspect about causation.
In Hungarian, there are gaps in multiple places in the paradigm of verbs. Hetzron
(1975:864) cites paradigmatic gaps in the indefinite imperative (marked by -j + person marker), the potential (‘may X’, marked by -hat/-het) and the verbal adverb (‘in doing’, marked by -va/-ve). Törkenczy (2002:314) cites these, and adds the definite imperative (- d). These gaps are frequently cited as being phonotactically-driven because both the
252
inflectional suffixes and the affected stems are definable in (mostly) phonological terms.
The inflectional suffixes above all begin with a consonant. This is not a requirement of suffixes generally in Hungarian. For example, the indefinite third person plural present indicative form is realized by the suffix -unk/-ünk. And, as Törkenczy (2002:315) notes:
“Defective stems are phonologically identifiable: (a) they end in exactly two consonants; and (b) they end in a consonant cluster which cannot stand at the end of a word...”
Consonant clusters at the end of the verbal stem also are not a general feature of
Hungarian verbs. Most of the verbs meeting these criteria belong to the so-called -ik class (e.g. csukl- ‘hiccup’), named after the form of the indefinite third person singular present indicative (e.g. csuklik ‘he hiccups’), although at least two belong to a different class (kétl- ‘doubt’, sínyl- ‘suffer’).
Thus, putting the two factors together, paradigmatic gaps appear in Hungarian verbs when the stem would end in a consonant cluster (Hetzron notes that for independent reasons, the second consonant of the cluster is virtually always /l/ or /z/), and the inflectional suffix would begin with a consonant cluster, resulting in a C{l,z}C cluster after concatenation. This has been argued to be a phonotactically illicit sequence in
Hungarian.
However, there is a significant caveat to the claim of phonologically-defined paradigmatic gaps, namely that many (perhaps even most) verbs with an “underlying”
C{l,z}C cluster in the imperative, potential or adverbial form surface with an
253
epenthetic vowel to break the cluster (Hetzron 1975:864).95 Thus, only a subset of the phonologically circumscribed forms actually have gaps.
(14) Sample verbs admitting epenthesis to break consonant clusters in the definite and indefinite imperative, potential, and verbal adverb forms.
base form potential form (‘he may...’) gloss kotlik kotolhat- ‘to brood’ feslik feselhet- ‘to become unstitched’ hajlik hajolhat- ‘to bend, incline, tend to’ kibicsaklik kibicsakolhat- ‘to become sprained’ botlik botolhat- ‘to stumble, trip’ verzik verezhet- ‘to bleed’
(15) Sample of Hungarian verbs with paradigmatic gaps in the potential, jussive and verbal adverb forms
base form potential form (‘he/it may...’) gloss csuklik *csuklhat- / *csukolhat- ‘to hiccup’ meghasonlik *meghasonlhat- / ‘to become disillusioned, *meghasonolhat- to conflict with’ özönlik *özönlhet- / *özönölhet- ‘to stream in large quantities’ piroslik *piroslhat- / *pirosolhat- ‘to shine red’ fogzik *fogzhat- / *fogozhat- ‘to teethe’ patakzik *patakzhat- / *patakozhat- ‘to gush’
Other verbs cited as defective:
kisiklik ‘to derail’ siklik ‘to slide, glide’ izlik ‘to taste good’ büzlik ‘to stink’ fénylik ‘to shine’ hámlik ‘to peel (intr.)’ hanyatlik ‘to decline (intr.)’ ködlik ‘to loom’ porlik ‘to pulverize (intr.)’ rajzik ‘to swarm’ elfajzik ‘to (become) degenerate’ morajlik ‘to rumble’ pará[Z]lik ‘to glow’ vedlik ‘to slough’ kétl- ‘to doubt’ sínyl- ‘to suffer’
95 Törkenczy is confusing on this point. He states that gaps appear because “...there is simply no epenthesis or deletion within the stem in Hungarian”, yet his own examples clearly show what Hetzron claims, and justifiably so, is an epenthetic vowel (one example from Törkenczy is reproduced in (16)). 254
(16) An example with “epenthetic repair”, and an example with a gap (Törkenczy 2002:314)
a. Addig fuldokol-hat-sz ameddig csak akarsz. ‘You may gasp for breath for as long as you want.’
b. *Addig csukl-hat-sz ameddig csak akarsz. ‘You may hiccup for as long as you want.’
c. *Addig csukol-hat-sz ameddig csak akarsz. ‘You may hiccup for as long as you want.’
According to Törkenczy, the few lexemes not belonging to the -ik class additionally have gaps in the indefinite third person singular present indicative (not marked by a suffix, the form being equal to the root), presumably because rising sonority of the consonant cluster at the end of the root makes the sequence illicit in word-final position. Thus, * kétl ‘he doubts’.
While most researchers have considered Hungarian verbal gaps to represent a straightforward case of synchronic phonological conditioning (esp. Rice 2005a, 2005b,
Törkenczy 2002), the use of epenthesis to break consonant clusters in some words but not others leads Hetzron to conclude that “...the limits of the phenomenon (blocking of epenthesis, leading to defectiveness) lie in arbitrary lexical specification or phonetic configurations which exhibit no clear-cut synchronic motivation” (Hetzron 1975:869).
A8. Icelandic
Language family: Indo-European, Germanic, North, West Scandinavian Source: Hansson (1999)
255
Icelandic has gaps among imperative verbs. Oddly, these gaps have received the least theoretical attention, despite presenting an intriguing (and complicated) distribution.
“Modern Icelandic has three different ways of forming the imperative of verbs...
The root imperative – the pattern inherited from Old Icelandic – is rarely encountered in spoken Icelandic except as a deliberate archaism, and will not concern us further... The full and clipped imperative are both formed by adding a coronal suffix, here analyzed as
{/T/,/Th/}, i.e. a morpheme with two input allomorphs, /T/ and /Th/, whose distribution is
(largely) phonologically driven” (Hansson 1999:106).96
(17) Normal imperative formation in Icelandic
a. Full imperative: root + {/T/,/Th/} + subject pronoun clitic /-Y/
taktu /thakh-Th-Y/ [thaxtY] ‘take!’ sýndu /sin-T-Y/ [sintY] ‘show!’
b. Clipped imperative: root + {/T/,/Th/} + full subject pronoun
tak-t ÞÚ /thakh-Th/ [thaxt Tu:] ‘YOU take!’ sýn-d ÞÚ /sin-T/ [sint Tu:] ‘YOU show!’
In an interesting sort of (apparently) paradigmatic connection, Hansson argues that the distribution of imperative gaps in Icelandic is crucially predicted by a combination of stem shape and the past tense formation of the verb. “As in most Germanic languages, verbs can be grouped into two major classes with respect to past-tense stem formation.
On the one hand, there are the strong verbs, whose past stem consists of the bare root,
96 There is not space here to justify positing different allomorphs, but the interested reader is referred to the original source. I follow Hansson in assuming that two different allomorphs are needed, although nothing regarding the form that those allomorphs take (a theoretical issue) is at stake for the following discussion of gaps. 256
which is subject to various ablaut alternations (mostly vocalic). The other major class is the weak verbs. Although ablaut-like alternations are also found among verbs of this class, the crucial property of weak verbs is that their past stem is formed by adding a coronal suffix...the past-tense suffix of weak verbs displays the very same kind of allomorphy alternations as does the imperative suffix. I therefore analyze both suffixes as consisting of {/T/,/Th/} in the input” (Hansson 1999:107). In different terms, one class of non-defective Icelandic verbs presents a syncretism. If the past tense is marked by a coronal suffix, {/T/,/Th/}, as is the case for all weak verbs, the imperative has the same form + a pronoun (clitic or full form). The syncretism is not universal, however, because where the past tense is strong, i.e., where it is marked by an ablaut pattern rather than the coronal suffix, the imperative is still marked by a coronal suffix.
However, there is a class of roots, those ending in /-ll/ and /-nn/, which behave unusually in the imperative. They display different imperative forms, depending upon whether past tense formation is strong or weak. In both cases, the phonetic
“...environments should cause /T/ to be selected as the allomorph of choice [for imperative formation], yielding output strings with [...lt] and [...nt]. But in fact, only a minority of verbs in /...ll/ and /...nn/ display the expected behavior” (Hansson 1999:112).
Specifically, only six weak verbs form both the past tense and the imperative as would be expected according to the morphophonological rules of Icelandic (i.e. with the /T/ allomorph). These are exemplified below.
257
(18) Weak verbs in /...ll/ or /...nn/ marked by the ‘expected’ imperative and past tense forms (N=6)
/fell-T/ [fElt-] /r9ell-T/ [r9Elt-] /tholl-T/ [thOlt-] /khenn-T/ [chEnt-] /prenn-T/ [prEnt-] /renn-T/ [rEnt-]
The remaining 27 weak verbs are still syncretic in the imperative and past tense, but are marked by the suffix /Th/. According to Hansson, this form is unexpected.
(19) Weak verbs in /...ll/ or /...nn/ marked by an ‘UNexpected’ imperative and past tense forms (N=27)
/fIll-Th/ [fIl9t-] /hell-Th/ [hEl9t-] /krIll-Th/ [cIl9t-] /prInn-Th/ [prIn9t-] /klenn-Th/ [klEn9t-] /krenn-Th/ [krEn9t-]
Other words: /kIll/, /hIll-/, /rIll/, /skell-/, /smell-/, /spIll-/, /stIll-/, /thrIll-/, /thIll-/, /vIll-/, /krInn-/, /l9Inn-/, /Inn-/, /khInn-/, /lInn/, /mInn-/, /nenn-/, /sInn-/, /spenn-/, /TInn-/
But most interesting or our purposes here is that “The handful of strong verbs whose root ends in /...ll/ or /...nn/, however, do not behave as expected. These verbs do not select the
/T/ allomorph, as dictated by the phonology, nor do they idiosyncratically select the /Th/ allomorph, as so many weak verbs of similar shape do in both imperative and past stem formation. Instead, these strong verbs simply lack an imperative form – in other words, we find a gap in their paradigm” (Hansson 1999:113). 258
(20) Strong verbs in /...ll/ or /...nn/ have paradigmatic gaps in the imperative (N=?)
/vInn-{T,Th}/ *[vInt-], *[vIn9t-] ‘work!’ /spInn-{T,Th}/ *[spInt-], *[spIn9t-] ‘spin (thread)!’ /fall-{T,Th}/ *[falt-], *[fal9t-] ‘flunk!’
And, there is a lone verb which meets all of the criteria to belong to the group with imperative gaps, but which nonetheless has a well-formed imperative.
(21) Strong verb in /...nn/ marked by the ‘expected’ imperative form (N=1)
/fInn-Th/ [fInt-] ‘find!’
“The explanation for the special behavior of this particular verb appears to lie in the fact that it has a unique root allomorphy, such that the (sub)string [f...nt] does occur elsewhere in its paradigm. This is not true of any of the other verbs in /...nn/” (see Hansson
1999:113 for details).
In summary, Hansson makes three significant generalizations about the distribution of imperative forms among /...ll/ and /...nn/ verbs:
• “The imperative stem is only exceptional when there is an exceptional past stem”
(Hansson 1999:114). When the past tense of a weak verb has an idiosyncratic
form, the imperative has the same idiosyncratic form. In the terminology of
modern Word and Paradigm theories, such as Paradigm Function Morphology,
this confirms that there is a systematic syncretism between the form of the past
tense and the form of the imperative, and not just accidental homophony.
259
• “In strong verbs, which have {T,Th} only in the imperative stem, exceptional
selection – subcategorizing for either /T/ or /Th/ specifically – is never found.
Paradigm gaps do not occur where a potentially ‘supporting’ surface string occurs
elsewhere in the paradigm. Imperative gaps are never found in weak verbs
(where the past tense forms can provide ‘support’). Nor do they occur in strong
verbs where the (otherwise expected) surface string happens to occur
independently in other inflected forms...” (Hansson 1999:114).
• “Paradigm gaps only emerge when the verb root has a phonotactic shape which
usually is indicative of exceptionality” (Hansson 1999:114), meaning that the
appearance of gaps among strong verbs correlates with an unexpected
imperative/past tense formation among weak verbs of the same stem shape.
A9. Mandarin Chinese
Language family: Sino-Tibetan, Chinese Source: Rice (2005a), citing Walker and Feng (2004)
Mandarin Chinese presents another example of defective reduplication. In Mandarin, the distributive meaning of classifiers is realized through total reduplication. However, this process only applies to monosyllabic roots. Disyllabic roots have paradigmatic gaps.
260
(22) Reduplication in Mandarin Chinese to form the distributive of classifiers (Rice 2005a:4)
a. Distributive is well-formed when composed from monosyllabic stems
dun ‘ton’ dundun ‘every ton, tons of’ bei ‘cup’ beibei ‘every cup, cups’ ping ‘bottle’ pingping ‘every bottle, bottles of’
b. Distributive is not well-formed for disyllabic stems
jialun ‘gallon’ *jialunjialun chabei ‘teacup’ *chabeichabei jiuping ‘wine bottle’ *jiupingjiuping
A10. Norwegian
Language family: Indo-European, Germanic, North, East Scandinavian, Danish-Swedish Original data: Rice (2003, 2005a, 2005b)
In three papers, Rice has discussed imperatives in Norwegian as having paradigmatic gaps (Rice 2003, 2005a, 2005b). I present here the basic facts of the case as it is presented there, but then discuss some distributional information, mentioned in passing in the last of these three papers, which leads to doubts about whether these are truly paradigmatic gaps.
Imperatives in Norwegian are identical in form to the infinitival stem. Infinitives have a suffixed schwa if the stem ends in a consonant and no suffix if it ends in a vowel.
Thus, the imperative form is equivalent to the infinitive with the schwa “removed”, if present.
261
(23) Infinitives and imperatives in Norwegian
å gå ‘to walk’ gå ‘walk!’ å be ‘to pray’ be ‘pray!’ å spis[´] ‘to eat’ spis ‘eat!’ å vis[´] ‘to show’ vis ‘show!’ å løft[´] ‘to lift’ løft ‘lift!’ å følg[´] ‘to follow’ følg ‘follow! å hopp[´] ‘to jump’ hopp ‘jump!’ å snakk[´] ‘to talk’ snakk ‘talk!’
However, there is a class of exceptions for which this principle does not work. If the infinitival stem ends in a consonant cluster with rising sonority, it does not have a corresponding imperative. Instead, Rice claims that many dialects have a gap in the imperative form.97
(24) Infinitives and defective imperatives in Norwegian (Rice 2003, Rice 2005a)
å apn[´] ‘to open’ *apn ‘open!’ å padl[´] ‘to paddle’ *padl ‘paddle!’ å sykl[´] ‘to bike’ *sykl ‘bike!’ å klatr[´] ‘to climb’ *klatr ‘climb!’ å vitn[´] ‘to witness’ *vitn ‘witness!’ å ordn[´] ‘to arrange’ *ordn ‘arrange!’ å kvikn[´] ‘to revive’ *kvikn ‘revive!’ å bobl[´] ‘to bubble’ *bobl ‘bubble!’ å kobl[´] ‘to connect’ *kobl ‘connect!’ å handl[´] ‘to shop’ *handl ‘shop!’
97 See Rice (2003) for details of other strategies used in Norwegian dialects, including epenthesis and syllabification of word-final liquids. 262
Other defective verbs: å takl[´] ‘to tackle’ å hekl[´] ‘to crochet’ å fikl[´] ‘to fiddle’ å hagl[´] ‘to hail’ å vagl[´] ‘to roost’ å ytr[´] ‘to express’ å saml[´] ‘to collect’ å sikr[´] ‘to secure’ å vikl[´] ‘to wrap’ å endr[´] ‘to change’ å forbedr[´] ‘to improve’ å hindr[´] ‘to impede’ å varsl[´] ‘to notify’ å fordr[´] ‘to require’
There is a significant caveat to the above claim, which suggests that these gaps are not, in fact, gaps, but instead result from a distributional restriction in the syntax. According to
Rice (2005a), the “gapped” forms may be produced if the following word in the phrase is vowel-initial. If the word is said in isolation or if the following word is consonant-initial, however, the sentence is ill-formed and a circumlocution must be used. This point can be most easily demonstrated with negated imperatives. “The negator ikke can come either before or after the imperative... However, when the imperative is of the type [where we expect a gap], only one of the two syntactic possibilities is well-formed. In this case, ikke, ‘not’, must follow the imperative if it is to be uttered, assuming that the next word is consonant initial. Otherwise, the imperative is followed by a consonant initial word and is therefore unutterable... Note that this is not a fact about negative imperatives; these simply form a convenient example because ikke, ‘not’, is vowel initial” (Rice 2005a:7).
263
(25) Imperatives followed by consonant-initial words (imperative gaps result in ill- formed constructions) (Rice 2005a:7)
a. Ikke hopp på møblene! not jump.IMPER on furniture ‘Don’t jump on the furniture!’
b. *Ikke klatr på møblene! not climb.IMPER on furniture ‘Don’t climb on the furniture!’
(26) Imperatives followed by vowel-initial words (well-formed for both gaps and non-gaps) (examples from Rice 2005a:7-8)
a. Hopp ikke på møblene! jump.IMPER not on furniture ‘Don’t jump on the furniture!’
b. Klatr ikke på møblene! climb.IMPER not on furniture ‘Don’t climb on the furniture!’
c. Ikke klatr opp på møblene! not climb.IMPER up on furniture ‘Don’t climb up on the furniture!’
Norwegian presents the best case for synchronically-motivated, phonotactically-induced defectiveness, but would seem to contradict the claim of a paradigmatic gap. Previous research has treated Norwegian imperatives as an example of paradigmatic gaps in particular, but it might be possible based on the data in (26) that the defectiveness actually resides outside of the morphology.
264
A11. Russian
Language family: Indo-European, Slavic, East
A11.1. First person singular non-past verbal gaps
Primary source: Halle (1973) Other citations: Albright (2003), Baerman and Corbett (2006), Baronian (2005), Fanselow and Féry (2002), Hetzron (1975), Iverson (1981), McCarthy and Wolf (2005)
See Chapter 5 for discussion.
A11.2. Genitive plural gaps in nouns
Main source: Pertsova (2005) Other citations: Baerman and Corbett (2006)
In addition to the verbal gaps, Russian has a number of gaps in the genitive plural of nouns. The distribution of these gaps may be precisely defined based on inflection class
(or, alternatively, the expected form of the genitive plural), stem shape, and the paradigmatic stress pattern.
For most nouns (i.e. excluding i-stem nouns), the genitive plural marker can take three forms: It can be realized by the suffix /-ov/, by the suffix /-ej/, or with a “zero ending”, meaning that the genitive plural has the same form as the noun stem. Examples are given below.
265
(27) Genitive plural formation in Russian
Nominative singular Genitive plural Gloss gorod gorod-ov ‘city’ drug druz-ej ‘friend’ knig-a knig ‘book’
Jakobson (1984) was the first to make the generalization which describes the distribution of these suffixes in the majority of cases: If the nominative singular is marked with a suffix (e.g. kniga, mesto), the genitive plural is not and will be the same as the nominal stem. If the nominative singular form is the same as the nominal stem,98 the genitive plural has either the suffix -ov or the suffix -ej, depending on palatalization of the stem- final consonant (-ov with non-palatalized stems, -ej with palatalized stems).
However, as noted by Pertsova (2005) among others, this means that when a zero- ending is expected in the genitive plural of non-palatalized stems,99 but the word is stressed on the ending in the plural oblique cases (the oblique cases in Russian are the genitive, dative, instrumental, locative/prepositional), stress usually retracts onto the stem.
98 Abstracting away from final devoicing of obstruents... 99 This generalization applies only to hard stems because soft stems have –ej, which violates the basic generalization of genitive plural distribution. 266
место ‘place’ singular plural nominative mje!st-o mjest-a! accusative mje!st-o mjest-a! genitive mje!st-a mje!st dative mje!st-u mjest-a!m prepositional mje!st-e mjest-a!x instrumental mje!st-om mjest-a!mi
Table 49: Russian nominal paradigm showing end stress in all oblique plural forms except the genitive plural
“Interestingly, it is a subset of precisely this group of nouns that have gaps in the genitive plural...” (Pertsova 2005:16-17). Pertsova argues that the crucial generalization is that the paradigms with genitive plural gaps have stress on the first syllable of the suffix throughout the paradigm (i.e. no stem stress at all).
мечта ‘dream’ singular plural nominative mječt-a! mječt-y! accusative mječt-u! mječt-y! genitive mječt-y! -- dative mječt-e! mječt-a!m locative mječt-e! mječt-a!x instrumental mječt-o!j mječt-a!mi
Table 50: Russian nominal paradigm - gap in the genitive plural
267
(28) Genitive plural gaps
nominative singular genitive plural gloss fata *fat ‘veil’ yula *yul ‘weasel’ mečta *mečt ‘dream’ kuma *kum ‘god-mother’ suma *sum ‘pouch, purse’
mzda *mzd / *mozd / *mzod / etc. ‘bribe (archaic)’ dno *dn / *dan / *don ‘bottom’ mgla *mgl / *mgol / *mogl / etc. ‘haze’
Other defective nouns: brjuzga ‘squeamish person’ duda ‘pipe (instrument)’ konura ‘dog house’ kljuka ‘cane’ yarmo ‘yoke’ baba-yaga ‘witch’ taxta ‘couch’ arba ‘bullock cart’ moljba ‘prayer’ korma ‘nose of a ship’ karga ‘hag’ kočerga ‘fire poker’ balda ‘brick-head’ čadra ‘chador’ baška ‘head (colloq)’
Three of these nouns (mzda, dno, mgla) lack stem vowels, making phonological illicitness a likely factor for their lack of a genitive plural form. However, the majority of the expected but non-attested forms would be pronounceable and would not violate the phonotactics of Russian. Thus, phonological ill-formedness cannot explain their distribution and existence; the issue seems to be morphological.
A12. Spanish
Language family: Indo-European, Italic, Romance, Italo-Western, Western, Gallo- Iberian, Ibero-Romance, West Iberian, Castilian
Primary source: Albright (2003) Other citations: Baerman and Corbett (2006), Baronian (2005), Pertsova (2005)
268
Albright (2003) distinguishes two types of present tense indicative gaps, what he calls anti-stress gaps and anti-egotistic gaps. Anti-stress gaps are demonstrated below on the left; anti-egotistic verbs are on the right.100
abolir ‘to abolish’ singular plural asir ‘to grasp’ singular plural 1st person -- abolimos 1st person -- asimos 2nd person -- abolís 2nd person ases asís 3rd person -- -- 3rd person ase asen
Table 51: Present tense indicative gaps in Spanish
(29) Sample of verbs with anti-stress gaps
abolir ‘to abolish’ agredir ‘to assault’ aguerir ‘to harden for battle’ arrecierse ‘to stiffen’ aterirse ‘to be numb’ colorir ‘to color’ denegrir ‘to blacken’ descolorir ‘to de-color’ empedernir ‘to harden’ garantir ‘to guarantee’ tra(n)sgredir ‘to transgress’ transhumar ‘to move pastures’
(30) Sample of verbs with anti-egotistic gaps
asir ‘to grasp’ balbucir ‘to stammer’ pacer ‘to graze’
Albright argues (convincingly) that these two patterns of gaps are related to morphophonological alternations. “The gap patterns... are extremely suggestive, since
100 In many cases there are parallel, fully inflectable verbs which are semantically very close to the verbs with gaps, which speakers may use to avoid the gaps listed above. For example, garantir may be replaced by garantizar. 269
they mirror exactly the distribution of unpredictable morphophonemic alternations. In particular, anti-stress verbs are missing forms where diphthongization and raising occur, while anti-egotistic verbs are missing the form where velar insertion occurs. Furthermore, verbs with gaps generally meet the structural description for alternations: anti-stress verbs mostly have mid vowels, while anti-egotistic verbs have stem-final [s]. In addition, virtually all of the defective verbs belong to class 3 ([i] theme), which is the most susceptible to alternations” (Albright 2003:4). Diphthongization and raising apply to mid vowels under stress. Velar insertion applies in the first person singular when there is a stem-final /s/. These alternations are exemplified below; diphthongization and raising in the left tables, velar insertion in the right side one.
sentir ‘to feel’ singular plural crecer ‘to grow’ singular plural 1st person s[jé]nto s[e]ntímos 1st person cré[sk]o cre[s]émos 2nd person s[jé]ntes s[e]ntís 2nd person cré[s]es cre[s]éis 3rd person s[jé]nte s[jé]nten 3rd person cré[s]e cré[s]en
pedir ‘to ask’ singular plural 1st person p[í]do p[e]dímos 2nd person p[í]des p[e]dís 3rd person p[í]de p[í]den
Table 52: Morphophonological alternations in the present indicative
Since the alternations do not apply uniformly especially in the classes where paradigmatic gaps are found, Albright argues that there is a synchronic, causative link between the “unpredictable morphophonemic alternations” and the existence of Spanish verbal gaps, although the gaps only appear in a subset of the “eligible” lexemes. 270
A bit of data not mentioned by Albright may also be relevant. Baronian
(2005:154) notes that those verbs with gaps in the singular and third person plural also fail to raise the root vowel in the present participle:
(31) Vowel raising in the present participle (does not apply to forms with first person singular present indicative gaps)
infinitive past participle 1st p. singular gloss d[o]rmir d[u]rmiendo d[we]rmo ‘to sleep’ s[e]ntir s[i]ntiendo s[je]nto ‘to feel’ ab[o]lir ab[o]liendo -- ‘to abolish’ ag[e]rrir ag[e]rriendo -- ‘to arrange’
Thus, while Albright argues for a probabilistic distribution of first person singular among those lexemes which meet the criteria for gaps, Baronian argues that a precise description of the distribution of the gaps is possible based on morphophonemic vowel alternation (or lack thereof) in the past participle.
A13. Swedish
Language family: Indo-European, Germanic, North, East Scandinavian, Danish-Swedish
A13.1. Swedish neuter adjectives
Major sources: Eliasson (1975) , Johansson (2003) Other citations: Fanselow and Féry (2002), Iverson (1981), Johansson (1999), Rice (2005a, 2005b)
271
In Swedish there are gaps among the neuter adjectives. There is significant disagreement as to the basic distributional facts for the gaps, so I present two contrasting descriptions here. I am not in a position to adjudicate between them.
A.13.1.1. Neuter adjectival gaps as a result of phonology (originally formulated by Eliasson (1975), and followed explicitly or implicitly by Iverson (1981), Fanselow and Féry (2002), McCarthy and Wolf (2005), and Rice (2005a)).
Eliasson (1975) was the first researcher to discuss these gaps in a theoretical linguistic work, although the issue of how to theoretically account for gaps was a side issue, and not the main focus of his study. Still, his description of the distribution of the gaps remains the best among those who argue that the gaps result from phonological restrictions. He states that there is a “...rather surprising prohibition against adding suffixes beginning with, or consisting of, a dental stop to adjective and verb roots ending in a double d in the underlying form. Disregarding various semantic and other restrictions, indefinite singular neuter forms may usually be generated for any Swedish adjective by adding the suffix /t/... The same is also true of many adjectives ending in a vowel, t, d, or tt, in their underlying forms. However, the indefinite singular neuter forms of adjectives ending in dd (fadd ‘flat, stale’ and rädd ‘scared’) are not used” (Eliasson
1975:427).101 Other researchers have followed Eliasson’s lead in positing a
101 Eliasson (1975) also notes that there is an interesting parallel among Swedish verbs. While the verbs do not have paradigmatic gaps, those with stem-final /Vdd/ have an unusual distribution. In the second and third conjugations there are three inflectional cells which involve adding suffixes consisting of dental stops (directly) to the stem (/-de/ for past tense, /-t/ for supine, and /-d/ for past participle). By contrast, in the first conjugation, a theme vowel /a/ “...intervenes between the verb root and the inflectional suffixes” (429). Interestingly, all of the verbs with a stem-final /dd/ sequence belong to the first conjugation class. Thus, the conditions which apparently lead to gaps among the adjectives are not encountered among verbs. Eliasson treats this as the static result of a gap-avoidance strategy. 272
phonological basis for the gaps, although not all assume that gaps appear only in words with stem-final geminate /d/ (e.g. Fanselow and Féry 2002)
(32) Productive Swedish adjectives (Eliasson 1975:428, and Fanselow and Féry 2002:manuscript p. 13)
a. en rysk pojke vs. ett ryskt barn a Russian.NON-N.SG boy.NON-N.SG a Russian.N.SG child.N.SG
b. en glad pojke vs. ett glatt barn a happy.NON-N.SG boy.NON-N.SG a happy.N.SG child.N.SG
c. en plat tallrik vs. ett platt hus a flat.NON-N.SG plate.NON-N.SG a flat.N.SG house.N.SG
(33) Swedish neuter adjectives with gaps
a. en rädd pojke vs. *ett rätt barn a scared.NON-N.SG boy.NON-N.SGN a scared.N.SG child.N.SG
b. en lat pojke vs. *ett latt barn a lazy.NON-N.SG boy.NON-N.SG a lazy.N.SG child.N.SG
c. en flat tallrik vs. *ett flatt hus a flat.NON-N.SG plate.NON-N.SG a flat.N.SG house.N.SG
(34) Neuter and non-neuter (common gender) indefinite singular forms according to stem type, including one gap (Eliasson 1975)
Type of root Non-neuter Neuter Gloss /V/ [slö] [slO·t:] ‘blunt’ /Vt/ [slE:t] [slEt:] ‘smooth’ /Vd/ [spE:d] [spEt:] ‘tender’ /Vtt/ [rEt:] [rEt:] ‘right’ /Vbb/ [snap:] [snap:t] ‘fast’ /Vgg/ [snyg:] [snyk:t] ‘tidy’ /Vdd/ [rEd:] -- ‘scared’
273
Note that the conditions for the adjectival gaps cannot be solely surface constraints.
Through independent processes of voicing assimilation and degemination, the expected neuter adjective form for ‘scared’ would be [rEt:], which is identical to the (perfectly acceptable form) for ‘right’, and very similar to [ret:] ‘prepared (neuter past participle)’.
A.13.1.2. Neuter adjectival gaps as a result of morphophonological and semantic factors (Johansson 1999, Johansson 2003)
By contrast, Johansson (1999, 2003) has argued that the distribution of gaps is the result of partly morphophonological factors, and partly semantic ones.102 “The essence of the problem is the existence of paradigmatic gaps. One case is the failure of an extremely frequent and simple regularity in the Swedish adjective paradigm – simply add t to make a neuter from a common gender form. Many neuter forms of adjectives are missing, regardless of the frequency of use for the common gender form. Two factors collaborate in causing this absence of neuters. First, all the problematic forms are monosyllabic, or have stress on the last syllable, and the gender marking is assimilated to that syllable...
The second necessary property is that the semantics of the adjective imply a property that cannot be verified by sensory experience only (such as internal states: fear, laziness, true happiness)” (Johansson 2003:1).
102 Johansson does not cite Eliasson or any of the work which followed his description, so it is unclear whether Johansson constructed his generalizations with reference to this previous work. 274
(35) Examples of well-formed Swedish adjectives in neuter form
lytt ‘handicapped’ satt ‘plump’ fett ‘greasy’ rått ‘raw’ hett ‘hot’ skitt ‘dirty’ insnöat ‘snowed-in’ stridigt ‘fierce’ vredgat ‘angry’ flackt ‘flat (landscape)’ lagt ‘put’ skrämt ‘scared’ grått ‘gray’ matt ‘faded’ nött ‘worn-out’ kavaljerskt ‘chivalrous’ brutet ‘broken’ rigitt ‘rigid’
(36) Examples of paradigmatic gaps in Swedish (non-neuter / expected neuter) (Johansson 2003)
kry / *krytt ‘well, healthy’ glad / *glatt ‘happy’ god / *gott ‘morally good’ ked / *kett ‘bored’ kåt / *kått ‘horny’ led / *lett ‘bored’ blid / *blitt ‘blithe’ pryd / *prytt ‘prudish’ snöd / * snött ‘sordid’ strid / *stritt ‘swift’ vred / *vrett ‘angry’ flat / *flatt ‘stale’ lat / *latt ‘lazy’ rät / *latt ‘right-angled’ såt / *sått ‘great (pals)’ fadd / *fatt ‘forced’ rädd / *rätt ‘afraid’ disträ/*disträtt ‘absent-minded’ kavat / *kavatt ‘plucky, cocky’ brydd / *brytt ‘puzzled’ gladd / *glatt ‘made happy’ gravid/*gravitt ‘pregnant’
A13.2. Indefinite genitive singulars of nouns
Source: Karlsson (2000)
Karlsson (2000:648) provides a very brief description of gaps among genitive singular nouns. “...in Swedish the ordinary genitive ending is –s. However, -s cannot be suffixed to stems ending in sibilants. Nouns such as hus ‘house’, svans ‘tail’, Max thus have no indefinite genitive singular forms like *hus-s, *Max-s (but they do have definite genitive singulars such as bil-en-s, Max-en-s).” Karlsson considers these examples to be
275
“borderline case[s]” because he does not accept phonological blocking as a legitimate cause of defectivity, reserving that term for the more quirky examples such as presented by Spanish and Russian. However, if we accept that paradigmatic gaps may arise from phonotactic restrictions, as we have for other descriptions in this Appendix, this case is parallel.
A.14. Tiene
Language family: Niger-Congo, Atlantic-Congo, Volta-Congo, Benue-Congo, Bantoid, Southern, Narrow Bantu, Northwest, B, Yanzi (B.80)
Main source: Orgun and Sprouse (1999), citing Hyman and Inkelas (1998)
Tiene has paradigmatic gaps in the definitive form of verbs. The definitive construction is normally formed by reduplicating the final VC of the stem.103
(37) Definitive reduplication in Tiene (Orgun and Sprouse 1999:211)
jOb- ‘bathe’ jOb-Ob- ‘bathe thoroughly’ mat- ‘go away’ mat-at- ‘go away once and for all’ jak- ‘believe’ jak-ak- ‘believe once and for all’ lON- ‘load’ lON-ON- ‘load once and for all’
However, problems potentially arise when the phonotactic constraints of the language would be violated by the definitive form. Orgun and Sprouse note four phonotactic constraints, of which two are relevant here:
103 With regard to the examples of the definitive construction, Orgun and Sprouse (1999:209) state “We do not show the final vowel morpheme, since it is not relevant to our purposes.” I simply replicate here the data that they provide, or, as the case may be, fail to provide. 276
• STEMSIZE: The base may not contain more than two syllables.
• STEMSHAPE 3: In C1VC2VC3 bases, C2 is coronal. C3 is either labial or velar.
As they argue, “While reduplicated forms may violate StemShape, StemSize is never violated in Tiene. For a CVCV stem, there is no possible morphologically expressed definitive form. A periphrastic form adding nkó mO⁄to to the verb conjugated in the neutral aspect must be used instead (Ellington 1977:93).”
(38) Paradigmatic gaps among definitive constructions in Tiene (Orgun and Sprouse 1999:212)
kótob- ‘chase’ *kótobob-/* kótob-/*kóbob- ‘chase once & for all’ vútek- ‘come back’ *vútekek-/*vútek-/*vekek- ‘return once & for all’ binem- ‘sleep’ *binemem-/*binem-/*bemem- ‘sleep thoroughly’ panam- ‘frighten’ *panamam-/*panam-/*pamam- ‘frighten thoroughly’
Framing their analysis within Optimality Theory, which makes little to no distinction between phonological and morphological processes, Orgun and Sprouse treat the above gaps as the result of phonotactic restrictions. Whether they should be considered phonotactically-induced gaps or lexically arbitrary gaps, however, is a question which can only be answered within a particular theory, according to its basic assumptions about the scope of morphological and phonological processes.
277
A15. Turkish
Language family: Altaic, Turkic, Southern
Main source: Orgun and Sprouse (1999) Other citations: Rice (2005a, 2005b), Törkenczy (2002)
Turkish apparently has gaps in the genitive singular of some nouns. The gaps seem to be related to a generalization that in some Turkish dialects, inflected words must be at least disyllabic. Words with monosyllabic CV roots, which are inflected via a suffix made up of only a consonant fail to meet this condition. “Phonologically ill-formed monosyllabic forms generated by regular inflection cannot be repaired by epenthesis. The violation of the minimality constraint results in absolute ungrammaticality: these stems have a defective paradigm” (Törkenczy 2002:313).
(39) do: ‘musical note C’ *do:-m ‘my C’ soly ‘musical note G’ soly-üm ‘my G’
However, the gap *do:m is the only example cited in these three works, so both the number of gaps that exist is unclear.
278
APPENDIX B
GENITIVE PLURAL GAPS IN MODERN GREEK
A search of two major Modern Greek dictionaries, Lexiko tis neas ellinikis glossas and the online version of Lexiko tis koinis neoellinikis104 (Babiniotis 1998, Triantafillidis
1998, respectively), uncovered genitive plural gaps in 2,141 distinct Modern Greek nouns. In this appendix I provide all of the genitive plural gaps, organized according to traditionally-defined inflection classes, with a comparison of how widespread that inflection class is in the language generally. See CHAPTER 3 and CHAPTER 4 for a discussion of the theoretical issues surrounding this distribution.
B1. Feminine nouns with nominative singular –α and plural –ες
Nouns in this inflection class represent two different stress patterns. Some nouns, like η
ελπίδα ‘hope’, have stress on the penultimate syllable in the genitive plural, but on any of the final three syllables in the other cells. Other nouns, like η θάλασσα ‘sea’, have stress on the final syllable in the genitive plural.
104 http://www.komvos.edu.gr/dictionaries/dictadv/DictAdvSea.htm 279
Example paradigms:
η ελπίδα η θάλασσα ‘hope’ SINGULAR PLURAL ‘sea’ SINGULAR PLURAL NOMINATIVE ελπίδα ελπίδες NOMINATIVE θάλασσα θάλλασες ACCUSATIVE ελπίδα ελπίδες ACCUSATIVE θάλασσα θάλλασες GENITIVE ελπίδας ελπίδων GENITIVE θάλασσας θαλασσών VOCATIVE ελπίδα ελπίδες VOCATIVE θάλασσα θάλλασες
η σήραγγα η ντοµάτα ‘tunnel’ SINGULAR PLURAL ‘tomato’ SINGULAR PLURAL NOMINATIVE σήραγγα σήραγγες NOMINATIVE ντοµάτα ντοµάτες ACCUSATIVE σήραγγα σήραγγες ACCUSATIVE ντοµάτα ντοµάτες GENITIVE σήραγγας σηράγγων GENITIVE ντοµάτας ντοµατών VOCATIVE σήραγγα σήραγγες VOCATIVE ντοµάτα ντοµάτες
Table 53: Examples of Modern Greek feminine nouns with nominative singular –α and plural –ες
Quick stats:
• Babiniotis and Triantafillidis combined: 1,841 nouns with genitive plural gaps
• Only Triantafillidis:
o 1,380 nouns with genitive plural gaps (88.5% of all genitive plural gaps)
o 8,022 nouns in this inflection class (29.4% of all nouns)
Nouns with genitive plural gaps:
Key to the tables of genitive plural gaps Gpl This column gives the dictionary which cited the relevant word as a genitive plural gap. G This column gives the dictionary (if any) which marked the relevant word as a genitive gap (both singular and plural), when the other dictionary marked it as a genitive plural gap. Pl This column gives the dictionary (if any) which marked the relevant word as a plural gap (genitive, nominative and accusative), when the other dictionary marked it as a gen. plural gap. µπ Lexiko tis neas ellinikis glossas (Babiniotis 1998) τ Lexiko tis koinis neoellinikis (Triantafillidis 1998)
280
Word Gpl G Pl Word Gpl G Pl αβάντα τ αβανταδόρισσα τ αβάντσα, αβάντζα τ αγάντα τ αγγειίτιδα µπ αγγουροντοµάτα τ αγγουροντοµατο- σαλάτα τ αγιαστήρα τ αγιαστούρα, αγιαστήρα µπ/τ αγκαλίτσα µπ αγκιτάτσια τ µπ αγκλίτσα τ αγκούσα τ µπ αγνωσία τ µπ αγνωσιαρχία τ µπ αγοραφοβία τ µπ αγορίνα τ αγουράδα τ άγρα τ µπ αγριόγατα µπ/τ αγριόγιδα τ αγριόκοτα τ αγριόπαπια µπ/τ αγριοφωνάρα µπ/τ αγριόχηνα τ αγρύπνια τ αδουατίντα τ αερόσκαλα τ αετοµάνα τ αετονύχισσα τ αζαλέα µπ αηδόνα τ Αιγαιοπελαγίτισσα τρ αιδοιοκολπίτιδα µπ ακονόπετρα µπ ακούµπα τ ακρίβεια τ µπ ακροφοβία τ µπ αλαλοµάρα τ αλάνα µπ αλαναρία τ αλατιέρα µπ/τ αλατίστρα τ αλεπότρυπα µπ/τ αλετροπόδα µπ/τ αλευρόκολλα τ µπ αλήθεια τ αλιάδα µπ/τ αλισίβα τ µπ αλκόβα τ αλµπάνισσα τ αλµύρα, αρµύρα τ µπ αλµυρήθρα τ αλόγα µπ/τ αλογοµούρα τ αλογόµυγα τ αλτάνα µπ αλτερνατίβα µπ αλφαβήτα τ µπ αλωνίστρα τ αµάκα τ µπ αµαξάδα µπ/τ Αµερικάνα τ αµµόπετρα µπ αµπάριζα τ µπ αµπούλα µπ αµυγδαλόφλουδα µπ/τ αµυγδαλόψιχα τ
Continued
Table 54: Genitive plural gaps among feminine nouns with nominative singular –α and plural –ες
281
Table 54 continued
Word Gpl G Pl Word Gpl G Pl αµυλόκολλα τ ανάβρα µπ αναγούλα τ µπ αναδουλειά µπ ανακατωσούρα µπ/τ αναµπουµπούλα µπ/τ ανανδρία τ µπ αναξιοπιστία τ ανάπαυλα τ µπ αναποδιάρα τ ανάσα µπ/τ ανατριχίλα µπ/τ ανδρεία τ µπ ανδροκρατία τ µπ ανεµοδούρα µπ/τ ανεµοθύελλα µπ/τ ανεµόσκαλα µπ/τ ανεµότρατα µπ/τ ανεµώνα τ ανεργία τ µπ ανηµπόρια τ ανηφόρα µπ/τ ανθρωπίλα τ αντάµισσα τ αντάρα τ µπ άντζα τ αντζούγα τ αντζούγια τ αντικάµαρα µπ/τ αντιπροπαγάνδα τ αντράκλα µπ αντρεία τ αντρέσα τ αντροπαρέα τ αντροχωρίστρα µπ απαισιοδοξία µπ απανταχούσα, πανταχούσα µπ απαξία τ µπ απενταρία τ απανωσιά µπ άπλα µπ/τ απλώστρα µπ αποκοτιά µπ απόπειρα µπ απροθυµία τ άργητα τ µπ αργυροχοΐα τ αριστεροχέρα τ αρλούµπα τ αρµόνικα µπ/τ αρµπαρόριζα µπ αρµύρα τ αρµυράδα τ αρµυρήθρα τ αροκάρια τ αρραβωνιάρα τ αρρώστια µπ/τ αρτίστα µπ αρχικλέφτρα τ αρχιτεµπέλα τ αρχιψεύτρα τ αρχοντοπούλα τ ασβεστίλα τ ασίκισσα τ
Continued
282
Table 54 continued
Word Gpl G Pl Word Gpl G Pl ασκήµια τ ασπλαχνία τ ασπρίλα τ Ασπροθαλασσίτισσα τ αστυνοµικίνα µπ ασχήµια τ ατάκα µπ/τ ατζέντα µπ/τ ατιµία µπ/τ ατρετσαρία τ αυγίλα τ αυγουλιέρα τ αυγουλίλα τ αυγοφέτα τ αυγόφετα τ αυγόφλουδα τ αυλόθυρα τ αυλόπορτα µπ/τ αυτοάµυνα τ µπ αυτοκινητάδα τ αυτοκινητάµαξα µπ αυτοπροσωπογραφία τ αφάνα µπ αφεντοπούλα τ αφρόκρεµα µπ/τ άφτρα µπ αχαµνάδα τ αχλάδα µπ άχνα τ µπ αχνάδα τ µπ αψάδα τ µπ βάβα τ µπ βαβούρα τ µπ Βαγγελίστρα τ βάνα µπ βανίλια µπ/τ βαρβατίλα τ µπ βάρδια µπ/τ βαρέλα µπ βαρεµάρα µπ/τ βαριεµάρα τ βαριεστηµάρα τ βαριοπούλα µπ βάρκα τ βαρκάδα µπ βαρκαρόλα µπ/τ βαρυγκόµια τ βασιλόπιτα τ βασιλοπούλα τ βασκαντήρα µπ βάτα µπ βατίστα µπ βατσίνα µπ βεγγέρα τ βέδες τ βεζιροπούλα τ βελάδα µπ βελέντζα µπ βενζινόκολλα τ βεντάγια τ βεντάλια τ βεντούζα τ βέρα τ βερβερίτσα τ βερµούδα µπ βία τ µπ βιόλα µπ βιρτουόζα τ
Continued
283
Table 54 continued
Word Gpl G Pl Word Gpl G Pl βιτσιόζα τ βλάχα τ βλαχάρα µπ βλαχοπούλα τ βοδάµαξα, βοϊδάµαξα µπ βόλτα µπ βοσκοπούλα τ βότκα τ µπ βούβα τ βουβαµάρα τ µπ βουκαµβίλια, βούκα τ µπουκαµβίλια µπ βουλγκάτα τ βουτυριέρα µπ/τ βραδυγλωσσία τ µπ βρακοζώνα µπ βράσσικα µπ βρατσέρα τ βρεχτούρα µπ βρόµα, βρώµα µπ βροµιά µπ βροµόγλωσσα µπ βροµοκουβέντα µπ βροµούσα τ βρυσοµάνα, βρόχα τ βρυσοµάννα µπ/τ γαβάθα µπ γαϊδούρα µπ/τ γαϊδουροκαβάλα τ γαϊδουροκαβαλαρία µπ/τ γαϊδουρότριχα τ γαϊδουροφωνάρα µπ/τ γαϊτανοφρύδα τ γαλαζόπετρα µπ/τ γαλατόπιτα τ γαλιάντρα τ γαλοπούλα µπ/τ γαλότσα τ γαµιόλα τ γάµπα µπ/τ γάµπια τ γαργάρα τ γαρδένια µπ γαρδούµπα µπ/τ γαρνιτούρα µπ/τ γάστρα µπ/τ γαστρεντερίτιδα µπ γαστρίτιδα µπ γάτα τ γατίλα τ µπ γειτονοπούλα τ γέννα τ γενναιοδωρία τ µπ γενναιοψυχία τ γεννήτρα µπ γητεύτρα τ γιάπισσα τ γιατρέσα τ γιάτρισσα τ γιάφκα µπ γιδόστρατα µπ/τ γκαβωµάρα τ γκαζιέρα τ γκαζόζα µπ/τ
Continued
284
Table 54 continued
Word Gpl G Pl Word Gpl G Pl γκαζόλαµπα τ γκάιντα τ γκάµα (γκάµµα) µπ/τ γκανιότα µπ/τ γκαρίλα τ γκαρνταρόµπα µπ/τ γκαρσονιέρα µπ/τ γκάφα µπ/τ γκιλοτίνα µπ γκίνια τ µπ γκιόσα µπ/τ γκλάβα µπ/τ γκλίτσα µπ/τ γκόµενα µπ/τ γκουβερνάντα, λουβερνάντα µπ γκουστερίτσα τ γκραβούρα µπ γκρανκάσα, γκραγκάσα µπ γκρέκα µπ γκριµάτσα µπ/τ γκρίνια, γρίνια µπ/τ γλαδιόλα µπ γλείφτρα τ γλίστρα µπ/τ γλίτσα µπ γλύκα τ γλυκάδα τ γλυκοπατάτα µπ γλυκόρριζα µπ γλυκοφιλούσα τ γλυφάδα τ γλωσσίτσα µπ γλωσσοκοπάνα τ γνώρα τ γνώστρια τ γόβα µπ/τ γόµα, γόµµα µπ γοµαλάστιχα τ γοµολάστιχα, γοµµολάστιχα µπ/τ γόνδολα µπ/τ γόπα µπ/τ γοργάδα τ γοργόνα µπ/τ γούβα τ γούλα µπ/τ γούνα τ γούρνα τ γουρούνα µπ/τ γουρουνότριχα τ γουστέρα µπ γουστερίτσα τ γουστόζα τ γουταπέρκα τ γραβιέρα µπ/τ γραιγοτραµουντάνα µπ γρεγοτραµουντάνα τ γρίλια µπ/τ γρίνια τ γρίπη τ µπ γυάλα µπ/τ γυαλάδα τ µπ γύµνια τ µπ γυναικοκρατία τ µπ γυναικοπαρέα τ
Continued
285
Table 54 continued
Word Gpl G Pl Word Gpl G Pl γυναικούλα µπ γύρα µπ/τ δαγκάνα τ δαµάλα µπ δασκάλα τ δασκαλίτσα µπ δαχτυλιδόπετρα µπ δειλία τ µπ δείξα τ δεκαεννιάρα τ δεκαεξάρα τ δεκαεφτάρα τ δεκαοκτάρα τ δεκαοχτάρα τ δεκαοχτούρα τ δεκατριάρα τ δεκοχτούρα µπ/τ δερµατόκολλα µπ/τ ∆ευτέρα τ διαβατάρισσα τ διαβόλισσα τ δίαιτα τ διακοσάρα µπ/τ διαµαντόπετρα µπ/τ διάρκεια τ µπ διάτα τ δίψα τ µπ δόγα, δούγα µπ δόµνα µπ δόνα τ µπ δόξα µπ/τ δούλα τ δράγα, ντράγα µπ δράκα τ δρεζίνα τ δροσοσταλιά µπ δρωτσίλα µπ/τ δυάρα τ µπ εβδοµηντάρα τ εβραιοπούλα τ εγγλεζοπούλα τ εγγυοδοσία τ Εγίρα τ εγκεφαλίτιδα µπ έγνοια, έννοια µπ/τ εθνοκαπηλεία τ εικοσάρα µπ/τ εικοτολογία τ εκατοστάρα τ εκθειάστρια τ ελαφρολογία µπ ελαφρόπετρα µπ ελληνικούρα µπ/τ Ελληνοαµερικάνα τ ελληνοπούλα τ εµπορία τ µπ έννοια, έγνοια µπ/τ εξάρα τ εξάστρα τ εξηντάρα τ εξυπνάδα µπ εξώθυρα τ εξώπορτα µπ/τ επαρχιωτοπούλα τ επιείκεια τ µπ επιπλοποιία τ µπ εποποιία τ εργένισσα τ
Continued
286
Table 54 continued
Word Gpl G Pl Word Gpl G Pl εσάρπα, σάρπα µπ/τ εσπέρα τ µπ εσχάρα τ εταζέρα µπ/τ Εύα τ ευγένεια τ ευκολοπιστία τ ευφυΐα τ εφταµηνίτισσα τ εφτάρα τ έχθρα τ έχθρητα τ εχθροπάθεια τ έχτρα τ ζαβλακωµάρα τ ζαβοµάρα τ ζαλίκα τ µπ ζάρα µπ/τ ζαργάνα µπ ζαρίφισσα τ ζαρντινιέρα τ ζαρτιέρα τ ζαφειρόπετρα µπ/τ ζαχαριέρα µπ/τ ζαχαροπλάσταινα τ ζέστα τ ζήλια µπ/τ ζηλιαρόγατα τ ζήτα τ µπ µπ ζιβελίνα τ ζιγκολέτα µπ ζιµπελίνα µπ/τ ζίνα τ ζορζέτα τ ζουζούνα τ ζούλα µπ/τ ζούργκλα µπ ζουρλαµάρα τ ζούρλια τ ζουρλοπαντιέρα µπ/τ ζυθοποιία τ ζωοκοµία τ µπ ζωολογία τ ζωούλα µπ ηθοποιία τ µπ ηµεράδα τ ηρεµία µπ ησυχία τ µπ θαλασσοκράτειρα µπ θαλασσοµάνα τ θέα τ µπ θεοµπαίχτρα τ θερµοφόρα µπ θολούρα τ µπ θράκα τ µπ θρούµπα µπ ίγκλα µπ ιντελιγκέντσια τ µπ ίντριγκα µπ/τ ινφάντα τ ίσκα τ ιχθυόκολλα µπ/τ ιχθυόσκαλα µπ κάβα µπ/τ καβαδούρα τ καβάλα τ καβαλαρία, καβαλερία µπ/τ καβαλίνα µπ/τ
Continued
287
Table 54 continued
Word Gpl G Pl Word Gpl G Pl καβίλια τ καβουροµάνα µπ/τ καγκάγια τ καγκελόπορτα µπ/τ καδρίλια τ καζάκα µπ καζάρµα µπ/τ καζούρα µπ/τ καθαρευουσιάνα τ καθεστηκυία τ καθίστρα τ καΐλα τ κακαράντζα τ κάκητα µπ/τ κάκια τ κακίστρα µπ/τ κακόγρια τ κακοκεφαλιά µπ καλαθούνα τ καλαµαριέρα τ καλαµοκάνα τ καλικατούρα τ κάλµα τ µπ καλντέρα, καλδέρα µπ καλοζωία τ καλοθελητής τ καλοκαιρία, καλοκαιριά τ µπ καλούµπα µπ καλοφαγία τ καλτσοβελόνα µπ κάµα µπ κάµαρα τ καµαριέρα τ καµαρίλα τ µπ καµαρόπορτα µπ/τ καµαροφρύδα τ καµέλια τ κάµερα µπ/τ καµηλιέρισσα τ καµιζόλα µπ καµπάνια µπ/τ καµπανούλα µπ κάµπια µπ/τ καµπούρα µπ καµπουροµύτα τ κανακάρισσα τ κανάρα τ κανέλα, καννέλα µπ/τ κάνουλα, κάννουλα µπ/τ καντάτα µπ καντέντσα µπ καντήλα µπ καντηλήθρα µπ/τ καντηλίτσα τ καντρίλια τ καούρα µπ/τ καπάντζα τ καπάτσα τ καπελαδούρα µπ/τ καπελιέρα µπ/τ καπελίνα τ καπιτάλα τ κάπνα τ µπ καπνικαρέα τ καπνίλα τ µπ καπνοσακούλα µπ
Continued
288
Table 54 continued
Word Gpl G Pl Word Gpl G Pl καπότα µπ καπριτσιόζα τ καραβάνα µπ καραβέλα µπ καραγκούνα µπ/τ καρακαηδόνα τ καρακάξα µπ καραµούζα µπ καραµπίνα µπ καραµπόλα µπ/τ καραντίνα τ µπ καραπουτάνα µπ/τ καράφα µπ κάργα τ κάργια τ καριέρα µπ/τ καριερίστα τ καρικατούρα µπ/τ καρίνα, καρένα µπ καριόλα µπ/τ καρµανιόλα τ µπ καρναβίτσα τ καρότσα µπ/τ καρούλα τ καρτίνα τ καρυδόπιτα τ καρυδόφλουδα µπ καρυδόψιχα τ καρφοβελόνα µπ κασάτα τ κασετίνα µπ κασίδα τ κάσκα µπ κασκαρίκα τ κασόνα µπ Κασσάνδρα τ καστανοµάλλης τ καστρόπορτα µπ/τ κατάντια τ κατάρα µπ καταφρόνια τ µπ κατεργάρα τ κατηγόρια τ κατηφόρα µπ/τ Κατίνα µπ/τ κατοστάρα τ κατρακύλα τ µπ κατραπακιά µπ κατσάδα µπ/τ κατσαρόλα µπ κατσίκα µπ κατσούφης τ καύλα µπ/τ καύτρα µπ καυχησιάρα τ καφετιέρα µπ/τ κάψα µπ καψάλα τ κάψουλα µπ καψούρα µπ/τ καψούρης τ κενοφοβία τ κεντήστρα, κεντήτρα µπ κερκόπορτα τ κεφαλογραβιέρα τ κιλοβατώρα µπ κιµαδόπιτα µπ κιτρινίλα µπ/τ
Continued
289
Table 54 continued
Word Gpl G Pl Word Gpl G Pl κιτσαρία µπ κλαδευτήρα τ κλάκα µπ/τ κλανιάρης τ κλαπάτσα τ κλάρα µπ κλασικούρα µπ κλάψα µπ/τ κλειδαρότρυπα µπ/τ κλεισούρα µπ/τ κληµατόβεργα µπ κλήρα τ κλούβα τ κόβα τ κόκα τ κοκάλα, κοκκάλα µπ κοκεταρία τ µπ κοκκινίλα µπ/τ κοκκινοµάλλα τ Κοκκινοσκουφίτσα τ κοκόνα τ κοκότα τ κόλα τ µπ κολαρίνα τ κολεκτίβα, κολεχτίβα µπ/τ κόλλα µπ/τ κολοκύθα τ κολοµπίνα µπ κολόνια µπ/τ κολορατούρα τ κοµµάρα µπ/τ κοµµούνα τ κοµότα µπ κοµπάρσα τ κοµπίνα µπ/τ κοµπιναδόρισσα τ κοµπογιαννίτισσα τ κοµπόστα µπ κόµπρα µπ κοµπρέσα µπ κοµούνα µπ κονκάρδα µπ κόνξα µπ/τ κονόµα µπ κονσόλα µπ κοντανάσα τ κοντεσίνα µπ κοντολαίµα τ κοντοστούπα τ κοντούλα τ κοντούρα µπ κόντρα µπ/τ κοτρώνα µπ κοπάνα µπ/τ κοπέλα µπ/τ κόπια τ µπ κόρα µπ/τ κορδελιάστρα µπ κόπιτσα µπ κόρνα µπ κόσα µπ/τ κότα µπ κοτόπιτα τ κοτσάνα µπ/τ κοτσίδα µπ κουβέντα µπ/τ κουβερτούρα τ µπ κουδουνίστρα µπ
Continued
290
Table 54 continued
Word Gpl G Pl Word Gpl G Pl κουζουλάδα τ κουΐντα µπ κουκουβάγια µπ κουκούλα µπ κουκουνάρα µπ/τ κουλαµάρα τ κουλούρα, κουλλούρα µπ κουλοχέρα τ κουλτούρα µπ/τ κουλτουριάρα τ κουνέλα µπ κούνια µπ/τ κουνίστρα τ κουνουπιέρα τ µπ κούπα µπ/τ κούρα τ µπ κουράδα τ κουραµάνα µπ/τ κουρέλα τ κουρελαρία µπ/τ κούρνια µπ κούρσα τ κούτα µπ/τ κουτάλα µπ/τ κουταµάρα µπ/τ κουτουράδα τ κούτρα µπ/τ κουτρουβάλα µπ/τ κουτσαµάρα τ κουτσαύτα τ κουτσοδόντα τ κουτσοµούρα µπ κουτσοµπόλα τ κουτσοµύτης τ κουτσονούρα τ κουτσουκέλα µπ/τ κουφάλα τ κουφαµάρα τ κουφοξυλία µπ κόφα τ µπ κοψοχέρα τ κράµπα µπ κρασίλα τ µπ κρασοκανάτα µπ/τ κρασοκανάτας µπ κρεατίλα τ κρεατόµυγα µπ/τ κρεατόπιτα τ κρεατόσουπα τ κρεβατίνα µπ/τ κρεβατοκάµαρα τ κρεβατοµουρµούρα τ µπ κρεµάλα µπ/τ κρεµµυδίλα τ κρησάρα µπ/τ κρουαζιέρα τ κρούστα τ κυράτσα µπ/τ κυρούλα µπ/τ κωλοπηλάλα µπ κωλοτούµπα µπ/τ λάβα τ µπ λάβρα τ µπ λαγάνα µπ λαγουδέρα τ λαγωφθαλµία τ λαγωχειλία τ λαδίλα τ
Continued
291
Table 54 continued
Word Gpl G Pl Word Gpl G Pl λαδόκολλα µπ λαθραλιεία τ µπ λαϊκούρα µπ/τ λαίλαπα µπ λαιµαργία τ λάκα τ µπ λακέρδα µπ λάκκα µπ/τ λακκούβα µπ/τ λαλαγγίτα, λαλαγγίδα µπ/τ λάµα µπ λάµια τ λανάρα τ λάντζα, λάντσα µπ/τ λαντζιέρα τ λαοθάλασσα µπ λαρυγγίτιδα µπ λατάνια, λατανία µπ λατινικούρα τ λάτρα τ µπ λάτρης τ λαύρα τ λαχτάρα µπ/τ λεβάντα µπ λεβεντογέννα µπ/τ λεβεντοµάνα τ λεβεντοπνίχτρα µπ/τ λεία τ λεµονίτα τ λεµονόκουπα µπ/τ λεµονόφλουδα µπ λέπρα τ µπ λέρα τ λετσαρία τ µπ λεχώνα µπ µπ λιακάδa µπ λιάστρα µπ λιβελούλα µπ λίγδα µπ λίγκα µπ λιγούρα µπ/τ λιγωµάρα τ λιθάγρα µπ λίµα µπ λινάτσα µπ λιοµαζώχτρα µπ λοβιτούρα µπ/τ λογικοκρατία τ µπ λοκάντα µπ/τ λοκοµοτίβα τ λόξα µπ/τ λόρδα µπ/τ λότζα τ λουίζα µπ/τ λουµπάρδα, λοµβάρδα, λούµπα τ λοµπάρδα µπ λουµπίνα τ λούπα µπ λουτροπετσέτα µπ λούτσα µπ/τ λούφα τ µπ λώβα µπ λωλάδα µπ µαγειρίτσα µπ/τ µαγεύτρα µπ/τ µάγια µπ/τ
Continued
292
Table 54 continued
Word Gpl G Pl Word Gpl G Pl µαγιονέζα µπ/τ µαγκιόρα τ µαγκίτισσα τ µαγκούρα µπ/τ µαγούλα µπ/τ µαεστρία µπ µαζούρκα µπ/τ µαζόχα τ µαζόχας µπ µαΐστρα τ µαϊστροτραµουντάνα µπ µάκα τ µακαρίτισσα τ µακροζωία τ µπ µαλαγάνα µπ/τ µαλάγρα µπ µαλάκα τ µαλαπέρδα τ µαλάρια τ µπ µαλαφράντζα τ µπ µαλλούρα τ µαλοτίρα µπ µαλτεζόπλακα τ µάνα τ µαναβέλα τ µανάβισσα τ µανέλα τ µανέστρα µπ/τ µάνητα τ µπ µανία τ µανιβέλα µπ/τ µανιέρα τ µπ µάνικα µπ/τ µανικιουρίστα τ µανιφατούρα µπ/τ µανόλια τ µανούβρα µπ/τ µανούρα τ µαντάµα µπ/τ µανταρίστρα τ µαντέκα τ µπ µαντζουράνα τ µαντόλα τ µπ µαντολινάτα µπ/τ µαντόνα τ µάντρα τ µαούνα µπ/τ µάππα µπ/τ µαργαρίτα τ µαρέγκα τ µπ Μαρία τ µαρίνα µπ/τ µαρινάτα µπ/τ µαριχουάνα τ µπ µάρκα µπ/τ µαρκίζα µπ/τ µαρµάγκα µπ/τ µαρµελάδα µπ µαρµίτα τ µαρουλοσαλάτα µπ µάσα µπ/τ µασέλα τ µασίνα τ µάσκα τ µασκαράτα µπ/τ µαστοειδίτιδα µπ µαστοράντζα τ µπ µαστόρισσα τ
Continued
293
Table 54 continued
Word Gpl G Pl Word Gpl G Pl µαστούρα µπ/τ µατζουράνα µπ/τ µατρόνα µπ/τ µατσαράγκα τ µατσόλα µπ/τ µατσούκα τ µαυραγορίτισσα τ µαυρίλα µπ/τ Μαυροθαλασσίτισσα τ µαυροµάλλης τ µαυροµάλλικο τ µαυροµαντιλούσα µπ µαυροφόρα τ µαυροφορούσα τ µαφιόζα τ µέγαιρα τ µεγαλοκοπέλα µπ/τ µεγαλουσιάνα τ µεγαλοφυΐα τ µεζονέτα τ µπ µειξοπαρθένα, µεζούρα µπ µιξοπαρθένα µπ Μέκκα τ µελίγκρα µπ/τ µελιτζάνα τ µελιτζανοσαλάτα τ µελόντικα µπ/τ µελόπιτα µπ/τ µένουλα µπ µέντα µπ/τ µερούλα µπ µεσόπορτα µπ/τ µεσσαλίνα τ µεταξότριχα µπ µετάνοια µπ µετζεσόλα τ µετρέσα µπ/τ µηδενικούρα µπ/τ µηλόπιτα τ µηχανότρατα µπ/τ µίζα τ µιζέρια µπ/τ µικροµάνα τ µικρόνοια τ µικροτυπία µπ µιλιόρα µπ µιµόζα τ µίνα µπ µινιατούρα µπ/τ µισµίζα τ µογγόλα τ µόδα µπ/τ µοιρολογήτρα τ µοιρολογίστρα τ µόκα τ µπ µολόχα µπ/τ µονέδα µπ/τ µονταζιέρα µπ µοντέλα µπ/τ µόρα τ µορταδέλα, µουρταδέλα µπ/τ µόρτισσα τ µόστρα µπ/τ µοτορόλα τ
Continued
294
Table 54 continued
Word Gpl G Pl Word Gpl G Pl µουβιόλα τ µούγγα τ µπ µουγγαµάρα τ µπ µούλα τ µουλάρα τ µούµια τ µουνόψειρα µπ/τ µούντζα, µούτζα µπ/τ µουντζούρα µπ/τ µούργα τ µπ µουργέλα µπ/τ µούρλα µπ/τ µούρλια τ µπ µουρµούρα µπ/τ µουρντάρα τ µουρούνα µπ µουρταδέλα τ µουρτζούφλα τ µουσαφίρισσα τ µουσίτσα µπ/τ µουσούδα τ µουστάρδα µπ µουσταρδιέρα µπ µουστόγρια τ µούτζα τ µουτζούρα τ µουτσούνα µπ/τ µουτσουνάρα µπ µούφα τ µουφλούζα τ µούχλα τ µπ µπαγαζιέρα τ µπαγαµπόντισσα τ µπαγαπόντισσα τ µπαγιαντέρα µπ µπαγιατίλα τ µπ µπαγιονέτα µπ µπαγκαζιέρα µπ/τ µπαγκατέλα, µπακατέλα µπ/τ µπαγκέτα µπ µπάζα τ µπάκα µπ/τ µπακάλισσα τ µπάλα τ µπαλαίνα, µπαναίλα µπ µπαλαλάικα µπ/τ µπαλάντα µπ/τ µπαλαντέζα µπ/τ µπαλαρίνα µπ µπαλάφα µπ µπαλκονόπορτα τ µπάµια µπ/τ µπαµπέσα τ µπαµπόγρια τ µπανανόφλουδα µπ/τ µπανιέρα τ µπάνκα, µπάγκα µπ/τ µπανκανότα µπ µπάντα µπ/τ µπαντάνα µπ/τ µπαντανόβουρτσα µπ µπαντιέρα τ µπάρα µπ/τ µπαρµπαρέσα τ µπαρµπουτιέρα µπ/τ µπαρόβια τ
Continued
295
Table 54 continued
Word Gpl G Pl Word Gpl G Pl µπαρούµα µπ/τ µπαρούφα τ µπ µπατάλα τ µπατζανάκισσα τ µπατίρισσα τ µπατονέτα µπ/τ µπατσαρία τ µπάφα τ µπαχατέλα τ µπεκιάρισσα τ µπεκροκανάτα µπ/τ µπελαµάνα τ µπελαντόνα µπ/τ µπέµπα µπ/τ µπεµπέκα µπ/τ µπερλίνα τ µπ µπέρτα µπ/τ µπετονιέρα τ µπιγκόνια, βιγόνια, µπετούγια τ βεγόνια µπ µπιζουτιέρα τ µπίλια µπ/τ µπιµπίλα τ µπίρα µπ/τ µπιρίµπα τ µπ µπιρµπίλα τ µπλούζα τ µπλόφα µπ/τ µπογιά µπ µποέµισσα τ µπόλια µπ/τ µπόµπα τ µποµπάρδα µπ/τ µποµπίνα µπ µποµπονιέρα τ µποµπότα µπ/τ µπονάτσα τ µπόρα µπ/τ µπορντούρα µπ/τ µπότα τ µποτίλια µπ µπουγάδα µπ/τ µπουγάτσα, µπογάτσα µπ/τ µπουγιαµπέσα µπ/τ µπούκα µπ/τ µπουκαδούρα τ µπουκάλα µπ µπουκαµβίλια τ µπουκαπόρτα µπ/τ µπουκίτσα µπ µπούκλα τ µπούλα τ µπουλντόζα µπ/τ µπουλούκα τ µπούµα µπ/τ µπουµπούκα τ µπουµπουνιέρα τ µπουνάτσα τ µπούρδα µπ/τ µπουρδελότσαρκα µπ/τ µπουρµπουλήθρα µπ/τ µπουρνέλα τ µπουτονιέρα τ µπουχάρα µπ µπόχα µπ/τ µπρατσέρα τ
Continued
296
Table 54 continued
Word Gpl G Pl Word Gpl G Pl µπροσούρα µπ µπροστάντζα µπ/τ µπροστέλα µπ/τ µπροστινέλα τ µπρουντζίνα τ µυγοσκοτώστρα µπ/τ µυΐτιδα µπ µυλόπετρα τ µύξα τ µυξιάρα τ µυρµηγκότρυπα µπ νέγρα τ νέκρα τ µπ νεκρόκασα µπ νεκροφοβία µπ νένα µπ/τ νεραγκούλα µπ/τ νεράιδα µπ νεραντζούλα µπ νερόκοτα µπ/τ νεροµάννα, νεροµάνα µπ/τ νεροµολόχα µπ νεροφίδα τ νησιωτοπούλα τ νίλα µπ/τ νοθεία µπ νοικοκυροπούλα τ νοµενκλατούρα µπ/τ νόνα τ νόρµα τ νότα τ νουβέλα µπ/τ νουγκατίνα τ νούλα τ µπ νταβανόσκουπα τ ντάλια µπ/τ ντάµα µπ/τ νταµιτζάνα µπ ντάνα µπ νταρντάνα µπ/τ ντελµπεντέρισσα, ντερµπεντέρισσα τ ντίβα µπ/τ ντιρεκτίβα µπ/τ ντουζιέρα, ντουσιέρα µπ/τ ντουζίνα τ ντουντούκα τ ντρίµπλα, ντρίπλα, ντρεζίνα τ τρίπλα µπ ντρίπλα τ νύστα τ νυφίτσα µπ ξαγρύπνια τ ξανθοµάλλης τ ξανθοµάλλικο τ ξανθούλα µπ ξάπλα µπ/τ ξαπλώστρα µπ/τ ξεβλάσταρο µπ ξεκούτα τ ξεµατιάστρα τ ξεµυαλίστρα τ ξενοµερίτισσα τ ξενύχτισσα τ ξεπατικωτούρα τ µπ
Continued
297
Table 54 continued
Word Gpl G Pl Word Gpl G Pl ξέρα τ ξεραΐλα τ ξεροφαγία, ξηροφαγία, ξερόλα τ ξηροφαγιά τ µπ ξετσιπωσιά µπ ξευτίλα µπ ξεφτίλα τ ξηρά µπ ξινήθρα τ ξινίλα µπ/τ ξόβεργα τ ξοµπλιάστρα µπ/τ ξούρα µπ/τ ξυλεία τ µπ ξυλόβιδα µπ/τ ξυλόκολλα µπ/τ ξυλόκοτα µπ/τ ξυλόπροκα τ ξυλόσοµπα τ ξωµερίτισσα τ οβίδα µπ Οβραία τ ογδοντάρα τ οδοντόβουρτσα τ οδοντόκρεµα τ οδοντόπαστα τ οδοποιία τ οδύσσεια τ µπ όζα τ οικοδέσποινα τ οινοποιία τ µπ οκτάβα τ οκτάβα, οχτάβα µπ οµελέτα µπ/τ οπαλίνα µπ όπερα µπ/τ οπερέτα µπ/τ οργαντίνα τ µπ οργκαντίνα τ όρκα µπ ορµήνια τ ορντινάντσα µπ/τ ορντινάτσα τ ορφάνια τ ορχιδέα τ ορχίτιδα µπ όστρια τ µπ ουάου τ ουβερτούρα µπ ούγια µπ/τ ουλίτιδα µπ ουρηθρίτιδα µπ ουρητηρίτιδα µπ όχεντρα τ οχτάβα τ οχταµηνίτισσα τ οχτάρα τ όχτρητα τ παγαπόντισσα τ παγκρεατίτιδα µπ παγόδα µπ/τ παγωνία µπ παιδούλα µπ πάλα µπ παλάβρα µπ/τ παλαβωµάρα τ
Continued
298
Table 54 continued
Word Gpl G Pl Word Gpl G Pl παλιατζούρα µπ/τ παλιάτσα τ παλιατσαρία, παλιατζαρία µπ/τ παλιοβρόµα µπ παλιογυναίκα µπ παλιοκουβέντα µπ/τ παλιοπαρέα τ παλιοσκρόφα τ πάµπα τ πάνα µπ πάντα τ παντζουρόβεργα µπ παντιέρα τ παντόφλα, παντούφλα µπ παντρολογίστρα, παντρειά µπ παντρολογήτρα µπ παπαγαλία µπ παπαδίτσα τ παπάρα µπ παπαρούνα µπ πάπια µπ/τ πάπρικα µπ/τ παραδουλεύτρα µπ παραµάνα, παραµάννα µπ παραπονιάρα τ παράτα µπ παρηγόρια, παρηγορία, παρηγοριά τ µπ παρκετέζα µπ/τ πάρλα µπ/τ παρλαπίπα µπ/τ παρλάτα τ παρµεζάνα µπ/τ παρoδοντίτιδα µπ παρόλα µπ/τ παρτιτούρα µπ/τ παρτούζα µπ/τ πάσα µπ/τ πασαρέλα µπ πασιέντσα, πασιέντζα µπ/τ πάστα φλόρα τ παστίλια µπ/τ πάστρα τ µπ πασχαλίτσα µπ/τ πατατούκα µπ/τ πατέντα µπ πατερίτσα µπ πατιτούρα τ πατούρα τ πατρόνα µπ πατσαβούρα µπ πεζούλα, πεζούρα τ µπ πείνα τ µπ πεινάλα τ πελελάδα τ πελότα µπ πελούζα µπ πενηντάρα τ πεντακοσάρα µπ/τ πεντάλφα τ πεντάρα µπ/τ πενταροδεκάρες µπ/τ πεπονόφλουδα µπ/τ
Continued
299
Table 54 continued
Word Gpl G Pl Word Gpl G Pl περγαµηνοποιία µπ πέργκολα µπ περδικούλα µπ/τ περηφάνια τ µπ περίσσεια τ µπ περιστερώνα τ πέρλα τ περπατούρα τ πεσκαντρίτσα µπ πέστροφα µπ/τ πετούγια τ πετούνια µπ πετρελαιόπισσα τ µπ πετροκαλαµίθρα τ πετροπέρδικα µπ/τ πηλάλα, πιλάλα τ µπ πήχτρα τ πιάτσα µπ/τ πίκα µπ/τ πίκρα µπ πικράδα τ µπ πικρίλα τ µπ πιλοτιέρα µπ πιπεριέρα µπ πιπίλα µπ πιρόγα µπ πισίνα τ πίτσα µπ πιτσιλάδα τ πιτσούνα µπ πλακόπιτα µπ πλατφόρµα µπ πλαφονιέρα µπ πλεµπάγια µπ/τ πνευµατολογία µπ πνιγούρα µπ ποδάγρα µπ ποδαρίλα τ µπ πόζα τ πολυδιψία τ µπ πολυθρόνα µπ πολυκαιρία τ µπ πολυκαρπία τ πολυτεκνία τ πολυτοκία τ πολυχρησία τ πολωνέζα τ µπ ποµάδα µπ ποµόνα τ πόµπα τ ποντικότρυπα µπ/τ πορφύρα µπ ποταµολογία τ ποτίστρα µπ πούδρα, πούντρα µπ πουδριέρα, πουντριέρα µπ/τ πουκαµίσα µπ πουλάδα, πουλακίδα µπ πούλια µπ/τ πούλπα τ πούντα τ πουντριέρα τ πουστιά µπ πούστρα τ πουτάνα µπ/τ πουτανιά µπ πουτανιάρα τ πουτίγκα µπ
Continued
300
Table 54 continued
Word Gpl G Pl Word Gpl G Pl πραλίνα µπ πραµάτεια µπ/τ πρασινίλα µπ/τ πρέζα µπ πρεµιέρα µπ/τ πρεµούρα µπ/τ πρεσβυτέρα µπ πριγκιποπούλα τ πρίζα, µπρίζα µπ πριµαντόνα µπ πριονοκορδέλα µπ πρόβα τ προβατίλα τ µπ προβατίνα µπ προβοκάτσια µπ/τ πρόγκα µπ πρόζα τ µπ πρόκα µπ προλετάρια τ προπαγάνδα µπ/τ προστασία µπ προστάτισσα τ προστατίτιδα µπ προστυχάντζα µπ/τ προστυχόφατσα τ προσφυγοπούλα τ πρυµάτσα µπ/τ πρωία τ µπ πρωτάρα τ πρωτευουσιάνα τ πρωτόγεννα τ µπ πυγµαία τ πύρα τ µπ πυράδα τ πυρόσφαιρα τ ραβδιστήρα τ ράδα τ ραουλιέρα τ ράσπα τ ράτσα µπ/τ ραχίτιδα µπ ραχούλα µπ ρεβερέντζα, ρεβεράντζα τ ρεβεράντζα µπ ρεβιθάδα µπ ρέγγα τ ρέγουλα τ µπ ρεκλάµα τ ρεµούλα µπ/τ ρεµούλκα τ ρέντα τ ρεντιγκότα τ ρεντινγκότα, ρεντικότα τ ρεντιγκότα µπ ρεπούµπλικα, ρεπλίκα τ ρεµπούµπλικα µπ/τ ρεπουµπλικάνα τ ρετσέτα τ ρετσίνα µπ ρετσινόκολλα τ ρέφουλα τ ρέχα τ
Continued
301
Table 54 continued
Word Gpl G Pl Word Gpl G Pl ρήγας τ ρηγοπούλα τ ρηµάδα τ ρητινόπισσα µπ ρηχία τ ριζάγρα τ ριζονευρίτιδα µπ ριµάδα τ ριµάτα µπ ρινίτιδα µπ ρόκα µπ/τ ροκάνα µπ/τ ροκιά µπ ροµάντζα µπ/τ ρόµπα τ ροµπόλα τ µπ ρότα µπ/τ ροτόντα µπ/τ ρούγα µπ/τ ρουκάνα τ ρουτίνα τ µπ ρουφήχτρα τ ρουφιάνα τ ροχάλα µπ/τ ρυµούλκα τ σακαράκα µπ/τ σακολέβα τ σακοράφα τ σακούλα µπ/τ σάλα µπ/τ σαλαµάστρα µπ σαλαµούρα τ µπ σαλατιέρα µπ/τ σαλιάρα µπ σαλιαρίστρα τ σαλτσιέρα τ σάµπα τ µπ σαµπάνια µπ/τ σαµπανιέρα τ σαµπούκα µπ σανιδόσκαλα µπ σαντακρούτα τ σαντέζα τ σάουνα τ σαπίλα τ µπ σαπουνόπερα µπ/τ σαπουνόπετρα µπ σαπουνόφουσκα µπ/τ σαπωνοποιία µπ σάρα τ σαρανταποδαρούσα τ µπ σαραντάρα τ σάρπα τ σαστιµάρα τ σατακρούτα, σαστισµάρα τ σαντακρούτα µπ/τ σάχλα τ σβάστικα µπ/τ σβελτάδα τ µπ σέντρα τ σεξουάλα τ σέπια τ µπ σέρα µπ σερβιτόρα τ σερενάτα µπ/τ σερπαντίνα µπ/τ
Continued
302
Table 54 continued
Word Gpl G Pl Word Gpl G Pl σέσουλα µπ/τ σηκωµάρα µπ σηµαδούρα µπ/τ σιγκούνα, σεγκούνα µπ σιγουράδα τ σιδερώστρα µπ σιλουέτα τ σιφονιέρα τ σιχαµάρα τ µπ σκακιέρα τ σκάλα τ σκάντζα τ σκαρλατίνα µπ σκαρταδούρα τ σκασίλα τ σκατίλα τ σκατούλα τ σκλαβοπούλα τ σκληράδα τ σκληρίτιδα µπ σκορδίλα τ σκορδοκαήλα µπ σκορδοκαΐλα τ σκορπίνα µπ σκότα τ σκοτεινάδα τ µπ σκοτισµάρα µπ σκοτούρα µπ/τ σκοτώστρα τ σκουληκαντέρα µπ σκούνα µπ σκουντούφλα µπ/τ σκουπιδιάρα µπ/τ σκούφια µπ/τ σκουφίτσα µπ σκύλα τ σκυλίτσα µπ σκυλοµούρα τ σµιχτοφρύδα τ σµπόµπα τ σνοµπαρία µπ/τ σόδα τ σοκακιάρα τ σοκολατίνα µπ σόντα µπ σοροκάδα µπ σούβλα τ σουετίνα τ σούζα µπ/τ σουίτα τ σουλτάνα µπ/τ σουλτανίνα µπ/τ σούµα µπ/τ σουµάδα µπ σούπα τ σουπιέρα τ σούρα µπ σουσουράδα µπ σούστα τ σούφρα µπ/τ σουφραζέτα τ σοφεράντζα τ σπαγγεταρία τ σπαγγετερία τ σπάθα µπ σπάλα τ σπανοµαρία τ σπαρίλα µπ/τ
Continued
303
Table 54 continued
Word Gpl G Pl Word Gpl G Pl σπασίκλα τ σπέκουλα µπ/τ σπεκουλάτσια µπ σπεράντσα, σπεράντζα τ σπιούνα τ σπονδυλαρθρίτιδα µπ σπόντα µπ/τ στάλα µπ/τ σταξιά µπ σταυροµάννα µπ σταφυλόρωγα µπ σταχτοκουλλούρα µπ Σταχτοπούτα τ στεατοπυγία µπ στέγνα τ στέκα τ στεναχώρια τ στενούρα τ στενοχώρια τ στέρνα µπ στοµατίτιδα µπ στραβωµάρα τ στράκα µπ/τ στρακαστρούκα τ στράτα µπ/τ στρατούλα µπ/τ στρατώνα τ στρούγκα µπ στρουκτούρα µπ στρωµατσάδα µπ συκοµαΐδα τ συµβία µπ/τ συµπεθέρα τ συµπιεστότητα µπ συµπόνια τ µπ συρµατόβεργα µπ συρταριέρα µπ/τ σφαλιάρα τ σφιχτοχέρα τ σφολιάτα µπ σφουγγαρίστρα µπ σχάρα, σκάρα µπ ταβανόπροκα, ταβανόσκουπα, νταβανόπροκα µπ νταβανόσκουπα µπ/τ ταχινόσουπα µπ ταβερνιάρισσα τ τάβλα µπ τάγια τ ταγκίλα τ ταλαιπώρια τ τάλια τ ταλιατέλες τ ταµπακιέρα µπ/τ τανάλια µπ ταξιδεύτρα τ τάρα, ντάρα µπ ταραντέλα µπ ταραντούλα µπ ταρταρούγα µπ ταστιέρα, ταστέρα µπ ταφόπετρα µπ/τ ταφόπλακα µπ/τ τέµπερα τ τενοντίτιδα µπ τερακότα τ τεσσάρα µπ/τ
Continued
304
Table 54 continued
Word Gpl G Pl Word Gpl G Pl τετρακοσάρα τ τεχνίτρα τ τζιριτζάντζουλα, τζίβα, τσίβα µπ/τ τσαριτσάντζουλα µπ/τ τζίφρα µπ τζούρα µπ/τ τζουτζούκα τ τιτίζα τ τοκάτα µπ τονοσαλάτα τ τοστιέρα τ τουβούλα τ τουλούµπα µπ/τ τούµπα µπ Τουρκάλα µπ τουρκογύφτισσα τ τουρκοµερίτισσα τ τουρκοπούλα τ τούρλα τ µπ τούρνα τ τουρτούρα τ τράβα µπ τραγάνα µπ τραγιάσκα µπ τραγίλα τ µπ τράκα µπ/τ τρακατρούκα, στρακαστρούκα µπ/τ τραµουντάνα τ µπ τράµπα µπ/τ τραµπάλα µπ τραπεζιέρα τ τραχηλίτιδα µπ τρέλα µπ/τ τρελαµάρα µπ/τ τρελάρα τ τρελοπαντιέρα µπ/τ τρεµεντίνα µπ τρεµούλα τ µπ τρεφιλιέρα τ τρεχάλα µπ/τ τριαντάρα τ τριάρα µπ/τ τριγυρίστρα τ τρίλια µπ/τ τρίπλα τ τριψάνα τ τρόικα µπ τροµάρα µπ/τ τρόµπα µαρίνα τ τροτέζα µπ/τ τρούφα µπ τρυγήτρα τ τρυπιοχέρα τ τρυφεράδα µπ τρυφερότητα µπ τσαγιέρα τ τσαγκαροδευτέρα τ µπ τσαγκίλα τ τσάκα τ τσακίστρα τ τσακµακόπετρα µπ/τ τσαλάκα µπ τσαµπούνα µπ τσαννάκα, τσανάκα µπ/τ
Continued
305
Table 54 continued
Word Gpl G Pl Word Gpl G Pl τσαντίλα τ τσαούσα τ τσαπερδόνα µπ/τ τσαρίνα µπ τσάρκα µπ/τ τσατίλα, τσαντίλα µπ/τ τσάτσα τ τσελιγκοπούλα τ τσιγγανοπούλα τ τσικουδία µπ τσίλια µπ/τ τσιµινιέρα τ τσίµπλα µπ τσίπα τ µπ τσιππούρα µπ τσιριµόνια µπ/τ τσίρλα τ τσίτσα τ τσιτσιµπίρα, τζιτζιµπίρα µπ τσίφτισσα τ τσιχλόφουσκα µπ τσοκαρία τ τσοµπανοπούλα τ τσόντα µπ τσοπανοπούλα τ τσότρα µπ/τ τσούλα τ τσούπρα, τσούπα µπ/τ τσόχα µπ τυπικούρα τ τύπισσα µπ τυράννια τ τυριέρα µπ/τ τύφλα τ µπ τυφλόµυγα τ µπ υγεία τ υπερεπάρκεια τ µπ υπόγα τ υποδηµατοποιία τ µπ φάβα µπ/τ φαγάνα µπ/τ φαγιάντσα µπ/τ φαγούρα τ µπ φαγωµάρα µπ/τ φάκα µπ φαλτσοστέκα τ φαµίλια τ φάµπρικα µπ/τ φανουρόπιτα τ φανφάρα, φαµφάρα µπ φάπα µπ φάρα µπ/τ φάρµα µπ/τ φαρµακίλα τ µπ φαρµακόγλωσσα µπ φάσα µπ φασίνα τ µπ φατσούλα µπ φάτσα µπ φελούκα µπ φευγάλα µπ/τ φθήνια τ φιγούρα µπ φιδότρυπα µπ φιλιέρα τ φιλιππινέζα µπ
Continued
306
Table 54 continued
Word Gpl G Pl Word Gpl G Pl φιλοτιµία µπ φινέτσα τ µπ φιοριτούρα µπ/τ φίρµα µπ φλογέρα µπ φλοκάτα µπ φλούδα, φλοίδα µπ φοβέρα µπ/τ φόδρα µπ φόλα µπ/τ φοντανιέρα τ φόρα τ µπ φόρµα µπ φορµάικα τ µπ φόρµουλα µπ φορτέτσα µπ φουκαριάρα τ φούµα τ φουµαδόρισσα τ φούξια µπ φούρια µπ/τ φούρκα µπ φουρτούνα µπ φούσκα µπ φουσκάλα µπ φουσκωµάρα τ φούστα µπ φουστανέλα µπ φραγκόκοτα µπ/τ φράντζα µπ φραντζόλα µπ φράξια µπ φρενοπαθολογία µπ φρίζα µπ φριτέζα µπ/τ φριτούρα µπ φρονιµάδα τ φρουτιέρα µπ/τ φρυγανιέρα µπ/τ φτήνια, φθήνια µπ/τ φτώχεια τ φτώχια µπ φτωχοµάνα τ φυσαρµόνικα µπ/τ φυσούνα µπ/τ φώκια τ φωλίτσα µπ φωνάρα µπ φωτοδότρα τ φωτοκόπια µπ χαβάγια µπ χαβούζα µπ χάβρα µπ χαζαµάρα τ χαζοβιόλα τ χαζοκουβέντα µπ/τ χαζοµάρα µπ/τ χαιρετούρα µπ/τ χαλάουα τ χαλάστρα τ χαµαλίκα µπ χαµηλοβλεπούσα µπ/τ χαµοκέλλα µπ χαµοπέρδικα µπ/τ χαµούρα τ χαραµοφάισσα τ χαρτοσακούλα µπ χασούρα τ µπ
Continued
307
Table 54 continued
Word Gpl G Pl Word Gpl G Pl χέστρα τ χηρεία τ χιλιάρα τ χιλιοχρονίτισσα τ χιονοθύελλα τ χλαπάτσα, κλαπάτσα τ µπ χλεµπόνα µπ χοντρέλα µπ/τ χορτόπιτα τ χορτόσουπα τ χούντα µπ/τ χριστοπαναγία τ χρυσόκολλα µπ χρυσόµυγα µπ χρυσοχέρα τ χρωµατοπυξίδα µπ χωµατίλα τ µπ χωνεύτρα µπ χωριατιά µπ χωριατοπούλα τ ψάθα τ ψάλτρια τ ψαρίλα τ µπ ψαρόβαρκα µπ/τ ψαροκασέλα µπ/τ ψαρόκολλα µπ/τ ψαροµάλλα τ ψαροπούλα µπ/τ ψαρόσουπα τ ψαρότρατα µπ ψηλογκαµήλα τ ψηλοµύτα τ ψηστιέρα τ µπ ψηφοθηρία τ µπ ψιλοκουβέντα τ ψίχα τ ψιχάλα µπ/τ ψιψίνα τ ψυχάρα τ ψυχοµάννα, ψυχοµάνα µπ/τ ψυχοπαίδα τ ψυχόπιτα τ ψύχρα τ µπ ψυχραιµία τ µπ ψωµιέρα µπ/τ ψωµόλυσσα τ µπ ψώρα τ µπ ψωροκώσταινα τ µπ ψωροπερηφάνια τ
B2. Feminine nouns with nominative singular –η and plural –ες
The nouns in this inflectional category have stress on the final syllable in the genitive plural, but may have stress on any of the final three syllables in the other forms.
308
Example paradigms:
η κόρη αντάµωση ‘daughter’ SINGULAR PLURAL ‘meeting’ SINGULAR PLURAL NOMINATIVE κόρη κόρες NOMINATIVE αντάµωση αντάµωσες ACCUSATIVE κόρη κόρες ACCUSATIVE αντάµωση αντάµωσες GENITIVE κόρης κορών GENITIVE αντάµωσης ανταµωσών VOCATIVE κόρη κόρες VOCATIVE αντάµωση αντάµωσες
Table 55: Examples of Modern Greek feminine nouns with nominative singular –η and plural –ες
Quick stats:
• Babiniotis and Triantafillidis combined: 192 nouns with genitive plural gaps
• Only Triantafillidis:
o 169 genitive plural gaps (10.8% of all genitive plural gaps)
o 1,455 nouns in this inflection class (5.3% of all nouns)
Nouns with genitive plural gaps:
Word Gpl G Pl Word Gpl G Pl αβροφροσύνη µπ/τ αγάπη µπ/τ αγαρµποσύνη τ αγιοσύνη τ αγκάλη µπ/τ αγνωµοσύνη τ µπ αγράµπελη τ µπ αγριοβρόµη τ αγριόβροµη τ αετοράχη τ αϊτοράχη τ άλµη τ αµασκάλη, αµασχάλη τ αµάχη µπ/τ άµπωτη τ µπ αµυαλοσύνη τ αναθύµηση τ αναξιοσύνη τ ανάπαψη τ ανέµη µπ ανεµοζάλη τ µπ ανεµώνη µπ/τ αντιλόπη τ αντίχριστη τ αντρειοσύνη τ απαλοσύνη τ
Continued
Table 56: Genitive plural gaps among feminine nouns with nominative singular –η and plural –ες
309
Table 56 continued
Word Gpl G Pl Word Gpl G Pl απελεύθερη τ απεραντοσύνη τ µπ άρµη, άλµη τ µπ ασηµόσκονη τ µπ ασχετοσύνη τ ασχηµοσύνη τ ατζαµοσύνη τ ατόλη µπ ατσαλοσύνη τ άχνη τ µπ βασιλοκόρη µπ βενεδικτίνη µπ βιασύνη τ βοή, βουή) τ µπ βουή τ βρύση τ γαλήνη τ µπ γενναιοφροσύνη τ γεροντοκόρη µπ γεροπαράξενη τ γιαούρτη τ γλυκόζη τ µπ δεξιοσύνη τ διαµαντόσκονη µπ διαµαρτυρόµενη τ διαµάχη τ διανοούµενη τ δουλοφροσύνη τ δραµαµίνη τ εγκυµοσύνη τ µπ ειδή τ µπ ελεηµοσύνη µπ/τ ελευθεροφροσύνη τ εµπορευάµενη τ εξαδέλφη, εξαδέρφη τ ερυθρόδερµη τ ευγνωµοσύνη τ µπ ζάλη τ ζάχαρη µπ/τ ζαχαρίνη τ ζέστη µπ/τ ζυµάση τ ζωγραφική τ ήβη τ µπ θρεψίνη τ θύµηση τ ιερόδουλη τ ιεροσύνη τ µπ κάδη τ κακοσύνη τ καλλιόπη τ καλοσύνη µπ/τ κάµαρη τ καπατσοσύνη τ κάππαρη µπ/τ καρβουνόσκονη τ κατηχούµενη τ κήλη τ µπ κουφόβραση µπ/τ κράµβη µπ κρεβατοκάµαρη τ κρέµαση τ κωλότσεπη µπ/τ λαδορίγανη τ µπ µπ λαίδη τ λακτάση τ λύπη µπ µαµµή µπ
Continued
310
Table 56 continued
Word Gpl G Pl Word Gpl G Pl µάµµη µπ µαργαρίνη µπ/τ µαρµαρόσκονη τ µασχάλη τ µαύρη τ µαυροδάφνη µπ/τ µεγαλοπραγµοσύνη τ µεγαλορρηµοσύνη τ µεγαλοσύνη τ µπ Μεγαλόχαρη τ µέγγενη τ µειξοπάρθενη τ µελλοθάνατη τ µέση µπ/τ µετριοφροσύνη µπ/τ µιλέδη τ µπ µνήµη µπ/τ µοναχοκόρη µπ/τ µούρη µπ/τ µπέµπελη µπ τ µπουνταλοσύνη τ µπριγιαντίνη µπ/τ µύλη τ µύτη τ µωαµεθανή τ νιόνυφη µπ/τ νιότη τ µπ νοικοκυροσύνη τ µπ ντραµαµίνη τ νύφη τ ξαδέλφη, ξαδέρφη τ ξερόβρυση µπ οδαλίσκη τ οδοντιατρική τ οδοντίνη τ οφθαλµαπάτη τ παιδίσκη µπ/τ πάλη τ µπ παππαδοκόρη, πανώλη τ παπαδοκόρη µπ παράνυφη τ παρατρεχάµενη τ πέραση τ πεύκη τ πεψίνη τ πικροδάφνη µπ πλώρη µπ πορσελάνη µπ προµάµµη τ µπ προσµονή τ προσοχή τ µπ πρωτοξαδέρφη τ ρητινόλασπη µπ ρίγανη µπ/τ σακχαρίνη τ σκανδάλη τ σκάφη τ σκόλη, σχόλη µπ/τ σκόνη µπ/τ σκορδόπιστη τ σµυριδόσκονη µπ τ σουλφαµιδόσκονη τ στάνη τ µπ στάφνη τ στάχτη µπ στέγη τ
Continued
311
Table 56 continued
Word Gpl G Pl Word Gpl G Pl στεφάνη µπ συγγνώµη τ συγνώµη τ σύφιλη τ µπ συχώρεση τ σχόλη τ ταπεινοσύνη τ ταπεινοφροσύνη τ µπ τεστοστερόνη µπ Τετάρτη τ τόλµη τ µπ Τρίτη τ Τσικνοπέµπτη µπ/τ υπακοή τ µπ φαρµακευτική τ φαρµακόγλωσση τ φέξη τ µπ φέρµελη µπ/τ φίλη µπ φλοκάτη µπ φορµόλη τ µπ φρουκτόζη τ µπ φτέρη µπ φυγόδικη τ φυγόποινη τ χάρη µπ/τ χάση τ µπ χειροπέδη τ χηρευάµενη τ χόβολη τ µπ χρυσόσκονη τ µπ ψυχοκόρη µπ/τ
B3. Masculine nouns with nominative singular –ας and plural –ες
Nouns in this inflection class represent two stress patterns in the genitive plural. Some have stress on the penultimate, while for other nouns the genitive plural has stress on the final syllable, regardless of stress in the other word forms.
312
Example paradigms:
ο φύλακας ο αγκώνας SINGULAR PLURAL SINGULAR PLURAL NOMINATIVE φύλακας φύλακες NOMINATIVE αγκώνας αγκώνες ACCUSATIVE φύλακα φύλακες ACCUSATIVE αγκώνα αγκώνες GENITIVE φύλακα φυλάκων GENITIVE αγκώνα αγκώνων VOCATIVE φύλακα φύλακες VOCATIVE αγκώνα αγκώνες
ο τουρίστας ο άντρας SINGULAR PLURAL SINGULAR PLURAL NOMINATIVE τουρίστας τουρίστες NOMINATIVE άντρας άντρες ACCUSATIVE τουρίστα τουρίστες ACCUSATIVE άντρα άντρες GENITIVE τουρίστα τουριστών GENITIVE άντρα αντρών VOCATIVE τουρίστα τουρίστες VOCATIVE άντρα άντρες
Table 57: Examples of Modern Greek masculine nouns with nominative singular –ας and plural –ες
Quick stats:
• Babiniotis and Triantafillidis combined: 57 genitive plural gaps
• Only Triantafillidis:
o No genitive plural gaps in this inflection class (0% of all gaps)
o 831 nouns in this inflection class (3.0% of all nouns)
313
Nouns with genitive plural gaps:
Word Gpl G Pl Word Gpl G Pl αλητάµπουρας µπ αµπλαούµπλας µπ άντρακλας µπ αρτίστας µπ γεροντόµαγκας µπ γεροξούρας µπ γκαβούλιακας µπ γρίβας µπ δηµοσιοσχετίστας µπ δόκτορας µπ ζήτουλας µπ ζουρλοµανδύας µπ θεσιθήρας µπ κάβουρας µπ καζανόβας µπ κανάγιας µπ κάπελας µπ κοντραµπασίστας µπ µαγαζάτορας µπ µάγκας µπ µαικήνας µπ µαλάκας µπ µάππας µπ µαριονετίστας µπ µπάκακας µπ µπαµπούλας µπ µπόµπιρας µπ µπουµπούνας µπ µπούσουλας µπ ξερόβηχας µπ ξερόλας µπ ξευτίλας µπ οµορφάντρας µπ οργανίστας µπ παπάρας µπ παπαρδέλας µπ πάφιλας µπ πεινάλας µπ πόντικας µπ πρόποδες µπ ραλίστας µπ ρούφουλας µπ σκούληκας µπ σπεσιαλίστας µπ σπόνσορας µπ τερορίστας µπ τζαζίστας µπ τζίτζικας µπ τζίτζιρας µπ τρελάρας µπ τυφλοπόντικας µπ φαταούλας µπ φίστουλας µπ φλαουτίστας µπ χρυσογέρακας µπ ψείρας µπ ψηλολέλεκας µπ
Table 58: Genitive plural gaps among masculine nouns with nominative singular –ας and plural –ες
314
B4. Neuter nouns with nominative singular –ι
The nouns in this inflection class have one more syllable in the plural and the genitive singular than in the nominative and accusative singular. Stress placement in the genitive plural is always on the final syllable, but stress in the rest of the plural forms and the genitive singular falls according to two patterns. If stress is on the final syllable in the nominative singular, it is on the final syllable throughout the paradigm, as in παιδί. If stress is on the on the penultimate syllable in the nominative singular, stress shifts to the final syllable in the genitive singular and genitive plural, but is otherwise columnar, as in
αγόρι. According to Holton et. al (1997:65) there is only one noun of this type with stress on the antepenultimate – φίλντισι ‘ivory’
Example paradigms:
το παιδί το αγόρι SINGULAR PLURAL SINGULAR PLURAL NOMINATIVE παιδί παιδιά NOMINATIVE αγόρι αγόρια ACCUSATIVE παιδί παιδιά ACCUSATIVE αγόρι αγόρια GENITIVE παιδιού παιδιών GENITIVE αγοριού αγοριών VOCATIVE παιδί παιδιά VOCATIVE αγόρι αγόρια
Table 59: Examples of Modern Greek neuter nouns with nominative singular –ι
315
Quick stats:
• Babiniotis and Triantafillidis combined: 15 genitive plural gaps
• Only Triantafillidis:
o No genitive plural gaps (0% of all genitive plural gaps)
o 1,794 nouns in this inflection class (6.6% of all nouns)
Nouns with genitive plural gaps:
Word Gpl G Pl Word Gpl G Pl γουλί µπ καζίκι µπ κρασοπότι µπ λιονταρίνα µπ µατζούνι µπ µέλι µπ µετάξι µπ νέφτι µπ νοίκι µπ σέσελι µπ σπόρι µπ τούλι µπ χαµαλίκι µπ
Table 60: Genitive plural gaps among neuter nouns with nominative singular –ι
B5. Neuter nouns with nominative singular –ος
In this inflection class, all nouns have stress on the final syllable in the genitive plural.
Nouns with stress on the penultimate in the nominative singular and most other cases have stress on the final syllable only in the genitive plural (το κράτος). Those nouns with stress on the antepenultimate syllable in the nominative singular have stress on the penultimate syllable in the genitive singular, and throughout the plural cases (το
πέλαγος).
316
Example paradigms:
το κράτος το πέλαγος ‘state’ SINGULAR PLURAL ‘sea’ SINGULAR PLURAL NOMINATIVE κράτος κράτη NOMINATIVE πέλαγος πελάγη ACCUSATIVE κράτος κράτη ACCUSATIVE πέλαγος πελάγη GENITIVE κράτους κρατών GENITIVE πελάγους πελαγών VOCATIVE κράτος κράτη VOCATIVE πέλαγος πελάγη
Table 61: Examples of Modern Greek neuter nouns with nominative singular –ος
Quick stats:
• Babiniotis and Triantafillidis combined: 12 nouns with genitive plural gaps
• Only Triantafillidis:
o 11 genitive plural gaps in this inflection class (0.7% of all gaps)
o 120 nouns in this inflection class (0.4% of all nouns)
Nouns with genitive plural gaps:
Word Gpl G Pl Word Gpl G Pl άγος τ άγχος τ άλσος τ βένθος τ έλεος µπ θάµπος τ θάρρος τ µπ κάλλος µπ/τ κόστος τ µπ µάκρος τ µπ µίσος µπ/τ σκότος τ µπ
Table 62: Genitive plural gaps among neuter nouns with nominative singular –ος
317
B6. Masculine nouns with nominative singular –ης and plural –ες
Masculine nouns with the nominative singular marker –ης fall into two inflection classes.
One class has the nominative plural marker –ηδες. The other has nominative plural form
–ες. Only the latter have paradigmatic gaps and in this class, stress in the genitive plural is always on the final syllable, regardless of stress placement in the rest of the paradigm.
Example paradigms:
ο κλέφτης ο καθηγητής ‘thief’ SINGULAR PLURAL ‘professor’ SINGULAR PLURAL NOMINATIVE κλέφτης κλέφτες NOMINATIVE καθηγητής καθηγητές ACCUSATIVE κλέφτη κλέφτες ACCUSATIVE καθηγητή καθηγητές GENITIVE κλέφτη κλεφτών GENITIVE καθηγητή καθηγητών VOCATIVE κλέφτη κλέφτες VOCATIVE καθηγητή καθηγητές
Table 63: Examples of Modern Greek masculine nouns with nominative singular –ης with the nominative plural –ες
Quick stats:
• Babiniotis and Triantafillidis combined: 12 nouns with genitive plural gaps
• Triantafillidis:
o No nouns with genitive plural gaps (0% of all genitive plural gaps)
o 1,468 nouns in this inflection class (5.4% of all nouns)
318
Nouns with genitive plural gaps:
Word Gpl G Pl Word Gpl G Pl αντεροβγάλτης µπ ασβέστης µπ γνώστης µπ καντηλανάφτης µπ λεβέντης µπ µαγκίτης µπ πολυτεχνίτης µπ πορφυρίτης µπ σκυλοπνίχτης µπ τυφλοσούρτης µπ φαµπρικάντης µπ γαϊτανοφρύδης µπ
Table 64: Genitive plural gaps among masculine nouns with nominative singular –ης and plural –ες
B7. Neuter nouns with nominative singular –ο
Nouns in this inflection class present three genitive plural stress patterns. If the other cases have stress on the final or penultimate syllable, the genitive plural will also have stress on that same syllable, i.e. stress is columnar. However, if stress is on the antepenultimate in the nominative plural, the inflection class presents two stress patterns.
Either stress will be columnar, including antepenultimate stress in the genitive plural (το
δάχτυλο), or stress will shift to the penultimate syllable in the genitive singular and plural
(το πρόσωπο).
319
Example paradigms:
το πρόσωπο το δάχτυλο ‘face’ SINGULAR PLURAL ‘finger’ SINGULAR PLURAL NOMINATIVE πρόσωπο πρόσωπα NOMINATIVE δάχτυλο δάχτυλα ACCUSATIVE πρόσωπο πρόσωπα ACCUSATIVE δάχτυλο δάχτυλα GENITIVE προσώπου προσώπων GENITIVE δάχτυλου δάχτυλων VOCATIVE πρόσωπο πρόσωπα VOCATIVE δάχτυλο δάχτυλα
το βιβλίο το βουνό ‘book’ SINGULAR PLURAL ‘mountain’ SINGULAR PLURAL NOMINATIVE βιβλίο βιβλία NOMINATIVE βουνό βουνά ACCUSATIVE βιβλίο βιβλία ACCUSATIVE βουνό βουνά GENITIVE βιβλίου βιβλίων GENITIVE βουνού βουνών VOCATIVE βιβλίο βιβλία VOCATIVE βουνό βουνά
Table 65: Examples of Modern Greek neuter nouns with nominative singular –ο
Quick stats:
• Babiniotis and Triantafillidis combined: 9 nouns with genitive plural gaps
• Only Triantafillidis:
o No nouns with genitive plural gaps (0% of all genitive plural gaps)
o 3,340 nouns in this inflection class (12.2% of all nouns)
Nouns with genitive plural gaps:
Word Gpl G Pl Word Gpl G Pl γύναιο µπ ζούζουλο µπ κοτόπουλο µπ µάγουλο µπ µεσούρανα µπ νυχτοκάµατο µπ παλιόφαγο µπ πράτα µπ τσούρµο µπ
Table 66: Genitive plural gaps among neuter nouns with nominative singular –ο
320
B8. Masculine nouns with nominative singular –ας and nominative plural -αδες
Nouns in this inflection class have a simpler stress pattern than most of those discussed above. When the nominative singular has stress on the final syllable or penultimate syllable, stress is columnar throughout the paradigm, including in the genitive plural. In the few cases in which the nominative singular has stress on the antepenultimate syllable, stress shifts one syllable towards the end of the word in the entire plural, including the genitive plural.105 This last pattern is explained by the fact that stress in Modern Greek can be no more than three syllable from the end of the word. Since the plural forms have one more syllable than those in the singular, stress shifts one syllable.
Example paradigms:
ο παπάς ο τσέλιγκας ‘priest’ SINGULAR PLURAL ‘shepherd’ SINGULAR PLURAL NOMINATIVE παπάς παπάδες NOMINATIVE τσέλιγκας τσελίγκαδες ACCUSATIVE παπά παπάδες ACCUSATIVE τσέλιγκα τσελίγκαδες GENITIVE παπά παπάδων GENITIVE τσέλιγκα τσελίγκαδων VOCATIVE παπά παπάδες VOCATIVE τσέλιγκα τσελίγκαδες
Table 67: Examples of Modern Greek masculine nouns with nominative singular –ας and nominative plural -αδες
105 Triantafillidis (1998) marks only nine nouns of this inflection class with antepenultimate stress, five of which are compounds with –παπας. 321
Quick stats:
• Babiniotis and Triantafillidis combined: 2 nouns with genitive plural gaps
• Only Triantafillidis:
o No nouns with genitive plural gaps (0% of all genitive plural gaps)
o 307 nouns in this inflection class (1.1% of all nouns)
Nouns with genitive plural gaps:
Word Gpl G Pl βαρέλας µπ γκιουλέκας µπ
Table 68: Genitive plural gaps among imparisyllabic masculine nouns with nominative singular –ας and nominative plural -αδες
B9. Neuter nouns with nominative singular –µα
Holton et al. (1997:66) describe neuter nouns with nominative singular –µα in the following way: “These nouns may be of two syllables, like κύµα ‘wave’ (with paroxytone stress), or of three or more syllables, like πρόβληµα ‘problem’ (with proparoxytone stress). The endings of the genitive singular and all plural cases involve an additional syllable, which has implications for the position of the stress. The genitive singular and the nominative and accusative plural of these nouns are always stressed on the antepenultimate; the genitive plural always has the stress on the penultimate.”
322
Example paradigms:
το κύµα το πρόβληµα ‘wave’ SINGULAR PLURAL ‘problem’ SINGULAR PLURAL NOMINATIVE κύµα κύµατα NOMINATIVE πρόβληµα προβλήµατα ACCUSATIVE κύµα κύµατα ACCUSATIVE πρόβληµα προβλήµατα GENITIVE κύµατος κυµάτων GENITIVE προβλήµατος προβληµάτων VOCATIVE κύµα κύµατα VOCATIVE πρόβληµα προβλήµατα
Table 69: Examples of Modern Greek neuter nouns with nominative singular –µα
Quick stats:
• Babiniotis and Triantafillidis combined: 2 nouns with genitive plural gaps
• Only Triantafillidis:
o No nouns with genitive plural gaps in this inflection class (0% of all gaps)
o 2,380 nouns in this inflection class (8.7% of all nouns)
Nouns with genitive plural gaps:
Word Gpl G Pl µάλαµα µπ σιγανοψιχάλισµα µπ
Table 70: Genitive plural gaps among neuter nouns with nominative singular –µα
323
APPENDIX C
SURVEY INSTRUMENT: GREEK PERIPHRASIS
Αυτό το ερωτηµατολόγιο έχει τρία µέρη. Στο πρώτο, θα σας ρωτήσουµε να δώσετε λίγες πληροφορίες γύρω από τον εαυτό σας και τη γλωσσική σας γνώση. Στο δεύτερο, θα σας ρωτήσµουµε να συµπληρώσετε µερικές προτάσεις που έχουν κενά. Τέλος, στο τριτο µέρος, θα σας ρωτήσω να ξαναγράψετε µερικές προτάσεις για να ακουστούν πιο φυσικές στα ελληνικά.
Πρώτο Μέρος: Προσωπικές Πληροφορίες
Ο∆ΗΓΙΕΣ: Όλοι οι άνθρωποι δε µιλούν µε τον ίδιο τρόπο. ∆εδοµένου ότι η ηλικία, το φύλο και ο τόπος καταγωγής είναι ορισµένοι από τους πολλούς παράγοντες που επηρεάζουν τον τρόπο οµιλίας, θα θέλαµε να ξέρουµε λίγα πράγµατα σχετικά µε σας. Αυτές οι πληροφορίες θα µας βοηθήσουν να συγκρίνουµε τις απαντήσεις σας στο ερωτηµατολόγιο µε αυτές άλλων συµµετεχόντων στην έρευνα. Όλες οι απαντήσεις θα παραµείνουν απόρρητες.
Α1. Ποιο είναι το φύλο σας; άντρας γυναίκα
Α2. Ποια είναι η ηλικία σας; ______χρονών
Α3. Ποιος είναι ο τόπος γεννήσεώς σας (πόλη, χώρα); ______
Α4. Ποιος είναι ο τόπος γεννήσεως του πατέρα σας (πόλη, χώρα); ______
Α5. Ποιος είναι ο τόπος γεννήσεως της µητέρας σας (πόλη, χώρα); ______
Α6. Καταγράψτε όλα τα µέρη στα οποία έχετε διαµείνει για τουλάχιστον ένα χρόνο και πόσο χρονών είσασταν όταν µείνατε εκεί.
το µέρος; πόσο χρονών είσασταν; Παράδειγµα: η Θεσσαλονίκη, η Ελλάδα 0-18 χρονών ______
324
Α7. Ποιο είναι το επάγγελµά σας; ______Α8. Ποιο είναι/ήταν το επάγγελµα του πατέρα σας; ______Α9. Ποιο είναι/ήταν το επάγγελµα της µητέρας σας; ______
Α10. Ποιο είναι το υψηλότερο επίπεδο µόρφωσης που έχετε; δεν πήγα σχολείο ή δεν τελείωσα το δηµοτικό απόφοιτος δηµοτικού απόφοιτος γυµνασίου φοίτησα στην τριτοβάθµια εκπαίδευση αλλά δεν πήρα (ακόµη) πτυχίο πτυχιούχος Πανεπιστηµίου κάτοχος Μεταπτυχιακού τίτλου σπουδών (π.χ. κάτοχος Μάστερ, κάτοχος πτυχίου Ιατρικής, κάτοχος ∆ιδακτορικού)
Α11. Ποια/ες είναι η/οι µητρική/ές σας γλώσσα/ες; Ελληνικά άλλη γλώσσα: ______
Α12. Γνωρίζετε άλλες γλώσσες; ναι όχι
Α13. Ποιες άλλες γλώσσες γνωρίζετε, και σε τι επίπεδο (Παράδειγµα: Αγγλικά (µε ευχέρεια) Γερµανικά (µόνο για απλή συνεννόηση) Ρωσσικά (µόνο διάβασµα)); ______
______
Α14. Σε ποια/ες γλώσσα/ες µιλάτε συνήθως µε τους γονείς σας; Ελληνικά άλλη γλώσσα: ______
Α15. Σε ποια/ες γλώσσα/ες µιλάτε συνήθως µε τους φίλους σας; Ελληνικά άλλη γλώσσα: ______
Α16. Σε ποια/ες γλώσσα/ες µιλάτε συνήθως στο σχολείο/στη δουλειά; Ελληνικά άλλη γλώσσα: ______
325
∆εύτερο Μέρος: Συµπληρώστε τα Κενά
Ο∆ΗΓΙΕΣ: Στο τµήµα που ακολουθεί θα σας δοθούν 54 ανολοκλήρωτες προτάσεις στα Ελληνικά. Ολοκληρώστε κάθε πρόταση µε τον τρόπο που θεωρείτε πιο φυσικό. Να γράψετε, παρακαλώ, το γράµµα που αντιστοιχεί στην απάντησή σας στο κενό. Μπορείτε να διαλέξετε µόνο µία απάντηση.
παράδειγµα: Η κατασκευή __Α__ πήρε ένα χρόνο. Α. αυτών των δρόµων Β. από τους δρόµους Γ. αυτών των δροµών ∆. από τους δροµούς ______
B1. Νόµισε πως τα ναρκωτικά ήτανε η λύση _____. Α. στα προβληµάτα του Β. των προβλήµατών του Γ. στα προβλήµατά του ∆. των προβληµάτων του
B2. _____ υπάρχουνε πολλές τροφικές διαταραχές. Α. Μεταξύ των αναβατών Β. Ανάµεσα στους αναβάτες Γ. Ανάµεσα στους αναβατές ∆. Μεταξύ των αναβάτων
B3. Ένας µεγάλος αριθµός _____ έσπασε κατά τη διάρκεια του σεισµού. Α. από βαθρά Β. βαθρών Γ. βάθρων ∆. από βάθρα
B4. Το πλήθος _____ κατέβαλλε τον κλόουν. Α. από τα παιδιά Β. από τα παιδία Γ. των παιδιών ∆. των παιδίων
B5. Οι καρτ ποστάλ είναι _____. Α. πιο αργές από τα γράµµατα Β. βραδύτερες των γραµµάτων Γ. πιο αργές από τα γραµµάτα ∆. βραδύτερες των γράµµατων
326
B6. _____ πήρε όλη τη µέρα. A. Η επιλογή των ελεκτόρων Β. Η επιλογή των ελέκτορων Γ. Το να επιλέξουµε τους ελεκτόρες ∆. Το να επιλέξουµε τους ελέκτορες
B7. Τα περισσότερα _____ έσπασαν κατά τη διάρκεια του σεισµού. Α. από τα βάθρα Β. των βαθρών Γ. των βάθρων ∆. από τα βαθρά
B8. Οι αίθουσες του σχολείου είναι µικρότερες _____ του πανεπιστηµίου. Α. από τις αίθουσες Β. των αιθουσών Γ. από τις αιθουσές ∆. των αίθουσων
B9. Στο όνειρό της, µια οµάδα _____ χόρευε γύρω της φωνάζοντας κατάρες. Α. δαίµονων Β. από δαιµόνες Γ. δαιµόνων ∆. από δαίµονες
B10. Ακόµα και _____ τα πορτοφόλια κλαπήκανε! A. των πελατών B. από τους πελάτες Γ. από τους πελατές ∆. των πελάτων
B11. Μεγάλος αριθµός _____ χαλάσανε αµέσως. Α. των µηχάνων Β. από τις µηχάνες Γ. των µηχανών ∆. από τις µηχανές
Β12. Αυτό το συγκρότηµα _____ οργανώνει ένα µουσικό φεστιβάλ κάθε χρόνο. Α. αιθούσων Β. από αιθούσες Γ. αιθουσών ∆. από αίθουσες
327
B13. Η λάθος τοποθέτηση _____ έκανε τη σκεπή να καταρρεύσει. Α. από τους δοκούς Β. από τους δόκους Γ. µερικών δόκων ∆. µερικών δοκών
B14. Τα στούντιο για τις πρόβες στο σχολείο είναι _____. Α. οκειότερα των συναυλιακών αίθουσων Β. πιο οκεία από τις συναυλιακές αίθουσες Γ. οκειότερα των συναυλιακών αιθουσών ∆. πιο οκεία από τις συναυλιακές αιθουσές
B15. Νοµίζω ότι τα φουντούκια είναι πολύ καλύτερα _____. Α. από τα καστάνα Β. των κάστανων Γ. από τα κάστανα ∆. των καστάνων
B16. Είµαι µεγαλύτερος _____ µου. Α. των αδερφών Β. από τους αδέρφους Γ. από τους αδερφούς ∆. των αδέρφων
B17. Τα µελάνια _____ είναι φτηνά. Α. από τις εφηµερίδες Β. των εφηµεριδών Γ. των εφηµερίδων ∆. από τις εφηµεριδές
B18. Κατά τη γνώµη µου µπορούµε να µάθουµε πολλά περισσότερα για την ιστορία της Ευρώπης µέσω των φρουρίων των Βαλκανίων παρά µέσω _____. Α. των ανακτόρων της ∆υτικής Ευρώπης Β. από τα ανακτόρα της ∆υτικής Ευρώπης Γ. των ανάκτορων της ∆υτικής Ευρώπης ∆. από τα ανάκτορα της ∆υτικής Ευρώπης
B19. ∆ιακοσµηµένες θήκες _____ εκτίθενται στο µουσείο. A. ξιφών Β. ξίφων Γ. από ξίφη ∆. από ξιφή
328
B20. Ένας µικρός αριθµός _____ υγείας δεν θεραπεύεται. Α. από τις διαταραχές Β. διαταραχών Γ. διαταράχων ∆. από τις διαταράχες
B21. Απολογήθηκε για τη ζηµιά _____. Α. στα αυτοκινήτα Β. των αυτοκινήτων Γ. στα αυτοκίνητα ∆. των αυτοκίνητων B22. Το µάρµαρο _____ ράγισε. Α. από τα βάθρα Β. των βάθρων Γ. των βαθρών ∆. από τα βαθρά
B23. Οι ώρες _____ θα ανακοινωθούν σήµερα. A. µαθηµάτων Β. από τα µαθηµάτα Γ. από τα µαθήµατα ∆. µαθήµατων
B24. Μερικές _____ χαλάσανε αµέσως. Α. από τις µηχανές Β. από τις µηχάνες Γ. των µηχάνων ∆. των µηχανών
B25. Μερικοί _____ δουλεύουνε στη βιβλιοθήκη. Α. από τους πανεπιστηµιακούς φοιτητές Β. από τους πανεπιστηµιακούς φοιτήτες Γ. των πανεπιστηµιακών φοιτήτων ∆. των πανεπιστηµιακών φοιτητών
B26. Οι περισσότεροι _____ στην πόλη ήταν ιεραπόστολοι. A. των διαµαρτυροµένων Β. των διαµαρτυρόµενων Γ. από τους διαµαρτυρόµενους ∆. από τους διαµαρτυροµένους
329
B27. Μεγάλη µερίδα _____ ψήφισε το νόµο. A. των βουλευτών Β. των βουλεύτων Γ. από τους βουλευτές ∆. από τους βουλεύτες
B28. Τα ποντίκια φάγανε _____. Α. τα µισά κάστανα Β. το ήµισυ των καστάνων Γ. τα µισά καστάνα ∆. το ήµισυ των κάστανων
B29. Τα υπόστεγα _____ καταρεύσανε µετά το σεισµό. A. των αεροπλάνων Β. για τα αεροπλανά Γ. των αεροπλανών ∆. για τα αεροπλάνα
B30. Η επιτροπή _____ ψήφισε κατά του νοµοσχεδίου. A. των εννέα γερουσιαστών Β. των εννέα γερουσιάστων Γ. από εννέα γερουσιάστες ∆. από εννέα γερουσιαστές
B31. Αυτό το συγκρότηµα _____ λέγεται ότι είναι στοιχειωµένο. A. νησιών Β. από νησιά Γ. από νησία ∆. νησίων
B32. Ένα µεγάλο αριθµός _____ βρεθήκανε στη νότια Ελλάδα. Α. από τα ξίφη Β. των ξίφων Γ. των ξιφών ∆. από τα ξιφή
B33. Άκουσα ότι ο πρωθυπουργός θα δώσει ένα λόγο _____από τους εργάτες σιδηροδρόµων. A. εναντίον των πρόσφατων απεργιών B. ενάντια στις πρόσφατες απεργίες Γ. εναντίον των πρόσφατων απεργίων ∆. ενάντια στις πρόσφατες απεργιές
330
B34. Αγωνίστηκε γενναία _____ του. Α. εναντίον των εχθρών Β. εναντία στους εχθρούς Γ. εναντίον των έχθρων ∆. εναντία στοθς έχθρους
Β35. Η στίβα _____ έπεσε στο κεφάλι του. Α. από βιβλιά Β. των βιβλιών Γ. από βιβλία ∆. των βιβλίων
B36. Μια τοπική οµάδα εθελοντών περιποιείται τους κήπους _____. Α. και των δύο των ανάκτορων Β. και των δύο των ανακτόρων Γ. και από τα δύο ανάκτορα ∆. και από τα δύο ανακτόρα
B37. Η λιγότερο δηµοφιλής έκθεση στο µουσείο είναι η αίθουσα _____. Α. ορύκτων B. ορυκτών Γ. µε τα ορύκτα ∆. µε τα ορυκτά
B38. Όλα τα κατάρτια _____ ταλαντεύονταν από τον άνεµο. A. των πλοίων Β. απ’τα πλοία Γ. των πλοιών ∆. απ’τα πλοιά
B39. Ο δικαστής απένειµε ακυρώσεις _____. A. και στους δύο γάµους Β. και στους δύο γαµούς Γ. και των δύο γάµων ∆. και των δύο γαµών
B40. _____ του Κώστα δεν υπάρχει ούτε ένας εµφανίσηµος άντρας! Α. Ανάµεσα στους απόγονους Β. Μεταξύ των απόγονων Γ. Ανάµεσα στους απογόνους ∆. Μεταξύ των απογόνων
331
B41. Οι δηµοσιογράφοι είναι πιο σεβαστοί _____. Α. από τους αρθρογράφους Β. των αρθρογράφων Γ. από τους αρθρογραφούς ∆. των αρθρογραφών
B42. Το χειµώνα, άνθρωποι στήνουνε φουφούδες _____ έξω απ’τους κινηµατογράφους. Α. για κάστανα Β. καστάνων Γ. κάστανων ∆. για καστάνα
B43. Τα αρχικά ανταλλακτικά ήτανε καλύτερα _____. A. από τις υποκαταστάσεις τους Β. από τις υποκαταστασείς τους Γ. των υποκαταστάσεων τους ∆. των υποκαταστάσεών τους
B44. Μερικοί _____ παραιτήθηκαν όταν δεν τους πλήρωσε το θέατρο. Α. των σκηνογραφών Β. των σκηνογράφων Γ. από τους σκηνογράφους ∆. από τους σκηνογραφούς
B45. Ακονίσανε τις κόψεις _____ τους πριν από τη µάχη. A. από τα ξίφη Β. των ξίφων Γ. των ξιφών ∆. από τα ξιφή
B46. Όλος ο κόσµος διαµαρτύρεται _____. Α. εναντίον των βρετανικών και αµερικανικών κυβέρνησεων Β. εναντίον των βρετανικών και αµερικανικών κυβερνήσεων Γ. εναντία στις βρετανικές και αµερικανικές κυβέρνησεις ∆. εναντία στις βρετανικές και αµερικανικές κυβερνήσεις
B47. Η ευτυχία των εργαζοµένων είναι πιο σηµαντική _____. Α. από τα κέρδη Β. των κέρδων Γ. των κερδών ∆. από τα κερδή
332
Β48. Η ένωση _____ οργάνωσε το φεστιβάλ. Α. πλοιοκτήτων Β. από τους πλοιοκτητές Γ. από τους πλοιοκτήτες ∆. πλοιοκτητών
B49. Οι χούλιγκαν σπάσανε τα παρµπρίζ _____ όταν ο Παναθηναϊκός έχασε το µατς. Α. από όλα τα αυτοκίνητα Β. όλων των αυτοκινήτων Γ. όλων των αυτοκίνητων ∆. από όλα τα αυτοκινλήτα
B50. Είναι δύσκολο να γίνει µια καλή µετάφραση _____ του Οµήρου. Α. των έπων Β. στα επή Γ. των επών ∆. στα έπη
B51. ∆ηµιουργήθηκε µία παρεξήγηση _____ που κρατούν τους οµήρους. A. µεταξύ των διαπραγµετευτών και αυτών Β. ανάµεσα στους διαπραγµετεύτες και σ’αυτούς Γ. ανάµεσα στους διαπραγµετευτές και σ’αυτούς ∆. µεταξύ των διαπραγµετεύτων και αυτών
Β52. ∆ηµιουργήθηκε µία παρεξήγηση _____. Α. ανάµεσα στους πελατές και το µαγαζάτορα Β. µεταξύ των πελάτων και του µαγαζάτορα Γ. ανάµεσα στους πελάτες και το µαγαζάτορα ∆. µεταξύ των πελατών και του µαγαζάτορα
B53. Η πλειοψηφία _____ κάνουνε κριτική στην κυβέρνηση. Α. από τις ανεξάρτητες εφηµεριδές Β. από τις ανεξάρτητες εφηµερίδες Γ. των ανεξάρτητων εφηµεριδών ∆. των ανεξάρτητων εφηµερίδων
B54. 100 ψήφοι _____ ήταν αλλοιωµένες. Α. στις εκλόγες Β. των εκλογών Γ. των εκλόγων ∆. στις εκλογές
333
Τρίτο Μέρος: Τι είναι καλύτερο;
Ο∆ΗΓΙΕΣ: Οταν δουλεύατε στο προηγούµενο µέρος, µήπως νοµίζατε ότι µερικάς προτάσεις δεν σας έδωσαν του τύπο που προτιµούσατε ή δεν ακούστηκαν πολύ «φυσικές» (όπως θα τις έλεγε ένας Ελληνόφωνος).
Να ρίξετε µια µατιά σε καθεµία πρόταση στο εκείνο µέρος, και να βαλετε ένα «Χ» στο κένο τετράγωνο ( ), αντιστοιχώς µε την αίσθησή σας που η απάντησή σας είναι ό πως θα εκθράσετε ‘σεις την έννοια της προτάσεως («Ναι» σηµαίνει ότι η πρόταση είναι «φυσική» και δεν πρέπει να αλλάξετε τίποτε. «Όχι» σηµαίναι ότι δεν είναι φυσική). Αν η απάντησή σας είναι «όχι», παρακαλώ να ξανά γράψετε την ολόκληρη πρόταση στη γραµµή κάτω αν υπάρχει άλλος τρόπος να εκφραστεί αυτή η έννοια. Σας παρακαλώ να ΜΗ σβύσετε τις απαντήσεις σας στο προηγούµενο µέρος.
παράδειγµα 1: Η κατασκευή __Α__ πήρε ένα χρόνο. Α. αυτών των δρόµων Β. από τους δρόµους Γ. αυτών των δροµών ∆. από τους δροµούς i. Ναι Όχι
Αν όχι, να ξανά γραψετε την πρόταση εδώ: ______
παράδειγµα 2: Η απώλεση __Α__ έγινε πριν τη σηµερινή εποχή. Α. µε τα δικαιωµάτα Β. ... Γ. ... ∆. ... ii. Ναι Όχι
Αν όχι, να διορθώσετε την πρόταση εδώ: Η απώλεση αυτών των δικαιωµάτων έγινε πριν τη σηµερινή εποχή.
______
Γ1. Ναι Όχι Αν όχι, να διορθώσετε την πρόταση εδώ: ______
(This is repeated for each of the 54 examples) Μπράβο σας! Τελειώσατε! Πολύ σας ευχαριστούµε.
334
APPENDIX D
SURVEY INSTRUMENT: GREEK GENITIVE PLURAL GAPS
Αυτό το ερωτηµατολόγιο έχει τρία µέρη. Στο πρώτο, θα σας ρωτήσω να δώσετε λίγες πληροφορίες γύρω από τον εαυτό σας και τη γλωσσική σας γνώση. Στο δεύτερο, θα σας ρωτήσω να υπολογίσετε µερικές λέξεις ελληνικές κατά την σικειότητα που έχετε για την καθεµιά. Τέλος, στο τριτο µέρος, θα σας ρωτησώ να γράψετε λέξεις σε διαφορετικού τύπου και µετά να υπολογίσετε αυτές τις λέξεις ανάλογα µε το αν σας φαίνονται γνήσιες λέξεις από τα νέα ελληνικά.
Πρώτο Μέρος: Προσωπικές Πληροφορίες
Ο∆ΗΓΙΕΣ: Όλοι οι άνθρωποι δε µιλούν µε τον ίδιο τρόπο. ∆εδοµένου ότι η ηλικία, το φύλο και ο τόπος καταγωγής είναι ορισµένοι από τους πολλούς παράγοντες που επηρεάζουν τον τρόπο οµιλίας, θα θέλαµε να ξέρουµε λίγα πράγµατα σχετικά µε σας. Αυτές οι πληροφορίες θα µας βοηθήσουν να συγκρίνουµε τις απαντήσεις σας στο ερωτηµατολόγιο µε αυτές άλλων συµµετεχόντων στην έρευνα. Όλες οι απαντήσεις θα παραµείνουν απόρρητες.
Α1. Ποιο είναι το φύλο σας; άντρας γυναίκα Α2. Ποια είναι η ηλικία σας; ______χρονών Α3. Ποιος είναι ο τόπος γεννήσεώς σας (πόλη, χώρα); ______Α4. Ποιος είναι ο τόπος γεννήσεως του πατέρα σας (πόλη, χώρα); ______Α5. Ποιος είναι ο τόπος γεννήσεως της µητέρας σας (πόλη, χώρα); ______Α6. Καταγράψτε όλα τα µέρη στα οποία έχετε διαµείνει για τουλάχιστον ένα χρόνο και πόσο χρονών είσασταν όταν µείνατε εκεί.
το µέρος; πόσο χρονών είσασταν; Παράδειγµα: η Θεσσαλονίκη, η Ελλάδα 0-18 χρονών ______
Α7. Ποιο είναι το επάγγελµά σας; ______Α8. Ποιο είναι/ήταν το επάγγελµα του πατέρα σας; ______Α9. Ποιο είναι/ήταν το επάγγελµα της µητέρας σας; ______
335
Α10. Ποιο είναι το υψηλότερο επίπεδο µόρφωσης που έχετε; δεν πήγα σχολείο ή δεν τελείωσα το δηµοτικό απόφοιτος δηµοτικού απόφοιτος γυµνασίου φοίτησα στην τριτοβάθµια εκπαίδευση αλλά δεν πήρα (ακόµη) πτυχίο πτυχιούχος Πανεπιστηµίου κάτοχος Μεταπτυχιακού τίτλου σπουδών (π.χ. κάτοχος Μάστερ, κάτοχος πτυχίου Ιατρικής, κάτοχος ∆ιδακτορικού)
Α11. Ποια/ες είναι η/οι µητρική/ές σας γλώσσα/ες; Ελληνικά άλλη γλώσσα: ______
Α12. Γνωρίζετε άλλες γλώσσες; ναι όχι
Α13. Ποιες άλλες γλώσσες γνωρίζετε, και σε τι επίπεδο (Παράδειγµα: Αγγλικά (µε ευχέρεια) Γερµανικά (µόνο για απλή συνεννόηση) Ρωσσικά (µόνο διάβασµα)); ______
Α14. Σε ποια/ες γλώσσα/ες µιλάτε συνήθως µε τους γονείς σας; Ελληνικά άλλη γλώσσα: ______
Α15. Σε ποια/ες γλώσσα/ες µιλάτε συνήθως µε τους φίλους σας; Ελληνικά άλλη γλώσσα: ______
Α16. Σε ποια/ες γλώσσα/ες µιλάτε συνήθως στο σχολείο/στη δουλειά; Ελληνικά άλλη γλώσσα: ______
336
∆εύτερο Μέρος: Η Οικειότητα
Ο∆ΗΓΙΕΣ: Παρακαλούµε να κρινετε κάθε λέξη κατά αυτή την κλίµακα. Να βάλετε στον κατάλληλον αριθµό.
1 ∆εν ξέρω αυτή τη λέξη 2 Λίγο ξέρω την έννοια ή µπορώ να µαντέψω την έννοιά της 3 Ξέρω την έννοια αυτής της λέξης αλλά δεν την χρησιµοποιώ 4 Έχω χρησιµοποιήσει αυτή τη λέξη µια ή δύο φορές 5 Χρησιµοποιώ αυτή τη λέξη πότε-πότε 6 Χρησιµοποιώ αυτή τη λέξη πολύ
δεν την δεν την την η λέξη χρησιµοποιώ ξέρω χρησιµοποιώ πότε-πότε η συστάδα 1 2 3 4 5 6 ο απόστολος 1 2 3 4 5 6 η γιαγιά 1 2 3 4 5 6 ο όφις 1 2 3 4 5 6 η λακκούβα 1 2 3 4 5 6
η µαµά 1 2 3 4 5 6 η σκλήθρα 1 2 3 4 5 6 η παγίδα 1 2 3 4 5 6 ο υπέρµαχος 1 2 3 4 5 6 ο λαθρέµπορος 1 2 3 4 5 6
ο κροκόδειλος 1 2 3 4 5 6 η καρότσα 1 2 3 4 5 6 η πιτζάµα 1 2 3 4 5 6 η µερίδα 1 2 3 4 5 6 η κροτίδα 1 2 3 4 5 6
το βιβάρι 1 2 3 4 5 6 ο επίσκοπος 1 2 3 4 5 6 η ροτόντα 1 2 3 4 5 6 ο υπήκοος 1 2 3 4 5 6
337
δεν την δεν την την η λέξη χρησιµοποιώ ξέρω χρησιµοποιώ πότε-πότε η αψίδα 1 2 3 4 5 6
ο πάσσαλος 1 2 3 4 5 6 ο τσάµικος 1 2 3 4 5 6 η κοπέλα 1 2 3 4 5 6 ο κλίβανος 1 2 3 4 5 6 ο διθύραµβος 1 2 3 4 5 6
το πάρτι 1 2 3 4 5 6 ο παπουτσής 1 2 3 4 5 6 η αρβύλα 1 2 3 4 5 6 η χαράδρα 1 2 3 4 5 6 η πραµάτεια 1 2 3 4 5 6
ο τραγέλαφος 1 2 3 4 5 6 ο µουσαµάς 1 2 3 4 5 6 η άσφαλτος 1 2 3 4 5 6 η τροχιά 1 2 3 4 5 6 η δίκη 1 2 3 4 5 6
η λαµπάδα 1 2 3 4 5 6 ο υπόνοµος 1 2 3 4 5 6 ο καναπές 1 2 3 4 5 6 o ρεπόρτερ 1 2 3 4 5 6 το πουλόβερ 1 2 3 4 5 6
η νουβέλα 1 2 3 4 5 6 η κοπάνα 1 2 3 4 5 6 η κολόνια 1 2 3 4 5 6 το δόρυ 1 2 3 4 5 6 η φυσούνα 1 2 3 4 5 6
το παιδί 1 2 3 4 5 6 ο δήµιος 1 2 3 4 5 6
338
δεν την δεν την την η λέξη χρησιµοποιώ ξέρω χρησιµοποιώ πότε-πότε η θυρίδα 1 2 3 4 5 6 η αρένα 1 2 3 4 5 6 ο βαγενάς 1 2 3 4 5 6
το ταψί 1 2 3 4 5 6 το εκκρεµές 1 2 3 4 5 6 η καµπάνια 1 2 3 4 5 6 το ήπαρ 1 2 3 4 5 6 ο ταξιτζής 1 2 3 4 5 6
η σαµπάνια 1 2 3 4 5 6 ο διάκοσµος 1 2 3 4 5 6 η πλαζ 1 2 3 4 5 6 η µουρµούρα 1 2 3 4 5 6 η κοµπίνα 1 2 3 4 5 6
ο κορβανάς 1 2 3 4 5 6 η κουλτούρα 1 2 3 4 5 6 το συµβάν 1 2 3 4 5 6 το µατς 1 2 3 4 5 6 ο κοµµωτής 1 2 3 4 5 6
ο µόδιστρος 1 2 3 4 5 6 η καρέκλα 1 2 3 4 5 6 η ταβέρνα 1 2 3 4 5 6 η σακούλα 1 2 3 4 5 6 ο αµανές 1 2 3 4 5 6
ο χορευτής 1 2 3 4 5 6 ο τύραννος 1 2 3 4 5 6 ο σπόνδυλος 1 2 3 4 5 6 η αρκούδα 1 2 3 4 5 6 το µπουρί 1 2 3 4 5 6
339
δεν την δεν την την η λέξη χρησιµοποιώ ξέρω χρησιµοποιώ πότε-πότε ο βαρύτονος 1 2 3 4 5 6 ο καφετζής 1 2 3 4 5 6 ο πάταγος 1 2 3 4 5 6 ο φοιτητής 1 2 3 4 5 6 η βελόνα 1 2 3 4 5 6
Τρίτο Μέρος: Σ’αυτό το µέρος, πρέπει να κάνετε δύο πράγµατα.
Πρώτο, για καθεµιά λέξη στις παρένθεσεις, να γράψετε τη λέξη στον τύπο που χρειάζεται για τη φράση εκείνη.
παράδειγµα: στα ____πανεπιστήµια_____ (το πανεπιστήµιο)
Μετά που γράψατε τη λέξη, παρακαλώ να γράψετε και εναν αριθµό που δείχνει το βαθµό εµπιστοσύνης σας ότι εκείνος ο τύπος που γράψατε είναι γνήσια ελληνική λέξη. Παραδείγµατος χάριν, αν έχετε µεγάλη εµπιστοσύνη, να γράψετε µεγάλο αριθµό (π.χ. 100) αλλά αν δεν έχετε µεγάλη εµπιστοσύνη, να γράψετε µικρό αριθµό (π.χ. 20).
Η κρίση που δίνετε στην αρχή για την πρώτη λέξη θα χρησιµεύσει σαν βάση για άλλες κρίσεις. ∆ηλαδή, αν η εµπιστοσύνη σας είναι δυο φορές πιο δυνατή µε τη δεύτερη λέξη απ’ ό,τι είναι µε την πρώτη ο αριθµός πρέπει να είναι δυπλάσιος. Αν η εµπιστοσύνη στη δεύτερη λέξη είναι µισή, ο αριθµός πρέπει να είναι µισός. ∆εν πειράζει αν οι κρίσεις φαίνονται να αλλάξουν καθώς δουλεύετε στο ερωτηµατολόγιο – ενδιαφέροµαι για τι κρίσεις κάθως αναπτύσσονται.
1 Ο λόγος του ______... Κρίση: ______(το συµβάν)
2 Το ______είναι εκεί. Κρίση: ______(το εκκρεµές)
3 Στον ______... Κρίση: ______(ο υπόνοµος)
4 Ο ______πρέπει να θερµανθεί γρήγορα. Κρίση: ______(ο κλίβανος)
340
5 Mεταξύ του ______και του αρουραίου... Κρίση: ______(ο όφις)
6 Οι ______βρίσκονται εκεί. Κρίση: ______(η πλαζ)
7 Mεταξύ του ______και του υπόλοιπου Κρίση: ______του καπνοδόχου... (το µπουρί)
8 Οι οργανωτές των ______είναι εκεί. Κρίση: ______(η κοµπίνα)
9 Βλέπω τον ______. Κρίση: ______(ο δήµιος)
10 Τραγουδώ για τον ______. Κρίση: ______(ο χορευτής)
11 Οι µιναρέδες διαφόρων ______... Κρίση: ______(η ροτόντα)
12 Mεταξύ του ______και του διχτυού... Κρίση: ______(το δόρυ)
13 Οι ______είναι οι ίδιες µε την αµφίεση Κρίση: ______του Γερµανού. (η αρβύλα)
14 Το ______βρίσκεται εκεί. Κρίση: ______(το πάρτι)
15 Τα κορδόνια των ______... Κρίση: ______(η πιτζάµα)
16 Το αγόρι είναι µεταξύ των ______... Κρίση: ______(η καρέκλα)
17 Βλέπω τους ______. Κρίση: ______(ο τσάµικος)
18 Το ρεφραίν του ______είναι αργό. Κρίση: ______(ο αµανές)
19 Ο ______βρίσκεται εκεί. Κρίση: ______(ο κροκόδειλος)
20 Το ______είναι απαλό. Κρίση: ______(το πουλόβερ)
21 Μεταξύ της ______και του παππού Κρίση: ______του... (η γιαγιά)
22 Οι ______βρίσκονται εκεί. Κρίση: ______(o ρεπόρτερ)
341
23 Οι ______θα έχουν µπλε χρώµα. Κρίση: ______(η σακούλα)
24 Η χρήση των ______... Κρίση: ______(η βελόνα)
25 Ο ______βρίσκεται εκεί. Κρίση: ______(ο όφις)
26 Τα ______αρχίζουν. Κρίση: ______(το µατς)
27 Η δηµιουργία διαφόρων ______Κρίση: ______χωριών... (η συστάδα)
28 Τα ______είναι γαλάζιο. Κρίση: ______(το πουλόβερ)
29 Βλέπω τους ______. Κρίση: ______(ο υπέρµαχος)
30 Η ______βρίσκεται εκεί. Κρίση: ______(η χαράδρα)
31 Βλέπω τον ______. Κρίση: ______(ο υπήκοος)
32 Η ______ήταν οδυνηρή. Κρίση: ______(η παγίδα)
33 Οι ______αρχίζουν. Κρίση: ______(η µουρµούρα)
34 Βλέπω τους ______. Κρίση: ______(ο υπήκοος)
35 Ο ______αρχίζει. Κρίση: ______(ο τσάµικος)
36 Ο ______στέκεται εκεί. Κρίση: ______(ο απόστολος)
37 Τα ρεφραίν των ______είναι αργά. Κρίση: ______(ο αµανές)
38 Οι ______ξεπορτίζουν εύκολα. Κρίση: ______(η κοπέλα)
39 Βλέπω τον ______. Κρίση: ______(ο επίσκοπος)
40 Η ______είναι εκεί. Κρίση: ______(η βελόνα)
342
41 Η πώληση των καλών ______... Κρίση: ______(η πραµάτεια)
42 Πρόκειται για τους ______. Κρίση: ______(ο λαθρέµπορος)
43 Ο ______είναι δυνατός. Κρίση: ______(ο πάταγος)
44 Οι ______βρίσκονται εκεί. Κρίση: ______(η λαµπάδα)
45 Βλέπω τον ______. Κρίση: ______(ο τραγέλαφος)
46 Με τον ______.... Κρίση: ______(ο µόδιστρος)
47 Η Ευρώπη βρίσκεται στην ______Κρίση: ______µεγάλων αποφάσεων. (η τροχιά)
48 Μία µετακίνηση των ______... Κρίση: ______(η φυσούνα)
49 Το ______στέκεται εκεί. Κρίση: ______(το παιδί)
50 Η παρότρυνση των άλλων ______... Κρίση: ______(ο καφετζής)
51 Η ______υπάρχει. Κρίση: ______(η κοµπίνα)
52 Οι ______του συγγραφέα διαβάζονταν Κρίση: ______από λίγους... (η νουβέλα)
53 Το ______είναι εδώ. Κρίση: ______(το ήπαρ)
54 Παίρνω γράµµατα από την ______µου. Κρίση: ______(η µαµά)
55 Το ξίφος είναι µεταξύ των ______... Κρίση: ______(η αρβύλα)
56 Βλέπω τον ______. Κρίση: ______(ο καφετζής)
57 Οι ______στέκονται εκεί. Κρίση: ______(η αρκούδα)
343
58 Η αρκούδα κοιµάται µεταξύ των Κρίση: ______... (η παγίδα)
59 Τα κλίµατα των ______στα Κρίση: ______πανεπιστήµια... (η µουρµούρα)
60 Ο αποκλεισµός µεγάλων ______της Κρίση: ______κοινωνίας... (η µερίδα)
61 Βλέπω τους ______. Κρίση: ______(ο διάκοσµος)
62 Ακούω τον ______. Κρίση: ______(ο βαρύτονος)
63 Ο ______στέκεται εκεί. Κρίση: ______(o ρεπόρτερ)
64 Εκτός από τους ______υπάρχουν και οι Κρίση: ______ιπποπόταµοι. (ο κροκόδειλος)
65 Στο ______. Κρίση: ______(το ταψί)
66 Η ______και τα άλλα οικοδοµήµατα... Κρίση: ______(η ροτόντα)
67 Μεταξύ του ______και του Κρίση: ______µηχανισµού... (το εκκρεµές)
68 Οι ______διαφέρουν. Κρίση: ______(η κουλτούρα)
69 Μεταξύ των ______... Κρίση: ______(ο παπουτσής)
70 Ακούω τους ______. Κρίση: ______(ο διθύραµβος)
71 Η γυναίκα κεντά την επιδερµίδα των ______... Κρίση: ______(ο µουσαµάς)
72 Βλέπω τους ______. Κρίση: ______(ο δήµιος)
73 Ο ______είναι εκεί. Κρίση: ______(ο φοιτητής)
344
74 Μεταξύ των ______στην Ευρώπης... Κρίση: ______(η κουλτούρα)
75 Στους ______... Κρίση: ______(ο υπόνοµος)
76 Οι ______στη χώρα µας είναι αταξικές! Κρίση: ______(η κοπάνα)
77 Μεταξύ των ______... Κρίση: ______(ο βαγενάς)
78 Στην ______... Κρίση: ______(η άσφαλτος)
79 Οι ______βρίσκονται εκεί. Κρίση: ______(η φυσούνα)
80 Μεταξύ του ______και της χορεύτριας... Κρίση: ______(ο χορευτής)
81 Οι φωνές των ______µας είναι δυνατές. Κρίση: ______(η µαµά)
82 Το χυδαιότητα των άλλων ______είναι Κρίση: ______αηδιαστικό. (η καµπάνια)
83 Βλέπω το ______. Κρίση: ______(το µπουρί)
84 Μεταξύ της ______και του καρδιού... Κρίση: ______(η σκλήθρα)
85 Η ρύθµιση των κινητών ______... Κρίση: ______(η θυρίδα)
86 Οι ______βρίσκονται εκεί. Κρίση: ______(η σαµπάνια)
87 Βλέπω τους ______. Κρίση: ______(ο σπόνδυλος)
88 Το ______άρχισε. Κρίση: ______(το συµβάν)
89 Τα βλέµµατα των ______µας... Κρίση: ______(η γιαγιά)
90 Βλέπω τους ______. Κρίση: ______(ο απόστολος)
345
91 Η ______είναι εκεί. Κρίση: ______(η θυρίδα)
92 Μεταξύ του ______και του πελάτη... Κρίση: ______(ο ταξιτζής)
93 Οι προοπτικές των ______... Κρίση: ______(η κοπάνα)
94 Η ______δέντρων είναι εκεί. Κρίση: ______(η συστάδα)
95 Το ______κρίθηκε πολύ νωρίς από το Κρίση: ______πρώτο µέρος. (το µατς)
96 Η ______βρίσκεται εκεί. Κρίση: ______(η πλαζ)
97 Κατά τους ______της Αµερικής... Κρίση: ______(ο επίσκοπος)
98 Μεταξύ των ______αυτού του Κρίση: ______συγγραφέα... (η νουβέλα)
99 Για τους ______... Κρίση: ______(ο τύραννος)
100 Η ______φαίνεται µεγάλη. Κρίση: ______(η µερίδα)
101 Τα µεγέθη των ______... Κρίση: ______(η καρότσα)
102 Τα δόντια των ______... Κρίση: ______(η αρκούδα)
103 Βλέπω τον ______. Κρίση: ______(ο κοµµωτής)
104 Μεταξύ του ______και των δέντρων... Κρίση: ______(το βιβάρι)
105 Οι ______βρίσκονται εκεί. Κρίση: ______(η κροτίδα)
106 Οι ______µυρίζουν... Κρίση: ______(η κολόνια)
107 Η διαχείριση των δηµοσίων ______... Κρίση: ______(ο κορβανάς)
346
108 Οι µυρουδιά των ______... Κρίση: ______(η κολόνια)
109 Ο γιατρός βλέπει το ______. Κρίση: ______(το παιδί)
110 Οι ______φαίνονται µεγάλες. Κρίση: ______(η λακκούβα)
111 Βλέπω την ______. Κρίση: ______(η σκλήθρα)
112 Ο ______υπάρχει. Κρίση: ______(ο λαθρέµπορος)
113 Ποτήρια των διαφόρων ______... Κρίση: ______(η σαµπάνια)
114 Ο ______είναι εδώ. Κρίση: ______(ο σπόνδυλος)
115 Το σπίτι βρίσκεται µεταξύ των ______. Κρίση: ______(η χαράδρα)
116 Οι λεπτοµέρειες για τη ______. Κρίση: ______(η δίκη)
117 Βλέπω τον ______. Κρίση: ______(ο φοιτητής)
118 Το ______υπάρχει. Κρίση: ______(το δόρυ)
119 Μεταξύ του ______και του τραπεζιού... Κρίση: ______(το ταψί)
120 Ένας από τους δύο ______... Κρίση: ______(ο µόδιστρος)
121 Η ______βρίσκεται εκεί. Κρίση: ______(η καρέκλα)
122 Η διαφηµιστική ______για το Κρίση: ______παπούτσια... (η καµπάνια)
123 Τα ______αρχίζουν. Κρίση: ______(το πάρτι)
124 Τα µεγέθη των ______... Κρίση: ______(η λακκούβα)
347
125 Τα ψάρια στο ______... Κρίση: ______(το βιβάρι)
126 Η ______είναι εκεί. Κρίση: ______(η πραµάτεια)
127 Βλέπω τον ______. Κρίση: ______(ο παπουτσής)
128 Σε αντίθεση µε τους κλασσικούς ______... Κρίση: ______(ο κλίβανος)
129 Στον ______... Κρίση: ______(ο κορβανάς)
130 Μεταξύ των ______... Κρίση: ______(η κροτίδα)
131 Η ______φαίνεται µεγάλη. Κρίση: ______(η πιτζάµα)
132 Μεταξύ των ______... Κρίση: ______(η ταβέρνα)
133 Μεταξύ του ______και του τραπεζιού... Κρίση: ______(ο µουσαµάς)
134 Η ______βρίσκεται εκεί. Κρίση: ______(η καρότσα)
135 Το τραπεζάκι είναι µεταξύ των ______... Κρίση: ______(ο καναπές)
136 Ο ______είναι εκεί. Κρίση: ______(ο κοµµωτής)
137 Η ______είναι µεγάλη. Κρίση: ______(η τροχιά)
138 Η ______είναι σήµερα. Κρίση: ______(η δίκη)
139 Ο πιανίστας πλαισιωνόταν µε τρεις Κρίση: ______. (ο βαρύτονος)
140 Ακούω τους ______. Κρίση: ______(ο πάταγος)
141 Μεταξύ των ______... Κρίση: ______(η σακούλα)
348
142 Μεταξύ του ______και του τραπεζιού... Κρίση: ______(ο βαγενάς)
143 Οι άλλοι κτυπήσαµε τους ______Κρίση: ______βαθύτερα. (ο πάσσαλος)
144 Βλέπω τον ______. Κρίση: ______(ο τύραννος)
145 Οι ______βρίσκονται εκεί. Κρίση: ______(η ταβέρνα)
146 Η αγόρα των ______... Κρίση: ______(η λαµπάδα)
147 Θα κουβαλούσα τον ______µόνος. Κρίση: ______(ο καναπές)
148 Οι µορφές των ______... Κρίση: ______(η αψίδα)
149 Τα ονόµατα των ______... Κρίση: ______(η κοπέλα)
150 Μεταξύ του ______και της καρδιάς... Κρίση: ______(το ήπαρ)
151 Μεταξύ των ______... Κρίση: ______(η αρένα)
152 Ο ______είναι εκεί. Κρίση: ______(ο πάσσαλος)
153 Στον πλαστικό ______σηµαντικών Κρίση: ______µνηµείων... (ο διάκοσµος)
154 Βλέπω τους ______. Κρίση: ______(ο τραγέλαφος)
155 Η ______γίνεται χωµατόδροµος. Κρίση: ______(η άσφαλτος)
156 Η ______βρίσκεται εκεί. Κρίση: ______(η αρένα)
157 Ο ______υπάρχει. Κρίση: ______(ο διθύραµβος)
158 Βλέπω τον ______. Κρίση: ______(ο υπέρµαχος)
349
159 Τα ταξιά των ______... Κρίση: ______(ο ταξιτζής)
160 Η ______βρίσκεται εκεί. Κρίση: ______(η αψίδα)
161 Κρίση: ______Tα πόδια των ______είναι κοντά. (η κότα)
Μπράβο σας! Τελειώσατε! Πολύ σας ευχαριστούµε.
350
APPENDIX E
RUSSIAN FIRST PERSON SINGULAR NON-PAST GAPS
Halle (1973) claimed that approximately 100 Russian verbs have paradigmatic gaps in the first person singular non-past, but he did not provide a list of these gaps. Based on a systematic search of the online version of Ožegov (1972),106 a less thorough search of eight other major Russian grammars and dictionaries (Avanesov 1983, Barxudarov et al.
1963, Graudina 2001, Okuntsova 2004, Rozenthal 1966, Švedova 1982, Ushakov 1974,
Zaliznjak 1977), and miscellaneous other sources, Maria Alley, Bryan Brookes and I were able to identify sixty-nine such words, after collapsing perfective and imperfective pairs, reflexive and non-reflexive pairs, etc. These words are given below. Where two words are given in the same cell, each was listed as having a gap, independently of the other. Other Russian words which share the same root, but which are not listed here, should not be assumed to have regular first person singular non-past forms. The correct interpretation would be that there is not enough information provided in the dictionary entries to know whether such forms have gaps.
The status of some of these words as first person gaps is questionable because it is doubtful that a person would have reason to use the first person singular form to any significant degree. For example, I would not expect шелестеть / šelestet’ ‘to rustle (of
106 http://starling.rinet.ru/cgi-bin/main.cgi?root=/usr/local/share/starling/morpho&morpho=1 351
leaves)’ to appear in the first person singular for purely semantic reasons, and without an expectation that there should be a form, this example would not meet the criteria for a paradigmatic gap. Nonetheless, only a few examples may be explained in this way, and all of the most commonly cited gaps are semantically plausible, even likely, in the first person singular. For the sake of completeness, I provide here all of the words which the sources listed as having first person singular gaps, regardless of potential semantic issues.
The forms marked with a √ were listed as having first person singular non-past gaps in at least five sources. On average 2.9 dictionaries listed any given gap (median 2).
WORD TRANSLITERATION GLOSS 1 басить basit’ ‘to speak or sing in a deep voice’ 2 √ бдеть bdet’ ‘to keep watch’ 3 бороздить borozdit’ ‘to furrow’ 4 √ бузить, набузить buzit’, nabuzit’ ‘to protest’ 5 выздвездить vyzdvezdit’ ‘to cover with stars (??)’ 6 √ галдеть, загалдеть, galdet’, zagaldet’, ‘to make a hubbub’ погалдеть pogaldet’ 7 гвоздить gvozdit’ ‘to hammer’ 8 гнусить gnusit’ ‘to speak in nasal tones’ 9 голосить golosit’ ‘to yell’ 10 грезить grezit’ ‘to dream’ 11 гудеть gudet’ ‘to honk’ 12 √ дерзить, надерзить derzit’, naderzit’ ‘to be imprudent’ 13 дубасить dubasit’ ‘to beat’ 14 √ дудеть, подудеть, dudet’, podudet’, ‘to play the pipe’ продудеть, дудить produdet’, dudit’ 15 елозить elozit’ ‘to go crawling about’ 16 √ ерундить, erundit’, naerundit’ ‘to do stupid or funny things’ наерундить
Continued
Table 71: A potentially complete list of the Russian 1st person singular non-past verbal gaps 352
Table 71 continued
WORD TRANSLITERATION GLOSS 17 желтить želtit’ ‘to turn yellow’ 18 застить zastit’ ‘to stand in someone’s way’ 19 √ затмить, затмиться, zatmit’, zatmit’sja, ‘to eclipse’ тмить tmit’ 20 зудеть zudet’ ‘to itch’ 21 капризить, kaprizit’, ‘to be capricious’ капризиться kaprizit’sja 22 колесить kolesit’ ‘to go/drive around’ 23 √ кудесить, kudesit’, nakudesit’ ‘to do magic’ накудесить 24 кучить kučit’ ‘to earth up (??)’ 25 ладить ladit’ ‘to get along well’ 26 лазить lazit’ ‘to climb’ 27 лебезить lebezit’ ‘to fawn’ 28 лисить lisit’ 29 лихорадить lixoradit’ ‘to have a fever’ 30 лямзить, слямзить ljamzit’, sljamzit’ ‘to steal’ 31 ляпсить, сляпсить ljapsit’, sljapsit’ 32 мерзить merzit’ 33 мутить mutit’ ‘to stir up’ 34 нудить nudit’ ‘to compel’ 35 обезлесить obezlesit’ ‘to deforest’ 36 обезлошадить obezlošadit’ ‘to steal a horse’ 37 обессмертить obessmertit’ ‘to make something immortal’ 38 облесить oblesit’ 39 обрусить obrusit’ ‘to Russify’ 40 обуржуазить, oburžuazit’, ‘to make bourgeoisie’ обуржуазиться oburžuazit’sja 41 отчудить otčudit’ ‘to alienate; to estrange’ 42 √ очутиться očutit’sja ‘to find oneself; to come to be’ 43 ощутить oščutit’ ‘to feel’ 44 парусить parusit’ ‘to sail’ 45 переубедить, pereubedit’, ‘to change one’s mind’ переубедиться pereubedit’sja
Continued
353
Table 71 continued
WORD TRANSLITERATION GLOSS 46 √ победить pobedit’ ‘to win’ 47 погалдеть pogaldet’ ‘to make a lot of noise’ 48 потчудить potčudit’ ‘to behave in a weird way’ 49 предубедить predubedit’ ‘to be prejudiced (against)’ 50 претить pretit’ ‘to sicken, to nauseate’ 51 приютиться prijutit’sja ‘to find shelter’ 52 пылесосить pylesosit’ ‘to vacuum’ 53 разубедить, razubedit’, to dissuade (from) разубедиться razubedit’sja 54 √ рысить, зарысить, rysit’, zarysit’, ‘to trot’ прорысить prorysit’ 55 сбондить sbondit’ 56 сбрендить sbrendit’ ‘to go crazy’ 57 √ соседить sosedit’ ‘to be a neighbor’ 58 √ убедить, ubedit’, ubedit’sja, ‘to persuade’ убедиться, бедить bedit’ 59 угобзить ugobzit’ 60 угораздить ogorazdit’ ‘to urge; to make (do s.t.)’ 61 √ умилосердить, umiloserdit’ ‘to take pity on’ умилосердиться 62 форсить forsit’ ‘to swagger, to show off’ 63 чадить čadit’ ‘to smoke, to emit fumes’ 64 чтить, почтить čtit’, počtit’ ‘to honor’ 65 √ чудесить, čudesit’, načudesit’ ‘to do magic’ начудесить 66 √ чудить, начудить čudit’, načudit’ ‘to behave in a weird way’ 67 √ шелестеть, šelestet’, zašelestet’, ‘to rustle (of leaves)’ зашелестеть, pošelestet’, пошелестеть, prošelestet’ прошелестеть 68 шерстить, šerstit’, perešerstit’ ‘to irritate (of a garment)’ перешерстить 69 √ шкодить, škodit’, naškodit’ ‘to misbehave’ нашкодить
354
APPENDIX F
SURVEY INSTRUCTIONS AND STIMULI: RUSSIAN GAPS
F.1. Instructions as they were presented to subjects
Добро пожаловать! В данном эксперименте мы зададим вам несколько вопросов о том, как вы используете русский язык. Эксперимент состоит из двух заданий. Чтобы продолжить, нажмите на любую клавишу.
_____
Задание 1: Насколько хорошо вы знаете это слово?
_____
В этом задании вы увидите на экране слово красного цвета, например ,,Привет”. Ваша задача – определить, насколько хорошо вы знаете это слово. Нажмите на любую клавишу, чтобы продолжить инструкции...
____
Вы также увидите цифры синего цвета от 1 до 6:
6=Я часто использую это слово. 5=Я иногда использую это слово. 4=Я использовал(а) это слово всего пару раз. 3=Я видел(а) это слово, но сам(а) его не использую. 2=Я могу угадать значение слова, но никогда его не слышал(а). 1=Я не знаю это слово.
Нажмите на клавишу с номером, соответствующим вашей оценке того, насколько хорошо вы знаете слово красного цвета. Постарайтесь отвечать как можно скорее и будьте как можно более внимательны и точны в своих ответах! Нажмите любую клавишу, чтобы продолжить инструкции...
355
____
У вас будет возможность выполнить два тренировочных упражнения, чтобы привыкнуть к этой оценочной шкале. В первом упражнении вы увидите полное описание каждой цифры в шкале. Во втором вы увидите сокращенные описания. Готовы? Нажмите любую клавишу, чтобы начать первое тренировочное упражнение!
____
Давайте потренируемся!
____
Насколько хорошо вы знаете это слово?
6=Я часто использую это слово. 5=Я иногда использую это слово. 4=Я использовал(а) это слово всего пару раз. 3=Я видел(а) это слово, на сам(а) его не использую. 2=Я могу угадать значение слова, но никогда его не слышал(а). 1=Я не знаю это слово.
____
Молодец! X секунд на один ответ. Постарайтесь отвечать быстрее!
____
Вы закончили первое тренировочное упражнение. Помните, что во втором упражнении оценочная шкала остается той же, но описания, которые вы увидите сокращены. Готовы? Нажмите любую клавишу, чтобы начать второе тренировочное упражнение.
____
Давайте потренируемся!
____
Насколько хорошо вы знаете это слово?
Использую его? 4=пару раз 5 6=часто Знаю его? 1=не знаю 2 3=знаю
356
____
Молодец! X секунд на один ответ. Постарайтесь отвечать быстрее!
____
Вы закончили тренировочные упражнения. Если у вас есть каки-то вопросы, задайте их эспериментатору. Готовы? Нажмите любую клавишу, чтобы начать эксперимент.
____
Поздравляем! Это конец первой части эксперимента. Вы выполнили 25% эксперимента. Пожалуйста, отдохните. Когда будете готовы продолжать, нажмите любую клавишу.
Задание 2: Назовите слово
В этой части эксперимента вы увидите предложение с пропуском, написанное зеленым цветом: ,,Я очень ____ яблоки”. После того, как вы прочитаете это предложение, нажмите любую клавишу и на экране под предложением появится слово красного цвета: “любить”. Назовите слово красного цвета в правильной форме, необходимой, чтобы заполниь пропуск в предложении: ,,Я люблю яблоки”.
Постарайтесь отвечать как можно скорее и будьте как можно более внимательны и точны в своих ответах! Нажмите любую клавишу, чтобы увидеть продолжение интсрукций...
____
После того, как вы назовете нужную форму слова красного цвета, вы должны будете ввести число, отражающее то, насколько вы уверены, что форма, которую вы только что назвали правильна. Введите большое число, если вы уверены в правильности данной формы и маленькое число, если вы не уверены в его правильности. Важно не конкретное число, которое вы выберете, а относительные значения, которые вы дадите различным словам по сравнению друг с другом. Если вы вдвойне уверены, что второе слово правильнее, чем первое, введите число, которое в два раза больше. Если вы в два раза менее уверены, введите число в два раза меньше и т.д. Во время эксперимента, вам может показаться, что ваша шкала меняется и что вы забыли, что вы говорили раньше. Постарайтесь не думать об этом. Вы можете выполнить это задание гораздо лучше, чем вам кажется! Если у вас появились какие-либо вопросы, задайте их экспериментатору. Нажмите любую клавишу, когда будете готовы начать тренировочное упражнение.
357
____
Давайте потренируемся!
____
Назовите слово в форме, необходимой для заполнения пропуска в предложении.
____
Молодец! X секунд на один ответ. Постарайтесь отвечать быстрее!
____
Насколько вы уверены, что эта форма слова правильна?
____
Вы закончили тренировочное упражнение. Если у вас появились какие-либо вопросы, задайте их экспериментатору. Готовы? Нажмите любую клавишу.
____
Поздравляем! Вы выполнили 50% эксперимента. Молодец! Пожалуйста, отдохните. Когда будете готовы продолжать, нажмите любую клавишу.
____
Конец эксперимента. Спасибо за ваше участие. Пожалуйста, найдите экспериментатора и получите у него (нее) компенсацию за участие и объяснение эскперимента. До свидания!
F.2. Instructions, English translation
Welcome! We are going to ask you some questions about how you use the Russian language. This experiment has two sections. Push any key to continue. ____
Section 1: How familiar is it?
____
358
In this section, you will see one word in red on each screen, for example: “Hello”. Your task is to decide how familiar you are with this word. Push any key to continue instructions...
____
You will also see blue numbers from 1 to 6:
6 = I use this word frequently. 5 = I use this word sometimes. 4 = I have used this word a couple times. 3 = I am familiar with this word, but don’t use it. 2 = I can guess the meaning of this word, but have never heard it. 1 = I don’t know this word.
Push the number on the keypad corresponding to how familiar you are with the red word. Important! Try to choose a number as quickly as possible while still being accurate! Push any key to continue instructions...
____
There will be two practice sessions so that you can become familiar with the scale. In the first, you will see the full labels for each number. In the second, you will see shorter labels. Ready? Push any key to start the first practice session!
____
Let’s Practice!
____
How familiar are you with this word? 6 = I use this word frequently. 5 = I use this word sometimes. 4 = I have used this word a couple times. 3 = I am familiar with this word, but don’t use it. 2 = I can guess the meaning of this word, but have never heard it. 1 = I don’t know this word.
____
Well done! X seconds per response. Can you go faster?
____
359
This is the end of the first practice. Remember, in the second practice the scale is the same, but the labels have been shortened. Ready? Push any key to start the second practice session!
____
Let’s Practice Again! ____
How familiar are you with this word?
Use it? 4 = a couple times 5 6 = a lot Know it? 1 = not familiar 2 3 = familiar
____
Well done! X seconds per response. Can you go faster?
____
This is the end of the practice sessions. If you have any questions, please ask the experimenter now. Ready to start? Push any key to begin.
____
Congratulations! This is the end of Section 1. You are 25% done with the experiment. Please take a short break. When you are ready to continue, push any key.
____
Section 2: Say the word
____
In this section, you will see a partial sentence in green: “I really ____ apples.” After you have read this sentence, push any key and a red word will appear below the sentence: “love”. Say the word, highlighted in red, in the form needed to complete the given sentence. Important! Say this form as quickly as possible while still being accurate. Push any key to continue instructions...
____
After you say the word, you will be asked to type a number which reflects how confident you feel that the form of the word that you just said is correct. Give a large number if
360
you are very confident, or a small number if you are not very confident. The absolute value of the number is not important. What matters is the relative ratings that you assign to different words. If you are twice as confident in the second word as you are in the first, type a number which is twice as large. If you are half as confident, type a number which is half as large. Etc. During the experiment, you might worry that your scale is changing or that you have forgotten what you said before. Don’t let that worry you. You can do this task better than you think you can. Please ask the experimenter if you have any questions. When you are ready to begin the practice session, push any key.
____
Let’s Practice!
____
Say the word in the form needed to complete the sentence.
____
Well done! X seconds per response. Can you go faster?
____
How confident are you that this form of the word is correct?
____
This is the end of the practice session. If you have any questions, please ask the experimenter now. Ready to start? Press any key.
____
Congratulations! You are 50% done with this experiment! Keep up the good work! Please leave this room and take a short break now. When you are ready to continue, push any key.
____
Congratulations! You are 75% done with this experiment! Keep up the good work! Please take a short break now. When you are ready to continue, push any key.
____
This is the end of the experiment. Thank you for participating. Please find the researcher to receive your pay and an explanation of the purpose of the experiment. Goodbye!
361
F.3. Stimuli
Russian word Russian sentence фузить Ты все время дуешься и ____ ! змлендить Они наверняка завтра ____ . мурасить Он к ней постоянно цепляется и ____ ее. надможить Когда вы купите дом, обязательно ____ его. забечиться Если они ____, я накажу их. встручить Мы пойдем в кафе и ____ это задание. огурить Вы ____ , если увидите ее? забунить Ты завтра ____ мне книгу? бурвать Мы всегда отдыхаем и ____ после работы. сдеговывать Когда ты дома, ты сам ____ ? чусать Она ____ каждый раз, когда приходит сюда. зилеть ____ можно и дома, но лучше на даче. мнетировать У нас в городе сейчас часто ____. ютать Когда у вас есть время, вы часто ____? фугать Что ты тут сидишь и ____ ? пражать Они всегда звонят друг другу и ____ . сшавать Завтра мы ____ все, что сможем. гучать Вы ____ , когда вам страшно? сивнуть Они ____ , если их попросить. мить Когда она звонит, мы все ____ . сжуреть Ты ____ со мной послезавтра? привать Вы здесь работаете или ____ ? понить Как только он приедет, сразу ____. свимать Они ____ , только когда спешат. слутать Хотите мы ____ ? навать Ты часто ____, когда путешествуешь? кухнуть Она ____ через два дня. защитить Кто ему поможет, кто его ____ ? струсить Не побоишься его, не ____? грузить Копаем. ____. Возим. кряхтеть Зрители покашливают, _____. превозносить Женщины любят скромных мужчин, ____ мужскую скромность. ходить Когда есть время, мы ____ в театр. тормозить Путин велел правительству ____ инфляцию. оповестить Если что-то случится, мы вас сразу ____. зависеть Здоровье ____ от состояния души. выпятить Бывает обидится, ____ губу, молчит. опустить Если что, вы ____ меня пораньше? наметить План составим сейчас, а собрание ____ на май.
362
процедить Лук пожаришь, а бульон ____ до осветления. штукатурить Красим, белим, ____. держать Сколько можно? Вы нас за идиотов ____? скандалить Он не ____, а протестует! сэканомить Вот ____ денег, и в отпуск съездишь. острословить Он всегда что-то рассказывает, ____. разрешить А если мы попросим, мать ____ нам погулять? утеплить К зиме автобусы отремонтируют и ____. кухарить Ты и в доме убираешь и ____. сандалить Сидит и ____ пол. сморщиться С годами кожа постареет и ____. тараторить Она много говорит, _____ без умолку. напылить Сейчас приедут с дачи, ____ везде! мыслить Почему мы чувствуем и ____ по-разному? тарабанить Перестань ____! дебоширить Он не сквернословит и не ____. грубить Почему ты пререкаешься и ____ взрослым? поделить Давай ____ яблоко поровну. молиться Мы ходим в церковь и часто ____ . храбриться Ты все ____, надеешься на чудо. обобщить Они встретятся и ____ свои наблюдения. соскоблить ____ с картины верхний слой краски и удивишься! помнить Вы ____ , что с вами произошло? прицениться Сначала узнаешь, где купить выгодно, ____. дарить Он ее балует, часто ____ подарки. таранить Машина выезжает на дорогу и ____ дерево. халтурить Мы никогда не ____, все делаем на совесть. гугнить О чем его не спросишь, он всегда ____ . приобрести Вот ____ машину, научишься водить. пырнуть Как он отреагирует? ____ ножом? мигнуть Увидишь его, тормознешь, ____. таять Весна, во дворе ____ снег. толкнуть Что происходит, что ____ людей на преступление? куковать Тишина. Только где-то вдали ____ кукушки. пропороть Если вдруг ____ колесо, позвонишь мне. сдуреть Будем сидеть на одном месте, пока не ____. хапать Вы же у простого народа воруете и ____! нашить Быстро ____ и продать много рубашек -- трудно. жадничать Ты же все время ____, всего жалеешь. важничать Она всю жизнь зазнается, ____. возражать Мы не просто сомневаемся, а категорически ____ ! вклеивать Сейчас в проездной обычно ____ фотографию. досаждать Реклама отнимает время и ____ зрителям. забраковать Он этот проект одобрит или ____ ? замерзнуть Если ____, наденете шапку.
363
скучать О доме думаешь? ____ небось? навлекать Работает он плохо, ____ на себя неудовольствие начальства. налиновать Я перепишу стихотворение, если ты ____ бумагу. обожать Она любит сладкое, просто ____ шоколад. коченеть Так холодно, ноги просто ____. окроплять Он берет святую воду, ____ ей ребенка. поручать Если ты ____ это Нине, она все сделает. поворковать Соберутся парочки: посидят, ____. подавлять Она не реагирует, ____ свои чувства. подражать Она берет пример с подруги, ____ ей. предвкушать Все ждут, с нетерпением ____ фейерверк. реветь Она сильно расстраивается, ____ . хаять Других ____ легко! спугнуть Не подходи близко к птице, ____ ! чихнуть Если он ____ , значит правду говоришь! сболтнуть Они не проговорятся, не ____ лишнего? вернуть Если найдем вашу книгу, обязательно ____ . заткнуться Когда ты каконец ____ ? рвануть Брошу все, ____ на море! забыть Вы сделаете то, что обещали, не ____ ? мешать Дети бегают и ____ взрослым. влиять Вы работаете с ней и на нее ____ . внушать Идея эта мне не нравится, не ____ доверия. вкушать Они ликуют, ____ плоды своего труда. вопиять Они жалуются, ____ к нему. промокнуть Если не возьмешь зонтик, ____ ! выбыть Если заболеете, ____ из соревнования. бледнеть Он всегда ____, когда волнуется. мешкать Давай быстрее, что ты ____ ? довлеть Ему не везет, над ним ____ злой рок. пьянеть Пьем много, но никогда не ____ . чокаться Будем говорить тосты, ____ . мужать Он быстро взрослеет, ____ . ветшать Идет время, наши дома ____ . изучать Она учится в институте и ____ химию. отрицать Что бы ни случилось, они всегда все ____ . воплощать Пора ____ мечты в жизнь! искажать Зачем ты врешь, ____ факты? затихать Заканчивается день, все ____ . затолкать Будешь сопротивляться, мы ____ тебя силой! ошалеть Сколько еды! ____ можно! донимать Если ____ изжога, сходите к врачу. размещать Принимаем заявки и ____ объявления. козырять Идет парад, военные ____ генералу. играть Когда собираемся, ____ в карты.
364
ерничать Не стоит ни плакать, ни ____ . стирать Она сама готовит, сама ____ . изменять Она не любит мужа и ____ ему. дать Если хочешь, мы ____ тебе этот фильм. жечь Ненавижу, когда ____ траву. пойти Завтра встанем пораньше, ____ на рынок. мёрзнуть Ну что ты ____ тут, заходи! сбрить Переоденешься, ____ бороду. кольнуть Бывает ____ сердце и дышать невозможно. сбавлять Поезд мчится и не ____ скорость! щипать Просыпаешься, ____ себя за руку -- неужели сон? сыпать ____ заварку в чайник, заливаем кипятком. хлестать Тоскливо. На улице ____ дождь. кудахтать Квохчут и ____ куры во дворе. махать Мимо проплывает пароход, пассажиры радостно ____ нам. пахать Они работают, на покладая рук, ____ как лошади. плескать Вокруг тишина; тихо ____ прибой. полоскать Когда горло болит, он его ____ ? хныкать Говорят, что все дети когда-либо ____. ощутить Когда приду домой, ____ огромное облегчение. защитить Сделаю все возможное, ____ страну! зачитить Сначала поем, потом ____ их. спускаться Выхожу из дома, ____ по лестнице. капать У меня насморк, капли ____ каждый день. победить Если захочу, всех врагов ____ . посадить Завтра куплю и ____ дерево. замедить Завтра ____ все, что надо. решаться Сомневаюсь, не ____ купить машину. кропать Сижу за компьютером, ____ статью. убедить Если постараюсь, ____ его в своей правоте. находить Я всегда ____ то, что ищу. кончать Все, ____ валять дурака, начинаю работать. клепать Я сейчас ничего не ____ такими клепками. дерзить Я вам не перечу и не ____ . сглазить Не буду тебя хвалить, а то ____ еще. керзить Я никогда не дуюсь и не ____ . пырнуть Пикнешь, я тебя ____ ножом! щипать Сижу на полянке и травку ____ . бузить Я часто кричу и ____ . струсить Ни за что не побоюсь и не ____! фузить Когда мне грустно, я всегда ____. боднуть Где ____, там кровь пролью. сыпать Напоминаю о прошлом, ____ соль на раны. зудеть Я не ____, я о тебе беспокоюсь. студить Варю яйца, ____ их в холодной воде.
365
мигнуть Посмотрю на тебя и ____. хлестать Хочу пить и потому ____ воду. лазить В походы не хожу и по горам не ____. сузить Подумаю и ____ число подозреваемых. базить Я всегда встаю и ____ во время. хаять Я никогда не жалуюсь и никого не ____. блистать Учусь плохо, успехами не ____. сбрендить Я скоро с вами совсем ____! ссадить Платите или я вас ____ на следующей остановке! змлендить Вот возьму и ____! куснуть Дай, я хлеба ____! кудахтать Я ____, а меня никто не слушает. колесить Я и сейчас ____ по свету. завесить Завтра куплю шторы и ____ окно. толесить Летом я отдыхаю и ____. колдовать Я каждый день гадаю и ____. махать Стою на берегу и ____ руками. ладить Соседей недолюбливаю, не ____ с ними. съездить Завтра позвоню и ____ к бабушке. надить Я всегда ставлю и ____ все на место. уткнуться ____ головой в подушку и заплачу. пахать Сам землю не ____ и ничего не сею. очутиться Растеряюсь, если ____ в незнакомом месте. охотиться Сам веду хозяйство, сам ____. оретиться Как только приду домой, ____. вливаться Чувствую, ____ в ваш коллектив. гудеть Сижу, ____ своим басом. твердить Настаиваю на своем, ____ одно и то же. шудеть Я по пустякам не шумлю и не ____. толкнуть Толкнешь меня, я ____ в ответ! грезить Думаю о нем, ____ наяву. грузить Сам ____, сам отвожу. дрезить Когда меня ругают, я ____. влипнуть Чувствую, сейчас ____ в какую-нибудь историю. плескать Отдыхаю на море, ____ в воде. приютиться Бывало, ____ на скамеечке, слушаю ее рассказы. пресытиться Я ем, пока не ____. куковать Сижу теперь в офисе, ____. полоскать Я сама стираю и ____ белье. дудеть На рояле не играю и в дудку не ____. кряхтеть Я ворочаюсь и ____ во сне. хапнуть Будет возможность, я ____. рыскать Ищу книгу, ____ везде. хныкать Сижу и ____. тыкать Закрываю глаза и ____ в экран.
366
внимать Смотрю на него, жадно ____ его слова. дубасить Играю громко, просто ____ по клавиатуре. пробасить Позвоню ему и ____ поздравление. мурасить Я редко сижу и ____. пропороть Нужно будет, я его штыком ____. стонать Лежу в кровати и ____ от боли. голосить Я никогда не кричу и не ____. загасить Буду уезжать, сам ____ костер. лапать Все порчу, ____ одежду грязными руками. елозить Все время верчусь, ____. заскользить Если я ____, обязательно упаду! омозить Сначала спрошу у него, потом ____. сдуреть Я скоро с вами совсем ____! мутить Я все время строю козни, ____ воду. вкатить Если пойдет дождь, ____ велосипед в гараж. нутить Стою в очереди, ____. стрельнуть Дай-ка я ____ разок! лебезить Я никогда ни хитрю, не ____. проказить Раньше я проказила, я сейчас я не ____. ремезить Часто думаю об этом, ____. хапать Я ____ все, что попадается на глаза. басить Говорю громко, ____. вмесить Возьму сахар, ____ его в тесто. васить Все делаю сама. И ____ сама. тухнуть Ты дома, а я на работе ____. чадить Занимаюсь хозяйством, ____ утюгом. прудить Деревья высаживаю, пруды ____. бадить Работаю много, ____. нашить Буду готовиться к лету, ____ юбок и сарафанов. гвоздить Пилю, крашу, ____. взводить ____ курок и стреляю. двоздить ____ редко, только когда мне плохо. выжать Пойду-ка я ____ сок из лимона. нудить Я никогда не жалуюсь и не ____. ссудить Приезжай ко мне, ____ тебе денег. вдунуть Бог: ,,Я ____ в человека дыхание." пылесосить Убираю и ____ каждый день. превозносить Люблю его, ____ его способности. задегосить Куплю продукты, ____. жадничать Я часто скуплюсь, ____. чудить ___ редко, только когда скучно. удить ____ рыбу и жарю ее. тудить Иду домой, ____. пухнуть Голодаю, просто ____ с голоду. лихорадить Болею, сильно ____.
367
омолодить Сниму заговор, ____! филенадить Просматриваю фотографии, ____. всовывать Одеваю пальто и ____ ноги в сапоги.
368
LIST OF REFERENCES
Ackema, Peter, and Ad Neeleman. 2000. Absolute ungrammaticality. In Optimality Theory: Phonology, syntax, and acquisition, eds. Joost Dekkers, Frank van der Leeuw and Jeroen van de Weijer. Oxford: Oxford University Press, 279-301.
Ackema, Peter, and Ad Neeleman. 2001. Competition between syntax and morphology. In Optimality-theoretic syntax, eds. Géraldine Legendre, Jane Grimshaw and Sten Vikner. Cambridge, MA: The MIT Press, 29-60.
Ackerman, Farrell, and Gert Webelhuth. 1998. A theory of predicates [=CSLI lecture notes 76]. Stanford, CA: CSLI Publications.
Ackerman, Farrell, and Gregory T. Stump. 2004. Paradigms and periphrastic expression: a study in realization-based lexicalism. In Projecting morphology, eds. Louisa Sadler and Andrew Spencer. Stanford, CA: CSLI Publications, 111-158.
Albright, Adam. 2002. The identification of bases in morphological paradigms, dissertation, UCLA.
Albright, Adam. 2003. A quantitative study of Spanish paradigm gaps. In West Coast Conference on Formal Linguistics 22 proceedings, eds. G. Garding and M. Tsujimura. Somerville, MA: Cascadilla Press, 1-14.
Alegre, Maria, and Peter Gordon. 1998. Frequency effects and the representational status of regular inflections. Journal of memory and language 40:41-61.
Andersen, Henning. 1973. Abductive and deductive change. Language 49:765-793.
Andersen, Henning. 1978. Perceptual and conceptual factors in abductive innovations. In Recent developments in historical phonology, ed. Jacek Fisiak. The Hague: Mouton, 1- 22.
Anderson, Stephen R. 1982. Where's morphology? Linguistic inquiry 13:571-612.
Anderson, Stephen R. 1988. Morphological change. In Linguistics: the Cambridge survey, v. 1: linguistic theory: foundations, ed. Frederick J. Newmeyer. Cambridge: Cambridge University Press, 324-361.
369
Anderson, Stephen R. 1992. A-morphous morphology [=Cambridge Studies in Linguistics]. Cambridge: Cambridge University Press.
Andrews, Avery. 1990. Unification and morphological blocking. NLLT 8:507-557.
Aristar, Anthony. 1997. Marking and hierarchy types and the grammaticalization of case- markers. Studies in language 21:313-368.
Aronoff, Mark. 1976. Word formation in generative grammar [=Linguistic Inquiry Monograph 1]. Cambridge, MA: The MIT Press.
Aronoff, Mark. 1994. Morphology by itself: Stems and inflectional classes [=Linguistic Inquiry monographs 22]. Cambridge, MA: MIT Press.
Aski, Janice. 1995. Verbal suppletion: an analysis of Italian, French, and Spanish to go. Linguistics 33:403-432.
Avanesov, R.I., ed. 1983. Orfoėpičeskij slovar′ russkogo jazyka: proiznošenie, udarenie, grammatičeskie formy: okolo 63,500 slov. Moskva: Russkij jazyk.
Avgustinova, Tania. 1994. On Bulgarian verbal clitics. Journal of slavic linguistics 2:29- 47.
Baayen, R. Harald. 1993. On frequency, transparency and productivity. In Yearbook of Morphology 1992, eds. Geert Booij and Jaap van Marle. Dordrecht: Kluwer Academic Publishers, 181-208.
Baayen, R. Harald. 2006. Experimental evidence for Word and Paradigm Morphology. Paper presented at The Annual Meeting of the Linguistic Society of America in Albuquerque, NM, January 5-8, 2006.
Baayen, R. Harald, and Fermín Moscoso del Prado Martín. To appear. Semantic density and past-tense formation in three Germanic languages. Language.
Babiniotis, Georgios. 1998. Lexiko tis neas ellinikis glossas. Athens: Kentro Lexikologias.
Baerman, Matthew, and Greville Corbett. 2006. Three types of defective paradigms. Paper presented at The Annual Meeting of the Linguistic Society of America in Albuquerque, NM, January 5-8, 2006.
Baerman, Matthew, Dunstan Brown, and Greville Corbett. 2005. The syntax-morphology interface: a study of syncretism [=Cambridge studies in linguistics]. Cambridge: Cambridge University Press.
Baker, C.L. 1991. The syntax of English not: the limits of core grammar. Linguistic inquiry 22:387-429.
370
Baker, Mark. 1990. Pronominal inflection and the morphology-syntax interface. CLS 26: Papers from the 26th Regional Meeting of the Chicago Linguistic Society 1:25-48.
Baronian, Luc V. 2005. North of phonology: Linguistics Department, Stanford University.
Barxudarov, S.G., S.I. Ožegov, and A.B. Shapiro. 1963. Orfografičeskij slovar' russkogo jazyka: okolo 104,000 slov. Moskva: Gos. izd-vo inostrannyx i nacional'nyx slovarej.
Beard, Robert. 1995. Lexeme-Morpheme Base Morphology: a general theory of inflection and word formation [=SUNY series in linguistics]. Albany, NY: State University of New York Press.
Becker, Michael. 2003. Lexical stratification in Hebrew: The disyllabic maximum. In Proceedings of the Israeli Association for Theoretical Linguistics 19, ed. Yehuda N. Falk. Be'er Sheva, Israel: Ben Gurion University of the Negev.
Berg, Thomas. 1998. The (in)compatibility of morpheme orders and lexical categories and its historical implications. English language and linguistics 2:245-262.
Bertram, Raymond, Robert Schreuder, and R. Harald Baayen. 2000. The balance of storage and computation in morphological processing: the role of word formation type, affixal homonymy and productivity. Journal of experimental psychology: learning, memory and cognition 26:1-23.
Blevins, James P. to appear 2006. Word and paradigm morphology. Stanford, CA: CSLI Publications.
Blommaert, Jan, ed. 1999. Language ideological debates. New York: Mouton de Gruyter.
Bloomfield, Leonard. 1933. Language. New York: H. Holt and Company.
Bobaljik, Jonathan David. 2001. Syncretism without paradigms: remarks on Williams 1981, 1994. In Yearbook of morphology 2001, eds. Geert Booij and Jaap Van Marle. Dordrecht: Kluwer Academic Publishers, 53-85.
Bochner, Harry. 1993. Simplicity in generative morphology [=Publications in language sciences 37]. Berlin: Mouton de Gruyter.
Booij, Geert. 1996. Inherent versus contextual inflection and the Split Morphology Hypothesis. In Yearbook of morphology 1995, eds. Geert Booij and Jaap Van Marle. Dordrecht: Kluwer Academic Publishers, 1-16.
Booij, Geert. 2002. Constructional idioms, morphology, and the Dutch lexicon. Journal of Germanic linguistics 14:301-329.
371
Börjars, Kersti, Nigel Vincent, and Carol Chapman. 1997. Paradigms, periphrases and pronominal inflection: a feature-based account. In Yearbook of morphology 1996, eds. Geert Booij and Jaap van Marle. Dordrecht: Kluwer Academic Publishers, 155-180.
Bresnan, Joan. 1973. Syntax of the comparative construction in English. Linguistic inquiry 4:275-345.
Bresnan, Joan. 2001. Explaining morphosyntactic competition. In The handbook of contemporary syntactic theory, eds. Mark Baltin and Chris Collins. Malden, MA: Blackwell Publishers, 11-44.
Brown, Dunstan, Greville Corbett, Norman Fraser, Andrew Hippisley, and Alan Timberlake. 1996. Russian noun stress and Network Morphology. Linguistics 34:53-107.
Browne, Wayles. 1993. Serbo-Croat. In The Slavonic languages, eds. Bernard Comrie and Greville Corbett. London: Routledge.
Butterworth, Brian. 1983. Lexical representation. In Language production, vol. 2: development, writing and other language processes, ed. Brian Butterworth. London: Academic Press, 257-294.
Bybee, Joan. 1985. Morphology: a study of the relation between meaning and form. Philadelphia: John Benjamins Publishing Company.
Bybee, Joan, and Paul J. Hopper, eds. 2001. Frequency and the emergence of linguistic structure. Amsterdam: John Benjamins Publishing Company.
Caramazza, Alfonso, Alessandro Laudanna, and Christina Romani. 1988. Lexical access and inflectional morphology. Cognition 28:297-332.
Carstairs-McCarthy, Andrew. 1994. Inflection classes, gender and the principle of contrast. Language 70:737-788.
Carstairs-McCarthy, Andrew. 1998. Comments on the paper by Noyer. In Morphology and its relation to phonology and syntax, eds. Steven G. Lapointe, Diane K. Brentari and Patrick M. Farrell. Stanford, CA: CSLI Publications, 286-301.
Carstairs, Andrew. 1983. Paradigm economy. Journal of linguistics 19:115-125.
Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris Publications.
Chomsky, Noam. 2000a. Minimalist inquiries: the framework. In Step by step: essays on minimalist syntax in honor of Howard Lasnik, eds. Roger Martin, David Michaels and Juan Uriagereka. Cambridge, MA: The MIT Press, 89-155.
Chomsky, Noam. 2000b. The architecture of language. Oxford: Oxford University Press.
372
Corbett, Greville. 1991. Gender [=Cambridge Textbooks in Linguistics 26]. Cambridge: Cambridge University Press.
Corbett, Greville, Andrew Hippisley, Dunstan Brown, and Paul Marriott. 2001. Frequency, regularity and the paradigm: a perspective from Russian on a complex relation. In Frequency and the emergence of linguistic structure, eds. Joan Bybee and Paul J. Hopper. Amsterdam: John Benjamins Publishing Company, 201-226.
Culicover, Peter, and Ray Jackendoff. 1999. The view from the periphery: the English comparative correlative. Linguistic inquiry 30:543-571.
Dixon, Robert M.W. 1982. Where have all the adjectives gone? and other essays in semantics and syntax [=Janua linguarum, series maior 107]. Berlin: Mouton de Gruyter.
Downing, Laura. 2005. Jita causative doubling provides optimal paradigms. In Paradigms in phonological theory, eds. Laura Downing, Christopher Hall and Renate Raffelsiefen. Oxford: Oxford University Press, 122-144.
Downing, Laura, T.A. Hall, and Renate Raffelsiefen, eds. 2005. Paradigms in phonological theory. Oxford: Oxford University Press.
Dressler, Wolfgang U., and Maria Ladanyi. 1998. On grammatical productivity of word formation rules. Wiener Linguistische Gazette 62:29-55.
Eliasson, Stig. 1975. On the issue of directionality. In The Nordic languages and modern linguistics, vol. 2, ed. Karl-Hampus Dahlstedt. Stockholm: Almqvist & Wiksell, 421-445.
Ellington, John Ernest. 1977. Aspects of the Tiene language: Linguistics Department, University of Wisconsin-Madison.
Embick, David. 2000. Features, syntax, and categories in the Latin perfect. Linguistic inquiry 31:185-230.
Fanselow, Gisbert, and Caroline Féry. 2002. Ineffability in grammar. In Resolving conflicts in grammars: optimality theory in syntax, morphology and phonology, eds. Gisbert Fanselow and Caroline Féry. Hamburg: Helmut Buske Verlag, 265-307.
Fertig, David. 1998. Suppletion, natural morphology, and diagrammicity. Linguistics 36:1065-1091.
Fodor, Janet D. 1972. Beware. Linguistic inquiry 3:528-535.
Frampton, John. 2001. The amn't gap, ineffability, and anomalous aren't: against morphosyntactic competition. CLS 37: Papers from the 37th Regional Meeting of the Chicago Linguistic Society, vol. 2.
373
Gazdar, Gerald, Geoffrey K. Pullum, and Ivan A. Sag. 1982. Auxiliaries and related phenomena in a restrictive theory of grammar. Language 58:591-638.
Geckeler, H. 1974. Le probleme des lacunes linguistique. Cahiers de Lexicologie 25:31- 45.
Graudina, L.K. 2001. Grammatičeskaja pravil'nost' russkoj reči: opyt častotno- stilističeskogo slovarja variantov. Moskva: Nauka.
Graziano-King, Janine, and Helen Smith Cairns. 2005. Acquisition of English comparative adjectives. Journal of child language 32:345-373.
Greenberg, Joseph H., ed. 1966[1963]. Universals of language. Cambridge, MA: M.I.T. Press.
Hall, Christopher. 1992. Morphology and mind: a unified approach to explanation in linguistics. London: Routledge.
Halle, Morris. 1973. Prolegomena to a theory of word formation. Linguistic inquiry 4:3- 16.
Halle, Morris. 1997. Distributed Morphology: impoverishment and fission. MIT Working Papers in Linguistics 30:425-449.
Halle, Morris, and Alex Marantz. 1994. Some key features of Distributed Morphology. MIT Working Papers in Linguistics 21:275-288.
Hansson, Gunnar Olafur. 1999. 'When in doubt...': Intraparadigmatic dependencies and gaps in Icelandic. In NELS 29: Proceedings of the 29th meeting of the North Eastern Linguistic Society, eds. Pius Tamanji, Masako Hirotani and Nancy Hall. Amherst, MA: GLSA Publications, The University of Massachusetts, 105-119.
Hawkins, Peter. 1979. Greek diglossia and variation theory. General linguistics 19:169- 187.
Hay, Jennifer. 2001. Lexical frequency in morphology: is everything relative? Linguistics 39:1041-1070.
Hay, Jennifer. 2003. Causes and consequences of word structure [=Outstanding dissertations in linguistics 23]. New York: Routledge.
Hetzron, Robert. 1975. Where the grammar fails. Language 51:859-872.
Hippisley, Andrew. 1999. Word formation rules in a default inheritance framework: a Network Morphology account of Russian personal nouns. In Yearbook of morphology 1999, eds. Geert Booij and Jaap Van Marle. Dordrecht: Kluwer Academic Publishers, 221-261.
374
Hippisley, Andrew, Marina Chumakina, Greville Corbett, and Dunstan Brown. 2004. Suppletion: frequency, categories and distribution of stems. Studies in language 28:387- 418.
Hirschbühler, Paul, and María Luisa Rivero. 1981. Catalan restrictive relatives: core and periphery. Language 57:591-625.
Hock, Hans Henrich. 1991. Principles of historical linguistics. New York: Mouton de Gruyter.
Hockett, Charles F. 1954. Two models of grammatical description. Word 10:210-231.
Hoeksema, Jack. 1985. Categorial morphology. New York: Garland Press.
Holton, David, Peter A. Mackridge, and Irene Philippaki-Warburton. 1997. Greek: a comprehensive grammar of the modern language. London: Routledge.
Hooper, Joan. 1976. Word frequency in lexical diffusion and the source of morphophonological change. In Current progress in historical linguistics: proceedings of the Second International Conference on Historical Linguistics, ed. William M. Christie Jr. New York: North-Holland Publishing Company, 95-106.
Hudson, Richard. 1977. The power of morphological rules. Lingua 42:73-89.
Hudson, Richard. 1985. The limits of subcategorization. Linguistic analysis 15:233-255.
Hudson, Richard. 2000. *I amn’t. Language 76:297-323.
Hume, Elizabeth, and Keith Johnson. 2001. The role of speech perception in phonology. New York: Academic Press.
Hyman, Larry, and Sharon Inkelas. 1998. Emergent templates: the unusual case of Tiene. University of Maryland working papers in linguistics 5:92-116.
Ilola, Eeva, and Arto Mustajoki. 1989. Report on Russian morphology as it appears in Zalizyak's grammatical dictionary. Helsinki: Department of Slavic Languages, University of Helsinki.
Ito, Junko, and R. Armin Mester. 1995. Japanese phonology. In The handbook of phonological theory, ed. John Goldsmith. Cambridge, MA: Blackwell Publishers, 817- 838.
Iverson, Gregory. 1981. Rules, constraints and paradigm lacunae. Glossa 15:136-144.
Iwakura, Kunihiro. 1981. No distinction between Trace and Pro? Linguistic analysis 8:325-341.
375
Jackendoff, Ray. 1975. Morphological and semantic regularities in the lexicon. Language 51:639-671.
Jakobson, Roman. 1984. Morphological observations on Slavic declension (the structure of Russian case forms). In Russian and Slavic Grammar: Studies 1931-1981, eds. Linda Waugh and Morris Halle. Berlin: Mouton Publishers, 105-133.
Janda, Richard. 1982. On limiting the form of morphological rules: German umlaut, diacritics, and the ‘cluster constraint’. NELS 12:140-152.
Janda, Richard. 1984. Why morphological metathesis rules are rare: on the possibility of historical explanation in linguistics. Proceedings of the annual meeting of the Berkeley Linguistics Society 10:87-103.
Janda, Richard, and Brian D. Joseph. 1999. The Modern Greek negator µη(ν)(-) as a morphological constellation. In Proceedings of the 3rd International Conference on Greek Linguistics, ed. G. Babiniotis. Athens: Elinika Gramata, 341-351.
Jannaris, Antonius N. 1987. An historical Greek grammar chiefly of the Attic dialect: as written and spoken from classical antiquity down to the present time: founded upon the ancient texts, inscriptions, papyri and present popular Greek. Hildesheim: Georg Olms Verlag.
Johansson, Christer. 1999. Learning what cannot be by failing expectations. Nordic Journal of Linguistics 22:61-76.
Johansson, Christer. 2003. How is grammatical gender processed? In Computational linguistics in the Netherlands 2002, ed. Tanja Gaustad. Amsterdam: Rodopi, 65-76.
Joseph, Brian D. 1997. On the linguistics of marginality: the centrality of the periphery. Papers from the 33rd regional meeting of the Chicago Linguistic Society:197-213.
Joseph, Brian D. 2005. Paradigms and inflexion: diachronic perspectives and synchronic consequences. Paper presented at 5th Mediterranean Morphology Meeting in Fréjus, France, September 15-18, 2005.
Joseph, Brian D. to appear 2006. On phonically based analogy. In Festschrift for Charles Gribble.
Joseph, Brian D., and Irene Philippaki-Warburton. 1987. Modern Greek. London: Croom Helm.
Joseph, Brian D., and Andrea D. Sims. 2006. Mechanisms of paradigmatic change: diachronic evidence for paradigms. Paper presented at Annual Meeting of the Linguistic Society of America in Albuquerque, NM, January 5-8, 2006.
376
Juge, Matthew. 1999. On the rise of suppletion in verbal paradigms. Proceedings of the annual meeting of the Berkeley Linguistics Society 25:183-194.
Karlsson, Fred. 2000. Defectivity. In Morphologie: ein internationales Handbuch zur Flexion und Wortbildung, eds. Geert Booij, Christian Lehmann and Joachim Mugdan. Berlin: Walter de Gruyter.
Kathol, Andreas. 1999. Agreement and the syntax-morphology interface in HPSG. In Studies in Head-driven Phrase Structure Grammar, eds. Robert Levine and Georgia Green. Cambridge: Cambridge University Press, 223-274.
Kendris, Christopher. 1996. 501 French verbs. Hauppauge, NY: Barron's Educational Series.
Kiparsky, Paul. 1978. Analogical change as a problem for linguistic theory. Studies in the linguistic sciences 8:77-96.
Kiparsky, Paul. 1982. From cyclic phonology to lexical phonology. In The structure of phonological representations, eds. H. van der Hulst and N.V. Smith. Dordrecht: Foris, 130-175.
Kiparsky, Paul. 2000. Analogy as optimization: 'exceptions' to Sievers' Law in Gothic. In Analogy, levelling, markedness: principles of change in phonology and morphology, ed. Aditi Lahiri. Berlin: Mouton de Gruyter, 15-46.
Kiparsky, Paul. 2001. Structural case in Finnish. Lingua 111:315-376.
Kiparsky, Paul. 2005. Blocking and periphrasis in inflectional paradigms. In Yearbook of morphology 2004, eds. Geert Booij and Jaap van Marle. Dordrecht: Springer, 113-135.
Koster, Jan. 1987. Domains and dynasties: the radical autonomy of syntax. Dordrecht: Foris Publications.
Labov, William. 1994. Principles of linguistic change: social factors. Cambridge, MA: Blackwell Publishers.
Langendoen, D. Terrence. 1970. Essentials of English grammar. New York: Holt, Rinehart and Winston.
Larson, Richard K. 2003. Event descriptions in Fon and Haitian Creole. In Recent development in creole studies, ed. Dany Adone. Tübingen: Max Niemeyer Verlag, 23 pp.
Lema, José, and María Luisa Rivero. 1991. Types of verbal movement in Old Spanish: modals, futures and perfects. Probus 3:237-278.
Lieber, Rochelle. 1980. On the organization of the lexicon. Bloomington, IN: Indiana University Linguistics Club.
377
Lieber, Rochelle. 1992. Deconstructing morphology: word formation in syntactic theory. Chicago: University of Chicago Press.
Lightfoot, David. 1979. Principles of diachronic syntax. Cambridge: Cambridge University Press.
Lloret, Maria-Rosa. to appear 2004. The phonological role of paradigms: the case of insular Catalan. In Selected proceedings of the 33rd Linguistic Symposium on Romance Languages, eds. Julie Auger, Clancy Clements and Barbara Vance. Amsterdam: John Benjamins.
Maslov, Jurij S. 1964. Zametki o vidovoj defektivnosti (preimuščestvenno v russkom i bolgarskom jazykax). In Slavjanskaja filologija, eds. Boris A. Larin and German I. Safronov. Leningrad: Izd. Leningradskogo universiteta, 82-94.
Matthews, P.H. 1972. Inflectional morphology: A theoretical study based on aspects of Latin verb conjugation [=Cambridge studies in linguistics 6]. Cambridge: Cambridge University Press.
McCarthy, John J. 2005. Optimal paradigms. In Paradigms in phonological theory, eds. Laura Downing, T.A. Hall and Renate Raffelsiefen. Oxford: Oxford University Press, 170-210.
McCarthy, John J., and Matthew Wolf. 2005. Less than zero: correspondence and the null output. Rutgers Optimality Archive http://roa.rutgers.edu/view.php3?id=998:45 pp.
McCawley, James D. 1988. Review of Noam Chomsky, 'Knowledge of language'. Language 64:355-365.
Morin, Yves-Charles. 1987. Remarques sur l’organisation de la flexion des verbes français. ITL Review of Applied Linguistics 77-78:13-91.
Morin, Yves-Charles. 1990. Parasitic formation in inflectional morphology. In Contemporary morphology, eds. Wolfgang U. Dressler, Hans C. LuschUtzky, Oskar E. Pfeiffer and John R. Rennison. New York: Mouton de Gruyter, 197-202.
Morin, Yves-Charles. 1995. De l'acquisition de la morphologie: le cas des verbes morphologiquement defectifs du francais. In Tendances recentes en linguistique francaise et generale: volume dedie a David Gaatone, eds. Hava Bat-Zeev Shyldkrot and Lucien Kupferman. Amsterdam: John Benjamins Publishing Company, 295-310.
Nevins, Andrew, and Bert Vaux. to appear 2003. Metalinguistic, schmetalinguistic: The phonology of shcm- reduplication. CLS 39: Papers from the 30th Meeting of the Chicago Linguistic Society.
378
Noyer, Rolf. 1998. Impoverishment theory and morphosyntactic markedness. In Morphology and its relation to phonology and syntax, eds. Steven G. Lapointe, Diane K. Brentari and Patrick M. Farrell. Stanford, CA: CSLI Publications, 264-285.
Okuntsova, E.A. 2004. Trudnosti ustnoj reči: slovar'-spravochnik. Moskva: Izdatel'stvo Moskovskogo universiteta.
Orgun, Cemil Orhan, and Ronald L. Sprouse. 1999. From MPARSE to CONTROL: deriving ungrammaticality. Phonology 16:191-224.
Ožegov, S.I. 1972. Slovar′ russkogo jazyka. Moskva: Sov. Ėnciklopedija.
Pertsova, Katya. 2005. How lexical conservatism can lead to paradigm gaps. UCLA working papers in linguistics 11:13-38.
Phillips, Betty. 2001. Lexical diffusion, lexical frequency, and lexical analysis. In Frequency and the emergence of linguistic structure, eds. Joan Bybee and Paul J. Hopper. Philadelphia: John Benjamins Publishing Company, 123-136.
Phillips, Colin. 1996. Order and structure, 306 pp.: MIT.
Pierrehumbert, Janet. 2001. Exemplar dynamics: word frequency, lenition, and contrast. In Frequency and the emergence of linguistic structure, eds. Joan Bybee and Paul J. Hopper. Philadelphia: J. Benjamins, 137-157.
Poser, William J. 1992. Blocking of phrasal constructions by lexical items. In Lexical matters, eds. Ivan A. Sag and Anna Szabolcsi. Stanford, CA: CSLI, 111-130.
Prince, Alan, and Paul Smolensky. 2002[1993]. Optimality Theory: Constraint interaction in generative grammar: Rutgers University.
Raffelsiefen, Renate. 1996. Gaps in word formation. In Interfaces in phonology, ed. U. Kleinhenz. Berlin: Akademie Verlag, 194-209.
Raffelsiefen, Renate. 2004. Absolute ill-formedness and other morphophonological effects. Phonology 21:91-142.
Raffelsiefen, Renate. 2005. Paradigmatic uniformity effects versus boundary effects. In Paradigms in phonological theory, eds. Laura Downing, T.A. Hall and Renate Raffelsiefen. Oxford: Oxford University Press, 211-262.
Revithiadou, Anthi. 1998. Lexical marking and dominance in Modern Greek. In Themes in Greek linguistics II, eds. Brian D. Joseph, Geoffrey Horrocks and Irene Philippaki- Warburton. Amsterdam: John Benjamins Publishing Co., 23-48.
Rice, Curt. 2003. Dialectal variation in Norwegian imperatives. Nordlyd 31:372-384.
379
Rice, Curt. 2005a. Gaps and repairs at the phonology-morphology interface: CASTL, University of Trømso.
Rice, Curt. 2005b. Optimal gaps in optimal paradigms. Rutgers Optimality Archive http://roa.rutgers.edu/view.php3?roa=753:18 pp.
Robins, Robert H. 2001[1959]. In defense of WP. Transactions of the philological society 99:1-36.
Rose, Sharon. 2000. Multiple correspondence in reduplication. Proceedings of the annual meeting of the Berkeley Linguistics Society 23:315-326.
Rozenthal, D.E., ed. 1966. Sbornik uprazhnenij po sovremennomu russkomu jazyku. Moskva: Izd. Moskovskogo Universiteta.
Sadler, Louisa, and Andrew Spencer. 2000. Syntax as an exponent of morphological features. In Yearbook of morphology 2000, eds. Geert Booij and Jaap van Marle. Dordrecht: Kluwer Academic Publishers, 71-96.
Sadock, Jerrold. 1991. Autolexical syntax: a theory of parallel grammatical representations. Chicago: University of Chicago Press.
Sapir, Edward. 1921. Language: an introduction to the study of speech. San Diego, CA: Harcourt Brace & Company.
Schenker, Alexander M. 1995. The dawn of Slavic. New Haven, CT: Yale University Press.
Schieffelin, Bambi, Kathryn Woolard, and Paul Kroskrity, eds. 1998. Language ideologies: practice and theory. New York: Oxford University Press.
Schreuder, Robert, and R. Harald Baayen. 1995. Modeling morphological processing. In Morphological aspects of language processing, ed. Laurie Beth Feldman. Hillsdale, NJ: Lawrence Erlbaum Associates, 131-154.
Sells, Peter. 1985. Lectures on contemporary syntactic theories: an introduction to government-binding theory, generalized phrase structure grammar, and lexical- functional grammar. Stanford, CA: CSLI.
Sells, Peter. 2004. Syntactic information and its morphological expression. In Projecting morphology, eds. Louisa Sadler and Andrew Spencer. Stanford, CA: CSLI, 187-225.
Sims, Andrea. 2005. Declension hopping in dialectal Croatian: Two predictions of frequency. In Yearbook of morphology 2005, eds. Geert Booij and Jaap Van Marle. Amsterdam: Springer, 201-225.
380
Skousen, Royal, Deryle Lonsdale, and Dilworth B. Parkinson, eds. 2002. Analogical modeling: an exemplar-based approach to language. Amsterdam: John Benjamins Publishing Co.
Smolensky, Paul. 1996. The initial state and 'richness of the base' in Optimality Theory: Rutgers Optimality Archive.
Soboleva, Polina A. 1979. Defektnost' paradigmy i semantičeskoe toždestvo slova. Voprosy jazykoznanija 5:37-47.
Spencer, Andrew. 1999. Analytic tenses in Slavic: an LFG approach. Paper presented at Generative linguistics in Poland in Warsaw, Poland,
Spencer, Andrew. 2004. Morphology – an overview of central concepts. In Projecting morphology, eds. Louisa Sadler and Andrew Spencer. Stanford, CA: CSLI, 67-106.
Spencer, Andrew. 2006. Characterizing deponency: implications for form-function mismatches. Paper presented at Deponency and morphological mismatches in Surrey, UK, January 16-17, 2006.
Steele, Susan. 1995. Towards a theory of morphological information. Language 71:260- 309.
Stemberger, Joseph Paul, and Brian MacWhinney. 1986. Frequency and the lexical storage of regularly inflected words. Memory and cognition 14:17-26.
Stewart Jr., Thomas W. 2004. Mutation as morphology: bases, stems, and shapes in Scottish Gaelic: Linguistics Department, Ohio State University.
Strunk, Klaus. 1977. Überlegungen zu defektivität und suppletion im Griechischen und Indogermanischen. Glotta 55:2-34.
Stump, Gregory T. 1991. A paradigm-based theory of morphosemantic mismatches. Language 67:675-725.
Stump, Gregory T. 2001a. Morphological and syntactic paradigms: arguments for a theory of paradigm linkage. In Yearbook of morphology 2001, eds. Geert Booij and Jaap Van Marle. Dordrecht: Kluwer Academic Publishers, 147-180.
Stump, Gregory T. 2001b. Inflectional morphology: a theory of paradigm structure [=Cambridge Studies in Linguistics 93]. Cambridge: Cambridge University Press.
Stump, Gregory T. 2006. Heteroclisis and paradigm linkage. Language 82:279-322.
Švedova, Julja. 1982. Grammatika sovremennogo russkogo literaturnogo jayzka. Moscow: Nauka.
381
Thomason, Sarah G. 1976. Analogic change as grammar complication. In Current progress in historical linguistics: proceedings of the Second International Conference on Historical Linguistics, Tucson, Arizona, 12-16 January 1976, ed. William M. Christie Jr. Amsterdam: North-Holland Publishing Company.
Thumb, Albert. 1964[1912]. A handbook of the Modern Greek language. Chicago: Argonaut Inc. Publishers.
Törkenczy, Miklós. 2002. Absolute phonological ungrammaticality in output-biased phonology. In Approaches to Hungarian, vol. 8: Papers from the Budapest conference, eds. Istvan Kenesi and Péter Síptar. Budapest: Akadémiai Kiadó, 311-324.
Touratzidis, L., and Angela Ralli. 1992. A computational treatment of stress in Greek inflected forms. Language and Speech 35:435-453.
Tracy, Rosemarie. 1986. The acquisition of case morphology in German. Linguistics 24:47-78.
Triantafillidis, Manolis. 1998. Lexiko tis koinis neoellinikis. Thessaloniki: Aristoteleio Panepistimio Thesssalonikis, Institouto Neoellinikon Epoudon.
Ushakov, D.N., ed. 1935-1940. Tolkovyj slovar' russkogo jazyka. Moskva: Gos. in-t "Sovetskaja enciklopedija".
Ushakov, D.N., ed. 1974. Tolkovyj slovar′ russkogo jazyka. Cambridge, MA: Slavica Publishers. van Marle, Jaap. 2000. Syntagmatics versus paradigmatics. In Morphologie : ein internationales handbuch zur flexion und wortbildung, eds. Geert Booij, Christian Lehmann and Joachim Mugdan. Berlin: Walter de Gruyter, 226-234.
Vincent, Nigel. 1987. The interaction of periphrasis and inflection: some romance examples. In Historical development of auxiliaries, eds. Martin Harris and Paolo Ramat. Berlin: Mouton de Gruyter, 237-256.
Vincent, Nigel. 2005. Defectiveness, inflection and derivation. Paper presented at 5th Mediterranean Morphology Meeting in Fréjus, France, September 15-18, 2005.
Vinogradov, V.V. 1969. The history of the Russian literary language from the seventeenth century to the nineteenth. Madison, WI: University of Wisconsin Press.
Vlasto, A.P. 1988. A linguistic history of Russian to the end of the eighteenth century. Oxford: Clarendon Press.
Walker, Rachel, and Bella Feng. 2004. A ternary model of morphology-phonology correspondence. In West Coast Conference on Formal Linguistics 23 proceedings, eds.
382
Vineeta Chand, Ann Kelleher, Angelo J. Rodríguez and Benjamin Schmeiser. Somerville, MA: Cascadilla Press, 787-800.
Wedel, Andrew Benjamin. 2004. Self-organization and categorical behavior in phonology: Department of Linguistics, University of California, Santa Cruz.
Weschler, Stephen, and Larisa Zlatić. 2000. A theory of agreement and its application to Serbo-Croatian. Language 76:799-832.
Williams, Edwin. 1981. On the notions "Lexically related" and "head of a word". Linguistic inquiry 12:245-274.
Woolf, H.B. 1980. The past participle of stride. American speech 55:298-301.
Wunderlich, Dieter. 1995. Minimalist morphology: the role of paradigms. In Yearbook of morphology 1995, eds. Geert Booij and Jaap Van Marle. Dordrecht: Kluwer Academic Publishers, 93-114.
Wunderlich, Dieter. 1999. Paradigmatic gaps and substitutions in correspondence- theoretical morphology. Paper presented at Workshop on conflicting rules in phonology and syntax in University of Potsdam, December 17-18, 1999.
Wunderlich, Dieter. 2001. How gaps and substitutions can become optimal: the pronominal affix paradigms of Yimas. Transactions of the philological society 99:315- 366.
Zaliznjak, A.A., ed. 1977. Grammatičeskij slovar' russkogo jazyka: slovoizmenenie. Moskva: Russkij jazyk.
Zwicky, Arnold M. 1985. How to describe inflection. Proceedings of the eleventh annual meeting of the Berkeley Linguistics Society:371-386.
Zwicky, Arnold M. 1989. Quicker, more quickly, *quicklier. Ohio State University working papers in linguistics 37:84-99.
Zwicky, Arnold M. 1990. Inflectional morphology as a (sub)component of grammar. In Contemporary morphology, eds. Wolfgang U. Dressler, Hans C. Luschützky, Oskar E. Pfeiffer and John R. Rennison. New York: Mouton de Gruyter, 217-236.
Zwicky, Arnold M. 2004. "Losing" "the subjunctive". Language log July 11, 2004: http://itre.cis.upenn.edu/~myl/languagelog/archives/001192.html.
Zwicky, Arnold M., and Geoffrey K. Pullum. 1983. Cliticization vs. inflection: English n’t. Language 59:502-513.
383