Adpositions in Distributed Morphology

Material Information

Adpositions in Distributed Morphology The Nature of Roots and Categorial Heads
Deacon, Robert
Place of Publication:
[Gainesville, Fla.]
University of Florida
Publication Date:
Physical Description:
1 online resource (319 p.)

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Committee Chair:
Committee Co-Chair:
Committee Members:
Graduation Date:


Subjects / Keywords:
Adpositions ( jstor )
Grammatical particles ( jstor )
Linguistics ( jstor )
Morphemes ( jstor )
Nouns ( jstor )
Phrases ( jstor )
Prepositions ( jstor )
Syntactics ( jstor )
Verbs ( jstor )
Words ( jstor )
Linguistics -- Dissertations, Academic -- UF
adpositions -- distributed-morphology -- dm -- late-insertion -- morphology -- particles -- post-positions -- prepositions -- roots -- semi-lexical -- subset-principle -- vocabulary-insertion
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Linguistics thesis, Ph.D.


This work examines the status of adpositions within the framework of Distributed Morphology, DM, demonstrating that adpositional items on the surface are a problem in DM because they are traditionally understood to have both lexical and functional properties and because the vocabulary insertion mechanism of DM requires a strict division between lexical and functional items. Upon closer inspection, however, this work shows that the lexical distribution of many adpositional items in English, Chinese, and other languages can be accounted for in DM by analyzing these forms as Root lexical items categorized by a little p head. Therefore, like nouns, verbs and adjectives, many adpositions consist of bare Roots selected by a category defining head, little p, which may come in several different flavors defined by different adpositional or relational features. Those adpositions that do not have a lexical distribution are then argued to be the representation of a little p head that has selected for a DP. These adpositional forms, like auxiliaries for verbs and pronouns for nouns, are viewed as functional insertions within a domain capable of framing lexical nodes. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis (Ph.D.)--University of Florida, 2014.
Electronic Access:
Statement of Responsibility:
by Robert Deacon.

Record Information

Source Institution:
Rights Management:
Applicable rights reserved.
Embargo Date:
LD1780 2014 ( lcc )


This item has the following downloads:

Full Text




2 © 2014 Robert Joel Deacon


3 To my Beautiful W ife Family and Friends May God K eep and Bless you In all Ways of L ife


4 ACKNOWLEDGMENTS If the wisdom of the east and west m ust collide then let it do so in complementary progression . We may then know that what is old is new and new old: will be again, what has been done will be done again; there is nothing new under the . That is w e do not ever learn anything new but rather remember what has been forgotten ( Socrates in Plato ). With every restatement, we discover/remember a littl e more of how and why things are . I want to thank all those who have helped me through this journe y: my parents Robert and Kathrin Deacon, my wife Yukari Nakamura Deacon, my daughter, who will be joining us soon , and the rest of my family (you know who you are). Without your support, I would not have had the will to press forward. I want to give thanks to my committee members Dr. Brent Henderson, Dr. Youssef Haddad, Dr. Fiona McLaughlin and Dr. Ozlem Yilmaz. This would not have been possible without your wisdom and guidance. I also want to give special thanks to my advisor and Chair Dr. Brent Henderson for guiding, encouraging, and helping to shape the ideas presented here. This would no t have been possible without him . Special thanks also to Dr. Youssef Haddad who likewise joined this project early, offering encouragement and insightful criticism. I wo uld not have exami ned things as deeply without him . I want to also thank my friends and colleagues that I spent time with in discussion, debate and foolishness: Sean and Wi young (Z oe ) Witty; Tyler McPeek; Dong yi L in; Lee Ballard; Jimmy Harnsberger; Chris tian Betzen; Demetris Athienitis; Laura Dawidziuk ; Rachel Organes ; Jules Gliesche; Sherry Tyrrell Gliesche; Andrew Wilson; Sean Park; Aazam Feiz; Yasuo


5 TABLE OF CONTENTS page ACKNOWLE DGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 10 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ABBREVIAT IONS AND TERMS ................................ ................................ ..... 12 ABSTRACT ................................ ................................ ................................ ................... 14 CHAPTER 1 ADPOSITIONS AND CATEGORIZATION ................................ .............................. 15 1.1 Overview ................................ ................................ ................................ ........... 15 1.2 Research Questions ................................ ................................ ......................... 19 1.3 The Lexical Category Dilemma ................................ ................................ ......... 21 1.3.1 Response to the Lexical Category Dilemma ................................ ............ 22 1.3.2 Functional and Lexical Categories ................................ .......................... 23 1.4 The Cat egory Adposition ................................ ................................ .................. 25 1.4.1 Traditional Methods for Classifying Adpositions ................................ ...... 28 1.4.2 Adpositions and Semi Lexicality ................................ .............................. 38 1.5 Unifying Adpositional Elements ................................ ................................ ........ 40 1.5.1 Syntactic Distribution and Function ................................ ......................... 41 1.5.2 Adpositional Stacking ................................ ................................ .............. 46 1.6 Semantic Syntactic Function: An Adpositional Unifying Feature ...................... 48 1.6.1 Semantic Func tion: Transitive Adpositions ................................ .............. 49 1.6.2 Semantic Function: Additional Meanings ................................ ................. 51 1.7 Particles and Prepositions ................................ ................................ ................ 54 1.7.1 Particles and Figures ................................ ................................ ............... 55 1.7.2 Semantic Function: Intransitive Adpositions ................................ ............ 59 1.7.3 Directional Particles and Results ................................ ............................. 62 1.7.4 Semantic Function and Stacking ................................ ............................. 63 1.7.5 Where Adpositions Stand ................................ ................................ ........ 66 1.8 Concluding Remarks ................................ ................................ ......................... 67 2 DISTRIBUTED MORPHOLOGY: THE LEXICAL FUNCTIONAL DISTINCTION .... 69 2.1 Distributed Morphology ................................ ................................ ..................... 69 2.2 The Lexicon ................................ ................................ ................................ ...... 70 2.2.1 The Traditional Lexicon ................................ ................................ ........... 70 2.2.2The Lexicon of DM ................................ ................................ ................... 75 2.3 The Vocabulary Insertion Mechanisms of DM ................................ .................. 78 2.3.1 Late Inse rtion ................................ ................................ ........................... 78


6 2.3.2 Associated and Contextual Features ................................ ....................... 79 2.3.3 Underspecification ................................ ................................ ................... 80 2.3.4 The Subset Principle ................................ ................................ ............... 81 2.4. L and F Morphemes and Categorial Heads ................................ ..................... 82 2.4.1 L and F Morphemes and the De rivation ................................ .................. 83 2.4.2 The Categorial Heads of DM ................................ ................................ ... 84 2.5 L and F Morphemes and Semi Lexical Categories ................................ ........... 85 2.5.1 L and F Morphemes and Adpositions ................................ ...................... 85 2.5.2 Adpositions in DM ................................ ................................ .................... 86 2.6 Concluding Remark s ................................ ................................ ......................... 87 3 MORPHOLOGY ................................ ................................ ................................ ...... 89 3.1 Introduction ................................ ................................ ................................ ....... 89 ................................ ................................ ................... 91 ................................ .......................... 92 3.3.1 Theoretical Reaso ...... 94 3.3.2 ... 96 ................................ ................................ ...................... 102 ...................... 102 3.4.2 Special Meanings: Inner v s. Outer Layers ................................ ............. 105 ................................ ..................... 109 Class Features ..... 109 Syntactic Features ................................ ................................ ................................ ...... 113 3.6 Categorial Gaps: Limitat ................................ .......................... 116 3.6.1 Categorial Gaps: Accidental Gaps ................................ ........................ 117 3.6.2 Categorial Gaps: Allomorphy and Gaps ................................ ................ 121 ................................ .. 123 3.7 Features vs. No Features ................................ ................................ ............... 126 3.7.1 No Features are Optimal ................................ ................................ ....... 127 3.7.2 Indices ................................ ................................ ................................ ... 132 3.7.3 Contextual Features ................................ ................................ .............. 134 3.7.4 Contextual Features versus Indices ................................ ...................... 136 3.8 The Encyclopedia ................................ ................................ ........................... 138 3.9 Early Insertion and Late Ins ertion ................................ ................................ ... 139 3.10 Concluding Remarks ................................ ................................ ..................... 140 4 ADPOSTIONAL FORMS IN ENGLISH AND MANDARIN ................................ ..... 142 4.1 Introduction ................................ ................................ ................................ ..... 142 4.2 Lexical Prepositions in English ................................ ................................ ....... 143 4.2.1 Nominal and Verbal Distributional Evid ence: Proposed LVI Prepositions ................................ ................................ ................................ 145 4.2.2 Adjectival Distributional Evidence: Proposed LVI Prepositions ............. 150 4.3 Functional Prepo sitions in English ................................ ................................ .. 154


7 4.3.1 Functional Prepositions: No Lexical Distribution ................................ .... 154 4.3.2 Implications of FVI Prepositions ................................ ............................ 155 4.4 Adpositions in Mandarin ................................ ................................ .................. 156 4.4.1 Adpositional Positions in the Mandarin Clause ................................ ...... 156 4.4.2 Determining Prepositions in Mandarin ................................ ................... 158 4.4.3 Determining Postpositions in Mandarin ................................ ................. 162 4.5 Lexical Adpositions in Mandarin ................................ ................................ ..... 168 4.5.1 Lexical Postpositions in Mandarin ................................ ......................... 169 4.5.2 Derived Lexical Postpositions in Mandarin ................................ ............ 172 4.5.3 Lexical Prepositions in Mandarin ................................ ........................... 173 4.5.4 Prepositional Forms and Compounding ................................ ................ 176 4.6 Functional Adpositions in Chinese ................................ ................................ .. 177 4.6.1 The Preposition Zài ................................ ................................ ............... 177 4.6.2 The Proposed FVI Prepositions ................................ ............................. 180 4.7 Concluding Remarks ................................ ................................ ....................... 183 5 ADPOSITIONS THROUGH THE LENS OF DISTRIBUTED MORPHOLOGY ...... 184 5.1 Introduction ................................ ................................ ................................ ..... 184 5.2 Potential Solution for Adpositions and Semi Lexical Forms ............................ 185 5.2.1 The Po nodes ................................ .................. 185 5.2.2 Revised Subset Principle ................................ ................................ ...... 186 5.3 Problems with Solution ................................ ................................ ................... 188 5.3.1 Overgeneration ................................ ................................ ...................... 188 5.3.2 Unidirectionality ................................ ................................ ..................... 190 5.3.3 Revised Subset Principle: Inde pendently Unmotivated ......................... 190 5.4 The Proposal for p_ ................................ ................................ ........................ 191 5.4.1 What is Little p_ ................................ ................................ ..................... 191 ................................ ........................... 193 5.4.3 The Functional Prepositions of English and Mandarin ........................... 196 5.5 Particles and Prepositions and Little p_ ................................ .......................... 200 5.5.1 Particle Forms and their Root node Distribution ................................ .... 201 5.5.2 Particles as Furthe r Evidence for this Analysis ................................ ...... 201 5.6 Revisiting to the Extended Projection of P ................................ ...................... 204 5.6.1 Revisions to the Extended Projection o f P ................................ ............. 205 5.6.2 Explaining Chinese Data with Revisions to P structure ......................... 206 5.7 Concluding Remarks ................................ ................................ ....................... 210 6 EVIDENCE FOR LITTLE P_ ................................ ................................ ................. 211 6.1 Introduction ................................ ................................ ................................ ..... 211 6.2 Evidence for Little p_ English ................................ ................................ .......... 212 6.2.1 A and Be Headed Forms ................................ ................................ ..... 213 6.2.1 A and Be Headed Particle Forms ................................ ........................ 216 6.2.2 The Suffix ward ................................ ................................ .................... 220 6.2.3 Evidence for Little p_ in Chinese ................................ ........................... 222 6.3 Evidence in other Languages ................................ ................................ ......... 223


8 6.3.1 Evidence from German ................................ ................................ .......... 2 23 6.3.2 Evidence from Dutch ................................ ................................ ............. 226 6.3.3 Evidence from Persian ................................ ................................ .......... 229 6.4 Concluding Remarks ................................ ................................ ....................... 236 7 LITTLE P_ IN A WORLD OF N_,V_, AND A_ ................................ ....................... 238 7.1 Introduction ................................ ................................ ................................ ..... 238 7.2 The Universality of Lexical Categories ................................ ............................ 239 7.2.1 Lexical Categories and Univers ality ................................ ....................... 240 7.2.2 Universal Lexical Categories and DM ................................ .................... 242 7.2.3 Universal Lexical Categories, DM and Little p_ ................................ ..... 244 7.3 The Behavior of the Categorial Heads ................................ ............................ 244 7.3.1 Little v_ in Relation to Little p_ ................................ ............................... 245 7.3.2 Little n_ in Relation to Little p_ ................................ .............................. 249 7.3.3 Little p_ in Comparison to Little n_ and v_ ................................ ............. 251 7.4 Structural Sim ilarities ................................ ................................ ...................... 253 ................................ ...... 253 7.4.2 A Possible Difference ................................ ................................ ............ 258 7.5 Concluding Remarks ................................ ................................ ....................... 260 8 THE ORGIN OF LITTLE P_ ................................ ................................ .................. 261 8.1 Introduction ................................ ................................ ................................ ..... 261 8.2 Old Language, Language Change and the Emergence of p_ ......................... 263 8.3 Adpositions as a Product of Grammaticalization ................................ ............. 265 8.4 Grammaticalization ................................ ................................ ......................... 266 8.4.1 Grammaticalization Understood via its Effects ................................ ...... 267 8.4.2 G rammaticalization and Direction ................................ .......................... 270 8.4.3 Grammaticalization Causes ................................ ................................ ... 271 8.5 Adpositions, Grammaticalization, and a General Model ................................ . 276 8.5.1 Historical Analysis of Proposed Adpositional FVI Prefixes in English .... 277 8.5.2 Historical Analysis of Adpositional Af .................. 278 8.5.3 Historical Analysis of Proposed English FVI Words .............................. 281 8.5.4 Historical Analysis of Proposed Engl ish Adpositional LVIs .................... 284 8.6 Modeling Adpositions with Grammaticalization ................................ ............... 286 8.6.1 Adpositions and Grammaticalization Stages ................................ ......... 286 8.6.2 Revisions to the Adpositional Cline ................................ ....................... 288 8.7 Grammaticalization and DM ................................ ................................ ............ 290 8.7.1 Contextual to Associated Features ................................ ........................ 291 8.7.2 The Proposed Model and Unidirectionality ................................ ............ 293 8.8 Concluding Remarks ................................ ................................ ....................... 293 9 CONCLUSIONS AND IMPLICATIONS ................................ ................................ . 296 9.1 Summary ................................ ................................ ................................ ........ 296 9.2 Implications and Future Research ................................ ................................ .. 301


9 9.3 Distributional Questions and Curiosities ................................ ......................... 303 9.4 Final Remarks ................................ ................................ ................................ . 306 LIST OF REFERENCES ................................ ................................ ............................. 308 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 319


10 LIST OF TABLES Table page 1 1 Differentiating lexical and functional material ................................ ..................... 24 4 1 Mandarin sentential template ................................ ................................ ............ 157 4 2 Mandarin post positions/relational nouns ................................ ......................... 164 4 3 Mandarin postpositional uses ................................ ................................ ........... 169 4 4 Derived post positions or relational nouns ................................ ........................ 172 4 5 Mandarin lexical prepositions ................................ ................................ ........... 173 4 6 Mandarin place prepositions ................................ ................................ ............. 178 4 7 Mandarin functional prepositions ................................ ................................ ...... 181 6 1 A headed forms ................................ ................................ ................................ 214 6 2 Be headed prepositions ................................ ................................ .................... 215 6 3 ................................ ................................ ........................... 227 6 4 Derivation with / en/ ................................ ................................ .......................... 228 6 5 Derivation with /b(e)/ ................................ ................................ ......................... 228 6 6 Persian prepositions ................................ ................................ ......................... 230 8 1 Etymological source and meaning of English adpositional prefixes ................. 277 8 2 Etymological source and meaning of English A derived adpositions ................ 279 8 3 Etymological source and meaning of English Be derived adpositions .............. 280 8 4 Etymological source and meaning of English functional prepositions ............... 281 8 5 Etymological source and meaning of English lexical adpositions ..................... 284 8 6 Etymological source and meaning of English lexical adpositions 2 .................. 285 8 7 Revision of adpositional cline ................................ ................................ ........... 290


11 LIST OF FIGURES Figure page 1 1 Extended projection of P. ................................ ................................ ................... 47 1 2 Semantic function of transitive a ................................ ....................... 50 2 1 Derivation in DM. ................................ ................................ ................................ 76 2 2 Associative and Contextual Features in tandem. ................................ ................ 82 2 3 Hierarchical product of merge. ................................ ................................ ............ 83 8 1 Contextual and Associative Feature swap. ................................ ....................... 292


12 LIST OF ABBREVIATIONS AND TERMS Adposit ion A term that captures a range of structural positions (pre, post, circum, and ambipositions) attributed to the category P. A headed Ps A multi morphemic preposition that begins with the prefix /a / Associated Feat . Features listed with FVIs that dete rmine which is most suitable given the Subset Principle Be headed Ps A multi morphemic preposition that begins with the prefix /be / Category P A set of features characterized by adpositions and perhaps Case markers Contextual Feat. Features that ref lect the usage of a VI with regards to the larger context. DM Distributed Morphology Figure An entity whose position is unknown or variable [ F LINK ] feature A feature responsible for linking a concept or other morpho syntactic features to a DP in such a way that the DP is understood as the Figure in relation to the meaning of the concept. F m orpheme A functional morpheme: a bundle of morpho syntactic features phonologically realized by a competing VI. F n ode A terminal node in the syntax that contains at least one morpho syntactic feature FVI A Vocabulary Item with Associated Features and perhaps Contextual Features Ground An entity whose position is known and not variable List A A collection of morpho syntactic bundles provided by UG from which t he syntax can be constructed via Merge. List B VIs listed with Associated and Contextual Features that are used to provide the terminal nodes of the syntax with phonological information List C A list of VIs in context with idiomatic meaning and real worl d knowledge.


13 L m orpheme Root dominated by a functional head. Root and not specific morpho syntactic features. L n ode The place in the sy ntactic structure that hosts a , a position void of morp ho syntactic features node. LVI A Vocabulary Item with no Associated Features but may be specified with Contextual Features P item The phonological forms that commonly represent both English prepositions and particles. This thesis argu es they are in the l morpheme class. p_ A functional head capable of licensing Root item in an adpositional domain . A place holder for what we traditionally call lexical items. UG Universal Grammar: A belief that all languages work from a set of innate biological principles. VI Vocabulary Item: Phonological strings used to realize f morphemes and l morphemes.


14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the R equirements for the Degree of Doctor of Philosophy ADPOSITIONS IN DISTRIBUTED MORPHOLOGY: THE NATURE OF S AND CATEGORIAL HEADS By Robert Joel Deacon August 2014 Chair: Brent Henderson Major : Linguistics This work examines the status of adpositions within the framework of Distributed Morphology (DM), demonstrating t hat adpositional items on the surfa ce are a problem in DM because they are traditionally understood to have both lexical and functional properties and because the vocabulary insertion mechanism of DM requires a strict division between lexical and functional items. U pon closer inspection, ho wever, this work shows that the lexical distribution of many adpositional items in English, Chinese, and other languages can be accounted for in DM by analyzing these forms as Root lexical item s categorized by a little p_ head. Therefore, like nouns, verbs and adjectives , many adpositions consist of bare Root s selected by a category which may come in several different flavors defined by different a d positiona l or relational features . Those adpositions that do not have a lexical distribution are then argued to be the representation of a little p_ head that has selected for a DP . These adpositional forms, like auxiliaries for verbs and pronouns for nou ns, are viewed as functional insertions within a domain capable of framing lexical nodes .


15 CHAPTER 1 ADPOSITIONS AND CATEGORIZATION 1.1 Overview This dissertation examines the category adposition and the theoretical model Distributed Morpholo gy (DM). It s hows how solving a DM internal problem advances our how this solution aligns the category adposition with other lexical categories and accounts for the traditionally described split functi onal/lexical behavior of the category . One of the few things that can be uncontroversial ly claimed about the putative category adposition is that it is difficult to classify. Even though adpositions have been gaining more attention ( oopman 2 000; Zeller 2001; Zwart 2005; den Dikken 2006; Svenonius 2004, 2006, 2007, 2008 ; Djamouri et al. 2012 among others ) , their classification identity continues to be a problem. Adpositions have been classified as a functional class (Grimshaw 1991, 2005 ; Baker 2003; DeLancy 2005 ) , a lexical class (Jackendoff 1973, 1977; Chomsky 1981; Koopman 2000) and as a semi lexical class (Zeller 2001; Mardale 2011). In addition to this, adpositions have been described as a word class in the middle (Kiparsky 1983), a split w ord class (Hudson 2000), and as a word class th at is simply difficult to cross linguistically describe (Haspelmath 2007). In some languages adpositional forms appear to only be functionally represented (Yoruba, here appear to be both functional and T his class thus appears to defy within many frameworks the unity found with word classes where meanin g, use and form all align with the traditio nal functional/lexical division of labor. Nevertheless, this grouping of linguistic items appears to be distinct from other linguistic


16 categories in semantic function and syntactic distribution. This therefore creates problems for grammatical models that depend upon a strict lexical/functional distinction for the organization of morphemes. Adpositions appear to be a problem for any grammatical system that must divide morphological components into two qualitatively different categories . This is especially true for framework of Distributed Morphology DM (Halle and Marantz 1993, 1994). DM traditionally demands that morphemes be categorized as either l morphemes (lexical morphemes) or f morphemes (functional morphemes). T he status of this category i n DM has no t been closely examined or defined within the model . There has been some work on the agreement of prepositions in Irish using the framework of DM in Legat e (1999) and Brennan (2007) where the status of prepositions appe ars to be functional , but the lexical / functional status of prepositions is not the focus of either work. Svenonius comments that prepositions in English might be [i.e. DM] T his view , h owever , is balanced with the opposite Svenonius (2004, 2007, and 2008) is working within the cartographic tradition and not the DM model, it is uncertain if he notices the theory internal problem that these items


17 create for DM, 1 leaving the status of semi lexical items ( cf. De Belder 2011, 2013; Deacon 2011) and adpositional items (Deacon 2011) unresolved within the DM model . Adpositional forms are often described as struct ura l c ase assigners (den Dikken 1995; Asbury 2006; Hagege 2010). This associates these items with at least the functional feature [ CASE ] and groups the category with other f morphemes. In some languages though, adpositional forms also appear in lexical dom ains without any functional features, suggesting that these items are l morphemes (Deacon 2011). The Vocabulary Insertion mechanisms of DM, however, prevent functional items from being inserted in lexical positions and make it unlikely for lexical items to be inserted in functional positions . problem, it fails to account for adpositions and has to introduce some independently (explained in Chapter 5) . If there is a Category P, moreover, there is no answer for why it appear s to be lexical (or more lexical) in only some languages. In addition to the strict division between l and f morphemes, DM proposes that the sour ce of lexical material (i.e. a Root node) has no category and must gain such a category by appearing in a local domain with a category defining f morpheme (Embick and Noyer 2005; Acquaviva 2008). Thus instead of adopting radical changes to DM, this work proposes that the categorial fea ture [+p] (i.e. a little p_ head ) be added to the list of possible categorial heads to categorize information in an adpositional context. Functional adpositions will be argued to be representations of this categorial p_ head 1 Chapter 2 demonstrates how the insertion mechanism for DM relies upon a qualitative distinction where lexical items cannot represent features in the syntax while functional items only represent such features.


18 while lexical adpositions will b e argued to be the result of a Root node categorized by this p_ head. Viewed in this manner, the category adposition is no longer problematic for the insertion mechanisms of DM. In fact, the findings suggest that the basic architecture of DM 2 acco unt s for the behavior of adpositions quite well . The inclusion of this head not only helps solve traditional problems within the perspective of DM, it opens the door for a range of investigations regarding how language features can interact with extra gram matical knowledge. It then can be suggested that so me languages do not categorize Root information with a p_ head (or perhaps it is very limited) much as it is argued that some lan guages do not categorize Root s in an adjectival frame (cf. Kim 2002 for a discussion of Korean) or if so it is very limited (cf. McLaughlin 2004 for a discu ssion of Wolof). The evidence, details, and implications of this proposal await your judgment. Chapte r 1 gives an overview of adpositions and provides a means of unificati on for adpositional elements. Chapter 2 gives an overview of DM (i.e. why DM is a help ful model and how adpositions are puzzling fit for the m odel); Chapter 3 , defines Root s in DM, discussing Root features and categorization, and what leads to gaps in a generative model ; Chapter 4 provides evidence in English and Mandarin Chinese that some adpositional elements have a lexical distributio n while others do not; Chapter 5 formally proposes little p_ as the categorizer of Root s in adpositional domains, 2 categorial and need to be framed by a grammatical head (discussed further in Chapters 2 and 3).


19 2013) solution for semi lexical forms in DM and shows how the little p_ analysis i s helpful for explaining the distribution of part icle forms in English; Chapter 6 gives phonological evidence for little p_ in English, Dutch and Persian and explains why it would be natural for there to be no evidence in Mandarin Chinese; Chapter 6 explai ns that there is no reason to exclude little p_ from the other more accepted categorica l heads: n_ , v_ and a_ ; Chapter 8 gives a historical account for the category P, noting that since many adpositional items begin as lexical items, it is not unreasonable to think that many have remained lexical and that those that have undergone grammaticalization are now inserted under the proposed little p_ head; Chapter 9 concludes, discussing remaining issues and suggesting points for further research. The organizatio n for the rest of Chapter 1 is given after the following section. 1. 2 Research Questions This dissertation aims to answer the question of how to treat adpositions in DM. In doing so it further explores the nature of the functional/lexical divide and the u niversality of morpheme classes. This dissertation also investigates the grammatical features proposed to frame content information in the structure of the grammar. By doing this, the question of what features can assimilate s into the grammar is rai sed. By justifying the inclusion of a [+p] feature for a proposed little p_ head, this work also explores the nature of the categorial heads n_,v_, and a_ to further answer the question of w hat exactly these basic morpheme types consist of. This is interes ting because it directly corresponds with the question of Universal Lexical Categories and the features of Universal Grammar. In summary, this dissertation aims to answer the following descriptive (1) and theoretical questions (2).


20 (1) Descriptive Research Qu estions a. Do adpositions form a natural class and how so? b. class? c. What features make up this class? d. Are some adp ositions truly lexical and others truly functional? e. How are adposit ions (form and function) created ? (2) Theoretical Research Questions f. How can adpositions be modeled in DM? g. Must the architecture of DM be radically changed for semi lexical forms? h. What features can categorize a node? i. Is there evidence for adpositional features categorizing lexical material? j. Are the features that categorize a Root node singular or a composite of several features? k. Can a more discrete inventory of combinatory categorial features be used to ex plain the contrary typological observations regarding the existence of universal categories? The remainder of Chapter 1 is organized as follows: Sectio n 1.3 introduces many of the problems faced when attempting to cat egorize the lexicon. Section 1.3 .1 exp lains how the lexical category dilemma shapes proposals made further in this work. Section 1.3.2 introduces the lexical versus functional distinction in language and shows how this is problematic. Section 1.4 shows how this is especially true for the categ or y adposition. Section 1.4 .1 show s how traditional methodology leads to classification iss ues for adpositions. Section 1.4 .2 discusses adpositions as semi lexical items. Section 1.5 begins the unifying description of adposit ions. Section 1.5 .1 provides th e attested syntactic distribution of adpositions and discusses the ir syntactic function (i.e. as c ase assigners) with regards to such distribution. Section 1.5.2 introduces adpositional stacking and discusses what this means for ideas regarding a dpositions and c ase. Section 1.6. introduces the semantic syntactic function of adpositions within the syntax and proposes that they can be unified via a common semantic syntactic feature [ F LINK ] . Section 1.6.1 specifically discusses this function with transitive a dpositions. Section


21 1.6 .2 then discusses additional meanings and features that are commonly added to the proposed [ F LINK ] feature. Section 1.7 introduces particle forms in English, noting the long standing connection between these for ms and prepositions. Section 1.7 .1 shows that what separates prepositions from particles is the fact that particles do not have a DP Ground. Section 1.7.2 then discusses the possibility of intransitive prepositions in contrast to anoth er type of particle. Section 1.7 .3 discuss es directional pa rticles and results. Section 1.7 .4 revisits stacking constructions, showing how particle form s can be integrated. Section 1.7 .5 concludes the discussion on adpositions forms by saying that both particle and prepositional forms can be unite d via the proposed [ F LINK ] feature. Finally a conclusion is given along with the organization of the remaining chapters. 1.3 The Lexical Category Dilemma Investigations of human language have persistently tried to describe, re describe, and account for wh at appear to be divisions in the lexicon (Sapir 1921; Robins 1952; Jespersen 1965: Chomsky 1970; Ross 1972; Jackendoff 1977; Hopper 1984; Emonds 1985; Abney 1987; Croft 1991; Ouhalla 1991; Hudson 2000; Co r ver and Riemsdijk 2001; Baker 2003; DeLancy 2005; M uysken 2008 among others ). These accounts of language generally divide linguistic items into lexical categories, 3 parts of speech or word/morpheme classes. Distinctions of this type can be found a s far back as 100 BC with Dionysius Thrax Tekhne Grammatike ó (Baker 2003:1). However when it comes to defining these lexical divisions, there seem 3 The term lexical category here should not be confused with the term lexical category when it is used to describe those items tha t are not functional. Here it would simply mean the categories of the lexicon.


22 to be as many disagreements as there are writings on the topic. That is despite obvious icon into (Hopper 1984: 703). 4 It seems that while there appear to be real differences between the units of the lexicon, we have failed to discover a consistent method to sati sfactorily standing problems in the study of lexical categories: how they are to be defined, and whether there is some set of 1.3.1 Response to the Lexical Category Dilemma The fact that lexical identity issues persist suggests that while something real has been discovered (i.e. one item is more or less similar to another), we are either overlooking something or assuming something incorrectly ab out the organization of human language and the categorization of its items . first question can only be provided via a particular stinctions can meaningfully transfer to another framework clearly depends on the assumptions that each framework is making. The superiority/validity of a featurally defined categorization system versus a contextually defined one versus a mixed system is le xical category dilemma. Thus 4 Classification issues have also resulted from concerns of bias from western grammatical tradition (Robins 1952; Jespersen 1965; Haspelmath 2007), a lack of focus on the dif ferences between categories (Baker 2003; Haspelmath 2007), the observation that class distinctions are actually gradable (Ross 1972), proposals for a semi lexical class of items (cf. Corver and Riemsdijk 2001), and simply because class distinctions have ye t to be clearly described or demonstrated (cf. Corver and Riemsdijk 2001; Baker 2003; Haspelmath 2007). With these concerns comes the suspicion that categories are merely the result ing real (Hopper 1984). Despite all these issues, however, linguists from opposite frameworks continue to pursue this topic with the belief that categorial questions are answerable and that such answers have important implications for understanding larger language phenomenon (Croft 1991; Corver and Riemsdijk 2001; Baker 2003; Muysken 2008).


23 as they are reported across the general literature. G iven the longevity of acknowledging and classifying items into categories, categorical distinctions will be taken to be an important realization of something definite abo ut language. It is also acknowledg ed that the labels for such categories (outside of the same defining framework) do not clearly entail a specific range of behavior . Thus w hile this work proposes a way to unify the category adposition (Section 1.5) , some of the conflicting information about adpositions is admittedly the result of different defining parameters , not a different understanding of the data itself . 1.3 .2 Functio nal and Lexical Categories In addition to the ind ividual, lexical classes mentioned in the previous section, the units of the lexicon are generally divided into two macro classes: a functional/grammatical class and a lexical/content class. The cross lingui stic observation has been that there are a group of morphemes that typically correspond with abstract/grammatical ideas/function and a group of morphemes that correspond with concrete ideas (Baker 2003; Muysken 2008). Those items that correspond with abstr act ideas are assigned to the functional class and those that correspond with concrete ideas are assigned to the lexical class. Evidence for this distinction has been found in practically every linguistic domain. That is it has been found by looking at pho nological, morpho phonological, lexical, morpho syntactic, syntactic, and discourse pragmatic behavioral differences in addition to semantic/meaning difference. Some of these are highlighted with Table (1 1 ).


24 Table 1 1. Differenti ating lexical and functional m aterial Domain: Test: Result: Indication: Phonology Stressed or Unstressed: Stressed Lexical Unstressed Functional Phonetic Weight: Heavy or Light/Polysyllabic or Monosyllabic Heavy/Poly Syl Lexical Light/Mono Syl F unctional Morpho phonological Natural Dependency: Free vs. Bound Free Lexical Bound Functional Optional Dependency: Contraction Lexical Can be bound Functional Lexical Membership Numbers: Large vs. Small Large Lexical Small Func tional New Members: Open vs. Closed Open Lexical Closed Functional Morpho syntactic Inflectional: Inflection vs. Non Inflection Can Inflect Lexical Functional Derivational: Forms a base for new words Yes Lexical No Functional S yntactic Position: Can occupy a subject, predicate, or object position Yes Lexical No Functional Selectional/Projectional: Can have a Complement/Specifier Yes Lexical No Functional Semantic Sense: Morpheme has a concrete sense Yes Lexical No Functional Discourse Use: Morpheme is used to reference a concrete sense Yes Lexical No Functional Table (1 1 ) provides a list of some common ways lexical material is distinguished from functional material from the perspective of different linguistic domains. In more recent works on the topic (Muysken 2008), the general status of an item is more likely to be scaled/determined via how that item behaves with regards to several parame ters (such as those from Table 1 1 ), meaning that some items are more p rototypically lexical or functional than others. 5 5 This is not necessarily helpful, however, for a Model like DM which makes an absolute distinction (cf. Chapter 2).


25 It should also be noted that the results from each domain are not equal. Certain domains are clearly peripheral to the distinction (phonology, morpho phonology, morpho syntax, and discourse ) 6 and represent tendencies, while the features of the other domains are more likely to be argued as the cause of the difference (cf. Chapter 8 to see the similarity within the grammaticalization framework) . In other words, no one argues that phonological, morpho phonolog ical, or morpho syntactic differences cause something to be functional or lexical but rather that such differences reflect an inherent difference. Moreover, if an item displays a lexical property in one domain, this does not require a lexical property to b e displayed in another domain because other language parameters may interfere with such manifestations. In many ways, even though this distinction is an old one, we do not fully understand the lexical/functional distinction. This point will prove relevant for the discussion of the behavior of adpositional items and how a particular model can view them . 1.4 The Category Adposition The purpose of this discussion is to show that the conclusions about the categorical status of adpositions heavily differ and th is is because the traditional methods of classification fail to provide consistent results. It should be clear from Section 1.3 that difficulties exist for determining individual categories and whether or not these categories can consistently be divided in to larger macro categories, functional/lexical categories. The adpositional category only intensifies this difficulty because, as mentioned in the introduction, adpositions on the surface do not behave as traditional lexical items or as traditional functio nal items. This 6 Those more involved with pragmatics and discourse may of course disagree.


26 thus causes them to be treated in a variety of ways : universal functional or universal lexical items or as a class of split items (also as sem i lexical items , cf. Section 1.4.2) . 7 As previously mentioned, many linguists from different appr oaches have treated adpositional elements as functional items. Some only go as far as to make language specific claims: Selkirk for one says that n ouns, verbs and adjectives constitute the class of lexical categories in English, fal l into the class of functional categories . However, others more strongly argue that adpositions are universally functional (Baker 2003; DeLancy 2005). Baker (2003) argues that the adpositional class is functional because it is a closed class, adp ositional items do not incorporate, and adpositional items do not parta ke in derivational morphology. Likewise Croft suspects that there are languages with no adpositions and that the category does (1991:144). In Croft (1991) and Baker (2003) universality is heavily associated with lexicality (i.e. lexical items are universal items). (19 91, 2005) system universally treats adpositions as a functional class. I based system , prepositions are treated as functional categories, being part of t he nominal extended projection. In this system, functional categories have a positi ve functional feature F (F1, F2, F3, etc. depending upon their place in the extended projection) whereas lexical categories (N and V ) have a negative or unvaluable ( [F0]) functional feature. P repositions are listed as F2 in the nominal extended projection, parallel to C in the verbal extended projection , a clearly functional position. 7 It may also be proposed that adpositions can only be categorized at the language specific level.


27 In contrast to the position above, i n some of the earliest generative work on categories, prepositions are treated as a syntactic category defined by negative nominal and ver bal features [ N, V] (Chomsky 1970, 1981) or a negative subject and positive object feature [ S,+O] (Jackendoff 1973, 1977). The binary [+/ N, +/ V] or [+/ S, +/ O ] feature matrix is thus proposed to be responsible for producing the four major lexical/syn tactic categories N, V, A, and P. In this way the category P is firmly grouped as a lexical category . 8 More recent proposals have also viewed adpositions as lexical items. Koopman (2000) for one clearly groups prepositions with the other traditional lexica l categories. ( Botwinik Rotem 2004; den Dikken 2006; Svenonius 2007, 2008) prepositions have been taken to project their own functional structure, a light p light v . Such proposal s align the category P with other major syntactic categories as Chomsky and Jackendoff proposed. Likewise, Muysken (2008), who finds fault with the arguments of Baker (2003), argues that it is better to universally view adpositions as not being a f unctiona l class because of the lexical behavior of adpositions in certain languages (i.e. they show a lexical like distribution and serve as the input for further morphological processes), because of particle verb incorporation as discussed in den Dikken (1995) an d because adpositions do not other clearly functional elements do (Muysken 2008: 240). 9 8 In Chomsky s account, however, prepositions were the last category added. 9 Assumingly , they do not provide the structure for Subjects, Verbs, and Direct Objects. However, it is uncertain what would be said about Indirect Objects.


28 In addition to the debate of whether the class is universally lexical or functional or i f the class is either lexical or functional at the language specific level, others have noted that the classification of adpositions is problematic at the language specific level. Adpositions are argued to be both functional and lexical in Dutch (Zwarts 19 95, 1997 ) and English (Deacon 2011). 10 the clearest example of a some prepositions are content words and some are FWs within the same language (2000: 17) . Given this, one could interestingly say that the split behavior seen at the language specific level (i.e. at least in some languages) mirrors the larger universal classification problem of unclearly being able to group items into functional or lexical categories . In summary, adpositions have eluded consistent treatment. One cause for this is that the authors of the various referenced works are relying on an indiscrete method of classification and the assumption that form+meaning are labeled/categorized before synta ctic integration. This method and assumption leads to confusion as can be seen when evaluating adpositions via the traditional parameters. 1.4 .1 Traditional Methods for Classifying Adpositions As outlined in Section 1.3 .1, the functional/lexical divide i s commonly justified by semantic, discourse pragmatic, phonological, morpho phonological, lexical, morpho syntactic, and syntactic behavioral differences. When the individual results of each method are compared, the conclusion is often inconsistent cross l inguistically. This section will show how differences in these domains fail to produce consistent results 10 This is also suggested for Mende, a Niger C o ngo La nguage, and Tikar, a Bantu Language (cf. Zwart 2005).


29 with the putative adpositional class. 11 For the sake of brevity, English examples are predominantly used with the argument being that if traditional me thodology fails for English it will also fail for many other languages. This argument help s set up the argument that while at first problematic, defining adpositions with the DM model provides more consistent results and perhaps enhances our understanding of adpositions and the types of items used as adpositions, as will be shown in Chapters 4 8 . A Semantic Definition: one semantic approach to defining linguistics items claims that items are organized according to whether or not they share a common point of meaning. As every grammatical description of a human language for describing the characteristic (1991: 62). Likewise Dixon argues t hat each From that locus other items may be judged as members or non members of that category (cf. Croft 1991). These prototypical senses thus serve as a semantic means for dividing categories and by ex tension the members of the lexical (concrete meaning) and functional categories (abstract meaning). Many so called English prepositions are claimed to have concrete meaning ( under, above, near while others are said to have abstract meanings or mere ly serve a function ( of, at, from, for, to, with and by ) . Accordingly Hudson argues that some Murcia and Larson Freeman point out the 11 This is done under the assumption that those items commonly referred to as adpositions are actually adpositions.


30 404). Ostensibly with in a semantic approach those with rich meaning would be considered lexical items and those with no or only abstract meaning would be considered fun ctional . In English, therefore, this approach splits the class of prepositions. S em antic meaning thus cannot be used alone to either cross or intra linguistically classify adpositions as a single class, much less a functional or lexical class. Discourse P ragmatic Approach: Like the semantic approach, a d iscourse pragmatic approach fails to consistently group adpositional items together . words (rather than prototypical meanin g categories) in situations to define the classes. For instance the nominal category is defined as that are general autonomous, discourse salient entities and the verbal category is defined as those words used to rt discourse (Hopper 1984: 726). Now while them from a discourse approach. as items that function as orientation m arkers in the discourse . In other words, akin to hey can be defined as those units that provi de orientation for discourse participants. However, this method would have to exclude items such as by [ AGENTIVE ] , of [ CASE , LINKER ] , for [ BENEFACTIVE ] , with [ INSTRUMENTAL ] fro m the list because their purpose is not clearly orientation al (assumingly defined by only space and time) . Another potential problem for defining adpositions in this manner is that this definiti on does not clearly distinguish adpositions


31 Of course some may then argue that actually some of these are members of the same category, but in doing so many differences would be undesirably glossed over (i.e. by other means, one form (a relational noun) is much more lexical than the other (a form such as at ). Assuming that verbs, nouns, and prepositions are lexical categories, items that prototypically function as entities, events, or orientation markers can be thought to be lexical. However, there is perhaps just as much reason to assume that adpositions are functional and thus all items that prototypically function as orientation marker s would be functional. Thus it seems unlikely that this perspective will solve the problem or best solve the problem. Phonetic Approach : While there has been nothing in language to suggest that certain phonetic features tend to correspond with certain cate g orical identities , linguistic descriptions have found a correlation between what is considered functional and what is Phonetic reductions are thought to cause or contribute to a functional it another word, causing it to be bound. Despite this, all traditionally described English prepositions are free words, as they can be stranded. Moreover, the existence of free functional prepositional items is not a surprising observation since many languages have both bound and free items that are semantically functional and bound and free items that are semantically lexical (i.e. bound s versus free s). Instead of solutions , problems generally arise when using phonological information to determine the status of adpositional items (i.e. the status of by in


32 comparison to in ) . epositions -for example, of, at, in, to, and by all qualify in of, at, in, to, and by are all monosyllabic and lack either a coda or onset. However, most, if not all, are going to semantically group by , the agent indicator, with the functional class while in may be considered functional (Hudson 2000) or lexical (Deacon 2011) as its meaning is arguably richer than the other mentioned forms (recall Celce Murcia and L arson Freeman 1999) . In other words, if the functional morphemes are phonetically reduced relative to the lexical ones in English, under should then be lexical while in would be functional. It is unclear, however, how the word under is any more concrete or semantically rich than in . Moreover, despite the fact that most functional material is not polysyllabic, phonetic reduction correlations like bounded correlations are also dependent on other language parameters. Phonetic reduction differences will be negl igible in languages were the concept of word is largely monosyllabic. In languages like Mandarin Chinese where words are largely monosyllabic and both putative functional and lexical forms are free, it is impossible to use morpho phonological information to determine lexical/functional category distinctions. This can be seen with (1). (1) rén zài kàndào le zhuàn person LOC library in see ASP biography In (1 ) the forms zài LOC le ASP would semantically constitute the func tional class and rén tú kàn zhuàn the lexical class. The forms dào are more similar to


33 under and in in English and thus a little more controversial to classify from a semantic perspective . Thus despite clear semantic divisions between some members of the adpositional class in both languages, all adpositional like forms appear to be free. A dpositions in Mandarin and English are not phonetically reduced as a class relative to items belonging to the less disputed lexical classes. Thus being a word does not clearly indicate that a morpheme is lexical or functional. 12 The correlation is interesting but non determining for any unit of the lexicon, especially adpositions. Lexical Approach : Another traditional way to support the lexical and functional division is by examining lexical inventory differences. This of course assumes that one has already established what is and is not lexical (as was the case with t he pragmatic approach) . Functional classes are small compared to the vast number of items found in the lexical/content classes. It is thus impossible to determine how many verbs and nouns a language has while it is relatively easy to count the number of it ems belonging to a given functional class. about seventy clear single 13 This number though is controversial for the reason being argued here: a dpositions are difficult to define and 0 (2005 :187). 12 The notion of a word is complicated and often differs cross linguistically. 13 The nu mber of forms used as adpositions in English is not tallied in this work since the category is determined to be like that of a verb and noun (cf. Chapters 4, 5, 6 and 7).


34 Dutch also has a large and productive adposition ca tegory (Muysken 2008). Other languages though have a very small number of adpositions: Yoruba and Igbo (Croft 1991: 144). Modern Lhasa Tibetan is said to have exactly five postpositions (DeLancy 2 005 mixed results. Like the number of distinction, t he open/closed class distinction provides inconsistent results for adpositions. 14 English prepositions form a relatively (2003: 304), but this c lass has allowed some new forms to arrive as loanwords: via , qua , pro , circa , vis a vis , per , and save (Hudson 2000: 19). Likewise, Kortmann and König point out that acknowledged that the class of prepositions in a language is by no means a closed as they discuss additional entries to the Germanic languages (1992: 671). I f the language is Igbo, however, with only one or two adpositional like members, and if those adpositional items form a class then this class must be a closed class. Given these lexical differences, some languages have either a closed group of adpositions or no adpositions while other languages have a class of adpositional like items that either form a large functional class or a small lexical class that is sort of c losed. Therefore neither t he class size distinction nor the open/ closed distinction clearly indicate the unive r sal status of adpositions because this putative category behav es differently along the s e parameter s in different languages. 14 Compare this method with some other putative lexical categories. There are languag es where a semantically defined adjective class is open (English and Spanish) and others (Mohawk and Yoruba) where it is closed (Baker 2003; Dixon 2004). In such cases no one makes generalization that adjectives should now be treated as a functional class.


35 Morphological Diff erences : Morpho syntactic tests are another way to try to determine lexical or functional membership. Morphological theories tend to agree that lexical word formation begins with a Root item that may combine with other affixes. Functional items on the oth er hand do not tend to have independent inflectional affix es or derivational affixes. I f a language has agreement, it should appear on lexical items or within a lexical projection within the syntax. This can be seen with (2). (2) a. Three dog s ran in the yar d. b. We jogg ed around the park yesterday. c. He is the tall est member on the team. In (2a) the lexical item dog inflects for number in the nominal position. In (2b) the lexical item jog inflects for tense in the verbal position, and in (2c) tall inflects for the superlative in the adjectival position. However, the adpositions in (2) ( in , around , and on ) do not inflect for anything. This could be a feature of English or indicate something more universal as it is generally thought that pre positions do not show agreement , 15 suggesting that they are universally functional items by this parameter . 16 It is also commonly thought that adpositions do not partake in derivational processes. O ne of the reasons that ad 3) proposed univers al categories (N, V, and A) is that these forms are not used derivationally to form new nouns, verbs, and adjectives . This conclusion, however, does not really appear to be true as can be seen with (3). 15 16 Postpositions do show agreement in some languages (i.e. German). However, such agreement may be argued to be comparable to the type of agreement seen between no uns and their determiners in many of the Romance languages.


36 (3) a. God's Aboveness and Closeness 17 ( Savoring Christ in Southern California by Dustin Smetona ) b. The underly are not without hope. c. Caffeine is an upper and t ime is a downer . d. There goes the arrogant and uppish fool. The purpose of (3) is simply to sh ow that the relationship between prepositions and derivational morphology is not as straightforward as Baker (2003) claims. 18 In fact adpositions (or the common form and meaning) do appear to be able to form the base for other word der ivation processes (cf. Chapter s 4 , 5, and 6). Given this discussion, inflectional morphological data suggests English prepositions are functional while derivational morphological data might suggest otherwise. Syntactic Differences: Syntactic distribution can also be used to determine categorical identity and the functional and lexical status of items. 19 Traditional lexical categories can be defined by positions within the syntax which are often identified by the presence of types of agreement features. Items that cannot or do not appear in these morpho syntactically defined domains are functional items. It is therefore interesting that a prepositional form will be inflected if it is used in a nomin al, verbal, or adjectival domain (4). 17 From: aboveness and closeness.html 18 It may be objec ted that similar uses could be found for other free word functional items such as pronominals if they are being discussed in terms of meta discourse. It is, however, here noted that the use of adpositional forms as the base for other derivational processes goes beyond such specialized usage as these derivations are much more a part of the natural language, obtaining dictionary entries in many cases. 19 Admittedly this method will be heavily utilized in other chapters, but in doing so meaning and morpholo gical agreement will also be consulted.


37 (4) a. The hungry kid downed the last bite of his sandwich. b. How many downs do we have left for this possession? 20 c. He is the downest gangster I ever met. 21 Example (4) complicates the claim that prepositions are functional because the same form that operates as a preposition also operates in other lexical domains, suggesting that these items are lexical. However, not all forms can be used in such a domain (i.e. forms such as with , by , and at ). 22 Therefore, syntactic distribution also cannot consistently define English prepositions as either a functional or lexical class. Looking specifically at the criteria for differentiating lexical and functional material, as discussed in Section 1.3.1, the functional/lexical nature of adpositions cannot be determined. At this point, one may argue that adpos itions need to be split as a category as mentioned in Section 1.4 ( cf. Hudson 2000; Zwart 2005) . T his problem , h owever , may have more to do with the traditional methodology for classification than the reality of categories themselves. Sometimes the data ha s driven the definition while at other times perhaps the definition has driven the data. Differences, moreover, at each level (semantic, phonological, morphological, syntactic etc ) , as mentioned in Section 1.3.1, do not necessarily indicate the same thing . Many of these differences are not the cause for such a distinction but rather the result of the distinction operating within a specific language. 20 Downs refers to possessions in American Football 21 Young %&*$ 22 Examples like (4) where a prepositional form is used elsewhere will be discussed in further detail in Chapter 4 where forms in both English and Mandarin Chinese can be seen operating in multiple domains.


38 The methodology for categorization is imperfect and these imperfections become magnified with adpositions. These issues can thus be suspected as the cause for the different positions taken by linguists as discussed in Section 1.4. 1.4 .2 Adpositions and Semi L exicality Even though evidence can be gleaned from all the approaches, as seen in the previous section , the traditional methodology for classifying adpositions fails to produce consistent , conclusive results. Thus many linguists have taken different sides on the matter, picking certain behaviors to be s true nature. On the o ther hand, others have abandoned the strict division, calling these items semi lexical. The observation that the functional/l exical distinction is insufficient to characterize all s yntactic elements is an old one . Ross (1972) instead proposed a continuum where category boundaries overlap by matters of degree. Given the mixed results from the traditional methodology for distinguishing lexical and functional material, it is not surprising that adpositions are labelled as a mixed class (Hudson 2000). Likewise it is unsurprising that adpositions are also described in some languages as being semi lexical or members of a semi lexical class 23 (Zeller 2001; Co r ver and Riemsdijk 2001; Mardale 2011). A semi lexical class, however, is not the same as a split class. A split word class seems to indicate that some members are truly functional while others are truly lexical. Instead a semi lexical class could be thought of as a group of items that occupy both 23 It is unclear if the semi lexical status is intended for all adpositional like forms.


39 functional and lexical domains. This description is problematic, however, because there turns out to be no consistent definition of semi lexicality in the literature. In s ome accounts (Emonds 1985) semi lexicality refer s to categories that have more semantic content than purely functional elements, but less than purel y lexical ones (cf. Butt and Geuder (2001) for a similar notion) . Others (van Riemsdijk 1998; features. Some researchers, such as Cardinaletti and Giusti (2001), reject the idea that designation. For them, semi lexical heads are lexical categories merged as functional heads in the syntax. Likewise , Haider (2001) seems to suggest that the term prototypical structural or lexical heads. From this it can be seen that the label semi lexical does not in and of itself make definitive claims about what semi lexi cal items consist of or how they interact with other items. As it should be clear, the category adposition is a problem that gets treated rather differently in different works and within different frameworks. To summarize, if adpositions form a lexical ca tegory, the category is probably not universal (Croft 1991; Baker 2003; DeLancy 2005). On the other hand, if adpositions are functional then they represent a rather different type of functional item (Muysken 2008). If adpositions are a split word class the n they are probably not really a word class and if they are semi lexical, it is unclear what that would mean. Major questions therefore remain regarding the natural class status of adpositions, their lexical/grammatical status, and their status as universa l or potentially universal items.


40 1.5 Unifying Adpositional Elements As outlined in Section 1.4 , the category adposition is a misfit category. Treating adpositions as a traditional single class is not readily supported by the lexical/functional distinctio n as adpositions have both lexical and functional properties. Moreover, there is difficulty in deciding which items belong to this class because like the traditional defining criteria for any lexical class, what exactly constitutes an adposition may rely o n inconsistent measures or tendencies. Adpositions are often described as elements that have spatio temporal meaning (DeLancy 2005:187). From this, the most basic meaning could be argued to be location . From a prototypical semantic viewpoint, this sense w ould most likely form the basis for this class. Howev er, as introduced in Section 1.4 , the split nature of adpositions would mean that many adpositional forms are actually members of a different class. If we wish to continue to label elements such as for [benefactive] , by [agentive] , and with [instrumental] as adpositions, there needs to be a better or more inclusive way to define this class, as these ideas are neither spatial nor temporal. This description would also create many homonymous forms (5 ). (5) a. John is at school. b. John is good at math. In (5a) the form at can be said to have a vague literal locational meaning, but this meaning is absent from (5b). meaning. However, classifying th e form at in (5a) and (5b) as homonyms seems incorrect because the function of each remains the same despite the non literal locational meaning in (5b). In this way meaning cannot be used to group these items. It would to provide a comprehensive and clear


41 m uch with regards to meaning (DeL ancy 2005: 187). There is thus as much reason to classify adpositional elements differently as ther e is to classify them together via the lexical semantic parameter . The following sections, however, will discuss the unifying function of adpositions to show that these items may be treated in a similar fashion if these items are defined by smaller discret e features and feature classes. 1.5 .1 Syntactic Distribution and Function The term adposition accounts for prepositions, postpositions as well as the less commonly reported ambiposition and circumposition subcategories (Hagege 2010). Ad positions can occur to the left ( prepositions ), the right (postpositions), on both sides (circumpositions), or either on the left or right side (ambipositions) of the DP they are said to govern (i.e. the DP Ground to be discussed in Section 1.6 ). Their distribution is simila r to affixes minus the infix. function marker license a DP via c ase assign ment ( Hagege 2010: 57). dps and case markers all serve the (2010:37). Likewise A sbury et al. (2006) argue that case markers among other things are all members of the Category P , suggesting that c ase is a defining feature of the adpositional class ( i.e. Category P). Accordingly, c ase assigning prepositions, postpositions, circumpositio ns and ambipositions can be found. One example of p r eposition assigning structural c ase can be seen with Mandarin Chinese (6) .


42 (6) *( ) zài kàndào le I LOC library see ASP book In (6) the form zài syntactically licenses the NP position as its absence renders the sentence ungrammatical. Likewise the English translation, I saw a book at the library , is ungra mmatical if the locative preposition at is removed: * I saw a book the library . One may here argue instead that this is not the result of a syntactic phenomenon but rather a semantic/conceptual incorporation problem. Further data, however, show that these a dpositions are not mandated for their putative semantic content. In addition to prepositional forms, Mandarin Chinese has postpositional items that convey meaning that should allow the prepositional form to be absent if the prepositional form is only need ed to semantically incorporate another DP into the larger structure (7). (7) * kàndào le I library in see ASP book Example (7) shows that one cannot introduce the adjunct noun, with just the postpositional item despite the fact that the postpositional form states the location of the book. In fact provides this information more specifically than zài does by itself in (6) . This suggests that zài is functionally serving a structural requirement. Moreover, the fact that the post position cannot serve this function is language specific. In Japanese postpositions operate similarly to Mandarin prepositions (8) .


43 (8) John wa teeb uru *(ni) aru hako o mimashita. John TOP table LOC BE box ACC see. POL . PST ( on ) In (8) the form ni LOC is required for the adjunct NP teeburu similar to the Mandarin form zài in that its locational semantic value is not precise. The form ni LOC merely gives the position of the box as being that of the table. Thus, similar to what was seen with Mandarin Chinese, another relational item can be added to make the locatio n more precise (9). (9) John wa teeburu no ue *(ni) aru hako o mimashita. John TOP table GEN top LOC BE box ACC see. POL . PST In (9) t he relational form ue s emantically specifies the loc ation of the box. The postposition ni LOC should not be needed because the DP could just as easily be related by the more specific meaning of ue without ni . This shows that ni LOC as an adpositional item is onl y fulfilling a grammatical /structural obligation. This is comparable to the English adposition at (10). (10) a. *I location library saw a book. b. *I saw a book location library. The sentences in (10) are easy to understand but remain ungr ammatical. Thus e xamples (6 10) serve to show that some adpositional items have a grammatical function in addition to a possible conceptual/semantic one. Circumpositions are another attestation of adpositional forms. In the literature, a prepositional form followed by a D P and another relational form are often referred to as circumpositions. However, in other accounts (Hagege 2010) the term is reserved for cases where both a prepositional and postpositional form is required to license a DP,


44 which appear to be rare. Hagege (2010:115) gives an exampl e of a possible candidate from Afrikaans (11) . (11) Afrikaans ( Suelmann 1 999:63) sy gann * (na) die stasie *(toe) 3sg.f go to the station to In the case of example (11 ), the same adpositional information is being repeated before and after the licensed DP. The forms na and toe semantically place the subject on a Path t hat ends at a station. The sen tence is ungrammatical if one or both adpositional forms are omitted. It thus might seem as if the c ase assigner undergoes some kind of fission over its derivation. Hagege (2010) , though , cautions one from absolutely accept ing (11) as a true circumposition because na and toe have other independent uses in Afrikaans. Fu r thermore, some have described some adpositional arrangements in English as circumpositions (12 ). (12) English ( Rooryck 1996:228 ) a. D ecisions were made ( straight) from the t op down . b. A partments can be found ( right) from the third floor up. In the case of (12 a,b) the function of the two relational forms is not the same as was the case in the Afrikaans. Moreover, the adpositional form to the right of the DP i s not required. Thus this is a case of something different as both forms cannot be syntactically justified as case assigners. I t has a lso be en argued that particular adpositional forms are ambiposition s , appearing either to the left or right of their DP, but not in both positions a t once (13). (13) Homeric Greek ( Hagege: 2010:116 117). a. apò hês alókhoio from poss.3sg.gen wife.gen


45 b. ápo from Example (13) shows that the adpositional form apò the licensed DP. Moreover (13 ) is interesting because i t would seem to suggest that a c ase assigner can be left or right adjoined in the same langua ge (i.e. assuming that it is a c ase assigner to begin with) . I t is unclear though if is a functional adposition at all. Not only is c ase represe nted o n the noun itself, the type of c ase on the noun (genitive case) is different than what is standardly associated with adpositions meaning c ase affix cou ld be the result of a separate c ase marking system (cf. Chapter 8 for a historical discussion of co occurring case markers an d adpositions) while apo is representing another grammatical feature or simply providing additional meaning as was seen with the Mandarin Chinese postposition example. 24 This section shows that adpositions and case markers can serve a similar functi on: syntactically licensing an addition al DP. However, given that adpositional like items may co occur with syntactically required adpositional items and case marking systems (in these cases providing addition relational meaning), it is best to not define this class as elements that assign case. While true circumpositions may be instances of c ase fission , 25 in Mandarin Chinese it was seen that only the prepositional item was structurally necessary. 26 Thus, while many adpositional items can assign c ase, the 24 It has been noted (see Hagege 2010) that if a language has both adpositional items and case markers that the adpositional marker will correspond with a more concrete notion. 25 In such cases both morphemes would essentially be performing the same function. 26 Compare this with the observation in Zwart that an a ll pu functional adposition will be a preposition if the language has both pre and postpositions (2005:691).


46 de fining f eature of adpositions is not a c ase feature. 27 This becomes explicitly clear when one views instances of adpositional stacking. 1.5 .2 Adpositional Stacking In the previous section it was seen that singular adpositions can appear as prepositions, p ostpositions, and ambipositions and that two adpositional forms can appear as circumpositions. In addition to circumpositions, multiple prepositions can co occur for a single DP. Ad positional like forms can be complex or stacked on top of each other , (14) . (14) John moved the box from under the table. In (14 ) only one new DP is introduced but yet there are two forms that generally get labeled as pre positions. Briefly returning to Japanese, similar facts are seen in (15) . (15) John wa teebu ru no shita kara hak o o ugokashita. John top table gen under source box acc move.pst hat was under the table to some The forms in English (14) and Japanese (15) appear to be mirror images of one another , involving two adjacent re lational forms . 28 In addition to this, however, the stacking or co occurrence of adpositional or adpositional forms can involve many more than two elements. Adpositional stacking seems quite prolific in Germanic languages. As mentioned i n Section 1.4 , Koopm an (2000), den Dikken (2006) and Svenonius ( 2004, 2006, and 2008 ) have argued that the category P (adpositional class) has functional projections 27 This is if one wants to continue to include non Case assigning adpositional like items in the class adposition. 28 I n Japanese , however, the s econd form in is often viewed as bei ng a type of noun and not an adposition . This difference will not be addressed here.


47 like the other lexical categories. Moreover, this projection h as been used to account for the types of stackin g phenomena discussed here . Svenonius for one (2008) , working within a cartographic approach to syntax, presents the following extended projection for English. Figure 1 1. Extended p rojection of P . To generate and order Fi gure (1 1 ), Svenonius (2008) places many of the English spatial prepositional forms into macro categories (Projective (Place), Bounded, Extended 29 or Particle Ps) where a special Path category dominates either a Projective or Bounded type and where Directio nal Particles may modify either Projective, Extended, and Bounded types. Each preposition al category links to a position in the extended projection and each position is defined by features such as [Source Path], [Goal Path] [D ir ection], [Path] and [Locatio n] among others. 30 These features are then 29 Extended prepositions are discusses as being the general Path prepositions while the forms to and from form a subset (i.e. canonical Source and Goal Path) 30 It should also be noted that many of the prepositional forms belong to more than one category type (i.e. those forms categorized as Particle forms may also be used as Extended and Projective forms). Whether


48 represented by appropriate prepositional items (i.e. those that have been categorized accordingly). Moreover, like the observa tion from Section 1.5.1 , not every P hea d is responsible for assigning c ase. In the arra ngement of Figure (1 1 ), c ase is shown to be assigned via a KP directly dominating the DP (not by one of the other heads eve n though in other arrangements c ase is not represented with of ). It should be clear that if a c ase feature defines adpositional ite ms, in situations where two or more forms are present, the category of the additional adpositional like forms is unclear . In cases of stacking, if the individual items of the projection are to be cons idered adpositional items then c ase assignment cannot b e the primary function for all the items/positions because c ase assignment should only need to be performed by one item/position as suggested by Figure (1 1 ) . 31 Thus calling all adpositional forms c ase assigners is invalid as it is clear that c ase assignmen t is ju st one of many potential roles. Instead it will be suggested that each adpositional form always performs a certain semantic syntactic function and that this function m ay sometimes be bundled with a c ase feature. 1.6 Semantic Syntactic Function: An Adpositional Unifying Feature DeL functions is thus one cause for the categorical problem s already discussed. Accordingly, it is argued here that the most unifying function of all adpositional like forms (to be termed P items) is that they can link an entity (a Figure) to another entity (a Ground) or these duplicitous forms come from a single source or multiple sources does not, however, create a serious issue in his discussion. 31 Adpositional stacking will be discusse d further in Chapter 5 .


49 to a spatio temporal vector (i.e. a positio n in time or space, a path through time or space, or a direction of motion in space and perhaps time). They may also link an entity to another entity through more abstract grammatical relations (i.e. [ G ENITIVE ], [ B ENEFACTIVE ], [ A GENTIVE ], [ I NSTRUMENTAL ], [ C OMMITATIVE ], [ P URPOSIVE features). Thus it is proposed that P items correspond with what will be called a Figure Link or [ F LINK ] feature. This feature may be bundled with the aforementioned features, creating different types of spatial, tempo ral, and grammatical adpositions. Furthermore, the claim here is that the fundamental difference between canonical adpositions and other P items (i.e. particles and perhaps some types of intransitive prepositions) is that these adpositions link a DP Figur e to a DP Ground while other P items do not. This is explained in the following sections. 1.6 .1 Semantic Function: Transitive Adpositions The common identifying semantic function of transitive adpositions is that they relate/connect two entities or two co nstituents (Zeller 2001). In English Svenonius ( 2007, 2008) de scribes spatial prepositions as morphemes that relate a Figure to a Ground. The Figure or Figure DP is an obje DP Ground the reference landmark for t he location of the Figure In Langacker (1987) the DP Ground is referred to as the landmark. In other words, the prepositional element gives the location of the Figure DP in relation to the Gr ound DP . This visually demonstrated with Figure ( 1 2 ).


50 Figure 1 2. Semantic f unction of t ransitive As Figure ( 1 2) shows, the category P relates or links an unknown position with a known one. The Figure DP in such cases can be thought of as the external argume nt and the Ground DP the internal argument (Zeller 2001) . The meaning of the adpositional form itself merely specifies how P links or relates these two entities. To account for more data, the semantic syntactic function can be extended to ite ms relating two conceptual entities even when these entities are verbally absent (either being elided or only implied as syntactic entities), as seen with (16). (16) [John Fig ] [wrote ] [ [in Rel [the house Gr ]]. In (16 ) the function of in could simply be describ ed as an element that relates the two DPs in accordance with its own meaning: John , the DP Figure, is the external argument and the house , the DP Ground, is internal argument as related by in . However, there is ambiguity, as (16) could mean that John wrote while he was in the house or that he often logically entails the first but the meaning is different: versus John was in the house writing. With the first inte rpretation, the unspoken object of wrote is acting as the DP figure. 32 In this way it appears that the adposition is capable of not only relating 32 There may be a third interpretation where the writing is actually carved into the house. In such a case the implied object of wrote would also be serving as the DP Figure.


51 the Figures verbalized in the syntax but also implied Figures, as we can assume that there is a writing if John indeed wrote. Likewise it can be seen that the DP Ground can also be elided or implied as seen in (17). (17) He turned off (the road) at second street. In (17) the DP ground is not pronounced but it is logically implied (i.e. to turn off entails exiting som ething . Moreover, this should not be confused with the phonologically identical verb+particle construction turn off Section 1.7) . Considering the facts of (16) and (17), adpositions can still be unified as items t hat relate Figures to Grounds with the understanding that a Figure and DP Ground need only be implied in some languages. 33 However, the range of meanings that this function needs to cover must be expanded to account for other adpositional forms. 1.6 .2 Sema ntic Function: Additional Meanings In addition to being able to spatially relate DP Figures to DP Grounds (i.e. providing a Locational/Place relation), adpositions also commonly relate these entities across a Path or in a Direction (as discussed with Sveno nius 2007, 2008) or via a Temporal extension of Place, Path and perhaps Directional frames (18) . (18) a. The man is in his house. b. The man ran in his house/up his ladder/ down his ladder. c. The man flew over/through/across his house d. The man was born in March. e. The dissertation will be finished in the future. f. We learned over/through the years 33 The ability to omit or ellipse the DP Ground is language specific as it will be shown in Chapter 3 to be impossible in Mandarin Chinese.


52 In (18a) the preposition in locationally relates the Figure NP, the man, with the DP Ground, his house. The function of this adpo sition is to statically place the DP Figure in relation to the DP Ground. Svenonius (2008) describes this as a Place preposition. It is this semantic function that seems most basic to the category. In (18b) there are two possible interpretations. The first is like (18a) where the Place of the running is the house. The second interpretation is that the man ran into the house. 34 With the second interpretation, in directionally relates the DP Figure, the man, with the DP Ground, his house. Furthermore (18c) is similar to the function of (18a), but in this case the P item positions the DP Figure along several connected points, creating a Path. This accounts for the three major spatial functions of transitive adpositions: [ L OCATION ], [ D IRECTION ], [ P ATH ]. On the ot her hand, (18d) temporally relates the DP Figure, the man, with the DP Ground, the month of March. Here the meaning also pertains to location or place, but instead of a physical location it relates a location in time. Likewise it is possible that in (18e) has a directional temporal meaning (i.e. towards the future). In the (18f) the P item, moreover, has a Temporal Path meaning. Thus in English it should be clear that at least Locational and Path features may be bundled with Temporal ones. Adpositions may also abstractly extend their potential Spatial and Temporal meanings when relating arguments (19). (19) a. The man is in a coma. b. The man stumbles through life. c. The trains arrive on time. In (19a) the man is not literally inside a com a. Rather the preposition abstractly relates his position with regards to what a coma means. In (19b) the preposition abstractly 34 Some have reported o


53 relates the position of the man along a path, the path of life, which of course is not a literal path. Finally (19c) abstractly relates the temporal location of the train as it would be impossible to literally be on time. In addition to the extension of Place, Path and Directional meanings to temporal frames and the extension of these meanings for abstract uses, there exists a se t of possibly unique Spatial features , [ SOURCE P ATH ] and [ GOAL P ATH ], which are demonstrated with (20). (20) Adpositions as Relators of Source and Goal. d. John saw a man from France. e. John left his belongings to his children. In (20) it can be see n that prepositions temporal features. It is unclear though if these features obtain the kind of abstract extension seen with ( 19). Furthermore, similar to [ SOURCE P ATH ] and [ GOAL P ATH ], there exists a set of items with features/meanings that also can be used to relat e a DP Figure to a DP Ground (21 ). (21) a. Mary bought this gift for Jane. [benefactive /target ] b. Mar y sailed for France. [ purposive/target] c. Mary sailed for hours . [durative] d . The edict was written by Mary [agentive], e. Mary hammered the Protestants with edicts . [instrumental] f. Mary stood with h . [comi tative] e. Lady Mary of England was a good Catholic. [genitive] In (21) the P items have nothing to do with spatial/temporal/or abstract extension of Place/Path/Directional features. Nevertheless, one can see that they function to lin k two entities or DPs (i.e. the function of for in (21a) is to connect a DP Figure the gift to a DP Ground Jane ) . Moreover, keeping with the observation that the DP Ground or DP


54 Figure does not need to be verbally expressed, the DP Ground for some of thes e forms can be dropped. (22) a. Would you like to come with (us)? b. He suddenly came to (his senses) c. ??While everyone is against them I am for (higher wages). In (22) it can be seen that the DP Ground for the adpositional elements with , to and perhaps for can be omitted if the DP Ground is already understood. All the examples discussed above can be said to have a DP Figure and Ground linked by the meaning of the prepositional element used. The meaning of the adposition is often spatio temporal , but they may also represe nt other grammatical relations. However, this description only gets one so far with regards to unifying a class of items commonly included in the Category P. The next section will discuss those forms that either do not have a DP Gr ound (Particles) and those that cannot license another DP but seem to semantically indicate a Ground (intransitive prepositions). 1.7 Particles and Prepositions Any discussion of English adpositions must address particles. In English particles and prepos itions are often argued to be members of the same class, category P (Asbury 2006; Svenonius 2007). The phonological form used to express particles is often the same as the phonological form used to express many prepositions. T erms such as P forms 98) or P words (Miller 2006) have been coined to capture this observation and the fact that particle preposition pairs often share a similar meaning (as done here with the label P item). Svenonius (2008) in a similar vein calls many of the adpositional for ms of English P elements , noting the large overlap between the same phonological form and different adpositional functions, including a particle domain within the extended prepositional phrase pr eviously discussed in Section 1.5.2 . However, the


55 category pa rticle is a lso controversial one as is the relationship between particles and prepositions. While t here is a plethora of literature describing complex verb constructions consisting of a verb followed by a particle (often referred to in the literature as p hrasal verbs, particle verbs, or verb particle constructions) and the difference between particles and preposition s Ramchand and Svenonius 20 02; Svenonius 2007 etc.), a general consensus on the numbe r and types of particles that exist as well as how to syntactically model phrasal verbs has not been reached. Nevertheless, the fundamental difference between transitive prepositions and particles appears to be the fact that particle forms lack a DP Groun d in comparison to prepositions which either overtly or implicitly have a Ground as discussed in Section 1.6 . 1.7.1 Particles and Figures The observation that particles are objectless appears to be a consistent defining feature of particles in contrast t o prepositions. Emonds ( 1972 ) calls particles intransitive prepositions while den Dikken defines them Case assigning, Both definitions, one semantic and one structural, capture the fac t that particle forms are not object taking or object licensing items. definition, instead of a [+/ c ase] distinction, the fact the particle forms lack an object can be further described as difference between DP Figure and DP Ground selection. Prepositions select a Ground (which may be omitted in certain circumstances) and a DP Figure


56 (which may b e only implied in certain circumstances) 35 while particles only select a DP Figure (Svenonius 2007 ). This difference can be seen with a variety of syntactic tests that can identify if a P item is operating as a particle or preposition . 36 Perhaps the most obvious or definitive (Cappelle 2004) test is the inversion test. For inversion, it is observed tha t the objects of phrasal verbs can optionally 37 precede the verbal particle (23 ) and generally must precede it when the object is pronoun (24 ) . 38 (23) Transitive Verbal Particle: Inversion Possible a. She blew up the house. b . She blew the house up. 39 c . The wolf blew down the house. d . The wolf blew the house down. (24) Transitive Verbal Particle: Inversion Obligatory a.*She blew up it. b . She blew it up. c.*He blew down it d . He blew it down. In (23) and (24 ) it can be seen that the DP b ehaves independently of the P item , allowing inversion or what is commonly ca lled discontinuous structure. This movement option will be taken to identify the P item as a particle . Here the internal object of the 35 It should be noted that never can both the DP Figure and DP Ground be only implied. 36 always render the same results. This is because most writers have already decided what a particle is and thus tests are only used to affirm their proposals. The issue of what is or is not a particle is also confounded by the fact that there are a variety of particle types. Our purpose here is not to describe the different types but rather to form the basis for which such descriptions could be made within a model like DM. 37 We do not claim that this movement is optional but rather point out that the motiv ation for such movement has not been clearly demonstrated causing it to appear optional. 38 Johnson (1991:594) 39 There are two particle interpretations for (18a): one is idiomatic (explode) and the other is resultative. These differences will not be analyzed here.


57 verb is the Figure for the particle. The sa me is not true, however, when the P item is in a prepositional domain (25) and (26 ). (25) Transitive Preposition: Inversion Impossible a. She ran up the house. b.*She ran the house up. c . The wolf ran down the house. d .*The wolf ran the house down . (26) a. She ran up it. b.*She ran it up. c. He ran down it. d .*He ran it down In (25) and (26 ) for the sentence to maintain it s original meaning , the object DP cannot be inverted . Likewi se, (26 a,c) show that it is grammatical for no movement to occur when the DP is pronominalized and that movement is ungrammatical , (26 b,d) , when trying to maintain the prepositional interpretation . 40 In these cases the Figure of the P item is the subject of the clause and its Ground is the internal object of the verb. Thus this test clearly shows that the same form is operating differently. It shows that structurally the DP is not tied to the P item as a particle and this reflects the different semantic relationship shared between the two elements. Another syntac tic test that identifies particles from prepositions is topicalization. A particle and DP cannot be topicalized together while a preposition an d DP can be fronted together (27 ). (27) Transitive Verbal Particle: Topicalization is impossible a. U p the house , t h e wolf blew . house/location of blowing is up the 40 Ungrammatical for the directional meaning (i.e. the direction of the blowing was down the house). If grammatical, the P item is being used a particle. The w olf may run the house down (let the house become dilapidated) but this reflects an idiomatic interpretation for a particle arrangement as the same meaning is available for ran down the house .


58 b . D own the house , t he wolf blew . house/ location of blowing is down the c. *Up the house, the wolf bl ew. d . * D own the house , t he wolf blew . As see n by the acceptable meanings (27 a, b) and unacceptable meanings (27 c,d), a prepositional phrase may be fronted but a particle belonging to the extended vP cannot move with the DP to the front, again showing that the preposition and Ground DP form a unit unlike the particle and potential Ground DP. That is, w hen the DP is fronted via Wh word movement , the part icle must remain with the vP (28 ). (28) W h movement without Particle 41 a. What did the wolf blo w up ? , it b . What did t he wolf blo w d own ? house , it The movement in (28) preserve the particle meaning (i.e. that meaning seen with structure s that al low for inversion). This, however, can also occur when the P element i s a preposition (29 ). (29) Wh movement without Preposition a. Where did the man co me from ? b. Where did the man travel to ? As can be seen with (29 ) a prepositional form does not have to move with its Ground. Nevertheless, it still relates the Figure to the fronted Ground. In (28) up and down do not directionally relate the wolf in any way. Thus the difference lies with the different semantic relationship between the fronted DP and part icle and prepositions. Structurally this surfaces as to whether the P item can occur at th e beginning of the clause as (27 ) shows they cannot when functioning as a particle. The reason for this is perhaps due to the stronger relationship between an interna l object (DP Ground) and e xternal object 41 Complicating this test is the fact that prepositions may als o be stranded, often preferably so.


59 (DP Figure). M oving the Ground is more likely to take the relating adposition along for the ride (as in pied piping constructions) . It is clear from this discussion that the difference between transitive prepositio ns and particles as designated by these tests is the fact that particles do not link to a DP Ground. This is what accounts for their different syntactic properties. Items that do not link to a DP Ground have no need to assign c ase. This observation, howev er, is further complicated by the possibility of intransitive prepositions. 1.7 .2 Semantic Function: Intransitive Adpositions In the literature surrounding adpositions, one encounter s another somewhat controvers ial classification, the intransitive adposit ion . This section discuss es some putative intransitive adpositional forms in comparison to particles which have just been defined as Groundless Ps. As seen in the pre vious section, Emonds (1972 ) referred to particles simply as intransitive prepositions. C appelle (2004) , however, argues that prepositions and particles are distinct categories. The purpose of his argument is to show that particles cannot be equated with prepositions (one reason being that not all prepositions can be particles and vice versa v ia the inversion test) 42 and thus should not be called prepositions and by extension intransitive prepositions. Nevertheless, the possi bility of intransitive prepositions is appealing because it aligns PP with VP, NP and AP ( all of which can be realized int ransitively ) (Capelle 2004). Capelle, however , provides no clear basis for distinguishing between intransitive prepositions and particles. One 42 This argument is only meaningful from a lexical perspective that groups syntactic identity with a form in isolation. Gaps in distribution are discussed in Chapter 3.


60 reason is because the dist inction between intransitivity and ellipsis is not clearly delineated in his account. Capelle (2004:6) says inside and under have intransitive uses, assumingly as in (30) . (30) Intransitive Preposition or Omission. a. John moved inside (something). b. John was pushed under (something). However, one may argue that the forms in (30 ) involve an implied semantic object as was seen in Section 1.6 , making them die . An argument is not made her e as it may be possible that inside and under are capable of both omitting their argument as in ( 30a) and (30b) and perhaps appearing in an environment where there is no implied DP ground (i.e. a position where inserting a DP Ground would be strange or ungrammatical) . 43 If so, t hese may be similar to some intransitive adpositions in German (31) . (31) Germ an Intransitive Prepositions (Reimsdijk 1990: 236) a. auf dem Burg oben on the mountain on top b. *dem Burg oben the mountain on top c. *oben dem Burg on top the mountain d. auf dem Burg unten on the mountain down e. * dem Burg unten the mountain down f. * unte n dem Burg down the mountain In (31 ) it can be seen that the forms oben and unter cannot license DP objects, making them more like a true intransitive form. These forms, moreover, are just like the post positio nal forms in M andarin, discussed in Section 1.5 , which also are intransitive from a structural perspective. H owever, it seems possible that in these cases the post position has the sam e DP Ground as the preposition. Even if they are not capable of structur ally 43 Such an environment, however, cannot be thought of here.


61 licensing a DP, 44 i t appears as if oben or unter semantically share a DP Ground with another prepositional form , (31a,d) , making these forms structurally intransitive but not semantically intransitive . In other cases, however, this is even less clear. As shown in Zeller (2001), putative intransitive prepositions also occu r after a verb in German (32 ). (32) Intransitive Use (Zeller 2001:4). Youri wohnt oben Youri lives above/upstairs In (32), the P item comes after a verb and since it cannot l icense a DP , as indicated in (31 ), it appear s as if this P item is merely linking a DP Figure to the meaning of the P item itself. 45 On the other hand, one may argue that the P item implies that Youri is above somethi ng and th at this something is the Ground . This would then be another case of a P item that is not structura lly capable of licensing a Ground even though semantically it implies one . 46 Lastly another type of intransitive adposition will be addressed. This t ype of intransitive preposition appears as a compounded form (33 ). (33) Left Headed Prepositions ( Boertien 1997: 689 690) 47 uphill, downhill, upstream, downstream, upstairs, downstairs, upwind, downwind, overhead, underfoot, underground, over seas, over land, offhand, off key, aboveboard, offshore, in shore, inhouse, on line, off line, on screen, onstage, off screen offstage, undercover, overnight, overboard, indoors, outdoors, between decks, belowdecks, af ter hours, after noons, 44 As further discussed in Chapter 5, this can be explained via the absence of a [ CASE ] feature. 45 In such cases the P item appears to only have a Place or Path meaning. 46 The correct analysis here might be that this form is a Place form that lac ks a Ground. Canonical particles are directional (also lacking a Ground). 47 Many of these forms mentioned by Boertien may involve a particle form incorporating with a verbal


62 The forms in (33 ) are argued to be a type of intransitive preposition because the prepositional object has incorporated with the preposition, absorbing its theta role (Boertien 1997: 690 ) . T he P item is thus incapa ble of taking another object, (34 ). (34) The girl paddled her boat downstream *DP. In (34 ) it is not possible for the P item to take another object. In cases such as this, it is here argued that the meaning of the adpositional ele ment compounded with another item serves as the semantic Ground. In fact b oth a Path (the boat moves downstream) and Locational interpretation (downstream is where the boat is paddled) is possible, meaning that this derivation can involve two different types of P heads merging w ith the compounded form downstream . 48 This section briefly discussed the notion of intransitive preposition in contrast to particles. Many so called cases of intransitive prepositions are just cases of omission. With other instances, it was seen that the P item semantically links to a Ground even though it lack s the ability to license a DP. Furthermore, in cases like downstream the compounded form semantically acts as the Ground. The presence of the Ground thus differentiates these uses from particle uses a s defined in this work. 1.7 .3 Directional Particles and Results In Section 1.7.1 , it was claimed that the difference between particles and prepositions is that particles only select for a DP Figure while prepositions also select for a DP Ground or Ground . It was also shown that the Ground need s to only be implied in some langua ges. Furthermore, in Section 1.7.2 it was shown that the meaning of a 48 For both interpretations, downstream functions as the Ground for either potential DP Figure, The girl or the boat .


63 Place or Path adposition can serve as the Ground. This is different than when Particles link to non VP internal object DPs (i.e. linking to either a subject DP or the VP). When this happens, one may only obtain a directional meaning (3 5 ) . (35) He looked down/up/over In constructions like (3 5 ) there is no Ground, only the subject DP. It is proposed that these P items d irectionally link the subject DP to the specific type of motion given with the VP . In other cases where the verb is stative, the direction is stative, making it appear more like a result (3 6 ). (36) a. The display is up b. The display stayed up c . The run ner is down d. The runner stayed down Examples like (3 6 ) are proposed to be exactly like (35), except that the verb is stative, giving the P items a stative interpretation. These also help to show that the P item is linking to the DP and is not just ad ding additional information to the v P. The types of P items in this section seem to also be differentiated by the fact that they involve no sense of Ground, making them as defined here a type of particle form. However, instead of linking to a verbal objec t, they are linking to the subject of the clause. 1.7 .4 Semantic Function and Stacking It was seen with Sec tion 1.5.2 that several adpositional items may be stacked. Thi s was another argument for why c ase is not the defining feature of adpositions. Howeve r, in this section it has been discussed that the function of adpositions is better understood by means of linking a DP Figure with a DP Ground and particles only linking


64 to a DP Figure or v P . This generalization is held when revisiting stacked constructio ns, (37 ). (37) The boulder rolled down from over up on the ledge. In (37 ) five P items occur together with only two overt DPs. As discussed in the previous sections, P items can link to implied or elided entities. This means that several interpretations are p ossible. This section will examine a few of these to show how the P items here link to Figures to Grounds or just Figures . In order to do this more clearly, however, the sentence will first be broken apart, beginning wi th only the subject and verb (38 ). (38) T he boulder rolled down . The interpretation of (3 8 ) is either of down operat ing as a directional particle, directionally linking the subject DP to the action of the Verb (i.e. t he boulder down wardly rolled ) or as directional preposition (i.e. the boulder r olled down something). 49 In that case the P item also has a Ground. This can now be compared with some of the addition al P items (39 ). (39) a. The boulder is [ up [ on [the ledge]] b. The boulder is [[ up on ] [the ledge]]. In (39 a), the P item on Places the DP Figure, the boulder, in proximity to the DP Ground, the ledge. On the other hand, the P item up either directionally positions the DP Fi gure relative to the speaker (39 a) , or forms a type of complex spatial relation up on or upon . This difference can be seen by putting a different emphasis on each P item. The P item in (39 a), could be considered a particle form as suggested in Svenonius 49 It seems perfectly possible for the following: the [boulder rolled down] [down the mountain].


65 (2008) and defined here, because it is seemingly only directionally relating the Figure, but if one interprets the P it em as relating the boulder to a position (i.e. up there) then perhaps it is a preposition. The difference is not a matter of concern here as all that is needed is to show that this item is linking a DP to a spati al/temporal vector. With the (39 b) up is act ing like a preposition as its DP Figure and Ground are the same as the P item on . These options can now be compared to possible interpretations after the P item over is added (40 ). (40) a. The boulder is [over [ up [on [the ledge]]]]. 50 b. The boulder is [ over up ] [on [the ledge]]]. c. The boulder is [ over [[ up on] [the ledge]]]. In (40 ) the P item over has the options that up did in (39 ). In (40 a) , the interpretation is that the DP Figure is in a direction that is first [over] then [up] relative to the speaker and connected to a DP Ground locationally with on . In (40 b) a complex direction [over up] relates the position of the Figure relative to the Ground and then the P item on places the Figure on the Ground. In (40 c) the Figure is in the direction of [over] and placed [up on] the Ground. Now the structure is ready for the P item from (41 ). (41) a. The boulder is [ from [ over [ up [ on [ the ledge ]]]]] b. The boulder is [ from [ over [[ up on ] [ the ledge ]]]] c. The boulder is [ from [[ over up ] [ on [ t he ledge ]]]] Since the P item fr om does not flexibly combine with the other P items as over , up and on can, it merely stack above the rest, marking the DP Ground as the origin of the DP Figure. This allows for the DP Figure to be placed on a Path that can agree with a verb of motion, returning us to the origina l clause that is re given as (42 ). 50 Some English speakers reported a strong preference for up over instead of over up . For the author, both sequences are fine.


66 (42) a. The boulder [rolled down (something)] [ from [ over [ up [ on [ the ledge ]]]]] b. The boulder [rolled down(something)] [ from [ over [[ up on ] [ the ledge ]]]] c. The boulder [rolled down(something)] [ from [[ over up ] [ on [ the ledge ]]]] d. The boulder [rolled downwards] [ from [ over [ up [ on [ the ledge ]]]]] e. The boulder [rolled downwards] [ from [ over [[ up on ] [ the ledge ]]]] f. The boulder [rolled downwards] [ from [[ over up ] [ on [ the ledge ]]]] As can be seen in (42 ) there are many possible interpretations when multiple P items adjoin. Nevertheless, the function of these forms is to spatially position a DP Figure either directionally, locationally, or along a path. 1.7 .5 Where Adpositions Stand The unifying semantic syntactic function of many adpositions is to relate something unknown (DP Figure) to something known (DP Ground) . In most cases where the adpositional element is claimed to be intran sitive, the DP Ground is somehow understood or incorporated into the meaning of the adposition itself. Thus the semantic function can still be stated as relate X to Y in the manner of Z where Z is the adposition. In the case of spatial adpositions, this ca n be static (i.e. place adpositions), dynamic, (path and directional adpositions), or temporal. Moreover, place, path, and temporal prepositions can be used non literally , creating what are often called ab stract adpositions (Djamouri et al. 2012). I n such cases a Figure is still linked to a Ground. Elements such as for [benefactive] , by [agentive] , and with [instrumental] merely represent specialized features that also link a Figure to a Ground. This appears to be the most consistent unifying analysis for t he putative class adposition. Accordingly particle forms are forms with similar meanings that only link to a DP figure. Thus adpositions and particle forms can be united via the proposed [ F LINK ] feature.


67 1.8 Concluding Remarks It should also be clear tha t the identity and nature of universal and or possible lexical categories has not been largely agreed upon within frameworks that consider this to be a possibility. When it comes to the issue of universality, the position of this work is that neither view is totally correct or incorrect. Unpredictability mixed with signs of systematicity suggests that smaller parameters are at work. Some of the aforementioned issues are seemingly insurmountable such as the ability to define or distinguish something without the use of the theoretical system that allows for a definitional analysis. However, others are not. The focus of one theoretical background may provide a different and more discrete analysis for this age old problem, helping to more clearly describe and d ifferentiate possible and or universal distinctions. In other words , more refined theoretical frameworks may enable us to better answer these class distinction issues by causing us to look at language in more detail. Thus maybe progress can be made with re gards to distinctions between the commonly accepted noun, verb, adjective classes as well as less accepted classes such as the adposition class. The remainder of this work is organized as follows. Chapter 2 discuss es the basic premises of Distributed Morp hology, showing how the model is helpful and showing how the traditional functional versus lexical problem with adpositions also appears to be a problem. Chapter 3 continues the discussion of DM, examining issues regarding the status of Root items in DM. It discusses how Root s are identified, how they are derived, what features compose a Root in List A , what features accompany a Root in List B and how they get interpreted. Chapter 4 will give a detailed account of functional and lexic al adpositions in English and Mandarin. Chapter 5 models adpositions in


68 Distr ibuted Morphology. Chapter 6 will present evidence for a categorial p_ head. Chapter 7 will compare this p_ head with the other prop osed categorial heads. Chapter 8 will give a h istorical account of adpositional for ms in English. Finally Chapter 9 will conclude this discussion of adposit ions and Distributed Morphology , highlighting how this work advances our understanding and showing what areas still need f uture research .


69 CHAPTE R 2 DISTRIBUTED MORPHOLOGY: THE LEXICAL FUNCTIONAL DISTINCTION 2 .1 Distributed Morphology Distributed Morphology, DM, (Halle and Marantz 1993, 1994 ; Marantz 1997, 2001; Harley and Noyer 1999; Embick and Noyer 20 05; Acquaviva 2008; Siddiqi 2010 among others ), is a generative model where all word derivation occurs within the syntax . Its main claim is that there is only one generative component in the grammar: the syntax. In this way the system is more elegant, containing fewer assumptions because the syntax i s the only mechanism for building structure. It is accordingly claimed that models utilizing a generative lexicon rely more on assumption than evidence. I n the early onset of DM, he underlying suspicion [of lexicalism] was wrong and that t Thus t he framework of dead, deceased, This was and is the foundation of DM. The remainder of this Chapter is organized as follows. Se ction 2.2 introduces the features from the lexicon. Section 2.2.2 explains what the non generative components of the lexicon contain and how these components now relate to the syntactic derivation. Section 2.3 explains the mechanisms used to supply phonology to the derivation. Section 2.3.1 discusses Late Insertion. Section 2.3.2 distinguis hes Associative and Contextual F eatures. Section 2.3.3 explain s Underspecification . Section 2.3.4 then shows how these other concepts work in terms of the Subset Principle. Section 2.4.1 discusses l an d f morphemes with regards to a strict division between lexical and functional items. Section 2.4.2 introduces the features that categoriz e traditionally


70 lexical information in DM as well as the basic motivation for such features. Section 2.5 discusses the l and f morpheme distinction with the idea of semi lexical ity , pointing out how this is a problem. Section 2.5.1 discusses the l and f mo rph eme distinction and shows how adpositions are a problem for the Subset Principle. Section 2.5.2 then summarizes the nature of adpositions, semi lexicality and how this can be addressed in DM . 2.2 The Lexicon The basic claim of the Lexicon is that it is a list of what is irregular and unpredictable (Kiparsky 1983 ; Di Sciullo and Williams 1987). The phonological representation of morphemes and the meaning of r oot s are not predictable, and thus (a) list(s) is/are needed to store the particular features and sound/meaning correspondences that belong to a particular language. 1 Beyond this basic claim, however, is the hypothesis that the lexicon is a locus for word construction. This is the lexicalist position. This section will discuss this position and show t the l exicon makes fewer assumptions. 2.2.1 The Traditional Lexicon Because there is no claim that all phonological and word formation processes are structure/meaning connections are derived in the lexicon, while other aspects of phonology and other aspects of structure/meaning relations are derived in (and after) the syntax . This division is generally described in terms of pre and 1 These items will of course differ cross linguistically as well as between individual speakers of the same language, making the lexicon a major source for language variation.


71 post syntactic phonological processes. It is this aspect of the lexicon that is fundamentally rejected in DM because of the lack of evidence to suggest a special generative component to grammar that is outside of the syntax. As Marantz argues (1997; 2001) , evidence for an independent component that produces lexical phonological rules, special meanings, and special meaning structure correspondences has not been demonstrated while counterevidence for these three domains is not difficult to produce. The remai nder of this section will recapitulate and add to this argument in order to demonstrate the theoretical advantage of DM. Lexical phonology has explored the possibility that some phonological rules are particular to lexical word derivation processes. Thes Ordering Hypothesis, which models the observation that some affixes trigger stress changes while others do not. Lexical phonological rules are deemed to be special to the lexicon in contrast to those rules that apply acro ss the board (cf. Kiparsky 1983 ). cases where a word or lexical item is not a large enough unit to explain a phonological process that would otherwise be classified as le xical. Hayes is forced to create two categories of post lexical rul quite muddy with regards to how these types of phonological processes justify lexicon specific phonology. If a phrase can look like a lexical item because it is subject to lexical phonological rules then perhaps lexical phonological rules are not actually indicative of


72 a special lexical domain. As Marantz (1997) argues, this phenomenon can just as easily be explained post syntactically in a la te insertion model such as DM. In addition to the fact that le vel ordering rules do not separate words from phrases, it is also clear that a phonological word is not the same as a lexical item (even though a lexicalist account suggests otherwise) . One might expect the two notions to better correspond if phonological information coexisted with morpho syntactic features in the lexicon. This mismatch can be demonstrated with a simple dialect joke about how different American speakers of Engli sh differently reduce the number of syllables that one would need to ask the sa m e question in Great Britain (3 ). (3) a. Have you eaten ? b. Did you eat? c. Jeat? In (3 c) the entire clause is reduced to a phonological word. However, we would not want to say that this clause c onsists of a single lexical word or a single syntactic word because of all the syntactic evidence to the contrary. What constitutes a phonological word is quite different than these other notions. This is what Marantz is getting at when with phonological words? Phonological word divisions do not tell us what the atoms of syntax are and since lexical items, from a lexicalist perspective, are argued to be the atoms of syntax, the mismatch at the surface is odd . That is unless phonology merely represents the structure that has been produced, as is claimed in DM , and thus may do so differently depending on other factors which may be best modeled in a model of competing constraints such as Optimality Theory .


73 In addition to the fact the phonological evidence does not indicate a separate generative component, the special meaning of derived words is no greater or different than the meanings that can be assigned to phrases. If special meaning is something that indica tes a lexical level of word formation then from a lexicalist perspective, derived words should be able to have the same special and unpredictable meaning that r oot s have while syntactic constructions should not. However, this is clearly not the case. Idiom atic meaning exists at the level of phrases as well as individual r oot s and derived he locality domains for special meanings are defined syntactically (Marantz 1997: 209). Thus there is no clear divide between the putative lexical word level and syntactic phrase level with regards to special meaning. An exam ple of this can be seen with (4 ). (4) a. I saw a cat yesterday. b. She was quite catty after the divorce. c. This puppy who was raised with cats is now acting a little catty himself. d. Marantz let the cat out of the bag e. Marantz let the cat out of the bag What (4 ) demonstrates is that special mean ing exists at the level (4a), the derived word level (4 b) and the phras e or vP level (4d). In (4 a) the meanings of cat memorized, making it a lexical item. Likewise s pecial meaning exists with ( 4b) where the specific meaning of the derived word catty making this too a lexical item by the same logic. However, as (4 c) attempts t o show, the form catty predictably derived from the meaning of cat . The meanings of (4b) and (4c) are


74 analogous to (4 d) and (4 e). That is idiomatic meaning and compositional mea ning is found at the level of words and phrase. Thus both derived words and phrases can have idiomatic and compositional meanings and this means that special versus predictable meaning cannot be used as a distinguishing criterion between the two domains. S ince it has been established that the lexicon is generally used to explain individual differences, 2 one would certainly not be misguided to say that both idiomatic word and phrasal meanings are stored in the lexicon. Instead what is perhaps misguided is th e unnecessary assumption that items are generated where their meanings are stored. Another area of contention is the belief that complex Lexical items assembled in the lexicon may not show complex meaning or grammatical properties while syntactically prod uced items must show complexity. However, if a word is complex, containing two or more active morphemes , 3 the evidence is that it will have grammatical qualities that a simplex word could not have. Marantz shows by arguing that words such as (1997: 2013). Likewise idiomatic phrases cannot have the same properties that simplex words have. Thus a word like transmission and the phrase let the cat out of bag cannot mean just anything if the individual parts ar e active morphemes in the language of the speaker. One may propose that the version of catty lexicon, listed with its special meaning, while the other version is produced in the syntax. The first version could be argued to be a simplex word meaning wise and the second a complex word. If catty is simplex in meaning then this would look like an item 2 It should also be noted that kn owledge of unpredictable meanings will differ from individuals of the same language while compositional meanings will be more consistent. 3 Perhaps the same word does not have the same number of morphemes for every speaker.


75 assembled in the lexicon, one that does not show complex meaning. However, the meaning of catty is derived from a mean ing of cat . If catty meant something unrelated then we would have to assume a different r oot . 4 In other words, the meaning always has something to do with the sum of the working parts, negating the argument that t he lexicon composes words out o f parts that lose their individual meaning in the larger syntax . Given that phonological, semantic, and compositional evidence do not clearly indicate a lexical generative domain, the null hypothesis is taken. No claim will be made for the existence of a generative l exicon until evidence is found for another generative grammatical component. Instead it is proposed that all phonology occurs post syntactically and thus the important observations and insights made with lexical phonology are instead reproduced post synta ctically (Halle and Vergnaud 1987). 2.2.2 The Lexicon of DM Give n the reasons from Section 2.2.1 , models such as Distributed Morphology have done away with the generative lexicon. Subsequently t he ungenerative parts of the l exi con have been split into 3 lists: List A, B and C or List 1, 2, and 3. The function of these lists is to provide the syntax with grammatical features (List A), phonological features (List B), and interpretations (List C). How this works is the focus of the following section. Lists A , B and C are distributed at different stages in a revise d Y Model. The placement of each list along the model is shown with Figure ( 2 1). 4 By making special meaning a feature of syntactic derivation, DM is forced to treat all instances of lexical items in context as polymorphemic.


76 Figure 2 1. Derivation in DM . As F igure (2 1 ) indicates, syntactic derivation begins with the assembly of items from List A. List A contains all the morpho syntactic , semantic features that a language has selected from all the available feature s provided by Universal Grammar and root(s) (Marantz 1997; Embick 2012). 5 List A is specified as [a given language such as] English memorize s such as CAT or abstract morp hemes such as [pl] and [past], ... are ). In other words, List A contains all the grammatical features that a language has selected from all the available feature s provided by Universal Grammar as well as language specific s. It also may be that List A only identifies a single r oot feature or placeholder (i.e. . This controvers y will be discussed in Chapter 3 . Neverthe less, the items of List A have uncontroversially been r oot 5 The notation form r oot , Root , or ) . For the sake of consistency the notation will largely be used here . are also often ref erred to as l morphemes (lexical morphemes) while abstract morphemes are often called f morphemes (functional morphemes) (Harley and Noyer 2000).


77 bundles of grammatical featur arantz 1997:203). In addition to this , it is also important to note that the syntactic terminal nodes consisting of features from List A have no phonological features. List A traditionally does not supply either l or f morphemes with any phonological feat ures. They are not phonologically realized until the derivation reaches List B. List B supplies the terminal nodes of the derived syntactic structure with phonological information right before the structure is linearized for pronunciation at PF . 6 List B accomplishes this by being composed of a list of Vocabulary Insertion Rules that consist of correspondences between sets of fe atures and phonological strings . These phonological strings are called Vocabulary Items, VIs, and are responsible for representing both f morphemes and l morphemes (Harley and Noyer 1999:5 6). VI s used to articulate l morphemes are known as L morpheme VIs ( LVIs) and those for f morphemes are known as F morpheme VIs (FVIs). Moreover while List B is not generative (Marantz 1997: 204) , new VIs can be added to List B . VIs ca n also be phonologically null. A fter List B has supplied the terminal nodes with VIs, this information is sent to PF and then List C. List C, the final lexical list, stores special meaning and real world knowledge. Af ter PF and LF, information is sent to List C, the Encyclopedia, for interpretation. It is argued that all r oot meaning is idiomatic (Marantz 1997), and thus t he Encyclopedia lists the special meanings of particular r oot s, relative to the syntactic context of the r oot s, within 6 Note that this diverges from 'lexicalist' minimalist models in which only phonologically relevant informatio n is transferred to PF at Spell out, and in which linearization might be taken to happen earlier, perhaps right at Spell Out. DM is characterized by ' hierarchical structure all the way down' (Harley and Noyer 1999).


78 local domains . T he Encyclopedia is therefore responsible for both linking idiomatic meaning to individual forms and recognizing phrasal idioms (Harley and Noyer 1999:8). Like List B, this list is not generative bu t information can be added 2.3 The Vocabulary Insertion Mechanisms of DM Section 2.2.2 gave a quick overview of the basic architecture of the model as it pertains to the distribution of the lexical com ponents of language. However, Vocabulary Insertion at List B depends upon a set of principles and specifications to ensure that the appropriate phonological form is used for a particular syntactic terminal node in the derivation. These are Late I nsertion, Associat ed and Contextual Features, Underspecification and the Subset Principle. 2.3.1 Late Insertion Distributed Morphology is a late insertion model. This simply means that ntz (1994:275 276). 7 Not only are phonological features irrelevant for syntactic operations a nd thus not needed during syntactic derivation, this principle is crucial for DM to describe the correct surface output. One must wait to insert phonology until after syntax has finished because other morphological processes need to occur to correct for mi smatches between the number of syntactic terminal nodes and the phonological 7 at the L


79 exponents seen representing those nodes at the surface. These other processes are called morphological operations and occur between Spell Out and List B. Since every morpheme form s a terminal node in the syntax, one might expect there to be a separate phonological realization or vocabulary item corresponding with every terminal node. T his of course is not the case (portmanteau morphemes). Thus phonology must wait to be inserted unt il after certain morphological operations change the terminal nodes to correspond with what is seen at the surface. Moreover, in order to guide the VIs of List B into the remaining correct terminal nodes, each V I must be listed with features. 2.3.2 Associ ated and Contextual Features In order for DM to explain how a piece of representational phonology finds a suitable syntactic terminal node, such phonology must come with directions. One such set of directions are Associated F eatures which are also called i dentifying features. T he se are the features a Vocabulary Item is sensitive to regarding insertion into a syntactic terminal node. Th is can be modeled as example (5 ). (5) /Vocabulary Item/ [Associated features] /a/ [+F1 ] /b/ [+F2 ] Vocabular y Items are commonly p laced between slashes // while Associated F eatures are noted by square brackets []. The set of vocabulary items above mean that if a terminal node contains [+F1], the vocabulary item /a/ will be inserted and if it contains [+F2 ] , /b/ will be inserted. However, o ften a VI's distribution cannot be fully defined by its Associated Features alone. In those cases, Contextual F eatures may be listed as well. Unlike Associated Features, Contextual F eatures need not be present in the terminal n ode targeted for insertion. Rather, they are features that must be in the immediate environment of the targeted node. Contextual Features are indicated with {}


80 brackets. A classic example of where they are used is with contextua l suppletive allomorphy: the plural morphemes / z/ and / en/ in English. In English b oth morphemes have the Associated F eature [pl]; however , / en/ also comes with Contextual F eatures which list the few Root s in English that it occurs with: / en/ [pl] / {ox__, child__, bre ther__} . The regular plural / z/, on the other hand, wou ld have no Contextual F eatures other than perhaps { n_ _}, requiring that it be in a nominal context . Contextual F eatu res are also relevant for LVIs. They are responsible for making sure an LVI is used in its appropriate contexts; e.g., ensuring that an LVI that is only used as an intransitive verb does not get inserted in a transitive context. Also if a speaker does not use a particular LVI as an adjective then that LVI would no t be listed with an adjectival Contextual F eature. Associated and C ontextual F eatures are used to limit where a VI can be inserted. However, as already m eatures add nothing to the actual meaning of the terminal node when inserted and thus the Associated F eatures listed with a VI are only necessary enough to provide a correct match bet ween the terminal node and a VI. 2.3.3 Underspecification T he idea that Vocabulary Items are not always fully specified to represent possible feature sets (synta ctic terminal nodes) is known as Underspecification . This means that the same Vocabulary Item could represent several different feature set s in a language as seen with (6 ). (6) Terminal Nodes: 1 [F1]; 2 [F1, F2]; 3 [F1, F2, F3] Vocabulary Item: /a/ [ +F 1 ]


81 Given three syntactic terminal nodes with the feature sets listed, vocabulary item /a/ would only be fully specified for terminal node 1. It would be underspecified for terminal nodes 2 and 3, but such underspecification would not forbid it from repres enting terminal nodes 2 and 3. This underspecification explains the common phenomenon of English present tense subject verb agreement paradigm, / ø / represents everything but [ [+3 rd ] [+Sg]]. This is also important for insertion because it explains why grammars would have an elsewhere morpheme. For a VI to be inserted its Associated F eatures need only be a proper subset of the terminal node. Given this underspecification , what co ntrols V ocabulary Insertion from List B when there is more than one valid VI, is the Subset Principle. 2.3.4 The Subset Principle The principle governing the insertion of VIs into terminal nodes is known as the Subset Principle. The Subset Principle states that a VI can be inserted if it meets all or a subset of the features of a terminal node during Vocabulary Insertion, thus allowing VIs to be underspecified with regard to the features in the target ed terminal node. A VI cannot have more or different A sso c iated F eatures than the terminal node it is being matched with. In other words, a VI can be associated with fewer but not more features than a potential insertion sight has to offer. This raises a corollary of the Subset Principle: when more than one VI q ualifies for insertion, the one that matches in the most features wins. In this way the Subset Principle relies on competition for each terminal node. Thus returning to the example of the English plural allomorph, Figure 2 2


82 shows how both associative and Contextual F eatures work to produce the correct representation. Figure 2 2. Associat ive and Contextual Features in t andem . As Figure 2 2 shows, there is competition between each VI vying for the terminal node in question. I n actuality this would include all the VIs in List B. In the case of / d/, it simply loses because listed with a different Associated F eature. As for / z/ and / en/, they tie based on their Associated F eature , [pl] . However, the Contextual F eatures of / en/ would cause it to win the competition for insertion. T he Subset Principle relies on competition for the insertion of FVIs at F nodes and allows for choice for the insertion of LVIs at L nodes. By means of associated and Contextual F eatures, the most suitable FVI wins and with LVIs there are many ties , allowing for choice amongst the suitable LVI candidates for the given context. 2.4. L and F Morphemes and Categorial Heads The features of List A were described as consisting of bundles of grammatical f eatures (f morphemes) and s (l morphemes). This section will discuss how distributed morphology thus separates all morphemes according to this distinction and how there exists a series of grammatical heads that are claimed to be responsible for catego rizing L morphemes.


83 2.4.1 L and F Morphemes and the Derivation L and f morphemes are proposed to they are the ultimate elements out of which words, phrases, and sentences are Therefore, to derive the simple sentence t hey walked to the store , the features [+3rd, +Pl], [+Pst], [ Root or s would be taken from List A. These features would then combine via Merge (Chomsky 1995) , starting from the bottom of the derivation. This leaves syntactic terminal nodes to only consist of Root s and Abstract Morphemes (Ha rley 2008: 3) , as can be seen with Figure (2 3) . Figure 2 3 . Hierarchical product of m erge . Figure 2 3 merely illustrates the morpho syntactic building blocks of the sentence t hey walked to the store in hierarchical order. H owever, it is not indicative of everything required to derive the sentence. It is generally proposed that derivations begin with a merging with some categorizing f morpheme. This is because s by themselves are argued to have no category and to only convey semantic/concep tual information (Marantz 1997; Acquaviva 2008). In other words, r oot s bear no syntactic features; in particular, they belong to no syntactic category 2001:10). Thus in order to gain a category and to be


84 licit within the grammar, must be categorized by some external grammatical feature. This principle is known as t he Categorization Assum ption (7 ). (7) CATEGORIZATION ASSUMPTION (Embick and Noyer 2005: 5) Root s cannot appear without being categorized; Root s are categorized by combining with ca tegory As describe in (7), the Categorization Assumption necessitate s that certain grammatical features (i.e. category defining f unctional heads) co occur with s in the grammar. Earlier in the DM model, Harley and Noyer say that Root whose near est c commanding f inal ident ity while is a Root whose nearest c commanding f morphemes are v, Aspect and Tense; without Tense such a Root i 7) . This has since evolved to having a node merge with a specific categorial head. This will b e di scussed further in Chapter 3 . 2.4.2 The Categorial Heads of DM As mentioned in the previous section, it is assumed that must first merge with a category defining head. From the list of grammatical features provided by List A, there is a subset of fe atures that are commonly thought to be res ponsible for categorization : s get categorized by verbal type features, nominal type features or adjectival type features: [+v], [+n], or [+a] Marantz (2001). The se categorizing features are thought to be some what different than the features that fo rm other f morphemes. As categorizes Root s: these are called category defining heads; by definition, varieties of v , n , and a that produce ve T he specific identity of these categorial heads as well as whether or not they are universal, however, remains a question as Embick (2012) claims they are universal but other linguists in other


85 frameworks strongly dis agree (Haspelmath 2007). Categorial heads will b e discussed further in Chapter 3 and Chapter 7 . 2.5 L and F Morphemes and Semi Lexical Categories As discussed in Chapter 1, adpositions are sometimes classified as a semi lexical category. It also pointed ou t that the idea of semi lexical class is not a coherent one. Nevertheless, if a coherent view is established, a semi lexical category would presumably neither be functional n or lexical . As discussed here, DM has no way of identifying or accounting for semi lexical items because a VI is either associated with features or not, making it problematic when an item with putative functional feature associations is found in a lexical domain. This is a problem raised in Deacon (2011) and De Belder (2011, 2013) and w ill be discussed further in Chapter 5. Nevertheless, this section will briefly make this issue clearer for the category adposition . 2.5.1 L and F Morphemes and Adpositions As discussed in Chapter 1, adpo sitions can be associated with c ase features. Moreo ver, it was proposed that all P items (particles and prepositions) can be unified by being associated with an [ F LINK ] feature. Translated into DM, t hese claims suggest that the Category P is a functional category (defined by [ F LINK ] ) and one that is comp osed of f morphemes. Accordingly, what ever item realizes this feature and other adpositional features has to be labeled an FVI as one might presume for the form down in (8). (8) [ The kids ] slid [p[ F LINK , G LINK , CASE , DIR ] -down ] [ the slide ] . In order for down to be inserted in the environment seen with (8), it needs to be associated with some feature that allows it to win over other FVIs (the identity of this feature does not matter for this argument). In addition to the prepositional use in (8), it was, however, briefly shown in Chapter 1 ( Section 1.4.1 E xample (4) ) that adpositional


86 with (9). 8 (9) a. The hungry kid downed the last bite of his sandwich. b. How many do wns do we have left for this possession? 9 Unlike (8), in (9) t he form down is used in a distributionally defined lexical position . This usage suggests that the form is an LVI, corresponding with Root nodes (l morphemes) categorized by little v_ (9a) and n_ (9b) heads (further argument for thi s analysis is given in Chapter 4 ). However, whichever feature allows down to win in (8) would necessitate that it would lose in (9) because the Subset Principle (Section 2.3.4) forbids inserting a VI into a terminal node if the VI has more featu re(s) than the terminal node. Re call that a Root node has no morph syntactic features. I f the origin of down is the same in (8) and (9), there is a categorical problem for adpositions in DM because forms like down are neither exclusively functional nor lexical and the Subset Principle as explained prevents insertion of the same Vocabulary Item into both functional and lexical domains . 2.5.2 Adpositions in DM As Chapter 1 demonstrated, adpositions have both lexical and functio nal properties, including a putative functional/lexical distribution. Given this, adpositions may be considered a semi lexical class. This is a problem for DM because the model creates a strict division of labor as discussed here. Moreover, as Chapter 1 s howed, the category semi lexical is an inconsistently defined notion. One definition, 8 Further examples of this adpositional form categorial flexibility are given in Chapter 4 for English and Mandarin Chinese. 9 This usage is from American Football.


87 nevertheless, is that semi lexical items involve a lexical head merging with a functional head (Zeller 2001). How this as well as the opposite position (a functional item being inserted in a lexical domain (cf. De Belder 2011, 2013) translates into DM is discussed in Chapter 5. However, a more discrete analysis (presented in Chapters 5, 6, and 7) shows that adpositions can be treated like the other major lexical categories in DM, solving the th eory internal insertion problem and avoiding the inconsistency of the semi lexical label for adpositions in DM . 10 2.6 Concluding Remarks One goal of this dissertation is to model the traditionally difficult category adposition within the Theory of Distributed Morphology. As can be seen, without a further analysis, the lexical functional distinction of adpositions creates a problem for Distributed Morphology. Moreover, general questions regarding the universality or possible universalit y of lexical categories remain within DM and in Language as a whole. In other words, while finding clear distinctions between lexical and functional categories and the parts of speech continues to be a problem, the fact that researchers have not abandoned such distinctions despite this difficulty suggests a real difference between the two types and the fact that researche r s still have not provided satisfactory definitions suggests we have not developed the tools to definitively capture these real distinctio ns. I t may also prove that the parameters of newer morphological theories such as Distributed Morphology yield insight into the problem by focusing on the relevance of smaller elements in the syntax. As will be discussed in Chapters 4, 5, 6, and 7, just as 10 As argued in Deacon (2011), the source of these forms should be viewed as the same because they ar e not only linked by the same VI but also by a similar meaning (i.e. a meaning similarity that can be Root framed in a different perspective).


88 there are both lexical and functional members of categorial features there are lexical and functional adpositions. That is u pon closer examination the hypothesis is that adpositional items begin as bundles of grammatical feature(s), explaining their funct ional behavior. Moreover, these bundles are capable of categorizing lexical information as the grammatical features of other categorical heads do, explaining their lexical behavior. In order to show this though, a deeper understanding of the components of the model is required.


89 CHAPTER 3 S, ENCYCLOPEDIC KNOWLEDGE, AND DISTRIBUTED MORPHOLOGY 3 .1 Introduction Chapter 2 discussed how l morphemes are classified differently than f morphemes. It is unclear, however, how one can identify individual l morphemes, if this is even possible, befo re insertion. It may also be unclear whether an item originates as a morpho syntactic feature or Root node when a lexical like item occurs in a position thought to contain at least one functional feature. This c hapter not only discusses how one may ide ntify Root items (l morphemes) in contrast to bundles of morpho syntactic features (f morphemes) , it also discusses how one Root item might be differentiated from another in the syntax . Accordingly it discusses possible Root features in accordance with the kind of distribution and mea ning potential associated with Root items (the meaning of a Root before context is uncertain) . Reasons to propose that Root items come with basic semantic information, phonetic features, diac ritic or class features, a Ro ot feature, and or indices in List A will all be discussed in contrast to the principle of Feature Disjoint ness and the Lexical Decomposition Hypothesis. In doing so, the advantages and disadvantages of proposing multiple Root types in List A (individuate d via one or more of the aforementioned ways) in comparison to having a single Root in List A will be discussed. Potential evidence for one approach over another is the presence of derivational gaps. Derivational gaps, however, can be grouped into at leas t four categories: accidental or possible word gaps, prescriptive word gaps, allomorphic gaps, and gaps of necessity (i.e. impossible word gaps). If impossible gaps truly exist , more motivation


90 exists for individuating Root s in List A because this means t hat there is a reason why a limited range of LVIs are only permissible at a given Root node. Moreover, as Chapter 1 discussed, LVI insertions may be limited with Contextual Features. In fact as will be di scussed, all gaps can be handled within the framew ork by listing LVIs with C ontextual F eatures. However, using Contextual Features to explain gaps that may be conceptually driven is unsatisfying. It will thus be suggested that in addition to C ontextual F eatures , indices ( cf. Acquaviva 2008; Harley 2011) m ight be used to account for impossible word gaps. It is the belief of the author that conceptual contradictions as appealed to in Deacon (2011) and alluded to in Embick in (20 12) play a role in addition to Contextual F eatures to explain impossible word ga ps despite the fact that there is no present way to model this in DM. Understanding these gaps is crucial in explaining why not every LVI can be in serted in every Root node . This is important here becau se it will be shown in Chapter 4 that preposi tional d omains can contain Root nodes and obviously not all LVIs are permissible as realizations for such domains . T here therefore needs to be a way to exclude undesirable LVIs . This c hapter is organized as follows: Section 3 .2 introduces the complexity of the te rm Root and why Root as an item of L ist A is pro blematic ; Section 3. 3 discusses the various proposals concerning what features can be present with a in List A and why DM generally wan ts s to be void of features ; Section 3 .4 then e xplains how Root s obtain their category and discusses this as being the locus for special meanings. It gives theoretical and empirical reasons to strip Root s of features and discusses how to identify s via their distribution since the features of Root s in


91 List A are unclear. It also discusses how distributions reveal that s are not spe cified for a category. Section 3 .5 on the other hand discusses why some have proposed features with s in List A and Section 3.6 discusses lexical gaps in deri vations and word derivations. Some of which might be explaina ble via positing features with s in List A. Section 3 .7 then discusses the solution of proposing indices in List A and C ontextual F eatures in List B in order to explain true gaps and i diosyncratic lexical behavior. Section 3.8 then discusses how some have called for the Encyclopedia to b e involved in the derivation and shows that no formal mechanism is in place to make meaningful statements. Section 3.9 rules out the idea of Early Inser tion. Finally it is concluded that s are best viewed as being featureless items, individuated by indices and controlled via Contextual F eatures. 3 .2 Defining the Term Root Before a discussion of s is possible, it is important to carefully define what one means when using the te rm . As Acquaviva asserts, the term needs to be clarified because a Root as node Root as exponent or as a Root as category (2008: 11) . Harley (2011) also breaks down the concept of a into three notions (i.e. Root as an element of List 1, 2, and 3). Combining these distinctions, one may arrive at five separate notions for the term (1). 11 (1) Notions of a : a. A as a Member of List A: a single individual a categorical item consisting of an unknown feature set, or multiple individual a categorical items consisting of unknown feature sets, or multiple featureless items individuated via indices, or an item that is not listed in List A 11 Notions (1a) and (1c) are themselves controversial or a mix of different notions.


92 b. A as a Category Free Domain: all the information to be merged with the first category defining head c. A as a node, the L node: a terminal node where VI insertion is free, or a terminal node where VI insertion is only partially free, or a ter minal node where VI insertion only appears to be free d. Root as an Exponent: The phonological exponent for a node, the LVI, (Could also be viewed as a node filled by an LVI.) e. Root s as an Interpretation: a set of ins tructions for how to interpret a within a specified syntactic context. Each of these five notions captures different aspects of the item in DM, reflecting t he distributed properties of a which are acquired at different stages of the derivation. However, within this breakdown, many disagreements can be found across the literature. Many of these disagreements or conflicting accounts involve the distinction in (1a): a as an item in List A. That is the information that is presen t with a at the beginning of the derivation is either unclear or controversial. Understanding the implications of the different proposals will be the focus of the discussion here as it is critical for the larger goal of this work: the proposal that Root s can be catego rized in adpositional domains. 3 .3 List A: The Features of a in List A The question of what composes a in List A has been difficult since the an important and open question [is] how much information about s is present in the narrow Lexicon As introduced in Chapter 1, List A contains all the morpho syntactic features of a language and it organizes these features into two morpheme classes: f morphemes and l morphe mes. While f morphemes were described as bundles of identifiable morpho syntactic features, the defining features, if any, of an l morpheme in List A were left unstated. It is unclear if List A only contains one copy of a generic item or if it contai ns a separate item for every a particular speaker knows. Moreover, if


93 the latter is true, it is unclear how much information is present with these items in List A. The prevalent view in DM is that much or all of the meaning of an l morph eme is not included in the item in List A and that the item functions primarily to link purely semantic features with a grammatical context. Accordingly items in List A are often called un analyzable atomic elements (Marantz 1997; Acquavi va 2008; Harley 2011). As such elements, they contain no morpho syntactic features or categorial features, being individuated from other bundles of morpho syntactic features by a formal/generic feature ( ) (Siddiqi 2010). However, in order to explain some idiosyncratic and otherwise unpredictable behavior, a number of proposals have associated the notion of in List A with various feature(s), contrary to view that a item is unanalyzable. These include semantic syntactic features (Harl ey and Noyer 1999 12 ; Embick 2000) phonological/phonetic features (Embick and Halle 2005; Embick and Noyer 2007), and diacritic features (Embick and Halle 2005; Embick and Noyer 2007). For similar reasons Pfau (2000, 2009) specified items in List A for individual /different concepts (via indices) , CAT MONKEY etc. , where the LVI /k æ t/ would spell out the item CAT . In response to these proposals, others have both empirically an d theoretically argued to keep Root s in List A separate from any pho nological features (Haugan and Siddiqi 2013 ) as well as separate from all features (Acquaviva 2008; De Belder 2011; Harley 2011 ; Acquaviva and Pana giotidis 2012 ). In one such analysis, items are 12 The model diagram shows the LVI /dog/ being associated with or contextualized by [+count] and List A.


94 not listed in List A, instead being epiphenomenonal fea tureless structural positions that result from the operation of First merge (De Belder, and van Craenenbroeck 2011; De Belder 2011). In other analyses, every item is stored in List A but only identified by indi ces (Acquaviva 2008; Harley 2011 ). 13 Thus while all of these analyses have abandoned the idea of a single item in List A, they have done so in different ways, each with its own cost. It should thus be clear that what identifies a in List A is inconsistent in the literature. However, t he implications that follow from a precise identity in List A are important. 14 What information comes with items in List A completely affects how Vocabulary Insertion works in List B. Moreover, one can imagine that misconceptions about the building bl ocks of language will lead to misconceptions about larger structures (i.e. morphology affects syntax because they are largely one and the same). In this way, it is important to know what information comes with a as this may affect larger description s. 3 .3.1 Theoretical Reasons to Associate s with no Features in List A The primary reason to strip s in List A of any features is one that is central to DM as a Late Insertion Model: the principle of Feature Disjoint ness (2) . (2) Feature Disjoint ne ss (FD) (Embick 2000: 188) A ll language spec ific and arbitrary features are subject to Late Insertion because such features play no role in the syntax. The implications of (2) are that if a feature does not play a role in the syntax, there can be no we are unaware of any 13 le (2005) to account for homophones. 14 The trend has been to move back to the strong assertion of the Lexical Decomposition Hypothesis.


95 arguments that language specific features are necessary at these [DS and SS] syntactic Thus by following such reasoning all arbitrary and la nguage specific features should not be included with the as an item in List A. This clearly excludes phonological features from the in List A because these play no role in the syntax. It also excludes most semantic features (those pure semantic features) from the in List A because the meaning difference between cat and monkey seemingly play no role in the syntactic derivation. Furthermore, not only are arbitrary/language specific features argued to be absent from the item in List A, all syntactic features are argued to be absent. This position reflects the stronger corollary of Feature Disjoint ness: The Lexical Decomposition Hypothesis (3) (3) The Lexical Decomposition Hypothesis (LDH) (cf. Aquaviva 2008: 3; Siddiqi 2010) A ll syntactic features are external to the This position is justified under the belief that the function of a is simply to provide a location where extra grammatical information (information that does not affect the syntax) can be added to the grammar (Siddiqi 2010). Thus Acquaviva st a consensus has emerged to the effect that s correspond to the non grammatically definable part of a word; that part, that is, which does not arise from the spell out of morpho syntactic features hosted on functional heads ). This m eans that the meaning of a item in List A either vastly or completely depends upon being provided elsewhere (i.e. the context the appears in and an entry in List C to interpret that context). 15 15 the discussion in Section 3.4.2 ).


96 Feature Disjoint ness and the Lexical Decomposition Hypothesis form the theoretical reasoning for postulating that a item cannot be identified by any grammatical feature. Following these ideas closely allowed for early proposals of a single all encompassing item in List A with all individuation being supplied elsewhere. This single item would be delineated from the rest of the bundles of morpho syntactic features by a stipulative formal ( ) feature as mentioned in Siddiqi (2010). Thus this ( ) feature is what would cause a particular node to be considered an L node and permit LVIs to be inserted. 16 3 .3.2 Identifying s: Empirical Reason to not L ist s with Features Section 3.2 and 3 .3 introduced the controversy surrounding what comp oses a in List A. Section 3.3 .1 gave the theo retical reasons for why (s) is/are to be viewed as a featureless position or one defined by a generic ( reliable way to distinguish items from all the other functional items at the surface. Given that the ( nnot be specifically observed at either PF or LF and since any other putative feature(s) which constitute a item in List A are unclear or controversial, items cannot be reliably identified by the presence of a type of feature. Instead they can best be identified as those items sharing a unique distribution. 17 This distribution in turn suggests that proposals where item(s) are featureless in List A are indeed correct. 16 This feature also could be used for the same purpose in re visions where multiple items are proposed to be in List A. 17 Perhaps this unique distribution is the manifestation of the ( feature.


97 The claim here is that in human language there exists a group of items t hat share a strong distributional tendency that differs from the distribution of those language items that would otherwise be deemed functional. In other words, items are those items that appear in a domain where other items appear and where no n items do not appear, a lexical domain. 18 This is the basis for Identification Guideline A (4). (4) Identification Guideline A 19 A item is an item that may appear in more than one environment (i.e. usually n_,v_ and a_ environments) . While a non item may appear in one similar environment, this same non item should not appear in two different environments that a item may appear in. Identification Guideline A provides the foundation for establishing the p resence of nodes in the derivation as opposed to strictly functional material. It is important to note that lexical domains might not always contain a node and in such cases the functional material that creates the lexical domain would be repre sented by only an FVI (e.g. /be/, /do/ 20 Identification Guideline A , evidence of the item appearing in two lexical domains should be found. 21 It of the traditional lexical categories, it need not be limited to them. As Chapter 1 made 18 This assumes that we have pro 19 s grammatical category features), and thus this behavior supports an aspect of the Lexical Decomposition Hypothesis. 20 This is the dummy auxiliary do , not the possible lexical do , as is evidenced by doer, deed. 21 Issues could be raised over what creates/i dentifies the different domains and whether the appearance of the same item in two or more variations of the same domain means the same thing as an item cla imed to exist (see Chapter 7 for more on this).


98 clear, there is no solution or clear consensus on what is lexical. Thus while the exact composition of features that constitute a lexical frame in a language may be language specific, the ability for items to appear in more than one of these frames is either universal or highly unmarked. items can appear in multiple contexts as is shown here with a language Isolate, Japanes e, (5) and a Sino Tibetan language, Mandarin, (6). (5) Japanese a. Taroo ga sutaba ni iku Taroo SUB starbucks LOC go b. Taroo ga sutabaru . Taroo SUB starbucks Example (5) shows the form sutaba verbal (5b) domain, indicating the presence of a node via Guideline A . This form flexibility is seen with Mandarin Chinese in (6) as well. (6) Mandarin a. méiyìsi dance very dull b. zài she DUR dance Like example (5), example (6) shows the same form operating in a nominal (6a) and verbal (6b) domain, indicating the presence of a node. However, this does not differentiate homonymous s or possible FVI LVI homonyms. This takes us to Identification Guideline B .


99 (7) Identification Guideline B If an item appearing in several domains can still be related via a shared core notion, a common source can be argued. In (5) the meaning of sutaba changes to fit the environment, but the meaning place while in (5b) the meaning includes the action that will result in one arriving at this same type of place. Li kewise the meaning of tiào shifts according to the frame it appears in. This shared meaning in addition to shared phonology 22 allows us to call tiào sutaba items via Identification Guidelines (A) and (B) . This approach is in no way controversial, nor does it offer a really novel observation. Traditionally (5) and (6) would be analyzed as cases of conversion, where a with a fixed category is derived via zero affixation. Likewise , conversion may best explain (5) and (6) i n the DM model. However, for the purpose of identifying s, the ability to appear as more than one category seems to be the nature of s whether by means of a featured undergoing a type of conversion or by means of a featureless element mergi ng in a featured environment. Furthermore, the reason to have both methods is that in addition to this flexible distribution, s also display flexible meaning beyond what was seen with (5) and (6). 23 Distribution al flexibility is not the only property of s. It is also possible for s to have underived meanings when appearing in different frames as is here seen 22 23 links to an idiom. Th

PAGE 100

100 with a Germanic, English (8), Austroasiatic, Mundari (9), and African Bantu Language, Swahili (10). (8) English a. John ran a mile. b. Th e Salmon make the same run every year. c. The sauce is too runny . In (8) we see that the morpheme run appears in three distinct environments: verbal (8a), nominal (8b) and adjectival (8c). Like (5) and (6) each environment derives a different meaning, but the relationship between the meanings is not as predictable as was in (5) and (6). In other words, there is no justification in saying that one domain (nominal, verbal, or adjectival) is derived from the other as a run in (8b) and runny in (8c) do not actually involve running and vice versa. This can also be seen with (9) and (10). (9) Mundari ( Evans and Osada : 354) a. buru =ko bai ke d a.4 mountain=3pl.S make compl tr indic b. saan=ko buru ke d a. re wood=3pl mountain compl tr indic (10) Swahili ( Hurshkainen 1992: 98) a. end a 'go' > end elea 'proceed' (double applicative) b. on a 'see' > on ea 'bully' (applicative) While related, the meaning of th e s in one domain in (9) and (10) do not contain the fixed meaning of the in the other domain. The meanings can only be related to a vague core meaning. In other words, for the forms in (10a) and (10b) to be related, the meaning of the mus t be highly underspecified and flexible. This property of s is described in Arad (2003:740) and wil l be discussed more in Section 3 . 4 .2. For now, though, it forms the basis for a revision to Identification Guideline B .

PAGE 101

101 (11) Revised Iden tific ation Guideline B If an item appearing in several domains is clearly related via a shared core meaning, a common source can be argued. If items appearing in several domains are only loosely related and share a phonological form, a common source can be argued. With Guideline A and Revised Guideline B , there are still problems with identification. A residual problem is that not every item that otherwise appears to be a it em (to be discussed in Section 3 .6) is attested app earing in multiple domains. This leaves us with Identification Guideline C . (12) Identification Guideline C If an item appe ars in only one domain and this domain is one where other flexible items occur, it must be semantically compared to o ther established items. If this item brings information that cannot be attributed to the grammar, a node can be argued for. The issue of Identification Guideline C will be discussed further but it should be explicit that flexible distribution does not cause an item to be a item. Rather the features of a must enable such a distribution. Thus a lack of attested flexible distribution, does not formally exclude an item from hood. Rather given the fact that item s are overwhelmingly flexible in where they appear, the more interesting question is why a particular item fails to be flexible. This is discus sed in great detail in Section 3 .6. This brings us to the conclusion of this section. The ability of an item to appear in a nominal, verbal, and adjectival domain suggests hood from a distributional identification standard. A strong form to meaning correspondence provides evidence that the items of question in the different domains are indeed multiple attestations of the same item. However, if one lacks a consistent fo rm, a consistent meaning must be found. Likewise if one lacks a consistent meaning, a consistent form must be found. If a flexible distribution is not found,

PAGE 102

102 semantics can be appealed to as a last resort. Furthermore, this flexibility property can be attri buted to the fact that s begin with no grammatical cate gory, as argued for in Section 3 .3. 1. Viewing s in this manner gives a reason for why this phenomenon is so prevalent and gives a reason for why some conversions in the old sense produce new words while others seem to just reframe established words. 3 .4 Categorization In the previous section, it was seen that items may appear in different domains without evidence for conve rsion. As mentioned in Chapter 2 , it has been proposed in DM that a Root obtains the identity of the grammatical feature( s) that most closely c command it (Harley and Noyer 1999). Thus a particular may be realized as a noun or verb depending on the morpho syntactic context it appears in, an insight going back to Chomsky (1970; se e also Marantz 1997). This has since been el aborated upon or refined by having a bare merge with one of the categorizing grammatical feature s t o give that Root its initial category (often listed as [+n] [+v] and [+a] features (Marantz 2001) as discu ssed in Chapter 1). 3 .4.1 Primary Merge : Featureless s entering the Syntax C ategorization is co mmonly modeled with Root begin ning in one of two structural positions. The first has a Root a s a sister to a categorial head (Embick and Marantz 2008 ; Lowenstraum 2008; De Belder and van Craenenbroeck 2011 ). The second has the Root begin in a Root phrase and head move to merge with the categorial head (Embick 2000; Harley 2005a). For example, the nominal and adjectival forms woul d have the following structure as it merges with a nominal and adjectival head (13 ) .

PAGE 103

103 (13) Categorization ( Embick and Marantz 2008: 22 ) . If it is a sister to the categorial head as in (13) , the will direct ly merge with the categorizing head (Lowenstraum 2008) . 24 If the Root is Root phrase, it merges with a phrase projected by a categorial head (Embick 2000; Harley 2005a) , as in (14) . (14) ( Embick 2000: 195) If the begins inside the phrase, it will head move and merge with the first c commanding categorizing head (Harl e y 2005a). The ( ) symbol in (1 4 ) could be the merging of a phrase or just a item. It sh ould also be clear that the difference in the categorizing environments shown between (1 3 ) and (1 4 ) is not necessarily dependent on the differences between the labels v_ , n_ , and a_ ( v_ structure , h owever , often requires a complement) . In fact we can simpl ify the 24 The difference between the two representations is not important for the proposals here, but a definition is s upposed to be featureless grammatically speaking and thus what it can project at the phrasal level is dubious. Logically it seems it should not project at the phrasal level. Acquaviva (2008) oot position does not act a like a phrase head but rather directly merges with a categorizing phrase head makes more sense theoretically.

PAGE 104

104 categorization model with the completely underspecified categorizing head little x_ as used to represent the derivation seen in ( 1 5 ) . (15) a. b. c. Example (1 5 ) underspecifies the model given by Embick (2000) by stating that x_ can Root becomes the category of whatever feature value is present under x_ . Example (1 5 b ) Root Root phrase and into the terminal node occupied by some valid categorizing feature. Example (15c ) then simply focuses on this position whe Root and categorizing head will merge. This of course could be the beginning of a derivation. Whether or not the is initially positioned as it is in (15a ) or as it is in (15c ) ostensibly depends on the s electional requirements of the particul ar categorial head. Following this, one may define a semantic features are merged with (a) feature(s) of the grammar, structuring unruly semantic features for interpretation in an orderly system. 25 It then follow s that when a node is first merged with a feature(s) of the grammar that obtains the category, grammatical frame, of the present feature(s). The rea son to list [+n] [+v] and [+a] features as the available categorial features appears to be a mat ter of convenience 25 here.

PAGE 105

105 in many descriptions, as they are referred to as being merely labels with no universal significance. On the other hand, these features are being listed in alignment with those three lexical categories that are agued to be universal as di scussed in Chapter 1 Section s 1.3 and 1.4 . There thus remains a tacit assumption that these features correspo nd to something substantial . Nevertheless, in theory, without further constraints, a could appear as all possible categories with this model. In addition to having a flexible distribution, items can be identified as those items that may correspond to several meanings. As seen in Section 3 .4.2 , items show unpredictable me anings in different contexts. This takes us to the next section which discusses how Primary merge is used to explain the kinds of s pecial meaning . 3 .4.2 Special Meanings : Inner vs. Outer Layers In Section 3 .3 .2 it was seen that the meaning of a can be unpredictable when appearing in different environments. Thus in addition to the idea that s are a categorical and need to merge with grammatical information to be incorporated with the g rammar , is the idea that special meanings are established duri ng first merge (Marantz 200; Arad 2003, 2005). 26 by no means predictable from the combination of the and the word creating head (2003:740). This is something that has to be memorized ( i.e. stored in List C). established between a and its outputs, one cannot predict the semantic outputs of a based upon the semantic output patterns of another . Compare (16) and (17 ). 26 Thi s is also proposed to be a phase (Marantz 2001).

PAGE 106

106 (16) Hebrew (Arad 2003:744) : BXN Meaning: Pattern: Category: a. baxan CaCaC (verb) b. hivxin hiCCiC (verb) c. mivxan miCCaC (noun) (17) Hebrew (Arad 2005:12) : GDL Mean ing: Pattern: Category a. gadal CaCaC (Verb) b. higdil hiCCiC (Verb) c. migdal miCCaC (Noun) In (16) and (17 ), the meanings of the different words forms of BXN and GDL are argued to come from a co mmon semantic core. It is also obvious that one could not predict what the actual meaning of GDL would be in the miCCa C pattern by looking at the meanings of the other patterns or by looking at all the meaning of BXN in this same pattern . In other words, a tower is one of many things associated with growing , enlarging or a place that is raised up . Based upon the pattern of BXN , one might anticipate the meaning of migdal to be instrumental. That is , one may discern or test something with an exam , but one does not grow or enlarge something with a tower . In this way, Hebrew provides examples for why items should be viewed as indefinite notions ( Identification Guideline B ) for which meaning only become s certain after categorization (i.e. in this c ase the consonant cluster is realized by the vowel pattern of a particular categorial head.) It should, moreover, be clear that t his is not unique to Hebrew s . E xample (9) in Section 3 .4 .2, showed that the buru in Mundari means he nominal subject context and struc ture with a mountai n shape) in a verbal context. The shared phonological form and

PAGE 107

107 the semantic connection between both specific meanings strongly suggest a common origin, but the relationship between the two meanings does not appear to be predictable. In other words, if someone were suddenly told to mountain firewood without having already learned a specific me aning, several possible interpretations would be available: bring firewood to a mountain, place the firewood on a mountain, make the firewood resemble a mountain, cause the firewood itself to have mountain like qualities meanings, a specific one is argued to be established at primary merge. Another exam ple of this can be seen with (18 ). (18) Mandarin a. ni de xiào hen mei you GEN smile very beautiful b. ni xiào le you smile/laugh ASP In (18 ) the phonology, hanzi (Chinese character), and meaning all point to the same . However, no one could predict that xiào used a s a verb or noun would specif Accordingly , it is argued that the links to a specific meaning at categorization. O nce a meaning is established , however, it will be maintained throughout the rest of the derivation. This explains the possibility for xiào to mean either smile or laugh in the verbal domain of (20b). In order to derive smile in the ve rbal domain, the argument would be that xiào gets categorized first by an n_ and then is categorized by a v_ head. At this stage a new meaning is not possible and merges with a head, it

PAGE 108

108 obtains a fixed interpretation. Likewise when on e categorial head is stacked on another in Hebrew, the meaning remains compositional. (19) Hebrew (Arad 2003:746) : SGR Meaning: Pattern: Category a. misgeret miCCeCet (Noun) b. misger CiCCeC (Verb) In (19 ) it was shown that the relationship between a and a word form (consonant+vowel pattern) is often unpredictable, but Arad shows that in examples like (19 ) where an affix attaches to an already categorized , the meaning is predictable (i.e. a frame to frame). The structure of (21b) is this: [[sgr]n miCCeCet] [[[misgeret]n]v CiCCeC] [[[misger]n]v]. More clearly the reason that this is called affixation is because the [m] at the beginning of the word can only come from the previous pattern. T his is also described as a productive process in Hebrew (Arad 2003) , indicating that there are two types of categorial derivational processes: from bare s and from already categ orized s . This brings us to the conclusion of this section. This section discussed why Arad (2003) argues that the first categorial head is the locus for special meaning s. Since the cannot be broken down further , the only way to explain th e range of meanings that can be given to a single Hebrew consonantal is via the and it s categorial head . This thus further explains the reason to revise Identification Guideline B , as one might not always easily see the semantic connectio n between a in one environment versus another. The meaning of a item can be better established via the sum of that item s appearances. All of this points to the fact that s are not only a grammatical but

PAGE 109

1 09 also unspecified for semantic features. Thus examples such as those shown in Section 3 . 3.2 and 3.4 .2 give empirical evidence for FD and the LDH. FD and LDH, however, create problems for how to handle those features that features). Questions regarding how to model specific lexical behavior (i.e. distinctions between lexical and syntactic causatives etc.), questions regarding how to model Vocabulary Insertion that is dependent on a s identity (i.e. the existence of Class features a nd the specific realization of F nodes that agree with such features) and questions regarding how to govern Vocabulary Insertion into a node (i.e. the belief/observation that certain s favor certain domains) remain if s have no features in List A. 3 .5 Associating s in List A with Features As discussed in the introduction, contrary to both theoretical and empirical reasons to keep features away from s in List A , there appear to be properties unique to some s which suggest that some s are different ly marked in List A. These features seem to be specific and to either influence the syntax or affect vocabulary insertion. This section will thus discuss some reasons to propose features with s in List A despite what has previously been claimed . 3 .5.1 Associating s in List A with Phonological and Class Features There are issues regarding to what extent s can be stripped of gender and class features even though these associations are often as arbitrary as the pho netic features used to pronounce them. 27 Embick and Halle (2005) and Embick and Noyer (2007) claim that 27 Some classifier associations may be less arbitrary.

PAGE 110

110 and H classes) (2005:46). Furthermore, if one has already listed phonological and diacritic/class featur es with s in List A, there does not appear to be any reason to not als o list s with their semantic features (or at least some of them) since phonological and class features represent the worst offenders for FD. Such proposals amount to fully spec ifying s in List A and this seems to be what Embick and Halle (2005) env ision as they also argue that no insertion takes place with nodes: Vocabulary Insertion only applies to abstract morphemes; (2005: 5). An early argument that supported proposals for full specification of Items in List A was the belief that s never exhibited true suppletion. 28 One rea son for this is conceptual with regards to language acquisition. It was presumed that children sh ould not expect completely different phonetic features to correspond with the same , preventing true suppletion (or at least a majority of it) (Harley and Noyer 2000). There was also the empirical evidence from better known languages that indicated t h at truly suppletive forms overwhelmingly involved what in the model would be considered f nodes and not nodes (i.e. English: be was, go dekiru bù méi T o explain this observation, one could propose that at least some phonological information was with the in List A. Thus when it came time for Vocabulary Insertion, the environment 28 The position is challenged in Section 3.7.1.

PAGE 111

111 that a ot was placed in could not cause a competing LVI to be inserted instead because the node would already contain phonological information. Vocabulary Insertion would thus only apply to f nodes, allowing for the allomorphy seen at these nodes to be expl ained via competition. In cases where s underwent weak suppletion or alternations (i.e. strong strength, goose geese), such differences w ere explained via Readjustment R ules that occur post Vocabulary Insertion. Another reason to claim that t s are specified for phonological and class features is perhaps a more obvious one. There are a number of non semantic/syntactic idiosyncratic properties associated with out of counters, gender/class features, con jugation classes and so on. Without other stipulations, these properties affect LVI insertion at List B and also present a challenge for a strict interpretation of FD and LDH. Acquaviva points out that while his LDH approach captures the a categorical natu re of not provide a formal expression of the stable association of most s with gen der, class, or other accidental or as properties that are inherently part of the . 29 Accordingly, designating s with a class membership feature has been used to indicate how certain agreement type morphemes get spelled out at List B. An example given in Embick and Halle (2005) shows a class feature being used to cr eate a rule for the choice of a particular thematic vowel in Latin (22). (20) Thematic Vowel Insertion Rule (Embick and Halle 2005:12) a. TH TH[X]/ [X] b. TH[I] a c. TH[IV] i 29 This is a position many linguists have assumed.

PAGE 112

112 The proposal was to have s like L AUD come with the diacritic feature [I] and come with the diacritic feature [IV] in List A, creating the items: , L AUD [I]. According to the rule in (22), for the thematic vowel / i / would get spelled out instead of another possible thematic vow el. Diacritic features may also be used to explain why new compounds in languages like German always involve the same gender agreement of a particular when that is the right most constituent of the compound ( the findings in Pfau ( 2000 : 151 ) fo r one suggest that gender comes from the or List 1 ) . The idea has been that this brings this information along, explaining the consistent association noted by Acquaviva above. In this description of DM, one can conclude that phonological and d iacritic features create the item in List A and perform the function of explaining the spell out of class features, and certain cases of allomorphy. Being supplied with phonological features early was also used to explain the difference between suppl etion at l nodes and f nodes. Moreover, one could also imagine diacritic features being used to limit the context a occurs in (i.e. if a has a class feat ure in List A then this feature can be active in the syntax, limiting the types of classifi ers or quantifiers a can merge with). This could be envisioned to prevent an inanimate from merging with the plurality feature that gets spelled out as men , which is disallowed in Mandarin Chinese. s would simply be listed as members of either an animate or inanimate class via a diacritic in List A. T his could also be done without the diacritic feature if certain semantic features also come with s in List A.

PAGE 113

113 3 .5.2 Reasons to Associate s in List A with Semantic Syntactic Feature s There are a number of syntactic/semantic features that have been found to have itself): causative features in some Japanese s (Harley 2005b), plurality features in some Greek s 30 (Alexiadou 2011 ), and passive features in some Latin s (Embick 2000). How to account for these features in a Lexical Decomposition Model is not fully agreed upon. Embick (2000) proposes that certain s inherently come with syntactic semantic f eature(s) in List A. 31 He does so in order to explain the behavior of deponent verbs in Latin. These verbs are active in meaning but receive passive morphology: the deponent verb hort or or generally marks pass ive first person agreement. In the active voice, deponent verbs follow the complete conjugation pattern of most s in the passive voice ( cf. Embick 2000 for full paradigmatic listing). Likewise while being active in meaning, deponent s show an analytic rather than synthetic pattern in the perfect aspect. Generally s only show an analytic pattern with a passive meaning: am a v am a t us sum hort a t us sum sum , which is often analyzed as not being part of the stem, is only present in passive constructions, except for depon ent verbs. Moreover, what is or is not a deponent in Latin is arbitrary. There is no supposed pattern (a semantic class) between which s are deponent and which are not. Moreover, this feature is argued to affect the syntax, it affects movement i 30 ility readings than coerced class readings. 31 In what Embick calls Early Insertion, discussed in Section 3.9.

PAGE 114

114 Because of t he need to explain the specific irregular spell outs of deponent verbs, the fact that this is seen with only an arbitrary subset of s, and the conclusion that the deponent causing feature also affects movement, Embick concludes that a type of [pass] feature is inherently present with the s that display deponent behavior. Thus the solution accounts for both a movement operation and the allomorphic spell out of a class of items that would otherwise have to be l isted. This is one way to explain why deponent s get spelled out with passive VIs even though the clause is not passive. Moreover, while Embick (2000) appears to be the only one to explicitly call for such a feature to be inherently tied to specific Root s in DM, the occurrence of lexical like features that affect the syntax could e asily prompt similar proposals. As analyzed in Harley (2005b), as well as others, Japanese causatives can be divided into two classes: productive causatives and lexical cau satives. Productive lexical causatives behave syntactically, semanti cally, and morphophonologically like sin gle The productive causatives would natura lly be the result of a merging with a causative feature in the syntax while the non productive , lexical causatives seem to be the result of a feature that is listed with the in List A. Moreover, membership in the class of lexical causativ es is arbitrary like deponent feature with the lexical causatives in Japanese to explain why only these s behave in this manner. One may alternatively propose that t hese s merge with this feature very early while productive causatives merge with this feature at a higher level in the syntax . This would, however, require an arbitrarily related group of

PAGE 115

115 exponents to be listed together in L ist B as available ca node that had merged with a causative feature early in the syntax (more of this in Section 3 .7). In addition to a class of lexical causitives in Japanese, h can be counted in Greek nominal domains indicates a similar pattern behavior . S ome s that are normally considered mass s can be made plural , as seen in (21 ). (21) I grabbed two coffees. I n ( 21 ) , h oweve r, the meaning of coffee changes (it und ergo type shifting) as it merges in the numb er en vironment. Essentially two coffees mean two conta iners of coffee. Alexiadou (2011 ), however, describes a subset of s in Greek that have a lexical plural quality to them that allows them to avoid type shifting seen in (22 ). (22) Greek (Alexiadou 2011 :36) a. hithikan ner a sto patoma dripped water pl on the floor b. hithike nero sto patoma dripped water on the floor Example (22 a) shows that when nero ner a s though can be used in this way. Some appear to always undergo the type shifting seen with (21 ) when appearing in countable domains. These facts could prompt an analysis where the difference is a property of the itself (i.e. some s contain a [plural] feature that allows them to be counted without type shifting). Alternatively the count/mass noun n_ head that projects for number vs. one that does not) as Alexiadou (2011) does.

PAGE 116

116 However, the solution for explaining LVIs can be used in one domain versus another appears to be a stipulative rule against identified in List A merging in certain environments or a stipulative grouping of val id exponents in List B . Deponent s, Causative s, and Lexical Plurals could initially lead one to give some s inherent semantic syntactic features in order to explain the type of behaviors discusses here. However, this analysis comes with a cost. In other words, the s begin to appear as if they can select their domain akin to the idea that a lexical item projects for its functional surroundings and selects complements in a more traditional syntactic model. Many of the positives of the DM model are the result of being rid of such assumptions. Nevertheless, the issues brought in t his s ection as well as Section 3 .6.1 emphasize that FD and LDH have issues. Moreover, while these Sections have shown how certain s behave differently than others with regards to features and spell out, there is also the issue of why not all s can appear in all domains. 3 .6 Categorial Gaps: Limitations on s In Section 3.3 the t heoretical advantage of keeping s featureless in List A was gi ven. In Section 3 .4 it was argued that s could be identified best via their flexible distribution. This flexibility was argued to be in part the result of the a categorical nature of items in List A. It was also shown that this flexibility produ ces both predictable (compositional) and unpredictable ( idiomatic) meaning (explaining how variable meaning can be connected to the same ) . In Section 3.4 .2 the difference between idiomatic meaning and compositional meaning was described as the differ ence between a first merging with a categorical head and subsequent merge s with other categorial heads. However, a possible problem for proposing that s lack any

PAGE 117

117 identifiable features in List A is that there exist distributional gaps that might suggest otherwise. Gaps will be defined here as instances where a does not participate in a particular derivational process that a com pletely generative system could create. In some cases these gaps would traditionally be explained by blocking, but within the DM model , blocking is translated into winners and losers via competition and listedness (Embick and Marantz 2008). Moreover, while many gaps ca n be explained as mere accidents , other cases are not so clear. The specific question here is whether gaps entail/suggest the presence of different properties at the start of the derivation, requiring s to be listed with feature s in List A contrary to the LDH? If so, s must be individuated in List A to control insertion patterns, explaining the gaps. If not, distributional differences can be explained via Contextual Features. To investigate this matter, gaps will be classified in several ways: unconventionally derived word ( accidental ) gaps, possible but unconventional first categorization (accidental ) gaps, allomorphic gaps (not reall y a gap), and impossible categorization word gaps. The discussion will begin with what appear to be the most common type of gap, accidental gaps. 3 .6.1 Categorial Gaps: Accidental Gaps Unconventionally derived word gaps are numerous and easy to find. One quickly finds that some affixes seem to avoid s or stems that would otherwise seem acceptable. For example the cause forms the base for to cause , the cause , causal , and causality while parent only seems to create the parent , to parent , and paren tal , failing to generate parentality . Likewise father creates the father , to father , but not fatheral or fatherality . If these forms are ungrammatical, they represent unexpected

PAGE 118

118 gaps in a system that claims morphology to be compositional like syntax, DM, a s seen in (23 ). 32 (23) Potential Derivational Gaps with / al/ and / ity/ a.[[[father]n]*a al] b.[[[[parent]n]a al]*n ity] c.[[[[cause]n]a al]n ity] Examples such as (23 ) might suggest that there is something about the s filled by the LVIs / father / and / parent / that prevents derivation from proceeding in a completely productive manner. 33 However, as mentioned in Embick (2012), unconventional forms like parentality acquire more acceptability when given the right conte xt. Cases such as these could then be similar to those syntactic structures that only seem licit when a certain context is provided. Can the fact then that people do not recognize fatheral or fatherality as words be explained by infrequency and or a lack o f context? If so then these gaps could be called accidental gaps because s do not create real world contexts for usage. The accidental explanation would seemingly also cover gaps seen with suffixes such as [ ess] that commonly attach to lion lioness , leopard leopardess but not panther pantheress as discussed in Embick a nd Marantz (2008). 34 These gaps have to be arbitrary because there is nothing about the semantics of one big cat versus the 32 If the derivation of words is like that of phrases then are there comparable unattested phrases? We do not expect new sentences that follow the established pattern/ rules of derivation to be ungrammatical. 33 established, preventing it from merging wi th unwanted affixes. However, this would cause problems for LVI insertion. 34 duck ling but not dog ling

PAGE 119

119 other to e xplain this pattern. Likewise it seems strange to say that one group of s is in a class that is different than the oth er. This explanation can also be extended to many first categorization gaps. Like fatheral , it is well known that some s are not generally ca tegorized in certain domains (24 ). (24) a. John purpled the walls. b. They greened their apartment complex. The use of the purple as a verb (24 quick Web search reveals that other English speakers do use this as a verb. Likewise the blue is not commonly u sed when making things blue, but it is used by steelers when bluing steel. Also given the fact that green in (24 b) is widely used and acceptable in the same environment further demonstrates that if (24 a) is a gap , it is an accidental one. In other words, t he distribution of one color term should indicate the potential distribution of all color terms. That is if one were to explain distributional differences via semantics. Otherwise one needs to claim that certain s have a categorial feature while many others do not. 35 Moreover, the use of the green as in (24 b) is a relatively new phenomenon tied to relatively recent concerns for the natural environment. M any gaps can thus be seen as possible words waiting for an eliciting context. In fact, it is actually difficult to find completely unattested zero derivational gaps in Eng lish, but (25) and (26 ) provide two examples. (25) a. We live near the sea. b. ?Today we seaed o u r boat. c. ?Thi s place seems seay. 35 It is unclear how such a proposal would be helpful as one would then have to explain gaps in the other

PAGE 120

120 (26) a. I have a cold beverage. b. ? I am feeling beveragy . c. ?Beverage me. The s + LVI sea and beverage are unattested in ver bal or nominal domains as in (25) and (26). In comparison to (24 ), the reason for this gap is perhaps less clear as the opportunity to use sea and beverage in such ways seems ample. These gaps, however, ca n also be explained as accidents . There is nothing about the semantics of these words that would prevent usage in other domains. It is also quite possible that these s are being used as verbs and adjectives and, if so, such usage would in no way be un grammatical. Thus while the re are some differences between (24 26 ), usage versus non usage appears to be either arbitrary or dependent on external circumstances (i.e. what is happening in the world). There is nothing obvious about the meaning of one that suggests usage versus non usage. Based upon this reasoning, accidents /convention will be used to explain why these gaps exist, requiring no modifications to be made to the Lexical Decomposition Hypothesis. Pure accidents , however, might not so easil y be the explanation for all derivational differences when considering what is t raditionally called blocking (27 ). (27) a. John steals for a living. He is a thief. b. Jo hn steals for a living. He is a ?stealer. Example (27 ) is a classic case of what is ca lled blocking where one is blocked in a context because another already occupies the same semantic space. If such blocking is the best explanation, then there must be semantic features with the to convey this information. 36 However, as Emb ick and Marantz (2008) argue, explanations of blocking via semantics are extremely inconsistent. This suggests that blocking is not 36 This may also be analyzed as a case of allomor phy.

PAGE 121

121 predictable (i.e. thief should also block robber and burglar and one could think of many similar examples). There is also pl enty of semantic space for stealer to emerge (28 ). (28) a. He always steals the ball. He is a good ?thief. b. He always steals the ball. He is a good ?stealer. 37 Since thief carries with it a meaning of breaking t he rules among other things, (28 a) seems s emantically off. This would ostensibly allow for the productive form stealer to emerge (28 b). However, instead of using it, people will rephrase the statement to avoid a word that they presumably think does not exist. The types of gaps seen in this secti on can be explained as the product of arbitrary societal judgment s and accidents , not by the grammar and not by a that gets realized by the LVI / steal /. to blocking [is] the difference between competition for grammaticality and competition Although (28 ) is different than (25 ) and (26 ), this type of gap can also be called an accidental gap because it can be explained as the result of competition for use. U se in this case depends on factors that are arbitrary for the g rammar (i.e. prescriptive judgments). 3 .6.2 Categorial Gaps: Allomor phy and Gaps In addition to the type of gap discussed in Section 3.6.1 , there also seem to be affixal derivational gaps that if produced create a more arguably ungrammatical output. An example of this can be seen with the prefixes in and un ( 29 ). (29) a. unimp ortant *inimportant / unhappy *inhappy b. impossible *unpossible / insane *unsane 37 As a child I thought the Pittsburg Steelers were so named because they were good at stealing the ball.

PAGE 122

122 The difference between the choice of which affix to use for the s in (29a) and (29 b) seemingly has nothing to do with a difference between the s themselves. I nstead it can be said that the different FVIs for this negation prefix are arbitrarily listed with different s just as the FVI / en/ is listed for a plural feature in the context of OX as discussed in Chapter 2 , Section 2.3.2 and similarly argued in Embick and Marantz (2008). This type of analysis opens the door to the possibility that many apparent derivational gaps result via competition and blocking amongst FVIs and have nothing to do with differences at List A. This, however, does always eas ily explain why certain affixes with possibly different meanings only seem to select for certain s. For example the following verbalizing suffixes might be thought to be in complementary distribution: / en /, and / ify / ( 30 ). (30) Putative (Derivational) Gaps a. Dark darken / ?darkify / b. Red redden / ?redify c. Ugly uglify / *uglien d. Simple simplify / *simplien One may ask whether such affixes represent different types of v_ heads or if they are allomorphs of the same thing. If the suffixes have different meanings, then it is clear that there are gaps. The possible meaning difference of / en/ in comparison to / ify/ is given with the following templates (31 ). (31) a. / en/ [X+en to Y = Give more of quality X to Y] ; b. / ify/ [X+i fy to Y = Give quality X to Y]. The difference in meaning can be seen with soften/darken/sweeten/quicken versus vilify/purify/classify/unify. Something can be softened without becoming soft, darkened without becoming dark, sweetened without becoming swee t and someone can quicken

PAGE 123

123 their pace without being quick. However, someone seemingly cannot be vilified without being a made a villain, something cannot be purified with becoming pure, and something cannot be classified or unified without be divided into c lasses or grouped into a whole. If these intuitions are generally correct, this means this difference cannot be explained as allomorphy because the suffixes seemingly have different meanings . If there is a pattern to the gaps this would suggest that some s have a property that inclines them to be favorable for the structure they appear in ( it is unclear, however, what that property is in the case of / en/ in comparison to / ify/ ) . As for now, it seems possible that the meaning of certain s limits how that wants to be categorized and with what features it will merge with . 38 3 .6.3 Categorial Gaps: Possible Limitations on s It has been established that it is advantageous to view s as a categorical because of flexibility. It has also been mentioned that the meaning of s is not fixed until the first categorial head merges with the . This helps explain instances like (3 2 ) . (32) A categorical Verbalization. a. John hammered the desk with his fists. b. The land was wate red with the blood of patriots . c. The U.S carpeted Baghdad with bombs. In (3 2 ) the fact that hammer does not necessitate the use of a hammer lends credence to the analysis that these forms involve an a categorical being merged with a verbal 38 What is really being said is that there is structure to concepts outside of language that hinder or negate the possibility of them being lingui stically framed in a certain manner. This thought by no means seems absurd and is surely to be supported by some theory unknown here.

PAGE 124

124 verbs which involve a verb containing a noun. Until this point, many gaps h ave been explained as accidents , however, reveals a plethora of gaps involving the verbalization of a categorical s (3 3 ). (33) Tr ue Denominals (Kiparsky 199 7: 15 16) a. #She taped the picture to the wall with pushpins. b. #They chained the prisoner with a rope. c. #Jim buttoned up his pants with a zipper. As (3 5 a c ) show, certain s are not capable of being categorized as a verb without first being categorized as a noun. s like tape , chain and button cannot be used as simple concepts of motion. One has to use a chain to chain something. More examples of these are not hard to find (3 4 ). (34) s derived Verbs Gaps. a. # John truc ked in the goods with a train. c. # John bagged the groceries with boxes. Like (3 3 ) the s in (3 4 ) resist interpretations that would allow them to be used in ways that did not involve the nominal interpretation acquired from the first level of categorization. Acquaviva (2008) talks about such occurrences to show the difference between an a categorical being first categorized and derivational stacking. He does so to demonstrate the adv antage of a model that includes both levels, but he does not explain why gaps exists with the types of s seen with (3 3 ) and (3 4 ). Embick of ungrammatical uses of certain s (2012: 80). It is thus unclear why some s

PAGE 125

125 are more pliable than others (i.e. why carpet allows for immediate verbal framing while s like button do not). In addition to this , it is quite clear that not just any concept can be coerc ed in any given way. This is explicit when a centered on an attribute, property or event is attempted to be coerced into an orientation frame (35 ) . (35) Forced Orientation Frame a. The box is located at the back/front/side of the store. b. #The bo x is located at the fast/huge/old of the store. c. #The box is located at the swim/jog/speech of the store. In (3 5 ) the semantic offness of (3 5 b , c) in comparison to (3 5 a) is because these s are not associated with anything to do with locations. T heir meanings cannot be stretched to this domain without an added orientational association (i.e. a change in meaning). This is also true of Japanese (3 6 ). (36) Japanese a. *Aka ni ki o ire ta. Red LOC wood ACC put PST b. *Suwaru/suwari ni inu ga iru. To sit/the sit LOC dog SUB be Given (35) and (36 ), it is argued here that while underspecified, s are connected to a meaning during the derivation (as in Pfau 2009) a nd that this meaning may not comply with the features of every categorial head or the larger context with regards to Encyclopedic knowledge. In many languages, use is reporte d to be more rigid than English . One may repeatedly be told that it is impossible for certain s to appear in certain domains even if a valid and useful meaning is available. Embick discusses this by citing Chung

PAGE 126

126 nouns that are expecte d to be verbalizable ( baina hayu mystery as to whether such words are truly impossible or if conventional use dictates prescriptive feeling more strongly in one sp eech community versus another . Thus while there are gaps for derived words, it is unclear if the properties of a ever cause these gaps at the derivational level. Section 3 .6 has shown that many gaps are not problematic for the proposal that s are a categorical and that morphology like syntax is completely generative. As is pointed out in Embick (2008) any theory that tries to account for conversion is going to have the same problem with gaps. However, there may be cases where items really cannot be conceptualized in certain ways just as there are phrases that cannot logically be conceptualized: colo rless green . This issue will be returned to in Chapter 9 . Many of these seemingly incomprehensible conceptualizations might, however, become better if the takes on a new meaning. This though requires more research. Nevertheless, while the Lexical Dec omposition Hypothesis helps explain categorical flexibility, it limits ones options on controlling insertion for semantically motivated gaps. 3 .7 Features vs. No Features The previous sections have provided reasons to propose features with s in List A. These reasons have led some to backtrack with regards to the strict im plications of Feature Disjoint ness and the Lex ical Decomposition Hypothesis. Some of this backtracking was minimal (Embick 2000) while other proposals completely violated the th eoretical elegance of FD and the LDH (Embick and Halle 2005). Nevertheless, taking FD seriously requires s to be separated from any arbitrary features while the LDH

PAGE 127

127 explains the categorical flexibility of s. 39 This section will show why the aforementioned proposals for phonological and diacritic features with s in List A are based on faulty reas ons or unnecessary assumptions. It will also show that the proposals for sema ntic features might similarly be unnecessary and too costly. 3.7.1 No Features are Optimal s ar e specified with phonological features violates everything about FD. Moreover, Haugan and Siddiqi ( 2013 ) show that s in languag es such as Hopi clearly undergo suppletion. These findings strongly suggest that s in List A are not specified for phon ological features. That is , a cannot be specified with phonological features prior to being merged with a syntactic environment, as that environment determines the spell out of that . Moreover, this suppletion cannot be explained by Readjustment Rules 40 because the forms share nothing phonologically in common as seen with (3 7 ). (37) Hopi Verb Suppletion (Hill and Black 1998: 866, 877) a. pitu b. öki c. puuwi d. tookya pl. subj. The phonological forms of (37 a , b) and (3 7 c , d) share nothing in common for Readjustments rules to work with. Given the fact there exist a plethora of examples like (3 7 ) from the Uto Aztecan languages means phonological features cannot reasonably 39 This seems, however, to leave open the possibility that influence the syntax. 40 As Haugan and Siddiqi (2013) note, the model is better without readjustment rules to begin with. Readjustment rules are not motivated by anything independent of achieving the right output.

PAGE 128

128 be included with s in List A. LVIs must be supplied at List B and be perhaps subject to competition like FVIs. This moves us to why diacritic features are problematic. The first argument against the use of diacritic features is that they are unneces sary. What they are being used to explain can already be explained by other mechanisms, namely allomorphy and the principle that vocabulary insertion targets the no de first (Legate 1999). If the F node bearing gender or class features is present in t he syntax, in many cases it can be proposed that this node is featurally the same in all occurrences (i.e. there is nothing syntactically different between a out of a gender/clas s feature has no real effect on the meaning. In such an analysis the spell out of such a node would be allomorphic, as an LVI would be inserted into the node first, giving it a phonological shape. This phonological shape/identity would then be the trigger for the specific class/gender spell out, much as we saw the LVI /ox/ causing the spell out of the plural node to be /en/ via Contextual Features in Chapter 2 . This could also be used to explain the specific appearance of a thematic vowel (albeit this node is argued not to be present in the syntax). Since the [TH] node does not seem to play a part in the syntax it is called a dissociative morpheme, b eing inserted during the morphological stage of the model. However, the diacritic feature used to match the correct FVI with the [TH] node has to pass through the syntax unnecessarily. Thus instead of being realized by the thematic vowel / i / because of the rule given in (38 ) we can have the rule in ( 39 ). (38) Thematic Vowel Insertion Rule (Embick and Halle 2005:12) a. TH ROOT [X] b. TH[I] a c. TH[IV] i

PAGE 129

129 (39) Allomorphy of Thematic Vowel a. / i / [TH] / {_ b. / a / [TH]/ {_ L AUD As can be seen with (40 ) in comparison to (39) , Contextual F eatures can be used just as diacritic features are without violating FD. In these cases diacritic featu res are not necessary, being just as stipulative as a random listing of exponents in List B. 41 Therefore these types of phenomena do not provide good reasons to include diacritics in List A. This takes us to the conceptual reason against the use of di acritics. s is conceptually either associates a with a category or simply gives it a category (2008:2) In other words, if the di acritic feature specifies for a type of noun class or verb class then it has also given the a has a feature that presupposes a category, then it is not really category 42 As discu ssed in Sections 3.3 .2 and Section 3 .4 .2, in DM the claim that s are a categorical is empirically well motivated. Thus the argument against the listing of diacritic features like phonological features with s in List A is both theoretically and em pirically motivated. The question then is if other features presuppose category membership as well. lexical causatives, that some s inherently come with a causative feature, this feature wo uld require the to be associated with the category verb because causative features are 41 Some of these issues might also be passed along to the phonology at PF. 42 expected.

PAGE 130

130 verbal features. In the case of Embick (2000) though, one may argue that the passive feature argued to be inherent to deponent verbs does not entail that these ot s be verbs since its presence does not affect the meaning and does not cause the clause to be passive. However, its putative presence in List A requires certain s to have (or have had) a verbal association 43 because why or how else would these s acquire such a feature. 44 ) description of Greek nominals, if a lexical plural type feature were associated with certain s, this would give such s a nominal quality and predetermine a type of categorical membership. A lexiadou thus argues against having this feature be a part of the for precisely this reason (i.e. to uphold the Lexi cal Decomposition Hypothesis). In short, it would pr oposals for semantic features with s in List A. This takes us to newer proposals that maintain FD and the LDH. A radically different proposal than that of Embick and Halle (2005) is that of De Belder (2011). De Belder proposes that List A only cont ains morpho syntactic features. s are created via the syntactic operation of Merge. As described, Merge requires the syntax to take a feature from the numeration so it can be merged with another already present feature. At the onset, however, Merge re quires two items to be taken from the numeration. De Belder and van Craenenbroeck (2011) make this process consistent by suggesting that First merge involves the numeration taking only one item and merging it with the already present empty set (i.e. the no thingness that exists 43 Presumably this association would be with pas sive structures. 44 See the discussion of Contextual Features and associations.

PAGE 131

131 before anything is taken from the numeration). This empty set forms the node. This proposal thus views L nodes as featureless place holders as was originally thought. This account , however, also removes the stipulative and generic ( from List A. Returning to this position helps to explain the kind of categorica l flexibility already discussed. Moreover, it helps to explain further observations regarding why s that are presumed to have certain features/properties can be coerced into frames lacking such features ( 40 ) and (4 1 ). (40) English Count to Mass Noun Coer cion (De Belder 2011: 20) a. Three dogs. b. There is dog in the soup. (41) Intransitive Experiencer to Transitive Agentive Verb Coercion a. They floated in the water. b. They floated me out the back of the car. 45 Furthermore, returning to the di scussion of nominals in Greek. Section 3.5. discussed The English realized by the LVI /wa / can optionally undergo this shifting (4 2 ) . (42) English Lexical Plural Coercion a. He hovered over the waters. b. I bought two waters. In (4 2 a) the undergoes type shifting while it generally does not in (4 2 b). Ho wever, if one really thinks about it, type shifting does not have to occur with (4 2 b) and can occur with (4 2 a). This option is claimed to result from different categorial features, not the itself (Alexiadou 2011 ). If the merges with the plural feature when it gets categorized (i.e. the plural feature is a part of the categorical n_ head), it gets the 45 Example is from a woman describing her abduction experience

PAGE 132

132 more standard meaning of (4 2 b) and if it merges with this feature after being categorized its gets the compositional meaning of (4 2 a). This can be taken as evidence that there is nothing about a at its core that requires it to be used in a certain way and that the difference in interpretation is always dependent on the environment. Stipulations, however, against s that appear to resist environments such as (4 2 a) are left unexplained. Al exiadou for one only mentions erger with categorization implies negotiated (apparently idiosyncratic) meaning of the and apparent semi productivity (bette r with some merger above functional heads :37). I f one adopts either the view that s are featureless in List A or the view that s are the result of First merge and the empty set, some of the same insertion/distribution problems remain. One still needs to explain why some s seem to be more suited for certain environments and the possible effects that semantic categories may have for notions. It is acknowledged here that proposing a single item in List A and then explainin g all exceptional or irregular properties as a product of post syntactic insertion rules seems to fail to capture the real reasons for some of these different properties. Thus a solution where s can be individuated without positioning feat ures in List A is a solution that appears to be optimal. 3 .7.2 Indices The argumentation above presented both theoretical and empirical reasons to keep a strict division between what is arbitrary (inserted late) and what is necessary in the syntax (prese nt from the start). By maintaining the LDH, one must keep items in List A featureless

PAGE 133

133 and 3.4.2 . issue is then explaining some of the gaps seen in Section 3 .6.3. This can be achieved by claim ing that s are listed with indices in List A , as discussed in this section, and or by claiming that LVIs are listed with Contextual F eatures as will be discussed in the next section. An alternative proposal (Pfau 2000; Acquaviva 2008 and Harley 2011) is to use indices instead of features to identify s through the derivation. This proposal has to reject the idea of a single in list A and instead proposes multiple featureless s in List A which are identified with (Harley 2011). These indexed items can then be selected for merge to form terminal node elements a phonological form and List C matches the same index with an interpret ation (43 ). (43) PF instruction (List B) LF instruction (List C) /k æ t/ As (43 ) shows, the serves as the linkage between a set o f instructi ons for phonological realization in context and a set of instructions for semantic interpretation in (2011:1). Thus while being identifiable units within the syntax, s are the most abstract of possible elements. One argument for this view is that the grammar cannot identify separate s via phonology or semantics. Harley concludes that (2011:14). In other words, without a context the meaning is unpredictable (special

PAGE 134

134 meanings) and the phonology is un predictable (true suppletion). T he only way to properly identify items in the derivation is to give them these deictic like markers. This approach is also seen as being desirable because it theoretically allows for s to be individuated without violating the LDH (Acquaviva 2008:16). Indices themselves have no meaning and do not consist of features that could influence syntactic operations. Thus everything remains external to the . This account gives each in the syntax an abstract identity, one that can be used to guide insertion at List B without being able to inf luence syntactic operations. The L node (Acquaviva 2008). Insertion of Dog in an L node intended for Cat would be prevented because the index for Cat would only allow the LV I / kaet /. The problem with this proposal is that this indexing would make the Vocabulary Insertion of s competition free. There could be no choice at Spell Out for LVIs unless several exponents where listed with the same index. 3 .7.3 Contextual Features The previous section discussed the advantages of using indices and how they could be used to control insertion. Another way to externally control fo r insertion is with the use of Contextual F eature s, as introduced in Chapter 2 . Contextual features cannot explain why different contexts exist but they do provide a mechanism for explaining distributional differences. Contextual features can be thought of as those features that record the conventional use of items. This convention could also be the result of something deeper, a real conceptual disharmony as discussed in Section 3.6.3. In such cases the semantics will have caused the context.

PAGE 135

135 Returning to the issue of using diacritic feature s, this section will show that C ontextu al F eatures can be used to explain insertion dependencies that depend upon the identity of the if indices are not used. Allomorphy may not always be the best analysis for when the spell out of external morphemes depends on the identity of the . Given the descriptions of s as either underspecified or meaningless abstract elements without a context, the prediction would be that any should be able to work in any context. 46 In Japanese and Mandarin, nouns must be merged with a counter feature to be m ade plural. These counters have a semantic association (albeit inconsistent at times) with the they classify. 47 If this relationship is explained as allomorphy, it would require all the counters to mean the same thing at LF, which in some cases appea rs untrue (i.e. many counters are derived from lexical nouns and can be argued to continue to have separate meanings). An op tion here would be to say that Contextual F eatures limit the LVI ( exponent) that can be inserted in a domain that differs in m eaning/f eatures as demonstrated with (44 ). (44) LVI Contextual Features at List B 48 a. Class: Flat [ ø ] /{ [mai [ n [__] /ti [ ø ] /{ [mai [ n [__] [ ø ] /{ [mai [ n [__] 46 mergi ng in an inappropriate frame because it has not been specified until it merges with such a frame. 47 An allomorphic explanation might be used as done with classifiers but counters are different than class markers as they generally have a semantic associati on with the meaning of the noun while classifiers and grammatical gender markers do not. 48 This example is problematic, however, because it assumes that there are grammatical features in the syntax that can differentiate the terminal nodes spelled out as mai versus ko . Further research on modeling counters in DM is needed

PAGE 136

136 b. Class: Small/round/compact [ ø ] /{ [ko [ n [__] [ ø ] /{ [ko [ n [__] Example (44 ) serves as a model for how exponents can be regulated for insertion into the corr ect domain (i.e. a domain w h ere semantic features will not contrast). s that are generally conceptualized as having flat qualities will be counted with mai and s that have a small/round and or compact quality will be counted with ko . Also , as dem onstrated , new loan words are divided according to their semantic properties. Cases such as these might be understood as semantic harmony where the use of the wrong counter would constitute strange or unconventional usage rather than ungrammatical usage . One may also utilize Contextual F eatures to explain agreement between a and an animacy requirement in languages where verbs agree with the animacy of their subjects. Likewise in the case of th e Greek nominals (Alexiadou 2011 ), it was seen that only a viable could be inserted in a domain categorized by a n n_ head wi t h a plural feature. Contextual F eatures can be used to explain why this process appears to be se mi productive. In other words, Contextual F eatures can be used to explain accidental gaps in all distribution patterns without violating FD or the LDH. 3 .7.4 Contextual Features versus Indices As explained in the previous two sections, one can determine LVI insertion with matching indices in List A and B an d or by limiting insertion via Contextual F eatures in List B. The questions is if one needs both. One could argue whether it is more costly to stipulate that only certain s merge with certain features at a low level or to stipulate that these s come with an identifying index. This discussion concludes that both are necessary to explain the ir regularities seen with Section 3 .6.

PAGE 137

137 It is unclear how indices would be used to explain gaps. Take for instance the insertio n rule given here with (45 ). (45) List A PF instruction (List B) /tejp/ Given (45 ) , nothing [the poster with pushpins]. It will spell out taped the poster with pu shpins which is semantically strange argued by Acquaviva (2008) and Harley (2011), the grammar has no way of knowing that this particular will be problematic in this domain until presumab ly an interpretation has been reached at List C. That is , in order for an index to block the merging of / tape / in the greater context , it must be listed as not viable f or this context. This is where Contextual F eatures come in (46) . (46) List B Context ual Features /tejp/ { n_, 49 In (46 ), the LVI /tejp/ is only listed as a verb that has framed an already categorized noun v_ . This prevents insertion into this domain. In ot her words, b y allowing an index to work as a deictic marker , we both to its to its copy in List C which lists all t he interpretations conventionally given. If the structure already created does not 50 In this way s in List A are like bundles of features (one does not merge them just anywhere) but different in that such 49 Crucially there is no listing for insertion into the following environment {v_} 50 A lot of further research and consideration appears to be needed here.

PAGE 138

138 merging is not formally disallowed by the Grammar as the item contains nothing in and of itself to make it ungrammatical in any given context. 3.8 The Encyclopedia The Encyclopedia, List C, is an und er d eveloped component in Distributed Morphology. Currently, however, it is unclear if any consensus has emerged from such debate. Uncontroversial, thoug Encyclopedia links non linguistic knowledge and meanings with s+ exponents morphemes are typically not idioms, but l (or n idiom (Harley and Noyer 1999: 4). It is thus responsible for storing the idiomatic meaning a , word, or phrase might have (Harley and Noyer 1999: 2). Accordingly, every in List A or every (+ LVI + Context) must have an entry in the Enc output of PF/LF, whi s interpretation can evaluate the context of the larger structure (2000:4).

PAGE 139

139 One could thus propose that the Encyclopedia has the power to influence insertion as suggested in the previous section . Ho wever, according to Siddiqi (2005: 8): The Encyclopedia p lays no part in the construction of a derivation, nor does it serve any role in determining whether the derivation is well formed. Rather, it simply assesses the interpretability of the sentence. may even appear in an environment that our real world colorless green ideas sleep furiously to reiterate the known distinction b we would never use (2000: 2). Thus the ability to explain gaps that appear to be based on conceptual disharmony is not established. It is uncertain how the En cyclopedia could be used to explain such gaps. 3.9 Early Insertion and Late Insertion T he timing of LVI insertion into a node has been a debate within DM that is correlated with the debate over what a in List A is compos ed of. As discussed in Chapter 2 , the foundational view of DM as a Late Insertion Model is that Late Insertion applies to all syntactic terminal nodes (Halle and Marantz 1993, 1994; Marantz 1995, 1997; Harley and Noyer 1999 etc.). This stems from the already discu ssed ideas of F eature Disjoint ness and the Lexical Decomposition, but different behaviors have also led to proposals for Vocabulary Insertion to be split between l nodes and f nodes with Early Insertion happening at l nodes. The arguments for Early Insertion thus largely stem from the same places as those for specifying s in List A with features. If s are inserted early, their properties can affect the syntax, limiting where they appear and, if necessary, determining how

PAGE 140

140 they are spelled out. However, it is unclear when and where a in List A acquires these early features. Descriptions of Early Insertion sometimes appears to be split, positioning some semantic syntactic features with s in List A and leaving phonological features to be supplied later by List B. Moreover, phonology might be supplied right befor e categorization in the domain, before morphology as entertained in Embick (2000), or phonology might always be present as discussed in Embick and Halle (2005). 51 Given this inconsistency (there does not seem to be anything that mandates when somethi ng has to be inserted early) and the fact that Haugen and Siddiqi ( 2013 ) showed that Late Insertion must apply to s because true suppletion is possible for s , and because Early Insertion prop osals violate Feature Disjoint ness, the possibility for Early Insertion will be rejected. Thus while Late Insertion theories may have to stipulate why a particular LVI cannot be inserted in a domain or why an already specified cannot be used in a certain derivation, Late Insertion maintains the original proposal, unites the process of Vocabulary Insertion, betters explain the empirical findings of suppletion, and is more conceptually appealing as it creates no exceptions for the Principle of Feature Disjoint ness. 3.10 Concluding Remarks Regardless of which of the LDH views of s in List A we adopt, such views agree that anything that could restrict LVI insertion at a node should start 51 In some ways everyone agre

PAGE 141

141 externally. Viewing s in this manner maintains the possibility for free choice at nodes and maintains th e elegance of Feature Disjoint ness. Versions of DM that propose phonetic and diacritic features as possible features, as in (Embick and Halle 2005) or (Embick and Noyer 2 007), violate Feature Disjoint ness. Moreover, while the size of list A is not a huge concern, the more s are specified, the more powerful List A becomes. If s are fully specified, then List A contains a copy of every the individual has at his or her disposal along with information about how that i s used. On the other hand, if s are featureless, we can explain the natural flexibility of s at the cost of h aving to stipulate indices and Contextual F eatures. If these costs are too high then perhaps a return to semantic features will be favore d. Considering all of this, it is taken that Roots contain no grammatical features at List A and that all derivation inconsistencies are best explained by either arbitrary groupings in List B and or an index marker that indicates how a particular concept t ype can be categorized.

PAGE 142

142 CHAPTER 4 ADPOSTIONAL FORMS IN ENGLISH AND MANDARIN 4 .1 Introduction Chapter 1 discussed how consistently categorizing linguistic items is a problem u nclearly defined. Chapter 2 introduced the framework of Distributed Morphology. I n DM it is believed that adjective have no universal significance ( Harley and Noyer 1999:7). These terms are merely labels that should be taken to refer to something more discrete in the grammar. Chapter 3 discusse d how the difference between categories cross linguistically can be captured in DM with the Lexical Decomposition Hypothesis and the fact that a Root takes the grammatical identity of the feature(s) it first merges with (variation to the precise identity of this/these feature(s) can be thought to explain cross linguistic differences amongst traditional categories as well as other lexical categorie s). The occurrence of traditional categories cross linguistically is the result of nodes co occurring with similar grammatical features (i.e. nodes being categorized by similar features). In other words, a noun in one language is not necessaril y the exact same thing as a noun in another language. Likewise the term adposition can be considered a label that captures a range of features with a similar function (i.e. a relation al item that links F i gures to Grounds ) . This c hapter is organized as fol lows. Section 4.2 shows that many forms that appear in adpositional domains in English also appear in several other lexical domains, indicating that these forms originate as node s as discussed in Chapter 3. It also shows that these forms vary in meaning as would be expected for items in

PAGE 143

143 different domains as also discussed in Chapter 3. Section 4.3 shows, however, that not all English adpositional forms involves a Root node, giving a list of FVI adpositions in English. Furthermore, 4.3.1 shows that this is expected given that the other major lexical categories also have associated FVIs . Section 4.4 introduces adpositions in Mandarin Chinese. Section 4.4.1 gives the major positions where prepositions and postpositions can occur in the clause. Section 4.4.2 then explains how prepositions can be differentiated from their closest lexical neighbor, verbs. Section 4.4.3 likewise differentiates postpositions from their clos est lexical neighbor, nouns. Section 4 .5 then shows that many of these forms have a lexical d istribution as was seen with English adpositions. Section 4.6 also shows that a smaller subset of these forms do not have a lexical distribution and thus can be co nsidered, FVIs. It is then concluded that the adpositional domain can contain a node in both English and Mandarin and that this requires a new type of categorial head in DM. 4 .2 Lexical Prepositions in English In Chapter 1, it was discussed that English has a rather large number of prepositional forms in compar ison to many other languages and that the exact number of prepositional forms in English is not agreed upon. A nalogy can be drawn to the fact that the precise number of verbs or nouns in any language is unknown because languages have a large and fluctuating number of s . 1 It is here argued that many so called adpositional forms are lexical, involving Root nodes categorized by a proposed little p_ head . 2 Thus the number of adpositional forms in English is not a 1 The proposed a on. 2 This position is categorized by an adpositional feature (little p_ as will be formally discussed in Chapter 5).

PAGE 144

144 concrete number as some new Root s may be appropriately used in such a frame. 3 This helps to explain the comparatively large number of adpositional forms in English 4 and helps to explain why certain forms in English may be used in adpositional domains as well as other lexical domains. The items to be discussed are given in (1). (1) Prepositional Items Found in Other Lexical Domains 5 Off, up, near, out, in, down, over, under, and on The items in (1) can clearly can be used as prepositions as d iscussed in Chapter 1 (i.e. they can be used to relate an adjunct DP Ground to a Figure DP and will license the adjunct DP by putatively assigning it case). Furthermore, most of these items get used in the particl e domain as defined in Chapter 1 . These items also arguably have richer semantic content than some other adpositional forms . However, the argument for their lexicality stems from the fact that the items in (1) appear in one or more verbal, nominal or adjectival domain, indicating that they are items as discussed with Identification Guideline A in Chapter 3 . T hus they are proposed to originate as nodes filled by LVI s and not bundles of morpho syntactic features, FVIs , because this best accounts for this lexical distribution. It is also important to say that the items in (1) are not proposed to be special s. What separates these items from other items is meaning. 6 The core/potential meanings of these s easily correspond with different relational positions, ma king them natural candidates for 3 What exactly makes it appropriate is still a matter of debate. 4 5 This li st leaves out the prepositional forms that begin with /a / or /be /. These forms will be discussed in Chapter 5. 6 This is a difficult position. As discussed in Chapter 2, Harley (2011) argues that neither phonology nor

PAGE 145

145 adpositional domains (most concepts do not have the potential to be used in this way). This explains how the item near , which more clearly resembles other lexical items, works as a prepositional item . The remainder of this section gives the distributional evidence for the proposal that many English adpositions involve a node. It will also be shown that some item s used in adpositional domains may at times correspond with several associated meanings in other lexica l positions . This extension of meaning or variable meaning corresponds with the property of s that allow s for a range of possible meanings in a given context ( c.f. Embick 2012), a property which is thought to belong to a item and its first categ orizing h ead as is discussed in Chapter 3 . It is thus argued that a lexical distribution followed by novel meanings in different frames strongly suggests that many adpositional forms are LVIs realizing nodes . 4 .2.1 Nominal and Verbal Distributional Evidence: Proposed LVI Prepositions This section examine s a subset of mono morphemic forms 7 in English, (1), that are traditionally described as prepositions but may also operate in other lexical domains. The first set of evidence that (1) is a list of oot items is that in addition to appearing in an adpositional domain, seven of the nine listed forms can be found in a transitive verbal domain (2). (2) Verbal Domain (Transitive) a. John downed the drink. b. The Heat down the Hawks 8 c. The farmer inned the hay d. The runner neared the finish point e. The mafia offed the FBI informant. 7 morphemic, consisting of node and category defining head. 8 These are the names of basketball teams.

PAGE 146

146 f. The newspaper outed them for lip synching . g. The horse and the r ider overed 9 the jump. h. Social media ups world awareness. Example (2) shows that all the items from (1) except under and on are used as transitive verbs in English. This non usage within the DM machinery can be controlled fo r with Contextual F eatures as discussed in Chapters 2 and 3 . Moreover, as also discussed in Chapter 3, this type of gap can be classified as an accidental gap. In other words, it is the resu lt of differential use, no t grammatical differences . The claim ther efore is simply that under an d on have not obtained a verbal usage (or common verbal usage), not that such usage would be ungrammatical or unexpected. In DM, those forms with this use would be listed with the contextual feature {v trans _} while under and on would not, as seen with (3). (3) OFF UP NEAR OUT IN DOWN OVER [ ] / {v trans _ , } It is important to recall that Associated F eatures are the features used during Vocabulary Insertion to pick the best VI candidate. Contextual F eatures on the other hand simply limit where a V I will be considered for insertion. This feature therefore does not mandate that these forms can only be used as transitive verbs. In addition to the lexical distribution seen with (2), the meaning associated with some of the forms in the provided context also supports a derivation analysis. E xamples (2a ) and (2 b) show the same form, down , with associated but different meanings in the transitive verbal context (i.e. the meaning of down in (2b) is a n abstract extension of the act of literally putting your opponent dow n . In this case it means to defeat ) . This meaning would not be expected if down was only associated with the 9 Attested usage found with Di

PAGE 147

147 meaning(s) it has as a preposition. Likewise the meaning of off predictable derivation of the prepositional meaning. An other fact supporting the proposal that these involve a node is that a subset of (1) is also used as intransitive verbs (4). (4) Verbal Domain (Intransitive) a. The end nears . b. Off before I call the sheriff. c. The truth will out . d. She upped and perjured her immortal soul. ( The Border Line Case by Margery Allingham ) e. (You) down. f. (You) up. In ( 4 ) it c an be seen that perhaps six of the nine items from (1) are used in an intransitive context as well. 10 C onsidering though that some uncontroversial s (hold, hit are only used transitively , 11 i t is interesting to note that there appear to be no forms from (1) that only appear in an intransitive context, unlike verbal s such as die . 12 This could be due to the smaller sample of items examin ed here compared to the larger number of items that are used in the verbal domain or something else. Presently , though, t his observation will also be explain ed as the result of Contextual F eatures limiting insertion (5) . (5) a. OFF UP NEAR OUT DOWN , [ ] / {v trans _ v in trans _ , } b. IN OVER [ ] / {v trans _ , } c. ON UNDER [ ] / {?} 10 One may argue that near is not really intransitive here and that its object is implicit. In addition, examples (e) and (f) may be deemed inappropriate in normal speech between two adults. However, these points do not greatly concern what i s being observed. 11 As for hit, this statement may not be true as it is easy to imagine a sentence such as H e hits akin to He bites . 12 die and kill are suppletive forms.

PAGE 148

148 As indicated with (5), the form s under and on are not used as verb s at all . However, it is being argued that nothing other than convention and accidents is p reventing such usage . I t thus seems perfectly conceivable that the sentence the lights will on tomorrow he lights will turn on tomorrow could be constructed in English. In addition to verbal domains, if the items in (1) are the product of a node, we would expect to find at least some of them in a nominal domain ( 6 ). (6) Nominal Domain (Count nouns) a. The team has two downs remaining b. I have an in with the boss. c. We own several inns. d. We just need one more out. e. This hatch is my only out. f. Rahul Dravid scored on that over. 13 g. She gazed at the ship in the offing 14 Example ( 6 ) shows that at le ast four items from (1) appear in nominal, countable domains . This additional distribution provides further evidence for proposing the existence of a node. It suggests that the source of these items is not specified for either a verbal or nominal dom ain, indicating the pre sence of an underspecified . Th e se four items would thus be listed as in (7). 15 (7) DOWN , IN , OUT , OVER [ ] / { [n_ NUM _], v trans _, } Moreover, as was seen with example (2), examples (6b) and (6c) and (6d) and (6e) show the same form interacting with different interpretations. In (6c) one may object to inn being considered a derivation of the proposed IN because of the different 13 A term in the game of cricket. Refers to a set of play where after six ball are bowled the fielding team switches places. 14 A n 15 It is tenuous whether some of these forms are ever tallied but the presence of indefinite article indicates that these forms are in a countable domain and th u s are available for counting. I t seems perfectly reasonable for someone to say now I have two ins with the boss.

PAGE 149

149 IN is the etymological source of as it simply is a place that is not outside. 16 The meaning is thus arguably not Root s, such a proposal is deemed justifiable. Moreover, certain meanings for some of these forms only appear to co me from being in a plural context ( 8 ). (8) Nominal Domain (Lexical Plural) a. LeBron James has ups . b. S he was supposed to be enjoying her offs . ( The Case of the Love Commandos by Tarquin Hall ) c. downs . 17 ( Title : University News of Saint Louis University by Laura Hicks) In ( 8 ) the forms ups , offs , and downs appear to be plural and this context seems to correspond with a particular meaning. 18 Furthermore, it is uncertain if any of th ese forms+meanings are ever used in a singular context (although nothing would prevent such usage in theory). One may then consider the possibility that [s] is not indicative of a number node but rather a plural little n_ head as discussed in Alexiadou (20 11 ). This would again suggest the presence of a node as the meaning of down in ( 7 a) is unpredictably different than ( 8 c). Furthermore , l ike the transitive/intransitive distinction, the forms in question are not equally dispersed (i.e. up and off are not used as singular nouns). Given this one cann ot argue that a prepositional head merges with a nominal head and then a number head because we do not have evidence for the intermediate 16 Compare this with the Japanese usage of uchi which means inside and the phonological form for house uchi . 17 From: learn from life%E2%80%99s downs/ 18 This discussion is igno ring set coordinations such as ins and outs .

PAGE 150

150 stage for the forms up s and off s . Instead it appears that there is a node and that its specification depends on Contextual F eatures (9). (9) OFF UP , DOWN , [ ] / { n plural _ , } The nominal context of off and up could be listed as one would list the s PANT and SCISSOR . 4.2.2 Adjectival Distributional Evidence: Proposed LVI Prepositions In addition to the v erbal and nominal domains discussed, the items from (1) also appear in adjectival domains as would be expected if the forms begin as Root items . Examining these forms, however, in adjectival domains is perhaps more complicated than the nominal and verbal domains. One reason for this is that these forms commonly compound or incorporate with other items ( 10 ). (10) Left Adjoined Compounds a. The off button *The very off button 19 b. The upside *The very upside c. The near east *The very near east d. The outfield *The very outfield e. The Infield *The very Infield f. The downtime *The very downtime g. The overtime *The very overtime h. The underworld *The very underworld i. The on button *The very on button j. The bystander *The very bystander In (10) the items from (1) are modifying another Root item, but since they themselves cannot be modified by very , as shown in the right column, they are not being used as attributive adjectives. Rather they are left adjoined to another Root item. These forms may be derived via primary compounding (more evidence for hood) or as 19 the very one you had mentioned.

PAGE 151

151 particle 20 incorporation. There are cases, however, where very can modify the forms in question , suggesting that in such cases the form is operating like an adjective (11) . (11) Attributive Adjectival Domain (Modified by Very) a. A very off day b. A very off individual 21 c. The very near future d. The very uppity man e. The very uppish crowd f. The very underly people g. ? The very in crowd h. ? A very down man i. ? A very up child As seen in ( 11 ), some of these forms seem able to be modified by very , suggesting that the forms in these examples are acting lik e normal attributive adjectives (12) . (12) OFF , NEAR , UP , UNDER , ? IN / { a _, The derivational morphology in (11d f) is further evidence and will be discussed in detail in Chapter 5. The reason, though, that some forms appear to resist very modificatio n is not entirely clear. ( 11 g i) do not seem ungrammatical but attested usage is scant at best. In addition to the attributive domain, some of these forms can be found in the predicate adjective domain (13) . (13) Predicate Adjectival Domain (Modified by very ) a. I am feeling very down and upset. b. We ar e so/very down for this trip. c. The store is very near. d. He is a little/very off. e. I am very out. not in the running 20 5. 21

PAGE 152

152 Example ( 13 ) sho ws the forms in question being used as predicate adjectives as they are here modified by the intensifier very . More of the forms can be modified by other so called intensifiers, but are awkward when modified by very (1 4 ). (14) Predicate Adjectival Domain (Modi fied by other intensifiers) a. I am so/?very on to what you are planning. on now. c. He was so out . The official must be blind. d. Faded jeans are so/quite/?very out these days. e. She is so/quite/?very over with bad relationships. f. I assure you that the patient was quite/?so/?very under during the procedure. unconsciousness g. After the in sult, the boy was so/quite/?very up for a fight. Without making further specifications, the contextual environment for these forms would be as (15). (15) OFF UP , DOWN NEAR , ? ON , ? OUT , ? OVER [ ] / { a pred _ , } The reason modification by very is often ungrammatical or awkward perhaps is from the fact that these items are here being categorized by features different than predicate adjective features. A few of the items from (1) are also used as manner adjectives, modifying the direction/location of the verbal action (1 6 ) . (16) Manner Adjectival Domain a. The subtle Fiend, though inly stung with anger and disdain, dissembled. ( John Milton Paradise Reg ained Book 1; Lines 465 466) b. H ow downly he look ed upon my behavior of the past 22 ( Not my Fairytale blog story) As (1 6 ) shows, two of the items from (1) are found in this domain , (17). (17) IN , DOWN , [ ] / { a manner _ , } 22 From: not my fairytale 11 the ball

PAGE 153

153 Example (17) is just another specific lexical domain that some of the adpositional forms are found in , adding more evidence to the argument that the forms in (1) are LVIs . This section has provided data to show that the fo rms o ff, up, near, out, in, down, over, and on may operate in the same domains as other unambiguous items: varieties of verbal, nominal, and adjectival domains. This simple fact is evidence that the vocabulary items that map the phonological strings of these elements to terminal nodes must be LVIs. After all, the VPs, NPs, and APs in the example sentences above must contain a of some kind to provide them with the conceptual meaning observed. The claim here , though, is not that all of the example s given are the product of a directly categorized by a little v_ , n_ or a_ head but that sometimes this is necessary to explain the derived meaning. In cases where the variation in meaning did not suggest compositionality, this section showed that th e se variations were comparable to the meaning variations seen with other bare a categorical s defined in different contexts as discussed in Chapter 3 . For instance the use of off in (2e) to mean kill does not incorporate a Figure/Ground association th at would be expected if off began as a type of P in this context. 23 In other cases, it may be best to analyze them as s categorized by a prepositional h ead (to be proposed in Chapter 5 ) and then another head. An alternative view would be to assume that all of the phrases above contain a null selected by the (functional) P head. It is unclear how this alternative would be helpful, however, as one would need to stipulate that anytime a preposition is used in another categorical context it must selec t a null . It might also alternatively be 23 The meaning of off in this context can be related to other uses of off as in we are d isconnecting this

PAGE 154

154 proposed that this phenomenon is the product of a less prepositional or particle node merging with a verbal head. Again it is unclear how this would be helpful (beyond trying to maintain that the members of the category adposition are always functional). Such a proposal would incur many problems such as explaining the variable like meaning seen with these forms in different contexts. 4 .3 Functional Prepositions in English In S ection 4.2 the lexical flexibility of English prepositional forms like up, in, off, and out was used as an argument that these prepos itions involve a node categorized by a variety of functional heads . As might be expected, there exists a set of strictly functional adposit ions that do not appear to involve a node. These forms are discussed in this section. 4.3 .1 Functional Prepositions: N o Lexical Distribution While the majority of English prepositions show lexical flexibility, some do not. This includes at least the following ten prepositional like items (18). (18) FVI Items Found in Prepositional Domains in English at, by(agentive), for, from, of, to, via, with, until, and as 24 These adpositions do not occur in verbal , nominal, or adjectival contexts (morphological or syntactic) (19) . (19) a. Verbal: *they atted, bied, fored, fromed, ofed, viad, withed, untiled, ased b. Nominal: *the ats, bys, fors, froms, ofs, tos, vias, withs, untils, ases, c. Adjectival: *very at, by, from, of, to, via, with, until, a s In addition to not having a lexical distribution, the elements in (18 ) are often noted as being strongly on the functional side of the continuum, mainly because they have less 24 It not debated in the work whether agentive by should be considered a type of preposition or as something else. The same applies to the form as.

PAGE 155

155 substantive semantic content than other prepositions and instead are empl oyed as largely functional elements. Some, such as of , are employed chiefly as case markers with no semantic content at all while others, such as at, to, and from , seem to encode for Location or Path. As Mentioned in Chapter 1, Svenonius (2008) divides Eng lish prepositions into Projective, Bounded, Extended or Particle forms with many of the forms being members of more than one category. However the aforementioned forms are not part of any of these general categories. Instead, from and to represent the spec ial canonical Path features Source and Goal and of is inserted under K in th e KP as a sort of last resort. The form at is largely absent from his analysis but it would make sense if it acted as default locative marker. Thus these forms seem both distributi onally and semantically different than those forms arg ued to be represented by LVIs. 4 .3 .2 Implications of FVI Prepositions The observation the category P can be split into lexical members and functional members should not be startling. This observation simply means that the category P is similar to the other lexical classes as it also has a closed set of functional items , (18). It is believed that there is a functional member or set of functional members for most if not all domains that define lexical ca tegories ( Schütz e exist significant closed classes of grammatical formatives that are simply subclasses of the head of such as auxiliaries for verbs and pronominals for nouns. The identity, however, of the grammatical subclasses for A and P is often less clear. In English Schütz e (2001) specifically identifies the functional member for these classes as one for nouns, be for verbs, so for adjectives an d with for prepositions, calling them semantically empty elements that correspond with a particular category. In the account described here, the

PAGE 156

156 other items from (18) would be added to with , forming the subclass of FVI members for the category P in English . S chütz do and of are not members of these lexical categories, but rather of associated to be based on the view that the categories N,V, A and P are absolute which is not the view taken here or the one generally advocated in DM. These other heads may be argued to represent varieties of n_, v_, and p_ . 4 .4 Adpositions in Mandarin U ntil now the discussion of adpositi onal forms has largely depended up on English data. L exically distributed adpositional forms , h owever , can also be found with other languages. T his section will focus on those adpos i tional forms in Mandarin Chinese that are found in lexical positions as wel l as discuss those forms that are likely to be the direct representation of grammatical features, FVIs. 25 I t should be clear , though, that by labeling item s as adposition s in Mandarin is not the same as featurally equating English and Mandarin adpositional domains . T his label , nevertheless, is argued to be appropriate as it capture s the fact that these items are used to relate Figures to Grounds in both languages. 4 .4.1 Adpositional Positions in the Mandarin Clause Items that appear to be prepositional in M andarin are described as appear ing in four different syntactic positions: First Topic Position, Second Topic Position, Post Subject Preverbal Adjunct Position and as part of the complement of a verb . 26 Items that 25 Many thanks to Dr. Lin , Dong Y i, Dr. Si , Chen and Yan, Le for their grammaticality judgments over the course of this work. If a mistake is present, it is entirely my own. 26 Arguably this is not always the same thing in this position.

PAGE 157

157 appear to be postpositional in Mandarin may appear with a DP in at least five different syntactic positions: First Topic Position, Subject Position, Second Topic Position, Post Subject Preverbal Adjunct Position, and as part of the complement of a verb (Li 1990; Cheung 1994; Djamouri et al. 2012) . 27 The template for these positions is given with Table 4 1. Table 4 1 . Mandarin sentential t emplate 28 { First Topic } Subject { Second Topic Position } { Adjunct Position } Verb Phrase [ { Complement } ] 1. zài LOC Shanghai dàji we zài LOC restaurant in eat breakfast In Shanghai we in a restaurant eat breakfast . 2. zài LOC Shanghai we eat breakfast In Table (4 1) the standard positions holding DPs are given. Positions given in {brackets} are places where both pre and postpositions can occur while the subject position can only have a po st position accompany its DP ( Djamouri et al. 2012 ). Moreover, while these adpositional positions share features with o ther lexical categories (verbs and nouns), most importantly, the prepositional position can be contrasted with 27 The range of positions have not been summarized as such in any of the cited work. However, with the data provided such a description can be achieved. 28 In Table ( 4 1), it is unclear if the Second Topic position and Adjunct position should not be conflated with regard s to the distribution of prepositional phrases. It is difficul t to tell if a prepositional phrase is ever in this Second Topic Position. However, this position clearly exists as one can move the verbal complement to it as was seen with (2) from Table ( 4 1 ). Also negation exists outside the adjunct position, giving it a boundary.

PAGE 158

158 always requiring an overt DP object. The prepositional doma Ground also naturally differentiates this position in comparison to nominal and adjectival positions . One the other hand, the postpositional position can likewise be contrasted with its closest lexical counterpart, the nominal pos ition , by its relationship with another DP and the inability for that DP to be omitted . 4 .4.2 Determining Prepositions in Mandarin T he s ection above describe d some of the major structural positions in the mandarin clause. If one is to argue that adpositio nal domains can contain Root nodes in Mandarin as was done for English in Section 4.2 , there is a need to distinguish this domain from the other lexical domains that uncontroversially contain Root nodes. Otherwise one may argue that what is witnessed i s just the categorization of a different type of verb or noun as is seen in the literature . 29 The following discussion will separate items in prepositional domains from items in verbal ones. Instead of coverb, t he term adposition or preposition will be used because the position in question will be shown to be different from both verbal and adjectival positions and because these form s have a similar function to other prepositions in other languages . 30 As discussed in Chapter 1, the defining behavior of the ma jor lexical categories differs from language to language, making universal claims difficult. Furthermore, a s far as disambiguating prepositional items from other items in Chinese, disambiguating prepositions from verbs is the most challenging because moder n prepositions are argued to historically be derived from Old Chinese verbs (Norm an 1988: 92; Li 1990). 29 In part they would be correct if such an argument contained the need to say that in such cases different features are framing the Root node , as is argued in Chapter 5. 30 T he label coverb is not meaningful in and of itself within a DM account .

PAGE 159

159 Verbs and prepositions are both transitive. Some linguists therefore classify at least some of the adpositional forms as coverbs (Packard 2000). 31 Moreov er, the distinction between predicate a djectives and verbs is not as clear in Mandarin as the distinction is in English. Predicate adjectives are argued to be a type of verb , creating overlap in the traditional terminology and thus somewhat linking this ty pe of item with prepositional items from a lexicalist perspective where items are marked with syntactic identities (i.e. a group of verbal elements versus a group of nominal elements) . E vidence for this similarity is found with the possibility of negation. In Mandarin , verbs, predicate adjectives and prepositions can all be negated by bù 32 while nouns cannot. (20) bù he NEG buy book In example (20 ) we see the negation of the matrix verb by the presenc e of bù d irectly preceding it. This ty pe of negation can also occur with predicate adjectiv es (33 ). (21) bù máng I NEG busy In (21 ) we see the MÁNG acting as the predicate of this sentence. If the Root in thi s context is being categorized differently (i.e. as a noun ) then it must appear with the copular verb shì . Moreover, if MÁNG appears in an attributive adjectival domain, then it cannot be negated with bù . For this reason and because there is no 31 This classification most notably applies when these forms appear directly before the matrix verb . In this position these forms can be said to aid verbs by specifying the location, direction, source or goal of the action . However, this is also the function of adpositional elements in many languages. 32 Some verbs such as méi .

PAGE 160

160 overt copular verb, some argue that predicate adjectives in Mandarin should be labeled as stative verbs ( Cheung 1994) . Likewise p reposit ional forms may also be negated with bù as i s shown with (22 ) . (22) Negation of Location/Adposition Marker (Cheung 1994:52) a . zài he LOC bookstore buy book b. bù zài he NEG LOC bookstore buy book In ( 22 b ), the negated version of (22 a ), the negation marker comes before the prepositional element zài and not the main verb In this position prepositional elements in Mandarin appear to have a verbal quality to them in addition to their transitivity, hence why the label coverb is sometimes used . N evertheless, while there are similarities between the two domains, the verbal and prepositional positions can be distinguished. Mandarin lacks t ense markers but it does have a spectual markers: le to show that an action is or will be completed and guò to show if something has been experienced . These markers appear either after a verb or after the verbal object . Elements , however, in pr epositional domains cannot head phrases that contain aspect markers (Her 2006) . That is le and guò cannot appear between a preposition and its object or after the preposition al object as is demonstrated with ( 23 ). (23) Aspect Markers and Prepositional domains. a. (* / ) (* / ) zài (* le/ guò ) s (* le/ guò ) qùnián zi LOC ASP shanghai ASP I last year eat dumplings

PAGE 161

161 b. (* / ) (* / ) zài (* le/ guò ) le/ guò ) zi I LOC ASP shanghai ASP eat dumplings c. (* / ) (* / ) fàng zài (* le/ guò ) zi (* le/ guò ) I put book LOC ASP box thing in ASP E xample s ( 23 a c) show definitively that aspects marker cannot be part of the prepositional phrase as their presence renders the sentences ungrammatical . The aspect markers cannot be before or after the DP ground. Thus this difference can be used to delineat e the prepositional position from the verb al position . I n addition to not being able to support aspect markers, the object of a preposition cannot be omitted (Djamouri and Pa u l 2009; Djamouri et al. 2012). On the other hand, the object of a verb may be om itted in certain circumstances. Compare the grammaticality of ( 24 ) , a verbal domain, and the ungrammatical ity of ( 25 ) , a prepositional domain . (24) Verbal Object Omission (Adapted from Djamouri et al. 2012:7) , ( ) guo shéròu, wáng péng guo (shéròu). 1 SG eat EXP snake.meat Wangpeng also eat EXP (25) Ungrammatical Prepositional Object Omission * ( ) Wáng péng , ( wáng péng ) bù shú Wa ngpe ng 1 SG with (Wangpeng) NEG familiar angpe In ( 24 ) the object of the verb is omitted in the second clause. It is possible to have the object of the verb be omitted if it is retrievable by context. However, this pattern does not apply to i tems licensed by prepositions. O ne cannot omit Wáng péng from the object of the preposition position in ( 25 ) despite the fact that it is als o

PAGE 162

162 retrievable via the context already provided. P repositions in Mandari n therefore need to have an overt DP. This also obviously separates the prepositional position form the predicate adjective and nominal positions as these two positions do not require another DP. As mentioned, prepositional items most closely resemble verb al elements as they are thought to be historically derived from verbs. If this is true, it helps explain why these forms have verbal chara cteristics such as negation and transit ivity . It also may help to explain the fact that these forms may still be used as verbs (a s will be discussed in Section 4.5.2 . Nevertheless , the lack of aspect markers and the inability for ob ject omission with prepositions shows that this domain is different than the verbal one in Mandarin Chinese . This now leaves one to delineate the other potentially confusing classification, separating postpositions from nouns. 4 .4.3 Determining Postpositions in Mandarin This section distinguish es postpositional elements from their closest lexical counterparts, nouns. The issue can be cloudy if one thinks of forms being absolutely linked to categories. This is because forms that can be nouns operate as postpositions or vice versa. If one views the distinction as a difference between syntactic domains, straightforward differences are then found. T he main difference between a form in postpositional domain and a form in a nominal domain is that the postpositional domain cannot itself be the object of the preposition (i.e. a DP Ground) and it cannot be stranded . S imilar to the classification issues fo r prepositional forms (prepositions or coverbs), postpositions are sometimes referred to as relat ional nouns (Li 1990; McCawley 1992) or as a t ype of substantive (Norman 198 8). On the other hand,

PAGE 163

163 Djamouri et al . ( 2012) argue that they are not nouns for the reasons that are similar to those mentioned above. For this reason and because the postpositional domain semantically helps to relate a DP Figure to a DP Ground, the forms defined in this section will continue to be referred to as postpositions . In other words, the forms in postpositional positions have a function that is different than other nominal positions . 33 Li (1990:4) makes the obvious point that even if we wish to call the elements to be discussed postpositions instead of nouns, they cannot be equa ted with prepositions. Postpositions are words with relational meanings that occur after the DP that they modify (26 ) unlik e viable prepositional forms (27 ) 34 or nominal forms (28 ). (26) zài he LOC bookstore in buy book (27) * / / / zài he LOC bookstore at/in/facing/from buy book (28) ?? 35 z zài chéng shì China sea LOC city north China sea is In (26 ) the form e n by the ungrammaticality of (27 ), not any of the prepositional forms can appear in this position. 33 The differ ence between a type of noun and a postposition will not be the focus of the argument here (although the fact that the controversy exists is pertinent to the idea that these forms are lexical). 34 That is, this refers to those forms that can be found in pre positional domains. 35 This work predicts that (40) could be grammatical for some speakers if north by itself is accepted for use in a postpositional domain because the meaning of north would theoretically allow it to be used in such a domain .

PAGE 164

164 Unlike their prepositional counterparts, the post positional elements are historically derived from nouns (Hagege 2010). However, not all nouns are used as post positions , as the use of (28 ) demonstrate s . N evertheless, the same forms that operate as postpositions can be found in some nominal positions and this creates some of the classification issues briefly discussed above. The morphemes tou can be added to forms that op erate as postpositions and in effect seemingly convert these forms in to nouns as listed with Table 4 2 . Table 4 2. Mandarin post positions/r elational nouns Pinyin Usage Character (Hanzi) Translation(s) b. hòu mian c. hòu tou 1. noun 2. postpo sition a. b. c. a. Backside/rear/behind b. backface 1. noun 2. postposition a. b. a. Inside/internal/lining 1. noun 2. postposition a. Lateral side a. qián b. qián mian b. qián tou 1. nou n 2. postposition a. b. c. a. front side/ the area in front b. fore/front c. at the head/in front/ahead a. wài bian b. wài mian c. wài tou 1. noun 2. postposition a. b. c. a. outside b. outside c. outside/out a. shàng bian b. shàng mi an c. shàng tou 1. noun 2. postposition a. b. c. a. up side b. atop/above mentioned c. above/ on the surface b. xià mian 1. noun 2. postposition a. b. a. underneath/underside b. below/next/under

PAGE 165

165 The morphemes / mian/ tou are often described as nominalizing suffixes. Forms such as can be used in a subject position (29 ) or object position (30 ) . (29) inside exists 5 CL people. (30) zài book LOC inside I n (29 ) t he form is clearly functioning as a noun as it can form the subject of t he sentence and as a noun in (30 ) as it operates as the DP object. Moreover, when / mian/ tou are attached , the form ostensibly can continue to function as a postposition as seen in (31 ) . (31) zài he LOC bookstore inside buy book In (31 ), the form seems to appear in the same place as a postposition, directly after a DP licensed by a prepositional form. It also has the function of semantically placing the DP Figure in a location, the DP Ground sh . However, its interpretation as a postposition when following th e DP Ground appears optional (32 ). (32) ( ) zài (de) book LOC library GEN inside In exampl e (32 ) is the object of the preposition and thus functions here as either a postposition or adjunct nominal . This is because t he linking element de , a grammatical element that links two noun phrases, a noun phrase and

PAGE 166

166 a clause, or a complement and resu lt, is optional in this context. 36 In other words, t he linking or genitive morpheme de in Chinese seems to optionally appear between a derived postpositional form and the object of the preposition. This would suggest that this form can appear as a n adjunct noun or as the postposition , the linker being present when it is interpreted as a noun. 37 Since can appear as postpositions , postpositions have been argued to be nouns or have nominal quality, but this simple analysis ignores several differenc es , some being obvious . O ne of the more obvious arguments presented by Djamouri et al. (2012) that post positions are not nouns is that in order to fulfil l the object requirement , they need to appear with one of the aforementioned nominalizing suffixes: b . Thus the use of ) is ungrammatical. (33) * S zài book LOC inside Example (33 ) shows that the form does not by itself serve as the DP object for zài . It can be concluded from this that by itself is not here being categorized as a noun or full noun as compared to the licit usage of in (31) and (32 ) . Moreover, the presence of this marker is ungrammatical if only a base p ostpositional form occurs (34 ). (34) Shu zài (*de) book LOC library GEN in(side) 36 There is a preference for omitting the linking element, especially in speech. Native speaker s reported that the presence so the linking elements made the phrase seem more formal. 37 This may be similar to English with sentences such as John is outside (of) the building/ John is at the outside *(of) the building in English .

PAGE 167

167 Unlike in (33), example (34 ) demonstrates that the linker de is not possible if the postpositional form is mono morphemic as is in this case. Thus if the marker is present we can conclude has the nominal identity that it does when it functions as the subject of a cl ause . The conclusion is that while and can be postpositions, only forms like can be noun s . Similar to prepositional forms , an other argument by Djamouri et al. ( 2012) for why unmodified postpositional forms are different from nouns is that they cannot be s tranded from their DP object (35 ). (35) Violation of Prepositional Stranding ( Adapted from Djamouri et al. 2012:8) * [e] [e] shàng zhe that CL car on lie DUR 1 CL cat As seen in (35 ), the sentence is ungrammatical if a postpositional element is stranded. In this case the DP element has been moved from the subject position , marked by [e] , to a topic position, leaving the post positional element behind. The resulting clause is ungrammatical as indicated. T his is another distinguishing characteristic of postpositions, making them appear more like prepositions in that their objects must be realized and local. From this data, as Djamouri et al. ( 2012) describe, the distinguishing features between post positions and nouns is quite simple. Postpositional elements that are not nominalized by , mian or tou are not nouns because they cannot be objects or subjects and they cannot be stranded via movement of the Ground DP to another position. However, this characterization implies that elements such as postpo sitional lexical identity. Given what is discussed in this work, it is argued to be

PAGE 168

168 advantageous to view these items as items behaving as nouns or postpositions. By doing so, other usages for so called post positional forms that are not discussed in Djamou ri et al. ( 2012) can be explained , such as (36 ). (36) Hòu Behind exist pursue enemy c hasing enemies who are in the rear Some unmodified forms that operate as postpositions can also appear as the subject of a clause. In (36 ) the form hòu as the subject of the clause. 38 This can be taken to be evidence that the form does not have a syntactic identity but is rather associated with certain syntactic positions. Thus it will be argued here that the base forms given in Table ( 4 2 ) are not defined by their inability to be nouns or be stranded from a DP Ground but rather this is what defines a postpositional domain, a lexical domain. The base forms them selves can be found elsewhere. 4 .5 Lexical Adpositions in Mandarin The sections above have served to demonstrate the difference between those positions that frame prepositions and verbs and post positions and nouns. C omparable to the English data, m any of the form s used for postpositions may also be used in other lexical domains, operating as verbs, n ouns and perhaps predicate adjectives. Likewise many of the forms used for prepositions can also operate in verbal domains . In the following sections evidence that they operate in other lexical domains will be given . This suggest s the presence of a Root node as is argued for many of the English prepositions in Sect ion 4.2 and Deacon (2011). 38 Usage of this type th ough seems less common then the type shown in (41).

PAGE 169

169 4 .5.1 Lexical Postpositions in Mandarin The discussion of lexical adpositional items in Mandarin Chinese will begin with the postpositional forms, as they are mo st clearly lexical items. Postpositions primarily indicate the feature Location or Place as is characterized by Svenonius (2008). The main postpositional forms and there uses in Mandarin Chinese are given with Ta ble 4 3 . Table 4 3. Mandarin p ostpositional u ses Pinyin Usage Character Translation(s) 1. hòu 1. postposition 2. noun 3. ??adjective 1. behind 2. noun 3. very behind 2. jìn 1. postposition 2. ?verb 3. adjective 1. near/ into 2. noun 3. adjective 1. postposition 2. ??adjective 1. in 4. nèi 1. postposition 2. noun 1. within 5. páng 1. postposition 1. n ext to/beside 6. qián 1. postposition 2. noun 3. ??adjective 1. In front of 2. noun 3. adjective 7. wài 1. postposition 2. noun 3. ??adjective 1. outside 8. xià 1. postposition 2. verb 3. ??adjective 1. below 2. to descend 9. shàng 1. postposition 2. verb 3. ?adjective 1. Top/on 2. to climb/ascend The following postpositional nominal usages can be found as defined by being the sole item in the subject position (37 40 ). (37) Hòu Behind exist pursue enemy hasing enemi

PAGE 170

170 (38) nèi yè wù huà inside exist business matter optimal ized aving optimized business as an internal condition (39) qián mái fú front exist bury ambush here is an ambush in the front (40) wài zhèng cè outside exist government policy spring wind h aving good policy as an external condition In addition to this nominal usage , some of the postpositional forms are used as v erbs . This is seen with example (41 ) a nd (42 ). (41) wo men shàng ke le ISG PL go to class ASP (42) xià le I down mountain ASP It is clear by examples (41) and (42 ) that these particular postpositional f orms c an be used as verbs. In (41 ) the relational idea is connected to the verbal idea in that the notion of up is tied to that of next and next is connected to the notion of going forward . Likewise the meaning of the form xià is realized verbally as conveying the direction of the motion . Moreover, all the postpositions can form the base for derived relational nouns that can continue to function as a postpositional form or as the object of the preposition as discussed in Secti on 4.4.3 .

PAGE 171

171 In addition to the nominal and verbal domain, at least one of these forms may be used in th e predicate adjective domain (43 ) while the other forms tend to rend er an ungrammatical judgment (44 ). (43) jìn She very close he is (44) ?? / / / / / hòu qián wài xià shàng She very behind/in/forward/outside/under/above S he is behind/in/forward/outside/under/above Most native speakers rejected the se ntences represented by (44 ). Howev er, it is unclear why the y are un grammatical in comparison to the grammatical sentence in (43). The idea attempted with (44 ) w as preferrably rephrased as (45), (46), or (47 ). (45) / / / / / hòu kào qián kào wài kào xià kào shàng She very by behind/by in/by forward/by outside/by under/by above She is behind/in/forward/outside/under/above (46) / / / / / hòu mian mian qián mian wài mian xià mian shàng mian He very behind side/ inside/forward side/outside/ underside/ above side He is behind/in/forward/outside/under/above (47) / / / / / zhàn hòu mian mian qián mian wài mian xià mian shàng mian He stop very behind side/ inside/forward side/outside/ underside/above side He stopped very behind/in/forward/outside/under/above Examples (45 47 ) show that forms that behave postpositionally may be modified to be used in the predicate adjecti ve domain. Interestingly the form mian , as discussed in Section 4.4.3 , is seen again being used to extend the usage of the forms in question. Perhaps unsurprisingly , this morpheme can also be used to allow new s to operate in postpositional domains.

PAGE 172

172 4 .5.2 Derived Lexical Postpositions in Mandarin In addition to the fact that elements joined with , mian, or tou do not have to be nouns , they also can be used to create new post positional forms. Some of t hese d erived forms constitute Table 4 4 . Table 4 4. Derived post positions or r elat ional n ouns Pinyin Usage Character (Hanzi) Translation(s) 1. noun 2. postposition a. b. a. northside/location to the north b. north/northern side b. nán miàn 1. noun 2. postposition a. b. a. location to the south b. southside/south 1. noun 2. postposition a. b. a. eastern side b. east side of something 1. noun 2. postposition a. b. a. Westside/ west of something b. Westside/ west In Table (4 4 ) we see that these forms after being modified by /mian/tou appear to be able to o perate as nouns (48 ) and postpositions (49 ) while the base form is not itself commonly used as a postposition (50 ) . (48) z zài China se a LOC northside (49) z zài chéng shì China sea LOC city north China sea is (50) ?? z zài chéng shì China sea LOC cit y north China sea is The data from S ection s 4.5.1 and Section 4.5.2 presents an interesting picture as it shows that many of the base postpositional forms in Mandarin have a lexical

PAGE 173

173 distribution beyond the postpositonal domain. This is similar to what was seen with English prepositions . This will be taken as evidence for t he a categorical nature of these Root Root Root items to form compounded items that then may be categorized by a functional head (to be discussed in more detail in Chapter 5) 4 .5.3 Lexical Prep ositions in Mandarin Prepositions in Mandarin are associated with the notion of Path 39 , as characterized by Svenonius (2008) . Moreover, i n Mandarin Chinese many prepositional forms may be used in a verbal domain in addition to the prepositional domain, Table 4 5 . Table 4 5. M andarin l exical p repositions The first item from Table 4 5 cóng is used as a preposition to indicate the source or starting point of a Path . This can be seen with (51 ). (51) ? xuéxiào qù I from school go cinema In (51 ) we see that the form in the preverbal adjunct position relating the DP figure xuéxiào In this position the form neither 39 In fact Djamouri et al. claim that only zài breaks the generalization of prepositions being path oriented. Pinyin Usage Character Translation(s) 1 .cóng 1. Preposition 2. Verb 1. Source (space) source to goal 2. Source Origination 2 . dào 1. Preposition 2. Verb 1. To (goal direction) 2. To arrive 3 . duì 1. Preposition 2. Verb 1. toward 2. to face, to 4 1. Preposition 2. Verb 1. To/for (goal for object)

PAGE 174

174 can be followed by an aspect marker nor have its object omitte d, making this a prepositional domain as discussed. However we can also see this form used as a verb in (52 ). (52) jiù le you right from ASP me PART In (52 ) we see the same form that was used as a preposition in (51 ) being used here as the main verb. We know it is operating as a verb because there is no other verbal form and because the aspect le directly follows it. Li kewise we see the form dào used as a preposition (53) and a verb (54 ). (53) From America to China very far a to Chi na is very far (54) le I arrived ASP In (53 ) we see the form dào licensing the goal DP preverbally as a preposition, the predicate in this sen tence being . Likewise in this position it can neither can be followed by an aspect marker nor have its object omitted. In (54 ) , h owever , the same form is clearly in a verbal position . Likewise the forms and duì can be seen operating in both domains (55 58 ). (55) . xìn I to me LINK friend write l etter.

PAGE 175

175 (56) le I give ASP him 1 CL book (57) duì teacher for me very good (58) duì zé I face PART him T he meanings of duì for in (57) and duì facing in (58), both have a basic directional meaning. Moreover, in addition to the verbal and prepositional domain, it may also be noted that the form duì can be used in an adjectival domain (59 ). (59) t duì s he very right In (59 ), though, the meaning of duì in this context does not appear to relate to the meanings of duì as a preposition or as a verb. I t thus appears as if the best analysis for (57,58) and (59 ) is to say these are homonyms. Likewise none of the other three forms are known to operate as adjectives or nouns. This therefore creates some difficulty for classifying these items within the g uidelines discussed in Chapter 3 . The items above can be argued to be lexical because in addition to a prepositional domain they also appear in a verbal domain. However, since we do not see these forms operating in other domains, one could also propose that these forms represent functional nodes. One would have to claim, however, that the verbal domain in the examples above does not contain a node and that the prepositional forms are

PAGE 176

176 underspecified and viable for both domains. Further research will thus be needed to determine if lexical prepositions in Mandarin are f irst categorized as prepositions and then verbs or if verbs can be formed with a bare filled by cóng , dào Moreover, while the meanings of these forms are more predictable than the meaning variation seen wi th English prepositions , these forms ar e seen adjoining to both the left and right of other items in compound forms , suggesting that they are indeed items . 4 .5.4 Prepositional Forms and Compounding If compounding distributions can be used as evidence for the lexicality of some Mandarin p repositional forms, it would appear that the forms discussed originate as nodes as evidences by (60 65 ). (60) dào Right Hand Constituent/Form. a. b. zhàndào c. wèidào d. wúdào (61) dào Left Hand Constituent/Form. a . dàotóu b. dàoàn (62) cóng Right Hand Constituent/Form. a. b. qín c ó ng capture and release (63) cóng Left Hand Constituent/Form. a. cóngrén b. cóngshàng c. cóngxià (64) duì Right Hand Constituent/Form. a. yuánduì b. duìqián in front of

PAGE 177

177 (65) duì Left Hand Constituent/Form. a. b. duìyì As it can be seen, these prepositional forms can appear on either side of other apparent items, indicating different derivational processes. This lends credence to the proposal that these are lexical items because one would only expect prepositional items to appear to the left of another item if the two items began as a preposition dominating a item (i.e. similar to forms like downstairs or by line in English). More investigation is needed though because some of these may actually be the result of historical compounds or homographs/homophones . Nevertheless , similar to English forms , there appear to be a subset of the prepositional forms that are FVIs. These forms do not appear as any other type of lexical item. 4 .6 Functional Adpositions in Chinese Similar to the En glish data given in sections 4.3 , not all the adpositional forms in Mandarin have a lexical distribution. In this way the findings of the analysis proved remarkably s imilar. There are some forms that are only used as prepositions. The following sections will discuss these forms beginn ing with the preposition that is the most difficult to classify, zài . 4 .6 .1 The Preposition Z ài There is prepositional item in Mandarin that can appear in multiple domains but the meaning/use of this form appears to be different than the prepositions di scussed

PAGE 178

178 above. The form zài appears to be the notable exception to the generalization that prepositions indicate Path features. 40 Table 4 6. Mandarin place p reposition s As mentioned in Chapter 1, t he meaning of zài is underspecified a s a preposition. It basically serves as a locative marker. Its meaning is either supplied via the relationship it most likely would have with the given DP G round or by a subsequent postposition. Th is can be seen with examples (66), and (67 ). (66) zài chúfáng They LOC kitchen eat something In (66 ) the location where the eating is taking place is the kitchen. However, zài , does not mandate that the eating be inside the kitchen. That interpretation would naturally be assumed via the context, but with opposite specification, the semantic meaning of zài is not contrad icted. This c an be seen with (67 ). (67) zài chúfáng wài They LOC kitchen out eat something In (67 ) the inclusion of wài , a locational post does not make the sentence semantically off because zài does not carry with it a specified locational meaning. Thus the presence of zài in (67 ) does not entail 40 In fact Djamouri et al. claim that only zà i breaks the generalization of prepositions being path oriented. Pinyin Usage Character Translation(s) 1. zài 1.preposition 2.verb 1. location 2. to exist

PAGE 179

179 that the figure be exactly in, on, below, or next to the ground DP it licenses. It could be any of these things. Instead, the DP just needs to be spatially oriented and the specifics of such orientation can be supplied via other lexical items. This could be from a logical relation between the Figure, the activity of the verb , and the Ground or specified by a locational postposition as seen with wài . Moreover, zài can be used as a s ingle verb, meaning to exist (68 ). (68) Verbal Zài (Adapted from Lü et al. 2000: 230). , ( ) Wo gangcai qu le yi tang, ta mei zai (jia) 1sg just go perf 1 time 3sg neg be h ome In (68 ) evidence for zài operating as a verb is given. The fact that it is being negated does not itself show that it is verbal, as it was seen that verbs, prepositions and predicate adjectives can all be negated. The fact that it is being negated, however, by mei and not bù , as it does when operating as a preposition, is revealing because the other existential verb in Mandarin you is always negated by mei and not bù . Additional proof that it is a verb is the fact that it can drop its ob ject as discussed in Section 4.4.2 . Furthermore, the meaning it has as this type of verb and the meaning it has as a preposition are related, eliminating potential homonym explanations. The form also appears with ot her characters to form exocentric like compounds (69 ). (69) zài Left Hand Constituent/Form. b. zàiwài c. zàidìng d. e. zàipí f. zàizài eve

PAGE 180

180 Example (69 ) perhaps support the idea that is a lexical item in Mandarin Chinese. It may also be that this is not indicative of the current status of the form used as a preposition and grammatical like verb. I ts lack of semantic m eaning and grammatical like function suggests that further study needs to be conducted as it may be that this item represents a type of locational feature that can select for a Path head. Evidence for this ide a comes from phrases such as (70 ). (70) j zài cóng r eport loc from open In (70 ) it appears as if two prepositional items are co occurring with both zài cóng linking the Figure to the Ground. More research needs, however, to be done to determine the s tatus of zài above cóng . As seen in this section, the status of the form zài is not fully clear. Given that this form when used as a verb appears to be a functional verb and out of all the prepositions it is the only one with a place meaning, it will be s uggested that this form is an FVI. Evidence taken from what appears to be compounding will require further research. Nevertheless, d espite the fact that the status of zài is not definitive , there are some clearly functional preposition in Mandarin Chinese. 4. 6.2 The Proposed FVI Prepositions As was seen with English , there exists a set of adpositional forms in Mandarin that are functional . They are argued to be functional because they lack the lexical distribution seen with other adpositional forms. These functional adpositional forms can be seen in Table 4 7.

PAGE 181

181 Table 4 7. Mandarin functional p repositions The first item from Table 4 7 , yú , appe ars to only be used in non lexical contexts. That is unlike the other prepositions, it never functions as a verb. The form yú appears to also be underspecified between a Place (71 ) meaning or a Path (72 ) type meaning. (71) yú huò yi ge ren police LOC CL mall captured 1 CL person (72) ( ) zhège yuán zì(yú) Táng cháo this CL saying originate from Tang dynasty 'This saying originated from Tang dynasty.' In (71 ) yú is clearly used as a locational marker , 41 working in a similar way as zài . In (72 ) , however, its older Path meaning can be seen. I n this sent ence yú can be omitted with the presence of zì . Moreover, the complaint that this form r eceives from learners is that i t has too many meanings out of 42 suggesting that it is really underspecified with regards to meaning a nd thus obtains the meaning of 41 It has been reported to be used more in formal speech. 42 From: Pinyin Usage Character (Hanzi) Translation(s) 1. yú 1. preposition 2. sub. conj. 1. in; at; to; from; by; out of 2. than 2 . lí 1. Preposition separation (needs duration) 3 . xiàng 1. Preposition 1. Direction/opposite to 4 .wàng 1. Preposition 1. Direction/toward 5 . zì 1. Preposition 2. anaphor 1. from/since 2. self, oneself

PAGE 182

182 the context rather than supplying any. Thus it will be proposed that this form is an FVI . This takes us to our next forms which can be seen in p repositional environments in (73 76 ). (73) lí xué xiào I home from school want 1 CL hour I n (73 ) one can see that lí is operating as a preposition. Furthermore, no data has been found to say it is operating as a different lexi cal item. Likewise the elements xiàng , 43 wàng and zì can only be u sed in prepositional domains (74), (75) and (76 ) . 44 (74) xiàng I direction east walk (75) wàng I direction east walk (76) zhège yuán zì Táng cháo this CL saying originate from Tang dynasty 'This saying originated fro m Tang dynasty.' This section has given examples of the form in Table ( 4 7 ) operating as prepositions. Interestingly, t he se form never appear as verbs , as adjectives or as nouns as the other adpositional forms discussed in Section 4.5 do . Given th is lack of lexical 43 I was told that xiàng Therefore it was excluded as the meaning is not related to that of the preposition and thus it could be a homonym. 44 Thanks to Dr. Lin, Dong yi for assuring that he also could not think of instances where these are used as verbs.

PAGE 183

183 distribution, t he se forms will th us be consi dered FVIs and be viewed on par with the English FVI adpositional forms discussed in Section 4.3. 4.7 Concluding Remarks This chapter has shown that many forms used in adpositional domains are also u sed in other traditionally lexical domains , making it difficult to insist that conceptual meaning and grammatical identity are packaged together. From a DM perspective, the overlap should not be entirely unpredicted or problematic. G iven the insight of the LDH as discussed in Chapter 3 , grammatical categories all correspond to syntactic environments which can be identified via word order and by the presenc e of local functional material. The issue in DM is then how to manage Vocabulary Insertion at prepositi onal domains which appear to frame items in both English and Mandarin. This will be the focu s of Chapter 5 .

PAGE 184

184 CHAPTER 5 ADPOSITIONS THROUGH THE LENS OF DISTRIBU TED MORPHOLOGY 5 .1 Introduction I n Chapter 1 it was introduced that the category adposit ion could be considered semi lexical in that this class demonstrates both lexical and functional properties as discussed in Co r ver and Reimsdijk (2001) . Likewise , i t was discussed in Chapter 2 that semi lexical items on the surface create a problem for Voc abulary Insertion and the Subset Principle in DM (Deacon 2011) and (De Belder 2011 , 2013 ). Chapter 3 then discussed ways to identify items and to c ontrol LVI insertion. Chapter 4 then provided data showing that some adpositional forms in English and Mandarin Chinese often behave as Root items as discussed in Chapter 3 while other adpositional items do not behave as Root items (a split in behavior that one might expect given its putative semi lexical or split class categorization) . Furthermore , it was discussed in Chapters 2 and 3 that all items must be categorized within the derivation. Given this the goal of this C hapter i s to explain how the archite cture of DM can account for th ese facts and how other attempts to solve semi lexical items in DM have fallen short . The proposal here is that adpositional forms with a lexical distribution (i.e. those shown in Chapter 4) origina te as Root nodes , explaining their lexical behavior /distribution . These nodes are proposed to be categorized by a little p_ head. On the other hand, those adpositional items that lack a lexical distribution originate as bundles of morpho syntactic feature s, explaining their predictable distribution . In these cases, the little p_ head is directly realized by an FVI. This chapter is organized as follows: Section 5.2 discusses one analysis to solving Semi Lexical items in DM from the literature. Section 5.3 e xplains how this analysis is

PAGE 185

185 undesirable . Given its weaknesses, Section 5.4 proposes little p_ as an alternative solution to the data raised in Chapter 4. Little p_ is proposed as the categorial head responsible for framing items in a lexical domain. Section 5.5 then shows how this analysis is helpful for uniting prepositional and particle forms in English (these items were argued to be the product of an [ F LINK ] feature in Chapter 1). Section 5.6 then shows how this an alysis can be translated into the extended projection of the category P discussed in Chapter 1. It suggests that this level of discreteness is able to solve problems raised by those working with the extended projec tion of P in Mandarin Chinese. 5 .2 Potent ial Solution for Adpositions and Semi Lexical Forms In Chapter 3 it was argued that the lexical flexibility of prepositions in English follows from the fact that these are LVIs that can be inserted into any of the standard ca tegorical contexts. It will be argued here that at least one type of semi lexical form, adpositional forms, can be accounted for in DM as a Root being categorized in a domain that is different than the standard nominal, verbal, and adjectival domains. This domain will be labelled by a little p_ head. H owever , an alternative proposal will be consider ed from the recent literature b efore explaining the details of this proposal . 5 .2.1 The Possibility of FVIs in nodes Instead of proposing that the semi lexical observation is often or always the result of nodes categorized in different domains, an alt er native approach is to propose that FVIs are being inserted into nodes, explaining the lexical like distribution of functional like items. For adpositions, this would be to as sume the traditional approach that they are functional items (FVIs) that can be exceptionally inserted into lexical domains. This would be similar to the approach taken by De Belder (2011 , 2013 ) who

PAGE 186

186 views some semi lexical forms as being FVIs inserted in a n L node. 45 De Belder (2011) gives examples of putative functional items appearing in lexical positi ons in Du tch (1). (1) FVI in Lexical Domain (De Belder 2011:42) Martha is mijn tweede ik . Martha is my second I In order to explain (1 ), De Belder first assumes that all vocabulary items are inserted through the same mechanism via competition, rejecting the need for a feature to distinguish nodes from other nodes. As briefly discussed in Chapter 2, De Beld er simply proposes that nodes are nodes w hich lack features altogether. The remaining problem, then, is how to insert F VIs with Associated F eatures in to terminal nodes that entirely lack features. This leads to the second part of the proposal, a tech nical redefinition of the Subset Principle . 5 .2.2 Revised Subset Principle By saying that competition insertion mechanism to allow FVIs to be inserted into nodes, the Revised Subset Principle : (2) The Revised Subset Principle (De Belder 2011: 62) Given a terminal node A with feature set F0 and vocabulary items (VIs) Fi F0 F0. When several VIs meet this condition, the one for which F0 Fi mo st closely matches F0 F0 is chosen. feature value is instead multiplied by itself in order to create a superset (in a 45 It should be noted though that De Belder is not specifically focused on adpositions or gives any examples of semi lexical adpositions, but it focused on more general problems related to semi lexical ity.

PAGE 187

187 F0). Since a node, as already proposed, has zero features, the value for the superset is zero ({ } x { }= ). This superset is now what the Associated F eatures of VIs are being matc he eatures are also multiplied by the feature value of the terminal node in question, F0 Fi, modifying the set of Associated F eatures used for that particular instance of insertion. In the case of a node, t he pos itive feature value of an FVI will be multiplied by the zero feature value of the terminal node. This yield s a set of zero Associated F eatures for t he purposes of insertion . The end result is an FVI with zero Associated F eatures. Since zero is a subset of zero, the Revised Subset Principle permits this FVI to be inserted into the node without adding features to the terminal node. Thus this revision only changes insertion at nodes. In other words, if a terminal node F0 has the features {x,y } and one is evaluating vocabulary item A {x} to B {x,y } for insertion into F0 , th e math for deciding which is inserted is shown with (3a c). (3) Revised Subset Principle Insertion Example for F node a. Value of Terminal Node = F0 x F0 = {x,y} x {x,y} = {,,, } b. Value of Associated Features for VI A = A x F0 = { ,} c. Value of Associated Features for VI B = B x F0 = {,,, } In (3) item B obviously wins (as it would also without the Revised Subset Principle) because its Associated Feature V alue is an exact match to the feat ure value of the terminal node. How ever, if there is a node then there are no features at terminal node F0 and the revision changes the results (4) . (4) Revised Subset Principle Insertion Example at L node a. Value of Terminal Node = F0xF0 = { } x { }= b. Value of Associated Features for VI A = A x F0 = {x} x { } = c. Value of Associated Features for VI B = B x F0 = {x,y} x { } =

PAGE 188

188 In (4) both items A and B tie with a value of { }, giving the appearance for choice at the insertion of a node . Anything s hould be able to be in serted and it would be left to Contextual F eatures to eliminate some candidates. Thus the Revised Subset Principle provides a way for FVIs to be inserted in featureless nodes as all VIs are now candidates for insertion . This po tentially solves the insertion problem of semi lexical classes in DM if these classes are viewed as consisting entirely of FVIs. In example (1), the FVI /a / is merely chosen over all other available candidates. 5 .3 Problems with Solution (2011 , 2013 ) solution provides a technical mechanism for the use of functional forms in the lexical domain, the revisions to the subset principle are both stipulative and costly. The following sections will discuss how the Revised Subset Principle overgen erates, contradicts historical patterns, and how the Revised Subset Principle is not independently motivated. 5 .3.1 Overgeneration One problem with the Revised Subset Principle is that it seems to drastically overgenerate the possibility for FVIs to be i nserted in nodes , suggesting that it should be possible for any FVI to appear in a lexical domain. Yet many functional items never appear in lexical positions, including many p repositional forms, as Chapter 4 Section 4.4 and 4.6 discussed . While lea v ing the specifics of example (1 ) unanalyzed, it will simply be point ed out that this example seems to involve a fairly specific and intentionally creative use of the pronoun. 46 46 We also leave aside the question of whether or not analyzing pronominal forms as LVIs inserted into particular n _ or D _ contexts might not be justified.

PAGE 189

189 are drawn from metalinguistic discour se or such discourse where linguistics items are being inserted in lexical domains because they are the topic of conversation. It seems just as easy to argue that in these cases LVIs have been created as would be assumed with the advent of any new proper n ame. The point here , however, is that aside from languages. 47 Thus the insertion mechanism shou ld not treat FVIs and LVIs equally for insertion into nodes. Returning, however, to first proposal, that nodes are void of features, it might also be argued to over generate in the other direction. G iven that LVIs never have Associated F e in this analysis, with or without the multiplication in (2), this analysis will in pr inciple allow any LVI to be inserted into an FVI node. This would be limited, however, by FVIs with Associated F eatures that form a subset of the value of the terminal n ode in question . If any such FVIs exist , no LVI will win the competition. This is potentially insightful, perhaps suggesting that grammaticalization is motivated by the absence of an appropriate FVI in the language to fill a certain f node (in which case a n LVI could be employed). However, one does not have to revise the Subset Principle to accomplish this. All that is needed is the elimination of putative features. This seems to be a good proposal. 47 Presumably this difference would have to be explained via Contextual Features. However, it seems as if other factors are preventing insertion.

PAGE 190

190 5 .3.2 Uni directionality Another problem for De Be , 2013 ) solution is that it disagrees with historical patterns regarding lexical and functional items. Historically it has been observed that grammaticalization is overwhelming ly uni directional. In other words, lexical elements grammaticalize i nto functional elements while functional elements seldom if ever go back to being lexical elements (Haspelmath 1999; 2004). Indeed, the o vergeneration problem noted above which makes it possible for any FVI to be inserted into a node suggests that the observation should be the opposite: functional elements should become lexical elements much more easily than lexical elements when examining change in language over time . 5 .3.3 Revised Subset Principle: Independently Unmotivated A third problem for the proposal by De Belder (2011, 2013) is that the Revised Subset Principle does not descriptively justify itself. In other words, this revision is not motivated by an attempt to describe something about how language is structured or functions. It merely solves a problem that a particular model has independently created. Thus the solution will be ignored by everyone else. There is nothing independent to suggest that the feature values of one list undergo multiplication with the actual values of the syntactic terminal nodes. It is hard to imagine how language could show us this. On the other hand, Underspecification and the original Subset Principle at least correspond with the observation that the phonological forms used to represent functional nodes often appear with different feature values in different syntactic positions. In other words, these mechanism attempt to capture why syncretism is so

PAGE 191

191 proli understanding of language structure and organization. 5 .4 The Proposal for p_ As discussed in Chapter 2 , all s must be categorized in DM. The reason for this, a s d iscus sed in Chapter 3 , is that the meaning of a is ungraspable without a syntactic context and this context is first identified via a categorial head . As already mentioned, t he proposal here is that there exist s a series of grammatical features that can c ategorize a differently than n_,v_ and a_ and they are to be labeled little p_ . In addition to providing an explanation f or the data shown with Chapter 4 , this proposal does not require major revisions to DM, showing instead that DM has the tools to account for adpositions and perh aps other semi lexical items. This proposal shows that adpositions can neatly be divided into F and L morphemes, aligning the category with other category defining heads as not all nouns (pronouns) and verbs (auxiliaries) i nvolve a node (as discussed in Chapter 4) . It also adds to our understanding of how concepts may be framed by linguistic features in the syntax from a non lexicalist perspective. Moreover, this analysis is more compliant with the general historical ontology o f adpositional forms (i.e. they come from s used for verbs and nouns) (Hagege 2010). It also allows us to account for the particle forms of English prepositional items. 5 .4.1 What is L ittle p_ As previously mentioned, the literature continually suppo rts the idea that s get categorized by verbal type features, nominal type features or adjectival type features: [+v], [+n], or [+a] Marantz (2001). T he se categorizing heads are merely labels

PAGE 192

192 whose value resides in identifying categorial feature classe s. 48 Thus the labels themselves are not actually present in the syntax (Harley and Noyer 1999). A specific feature in the syntax merges with the unspecified item, as d iscussed in Chapter 2, and directs how that item is interpreted , Chapter 3 . A s mentioned , an obvious solution to the observation s presented in Chapter 4 is to propose another type of categorizing head, p_ , a label for categorial adpositional features . This head thus represents potential relational features (Source, Path, Place, Goa l, and Direction) where a relational feature links a Figure DP to either another DP (a prepositional context) or links a Figure DP to the meaning of the it categorizes (a particle context). Moreover, the thought of having p_ projects a (so n_ , v _, and a_ is not entirely novel as it is briefly mentioned in Bo rer (2009:1) . 49 Moreover, others such as van Reimsdijk (1990) and Svenonius (2004, 2008) have proposed light p_ as a head in the extended projection of the Category P. In these other pr oposals, however, light p_ is not responsible for categorizing a categor ical s as is proposed here . 50 W outlandish if different grammatical features or a set of features can de fine the grammatical role of a item. In this way, t he proposal for l ittle p_ is similar to proposals for other types of n_, v_ , (or a_ ) , 51 which as will be discussed in Chapter 7 48 Perhaps this is better Root to be viewed/conceptualized in a particular way. 49 It should be noted that Borer is not directly discussing the DM model. Instead Borer is discus sing her own model, the Exo Skeletal Model (XSM), which resembles DM in many ways. 50 The purpose of p_ in these works has putatively been to introduce the Figure DP. 51 It would be interesting to discuss a_ as a categorizer but given the prompt, time, and space it will be mainly excluded from any analysis.

PAGE 193

193 are not uncommon in the literature . The label p_ will be used only as long as it remains a helpful descriptor for another range of features that c an categorize a . In further study one may want to relabel some types of little p_ as a type of little n_ and others as types of little v_ or vice versa . 5 .4.2 Categorization by Little p_ Categorization by little p_ is argued to be the same as categorization b y any other categorial head in DM. of the kind seen in Chapters 3 and 4 have been handled in DM by referring to the fact that in DM s (LVIs) are a categorical; they can therefore in principle be realized as any cat egory so long as they occur with the proper categorizing head as discussed in Chapter 2 and 3 . The proposal here is that prepositions with lexical uses be analyzed in precisely the same wa y, as LVIs inserted into a node. They therefore, in principle, may be inserted in a node governed by any categorizing head: a_, n_, or v_ as well as the proposed fourth categorizing head, little p _ . Thus t he use of down in We walked down the street and John downed the ball would have the following two structures as sh own in (5). (5)

PAGE 194

19 4 In (5 ), the DOWN is identified as a preposition by merging with the little p_ head in the left structure while in the right structure it is identified as a type of verb by merging with little v_ . 52 Like all LVIs, we assume t his form has no Associated F eatures, b ut since it is found in these contexts we can also assume that it has these Contextual F eatures as shown in (6 ). 53 (6) DOWN [ ] / { P DIR _, V TRANS _ , } Likewise based on the observations of Chapter 4 , the other lexical prepositional forms can be listed with the lexical contexts in which they have been attested (7 ). (7) Other LVI Prepositional Form Insertion Rules English OFF [Ø] {p_,v_,n_,a_} OVER [Ø] {p_,v_,n_,?a_} UP [Ø] {p_,v_,n_,a_ } UNDER [Ø] {p_,a_ } NEAR [ ] {p_,v_,a_ } ON [ Ø] {p_,?a_ } OUT [Ø] {p_,v_,n_, a_ } BY [Ø] {p_, } IN [Ø] {p_,v_,n_,?a_ } This same analysis applies to those prepositional forms in Chinese from Chapter 4 that showed in a p repositional and verbal domain is given in (8) . 52 In the structure in ( 5 ), it is assume d that the functional c ase checking property of prepositions is performed by the functional head p _ in the same way it is standardly assumed that the head v _ che cks the case of verbal complements. 53 T he possibility that de adpositional verbs, nouns and adjectives in fact involve two categorizing heads with the outer head defini ng the surface category and p_ selecting the . This is the an alysis assumed elsewhere, for example, for gerundive nominals or de adjectival nouns that use the suffix ness (Marantz 1997; Embick and Marantz 2008). One fact in favor of this more complex analysis is that prepositional forms never seem to undergo allomorphy when they appear as verbs, nouns, or adjectives. If categorizing heads, including little p, are all phases as Marantz and Embick (2008) argue, then this fact would follow if prepositions used in other contexts still involve a p_ head selecting t he . By the same logic, however, Marantz (2001) and Arad (2003) have argued that the domain of specialized meaning is the domain of the and its first categorizing head. Many of the non preposition al uses of prepositions in Chapter 4 seem to have me anings at least somewhat unpredictable from the meaning of the prepositions that they are derived from (e.g., ups ability to jump high). T his question is thus left for future research.

PAGE 195

195 (8) As for the postpositional forms, in Djamouri et al . ( 2012 ) , they are analyzed as prepositions that lack a structural c ase featur e . Th e DP object thus moves to a higher position to receive case from a preposition . Thus we can model the lexical postpositions exact ly as we did the prepositions , as in (9 ). (9) The Contextual F eatures would be l isted with the following rule (10 ). (10) LVI Adpositional Form Insertion s Rules for Mandarin SHÀNG [Ø] { p intr _,v_,n_, ?? a_} NÈI [Ø] { p intr _,n_} XIÀ [Ø] { p intr _,v_,n_, ?? a_} PÁNG [Ø] { p intr _,n_,} J ÌN [Ø] { p intr _, ? v_,n_,a_} C"NG [Ø] {p tran _,v_} H"U [Ø] { p intr _,n_, ? ?a_} DÀO [Ø] { p tran _,v_ } [Ø] { p intr _,n_, ?? a_} DUÌ [ Ø ] { p tran _,v_,} QIÁN [Ø] { p intr _,n_, ?? a_} [ Ø] { p tran _,v_,} WÀI [Ø] { p intr _,n_, ?? a_}

PAGE 196

196 It is therefore proposed that these forms can be inserted in a Root node that is in a local re lationship with any one of the Contextual F eatures listed ( v _ 54 To explain the f act that the postpositional forms never operate as prepositional forms and vice versa, postpositional forms are listed for intransitive contexts while prepositions are listed with transitive contexts. Note that the Contextual F eatures do not determine whic h of the se forms will be inserted in any given derivation. This is desirable in the same way that it is desirable that Contextual F eatures do not determine the insertion of the nominal DOG over the CAT . As discussed in Chapter 2, LVI insertion traditionally involves choice. O ne may explain the choice of adpositional forms as one explains the choice of any item (11 ). (11) The n [cat/dog/gorilla] v [saw/watched/observed] a a [blue/green/turquoise] n [monster] p [near/around/by] the n [ river ] . As (11 ) attempts to show, there is a choice at the established lexical domains and in the proposed lexical domain, the prepositional domain. This thus helps to explain the a pparent choice at this domain, a choice that would be difficult to describe with grammatical features. However, this does not mean that all the p repositional forms do not have Associated F eatures. 5 .4 .3 The Functional Prepositions of English and Mandarin A s discussed in Chapter 4 S ection 4.3 , there were a number of adpositional forms in English that did not show a lexical distributio n. These are regiven with (12 ). (12) at, by(agentive), for, from, of, to, via, with, until, and as 54 A possible exception is by , which resi sts use as a noun or verb. However, being an LVI does not guarantee category flexibility. Moreover, we argue below that by status.

PAGE 197

197 Likewise in Chapter 4 S ection 4.8 a number of prepositional forms in Mandarin were shown not to have a lexical distributio n (13 ). (13) yú lí wàng , zì and de 55 U nlike the lexically flexible prepositions discussed in Section s 4.2 and 4.5 , the forms in (12 ) and (13 ) are argued to be FVIs inserted directly into the functional node p _. As such, it is assume d that they have Associated F eatures that must match a subset of the features in p_ . Presently, English FVIs will be argued to be associated with the features of (14 ) and Mandarin (15) . (14) English F VIs: a. at b. by c. from d. to e. with f. via g. until h. of i. for j. for goal, volition ] k. as (15) Mandarin FVIs: a. zì b. xiàng c. wàng d. lí e. de f. yú The forms from (14) and (15 ) are thus argued to be inserted as direct representations of p_ . Take for instance, at , which is argued to be inserted when it con tains th e feature [loc] as in (16 ). 55 It is likely that zai has replaced yu as the FVI representation of the [loc] feature.

PAGE 198

198 (16) As (16 ) de monstrates if there is no pres ent for categorization, the elsewhere rule from (14) applies (in this case inserting /æt/ ) . The same could be said wi th yú in; at; to; from; by; out of (17 ) . (17) Note that it is assumed that p_ selects f or a DP in (16 ) and (17 ) above. There is no position associated with the p _. In such structures p_ recategorizes the DP it has selected much like, say a kind of n_ may recategorize a verb phrase, creating deverbal nominals such as English gerunds. A potential problem though arises when tryin g to block insertion at p _ after one of the forms from (14) or (15) is inserted . I n principle it would be possible to insert one of th e LVI prepositional forms in (7 ) or (10) into the node and also insert the FVI /æt/ directly into p_ , yielding non e at down, at by, at on

PAGE 199

199 etc The problem is easily solved, however, by assuming that when an LVI is inserted into the node, a null allomorph of P loc is inserted in p_ by the following rule (18). (18) Realization of P A fter Insertion /Ø/ IN OFF BY OUT ON NEAR , As (18 Root is pres ent for categorization , a null morpheme is inserted . This null morpheme is ins erted under the p_ head to satisfy the features of p_ . Moreover the rule follows the established insertion order as makes crucial use of out insertion, that is Vocabulary insertion must affect the verbal or nominal befor e it affects any functional head adjoined to the (Legate 1999:15). Thus Root item merged with p_ , then vocabulary insertion occurs like any other f morpheme. If a null element were always inserted into p_ when p_ selects for a t LVI, our analysis might appear to exhibit a convenient conspiracy since nothing would explain why p is null only in cases where it selects a . In Chapter 6 , however, evidence is presented to show that p_ is not always null in this environment and th is analysis is not a conspiracy . Also they are never used as the right hand member in compounds or incorporations. They are occasionally, however, found attached to the left of another (19).

PAGE 200

200 (19) Left Adjoined FVI Compounds a. your at home vet b. In general whe n the forms in (1 4 ) get used in word like derivations, they look like (19a). (19a) appears to involve the prepositional phrase at home moving to left adjoin and mod ify the VET . Accordingly this does not provide evidence that at is lexical , as functional items are often trapped in compound formations. (19b) on the other hand is a little more interesting as this composition seems to be happening at a lower level in the derivation. Furthermore, the compound byline is used as a verb as in bylining an article . This is the only known instance of this type. For now, it will be taken that this alone is not a good argument to claim a lexical status. 5.5 Parti cles and Prepositio ns and Little p_ It was discussed in Chapter 1 how the category adposition and particle P items often share a common and form and similar meaning. 56 However, while traditional models have recognize d the overlap, they do not synchronically or explicitly acc ount for why adpositional forms have been used in this manner . 57 This commonality suggests that each pair not only historically comes from a common source but that the source is the same for modern derivations (i.e. the difference is similar to the differen ce between to run and a run ), but traditional lexical models have not accounted fo r this with any great insight. 56 One may wish to argue that the connection is solely based on the phonological form but the semantics of prepositions and particles overlap quite a bit as well. 57 This may not always be the case in every language though (cf. Svenonius 2007: 20)

PAGE 201

201 5.5.1 Part icle Forms and their Root node D istribution Within the mechanism of DM , the shared phonological form and meaning can be captured in two ways. The first is to propose that these P elements are represented by underspecified FVIs where the same FVI can be inserted in both a prepositional and particle environment. The second is to propose that the overlap is the result of a node in different environments being realized by the same LVI as was proposed for the other lexical distributions seen in Chapter 4. It is argued here that the forms that can double as prepositions and particles are the result of an LVI inserted into node. The primary reason here to support the node explanation is that the distribution of those P items that overlapped with nodes in the previous section may also be found in particle positions. It is thus argued that the particle position contains a node in accordance with the Identification Principle discussed in Chapter 3 . This explanation accounts for the overlap between the phonological form of many prepositions and particles and the fact that all the forms argued to be LVIs in the p revious section, except for the form near , can also be used as particles. 58 The proposal also accounts for the fact that other lexical items appear to operate in the particle position because it proposes that particles begin with a node. 5.5 .2 Partic les as Further Evidence for this Analysis The purpose of this section is twofold: to show which forms may operate in the particle position and to show that the forms that operate in the particle position are 58 Near m ay be used as a particle by some English speakers. Its absence is apparently accidental. This will be explained as another accidental gap.

PAGE 202

202 th ose forms argued to be L VIs in Chapter 4 S ecti on 4.3 . Accordingly, it is proposed that the particle position contains a Root node . Utilizing Inversion and Fronting to identify if a P item is a particle or preposition, the forms in (20 ) may be found in a transiti ve verb particle arrangement (21 ). (20) Up, down, on, off, in, out, and by, over, and through (21) P elements in vP Par ticle Domain. Inversion a. He passed by the opportunity. He passed it by. b . He thought over the problem. He thought it over. c . She thought through the problem. She thought it through. d. She looked up the information. She looked it u p. e. He brought down his opponent. He brought him down. f. She put on her hat. She put it on. g. She took off her hat. She took it off. h. He moved in his stuff. He moved it in. i. She moved out her stuff. She moved it out. Front ing j. *By the opportunity, h e passed. The opportunity, he passed by k.*O ver the problem , h e thought. T he problem , h e thought over l . *T hrough the problem , s he thought. T he problem , s he thought t hrough m.*U p the information, she looked. The information, she looked up n. *Down his opponent, he brought. His opponent, he brought down o. *On her hat, she put. Her hat, she put on p. *Off her hat she took. Her hat she took off q. *In his stuff he moved. His stuff he moved In. r. *Out her stuff she moved Her stuff she moved out As can be seen with (21 ) the forms u p, down, on, off, in, out, and by, over, and through appear in the vP particle domain . 59 Since all of these forms except through have been shown to have a lexical dis tribution, it is proposed that the particle position contains a Root item. 60 Accordingly , the prepositional forms that were shown not to have a lexical distribution also fail the inversion test (22 ). 59 This is true if tests such as Inversion and Fronting as discussed in Chapter 1 define the distinction between par ticles and prepositions in a transitive verbal context. 60 It can then be argued that through and perhaps by are LVIs with a very specific contextual listing.

PAGE 203

203 (22) FVI Prepositions Failing to be Verbal Particles a . John looked at the problem. *John looked it/the problem at. b . John looked for his team. *John looked them/his team for. c . John came from England. *John came it/ England from. d. John came to his senses. *John cam e them/his senses to. e . John heard via his friend. *John heard him/his friend via. f . John came with his friends. *John came them/his friends with. g . John learned of his error. *John learned it/his error of h. John worked until 12. *John worked 12 until. i . T he book was written by John. * The book was written him/John by. Example (22 ) shows that none of the proposed FVI forms work as particles as defined by the inversion test. Furthermore, forms argued n ot to be used as particles c an be fronted with the DP with more success than particle verb arrangements as seen with (23 ). (23) Fronted Prepositions a. John looked at the problem. At the problem John looked b . John looked for his team. For his tea m John looked c. John came from England. From England John came d. John came to his senses. To his senses John came e . John heard via his friend. Via his friend John heard f . John came with his friends. With his friends John came g . John learned of his error. Of his error John learned h. John worked until 12. Until 12 John worked. i . The book was written by John. *By John the book was written 61 As can be seen with (22) and (23 ), even pairings such as look at , come to , and come with , which can drop their DP object, do not behave as particles as defined here . This further suggests that these prepositional forms are categorically different from those that are argued to be lexical. 61 Merely shows that agentive by is different than other functional prepositions.

PAGE 204

204 Since this analysis has proposed that a Root node is present in particle constructions, the presence of other the lexical items such as back and home in the particle domain can be explained (24 ). (24) B ack and H ome as Particles a. He paid back his debt. He paid it back. b. He brought home the gi ft. He brought it home. c. ?Back his debt he paid. His debt, he paid back. d. *home his gift he brought. His gift, he brought home. This is proposed to be further evidence that those adpositional forms that are argued to be FVIs are qualitative ly different than those argued to be LVIs. Given the overlap in distribution between P items in lexical and particle domains and the fact that a few other lexical items also seem to operate as particles, it is reasonable to propose that the particle positi on contains a node . This section has shown that there is a difference between the set of prepositional items that can be used as a particle and those that cannot. Those items that were shown to have a lexical distribution in Chapter 4 were found in the particle d omain described here. Those prepositional items that were not shown to have a lexical distribution in Chapter 4 were shown to be incapable of operating in the particle domain described here. The natural conclusion is that the particle domain involves a ot node. As for the difference between s in a particle domain versus a prepositional domain, it is maintained that the particle only selects for a Figure while a preposition semantically must have a Figure and a Ground as explained in Chapter 1 . 5.6 Revisiting to the Extended Projection of P It has been proposed that there is a little p_ head that performs the function of categorizing s in adpositional domains. Moreover, a s previously alluded to in Chapter 1 , there appears to be a strong tendency for co occurring prepositional heads

PAGE 205

205 to be arranged in a certain order cross linguistically . T he strongest tendency observed cross linguistically is to have Path head dominate a Place one (Koopman 2000; den Dikken 2006; Svenonius 2004, 2006, 2007, and 20 08) . T his section will examine this hypothesis further by modeling these elements as items under varieties of p_ head s to see if the proposal of the dissertation is compatible with this idea. 5.6.1 Revisions to the Extended Projection of P Here the extended projection discussed in Chapter 1 will be explained as separate prepositiona l type phrases that may or may not merge with other prepo sitional phrase types (25 ). (25) Extended P in DM Th e structure given with (25) essentially adapts the extended projection of Svenonius (2008) into the framework of Distr ibuted Morphol ogy (minus a few projections). Each prepositional phrase consists of a set of defining feature(s), forming a p_ head. 62

PAGE 206

206 Following the framework of Distributed Morphology this head can be the locus for the insertion of Functional Vocabulary Ite ms, FVIs, as is the situation with the FVI from being inserted under the [+ SOURCE ] head. Moreover, as proposed , these p_ heads may Root node local to the p_ h ead . This allows one to insert prepositional items a s LVIs in nodes instead of under the p_ heads themselves. The advantage of this is what Svenonius himself notices, being that within with several heads, without head (2008:11). In his current model there is no strong reason given for such head movement . This revision can also be used to model Chinese adpositions and to advance an argument made by Djamouri et al. (2012) that post positions begin as preposition s in Mandarin. 5.6 .2 Explaining Ch inese Data with Revisions to P s tructure Chapters 1and 4 showed that in Mandarin Chinese a preposition and postposition can relate/link to the same Figure DP . This construction is re given with (26). (26) shàng From table on One can utilize the stacked construction seen in (25) to help explain why the items have this arrangement by following the argument of Djamouri et al . (2012) within this more di screte model . I t has been shown that the prepositional form C"NG Root item because of its ability to be categorized in more than one environment (Chapter 4) . Moreover, as a

PAGE 207

207 preposition it can license a DP in a c aseless position as argued by Djamouri et al. C"NG will be categorized by a little p_ head that is created by the features [ F LINK ] and [ CASE ] . Moreover, it will be argued that it need not be categorized by a p_ head with a [ SOURCE ] feature because the Root itself holds that meaning, as it has that meaning when used as a verb. As for the other form SHÀNG , Root item. It will also have to be categorized by a p_ head . In this case it contains a [ LINK ] and [ LOC ] feature. This SHÀNG does not inherently mean location . This is represented with (27 ). (27) As shown in (27 ) both the preposition and postposition will begin above the DP. This makes the account resemble the structure of English and the pattern for pre positional structure . Moreover, as Djamouri et al. (2012) argue, the DP moves to a higher position, one above the postposition, to check case with the feature under P . A modified version of tha t account is given in (28 ) .

PAGE 208

208 (28) In (28 ) moves to Spec PlaceP to check case from PathP. In this example since C"NG SHÀNG are argued to be s, the realization of p_ in both these cases is Ø. Moreover, crucially the difference here between English and Chinese word order is a m atter of ca se feature strength ( a strong case featu re that is positioned with the P ath p_ head ) . In Chinese the feature is strong and thus checking must occur in a Root s to merge with th e functional heads for categorization as outlined in Chapter s 2 and 3 , it is here noted that it is strange to have a copy of a featureless unit (i.e. Root ). 63 Never Root s in essence get to represent the functional features under p_ . T he more discrete analysis presented here is also able to help solve other issues regarding the distribution of adpositional forms in Mandarin Chinese. It is addressed in Djamouri et al. (2012) that sometimes a post positional item appears to be able to 63 This observat ion will be left for further research.

PAGE 209

209 license a DP , contrary to everything that has been said up until this point. This can be seen with (29 ). (29) Post Positional Case Marker ( Djamouri et al. 2012: 12) yuánzé *(shàng) zhèyàng zuò. 2 PL principle on can this.way do The sentence in (29 ) would be deemed ungrammatical if the postpositional form is not present. It see ms that this form is necessary for the DP to be understood in this adjunct position. The thought will be entertained here that there actually is a preposition assigning C ase and in this situation it is being represented by a null item (30 ). (30) The analysis entertained here is that yuánzé with the p_ that here gets realized a s null. As is shown earlier in this chapter , not all p_ heads come with a as in cases where p_ heads are represented by words such as from . It would have to be further specified that the insertion of certain s (i.e. those corresponding to abstract concepts such as yuánzé ) is causing null allomorphy in these cases . One may also propose that yuánzé is actually being categorized as the preposition in this case (making the allomorphy more similar to what was seen with

PAGE 210

210 English LVIs in , out Without th is level of discreteness, there does not seem to be a way to explai n why a putatively c aseless postposition can license a DP in a c aseless adjunct position. This section has thus demonstrated how viewing certain adpositional forms as Root items categorized by a little p_ head works and how this does not need to conflict with modern accounts of achieving the surface structure of Mandarin adpositions or the mor e general structure of spatial adpositions. 5.7 Concluding Remarks lexical items in DM is flawed, costly, and does not account for adpositions. Instead this chapter has proposed tha t a little p_ head can categorize information in adpositional contexts. This along with listing s with Contextual F eatures can account for the data discussed in Chapter 4 without requiring any major revisions to the DM model. Moreover , this chapter proposed that a null marker be inserted under the p_ head whe n one of the LVIs from Chapter 4 is inserted. This is argued not to be a conspiracy becau se it will be shown in Chapter 6 that this head is realized with certain s. This thus aligns the p_ head with other categorical hea ds that get realized differently depending upon the identity of the LVI inserted. Lastly this chapter showed how a proposal for a p_ head can not only work with previous accounts of adpositional stacking but how this analysis may help solve some of the iss ues encountered within the cartographic tradition.

PAGE 211

211 CHAPTER 6 EVIDENCE FOR LITTLE P_ 6 .1 Introduction In Chapter 4 , it was seen that several adpositional forms in both English and Mandarin Chinese appear in lexical domains other than an adpositional doma in. G iven the discussion in Chapter 3 , it was argued that this distribution (along with some unpredictable meanings) indicated that these forms originate with a node. In Chapter 5 , a little p_ head was formally proposed to be the categorizer of s in adpositional domains. It was also proposed that those adpositional forms from Chapter 4 that are never found in a lexical domain are a direct representation of this p_ head . However, w hen little p_ categorizes one of the node s discussed in Chapt er 4 as an adposition , it was proposed that a null FVI is inserted to represent its features. As noted, this would appear to be a convenient conspiracy if no other evidence existed for this categorizing head. M orphophonological evidence , however, for the e xistence of p_ comes from large numbers of prepositional elements in English that have complex morphological structure. In particular, many previously undiscussed prepositional items are composed of a element ( i.e. one that may appear in a different category frame elsewhere in the language) together with the proposed prefixes /a / and /be /. It is thus argue d that these two morphemes should be understood as realizations of the proposed p_ head muc h as the suffixes / tion / or / ness / are understood to be the manifestation of n_ heads , / ify/ or / ize/ v_ heads and / ive/, / ous/ a_ heads . In addition to these /a / and /be / prefixes, evidence for a n adpositional creating head can be found in Dutch a nd other languages.

PAGE 212

212 This chapter is organized as follows: Section 6.2 gives phonological evidence for p_ in English. It shows that /a / and /be / can be taken as evidence for p_ and when done so, this accounts for the different behavior seen between a hea ded and be headed prepositions in English. Section 6.3 then gives similar evidence for German, Dutch and Persian. It is thus concluded that an analysis of p_ as a categorical head is not only helpful for explaining lexical items being used as adpositions b ut also phonologically supported by the presence of extra phonological information when a lexical item is framed as an adposition. 6 .2 Evidence for Little p_ English Not every categorial head is represented by phonological material in all environments acr oss the w O ften a null categorizing morpheme is argued to be present in what are traditionally called zero conversions. Nevertheless , one piece of evidence for the existence of categorial heads is seen by affixes that gory . In English, there exists affixes that can be attributed to the three main lexical categories. This would leave little p_ as the odd head out as it so far has only been represented by /Ø/ or a functional prepositions such as /of/ . In Chapter 4 , howev er, adpositional forms in English that begin with /a / and /be / as in around and be low were not discussed. I f we analyze forms such around and below as consisting of two morphemes /a ROUND and / be LOW , we obtain phonological evidence for the lit tle p_ head. The rest of this section will show how this is not an ad hoc argument as there is a systematic correspondence between form and meaning and the behavior of a headed prepositions is different than that of be headed prepositions. These fact s can only be ex plained if forms such as around and below are viewed as being bi morphemic.

PAGE 213

213 6 .2.1 A and B e Headed Forms The proposal here is that evidence for the p_ head in English can be found if some or all a headed and b e headed prepositions are analyzed a s bi morphemic units. Historically the source of /a / and /be / is a free prepositional form (Barnhart). 1 Generally, however, forms such as around and below are treated as simplex prepositional items in modern English. Returning to the extended projection of P discussed in Chapter 1, Svenonius (2006) for one treats around and below as monomorphemic units positioned under what he calls an Axial Part head , as seen in Figure 1 1 in Chapter 1 . 2 He does this because these forms play a locational role and do not act like full DPs (i.e. in his account around is occupying the same head as front does in the phrase in front of ) . This treatment, however, completely ignores the lexical ontology of these forms, leaving the facts of their prefix + structure unaccou nted for. To not treat them as bimorphemic units makes the phonological and semantic correlations between the s in the prepositions and the use of the s elsewhere in the language completely accidental. Interestingly, within the DM framework, Acqu aviva makes a brief note about forms such as these, suggesting that the a in a side class side forms a preposition and not, as is not offer details or explain why this would cause the to change categories as his analogous particle de in destroy is 1 This will be discussed more in Chapter 9 . 2 This structure is modified for the DM framework in Chapter 5

PAGE 214

214 not categorizing the putative STROY as a verb. Instead, it seems to make more sense to analyze t he phonological appearance o f /a / and /be / when certain LVIs get categorized as prepositions as the phonological spell out of a p_ head , a head that turns s into adpositions (1) . (1) Examples of a headed pre positions are seen with Table ( 6 1 ). Note ther e are a few prepositions ( against, amongst, etc.) which could be understood as morphologically simple, or as involving the a prefix and a bound . 3 Some a headed particles which were not discussed in Chapter s 1 or 5 and a headed results are also repres ented . Table 6 1. A headed forms Item a headed prepositions Item a headed particle s/results cross across side aside long along stray astray round around ground aground top a top part apart board aboard ghast aghast mid am id/amidst way away side aside shore ashore fore afore sleep asleep B ound s new anew against float afloat among/ amongst drift adrift above head ahead about cross across after go ago 3 For other adpositional forms, where for the a morpheme to attach to,(e.g. after where we do not have the fter ) , it is probably best to analyze these as mono morphemic items represented by a single LVI or FVI. This is because there is no way to test the effects of a possible a morpheme attaching to an otherwise non fter . (This analysis is also supported historically.) If a form like after gives no evidence that it is lexical than we can assume that it is an FVI, as shown in Deacon (2011) not al l prepo item.

PAGE 215

215 Table ( 6 1 ) gives several ite ms that can be used as prepositions in column one. A t least eight non prepositional LVIs become prepositions with the addition the affix / a / as shown in column two . Likewise the items in column three can be used as particles or results with the addi tion of the affix / a / as shown in column four. Since a systematic correspondence between form and function is seen , this / a / morpheme can be argued to be a phonological manifestation of a p _ head when certain LVIs are merged with it (i.e. items li ke those in columns two and four). This phonological realization can be taken as evidence that a p _ head exists independently from the relational LVIs that carry the conceptual meaning. Similar observations can be made with prepositions th at begin with th e prefix /be /. These f orms can be found with Table 6 2 . Table 6 2. Be headed p repositions 4 In Table 6 2 , one can see a patt ern where /be / can be added to a item to form a morphologically complex preposition. As with /a /, the proposal is that / be / is a manifestation of p _ when ce rtain s are merged with it. These realizations are thus in addition to the insertion o f / / , as discussed in Chapter 4 , when the items 4 Those of you looking for a linguistically related Halloween costume may consider this as an option. item be headed prepositions be headed particles /results fore before GAP hind behind GAP low below GAP yond beyond GAP side beside GAP neath two Beneath Between ? GAP GAP B ound GAP Between GAP

PAGE 216

216 DOWN , IN , OFF , OVER , UP , UNDER , NEAR , ON , OUT , and perhaps BY are categorized by little p_. To control for this, however, the different realizations of the p_ head need to be listed with bo th A ssociat ed a nd Contextual F eatures. The morphemes /a / and /be / and / / are thus listed with the following Associated and Contextual F eatures ( 2 ). (2) Basic Insertion Rules for /a / and /be / FVIs. /a / [ L INK BOARD LONG SIDE ACROSS __ e tc. /be / [ L INK , T RANS LOW HIND YOND etc..} / / [ L INK ] 5 /{ __ DOWN __ IN NEAR The Associated F eatures in the rules above represent the fact that /a / and /be / do not seem to be merely allomorphs, but are inserted into differe nt types of p_. One argument for this comes from the fact that some s ( side and fore ) may occur with either prefix , resulting in different meanings . T he main argument for this, however, comes from the fact that only a headed forms may operate as p art icles as defined in Chapter 1 . On the other hand be headed forms always need a Ground. 6 .2.1 A and B e Headed Particle Forms The discussion o f parti cles in Chapter 1 and Chapter 4 resulted in the proposal that contrary to prepositions, particle forms alw ays involve a node . As T able 6 1 from the previous section indicate s , many items can function as particles after being prefixed with / a / . 6 5 Recall that the [link] features is merely proposed to be the most basic adpositional feature, one that conne 6 Some speakers seem to strongly prefer for certain a headed P items to come after the DP object (i.e. I moved the furniture arou nd and not moved around the furniture ). Also recall that a Place adpositional interpretation is also often available when the P item precedes the DP object, but if the P item is acting like a particle (as claimed here), the interpretation should be the same whether it is pre or post DP.

PAGE 217

217 (3) A headed Particle Forms a. She got across the information. She got it across. b. He took apart the engine. He took it apart. c. I moved around the furniture. I moved it around. d. I set aside some money. I set it aside. e. She pushed away the man. She pushed him away. f. They ran aground the ship. They ran it aground. As can be seen with (3 ), a headed forms may operate as particles as defined by the inver sion test disc ussed in Chapter 1 and Chapter 5 . These forms are only linking to a DP Figure (i.e. the internal argument of the VP). In this arrangement, the FVI /a / is argued to represent t he p_ head in the vP just as it does for a p_ head in the pP elsewhere . It is thus argued that these particles are the result of a lexical categorized by a p_ head which gets spelled out with the FVI /a / CROSS , PART , ROUND , SIDE , WAY , GROUND are categorized in this position , they are realized by /a / while OVER OFF , ON , IN , etc. are categorized in this position, p_ is realized by a null FVI) . I nterestingly though, as Table ( 6 2 ) indicates, be headed forms never function a s particles . In contrast to the forms found in (3), be headed forms never surface as particles (4 ). (4) Be headed Forms Fail Inversion Test a. John played before the guitar *John played it before b . John played behind the computer *John played it behind. c . John played below the guitar . *John played it below. e . John played beside guitar *John played it beside f . John played beyond guitar *John played it beyond g . John played beneath guitar *John played it beneath. While some of the senten ces in exa mple (4) may appear strange, the purpose of (4) is to demonstrate that a DP cannot appear both between the verb and P item and after the P -

PAGE 218

218 item if the P item begins with the proposed morpheme / be / . In other words, one cannot find a verb+(be form ) that functions as particle and DP as was defined in Chapter 1 . This contrast is made even more explicit with (33). (5) a. John pushed his dinner aside. b. *John pushed his dinner beside. The fact that be headed forms never occur as particles is fur ther evidence that not only are the / be / and / a / heads active morphemes in English but that they represent different features. Given that be headed forms do not appear as particles and that particles are argued to be case less p_ heads (den Dikken 1995) , one might conclude that /be / is associated with a case feature. However, this is problematic because it was demonstrated in Chapter 1 that not all adpositional heads (or forms associated with heads) need to be case assigners. If all FVIs forms had assoc iated case features, sequences such as from behind the house would not make sense because two p_ heads would have to have a case feature while there is only one DP. Instead the Associated F eatures of the b e headed form seem to restrict the be head to p_s w ith a locational (Place) or a path (Extended) meaning as discussed in Chapter 1 . Therefore to explain the distribution difference between the /a / and /be / head, it is propose d that the FVI /be / is not associated with a case feature but instead with a fe ature that is common to Projective and extended Ps and absent with Particle Ps. 7 Ostensibly this feature is a type of transitive feature that entails a Ground, implied or otherwise, in a Place or along a Path. The FVI /a / on the other hand is less specifi ed than the /be / FVI , allowing it to represent both transitive and intransitive contexts (6) . 7 Of course the terminal node realized by /be / can contain a c ase feature.

PAGE 219

219 (6) The description in (6) will work as long as one ranks Contextual F eatures over associative features as in (7) . 8 (7) /a / [ Link BOARD LONG SIDE ACROSS __ etc. /be / [ Link, Trans LOW HIND YOND etc..} / / [Link] 9 DOWN IN NEAR If this is not desirable, perhaps a more discrete analysis of the features available will reveal a uni que feature set for /be / to capture its different usage pattern. T his section shows why the /a / and /be / prefixes cannot be considered contextually select ed allomorphs . Instead they must be treated as the realization of two different feature values . Thu s these phonological realizations can be taken as evidence that a p_ head exists independently from the relational LVIs that carry the conceptual meaning. Moreover, there appears to be other morphology in English that corresponds only with the category P. 8 The literature is unclear with regards to this. 9 Recall that the [link] features is merely pro posed to be the most basic adpositional feature, one that

PAGE 220

220 6 .2.2 The Suffix ward The suffix / ward / is another pontential piece of evidence for p_ as a categorial head. Forms suffixed with / ward / traditionally are categorized as a type of adverb, but such a categorization does not seem descriptively helpful as forms suffixed with / ward / do not show the distribution, morphology, o r semantics of typical adverbs. Instead, t he suffix clearly has a direction al meaning. It indicates the directi on of the place meaning of the it attaches to. The suffix / ward / attaches to forms argued to be s (8) but it does not at tach to derived prepositions (9 ) and those prepositional forms argued to be FVIs (10 ) . (8) skyward, homeward, schoolward, sideward, backward, forward 10 , afte rward, upward, downward, inward, outward, onward, toward 11 (9) *beforeward, *beloward, *besideward, *aforeward, *asideward *alongward, *aboveward, ? (10) *atward, *fromward, *ofward, *viaward, *withward, *byward, *asward In (8 ) ward attaches to several words that are clearly s: sky+ward, home+ward, and school + ward. This is further evidence that prepositional forms such as up and down Root s just as SCHOOL and SKY Root s . In other words, the se Root s +/ ward / must be categorized by something othe r than v_, n_ or a_. That is *the homewards, *homewarded, *very/really homeward are all ungrammatical, leaving Root s +/ ward / forms to be a type of p_ . The argument for viewing these elements as types of p_ is first based upon distribution and meaning, (1 1 ) and (12) . 10 Composed of fore and ward. 11 To be argued as not a composition of to + ward but rather a single FVI as its history comes from the lexical to which has grammatic alized to the functional variety.

PAGE 221

221 (11) a. John (*into the cave) walked . b. John walked into the cave. c. John (*inward) walked . d. John walked inward. (12) a. John (*up the ladder) walked up the ladder. b. John (*upward) walked upward (*the ladder) Examples ( 11 d ) and (12 b) show that a Root suffixed with / ward / occurs in the same place as a corresp onding adjunct PP. Both the pPs in (11b) and (12a ) have a directio nal meaning like the forms suffixed with / ward / in (11d) and (12b), suggesting that these are two instantiations of a simil ar category. In addition to the overlapping phrasal distri bution and semantic parallels, s suffixed with / ward / compete with other prepositional morphology , suggesting either that ward is the FVI representation of a type of p_ or that it occurs with a specific type of p_ . Forms modified by what has been argued to be a prepositional realization /a / or /be / can no longer attach to / ward/ as was seen with (8) (i.e. sideward but no asideward). This suggests that / ward/ is an exponent of p_ or that the type of p_ th at project s the feature represented by / ward/ is in complementary distribution with other types of p_ . I f the is suffixed with / ward / , the form no longer takes an object. These forms now appear much like directional particles, linking to only a DP Figure. T here is , h owever, o ne counterexample given with (1 3 ). (13) a. John flew to/toward the sky. b. John fle w skywards. In (1 3 ) the form toward breaks the pattern set above as it can take an object. In this case, it will be argued that the form toward is an FVI representing a p_ head that has the features [+ GOAL ] and [+ DIR ]. This is because to has been shown t o no longer be a

PAGE 222

222 in the language and this would explain why it can take an object while all the oth er forms cannot . Moreover , in addition to not being able to take a DP object, forms suffixed with suffix / ward / cannot be shifted with a verbal ob ject. Compare (14) an d (15 ). (14) John threw th e ball out . John threw out the ball. (15) John threw the ball outward. *John threw outward the ball. This appears to be the consistent pattern with these forms. Given this is appears that / ward/ is provides more evidence for p_ as a t ype of categorial head. This particular p_ head appears in complementary distribution to /a / and /be / and the FVI prepositions of (10 ). Thus / ward/ is either another phonological realization of a type of p_ head or a feature of the prepositional domain . Either way, this suffix provides more evidence for p_ as a categorial domain. 6 .2.3 Evidence for Little p_ in Chinese It may have been noticed that unlike English in Mandarin there is no phonological evidence for a p_ categorizing head (In English we fou nd /a /, /be / and maybe / ward/ ). However, this should not be surprising because in Mandarin there is little evidence for v_ either. Moreover in the case of n_ other than , a suffix that attaches to nouns tou , there are not many morphemes that could be used to indicate the existence of an n_ head either. Little a_ would be missing a phonological realizat ion as well. In this way it would be strange to find a bound morpheme realization of p_ . With this is mind, perhaps one could argue that a lack of phonological evidence is evidence that s are being categorized in this domain in the same fashion as th e y are in other environments.

PAGE 223

223 6 . 3 Evidence in o ther Languages When categorizing a node, positive phonetic evidence for the proposed categorial p_ head has only been seen with English. In Mandarin Chinese there were no cases of complex prepositions wh ere part of the adposition could be claimed to be the representation of a little p_ head . This was not deemed troublesome because in Mandarin Chinese a vast majority of all putative categorial heads must be labelled null. This does not mean though that Eng lish is the only language to show phonetic evidence for a categorial p_ head. Phonetic evidence is found in other Germanic languages such as German and Dutch. Furthermore, an analysis of Japanese postpositions shows that they too provide phonetic evidence for a functional p_ head categorizing a lexical relational item. The discussio n below will begin with German. 6 .3 .1 Evidence f rom German In German, postpositional forms ( 16 ) are often accompanied with the prefix her or hin as opposed to prepositions as s een in ( 17 ) . (16) a uf die L eiter o n/up the ladder (17) die Leiter hin auf the ladder DIR up In German, postpositions generally mark Directional/Path features. Her and hin function as deictic items ( her is used when the path is towards or with the speaker and hin is used when the path is away from the speaker) and are referr ed to as h prefixes (Zeller 200 1 : 5) . This can be seen in (1 8 ) and (1 9 ) . (18) Wulfila schickt den Hund hin aus. Wulfila sends the dog awayfrom out Wulfila sends the dog out

PAGE 224

224 (19) Wulfila jagt den Hund her aus. Wulfila chases the dog towards out Wulfila sends the dog out In example ( 18 ) and ( 19 ) it can be seen that the German prepositional form aus appears with either hin or her as a po stpositional form . Within a more traditional lexical model, Zeller (2001) analyzes h prefixation as a lexical preposition being converted into a functional element via suffixation w ith a functional zero morpheme ( 20 ). (20) German Post position (Zeller 2001: 42) As seen with ( 20 ) analysis the preposition aus is merged with a functional suffix realized by nothing and this suffix projects for the node realized by either hin or her . One of the primary reasons for this is the Right Hand Head Rule whereby Zeller reasons that categorical information cannot be given by a prefix in German. This thinking, however, appears flawed as it has been shown in English that the Right Ha nd Head Rule is not absolute, the exception being with prepositional items (i.e. downstairs , upstream Boertien: 1997) . The observation that the Right Hand Head Rule appears to capture is likely the result of another phenomenon. Moreover, the basis f or such a rule is unclear in a model like DM. Instead, the picture is simplified if prepositions and postpositions are modeled as demonstrated in ( 21 ).

PAGE 225

225 (21) German Preposition in Node In ( 21 ) a node realized by A US is categorized by a little p_ head defined by the feature [ LOC ]. This node is realized by / /. However, the FVI /hin/ or /her/ is inserted if the terminal node contains a directional feature ( 22 ). (22) German Postposition in Node Thus i nstead of claiming a zero functional suffix that projects for an F node that is realized as hin or her , it is here claimed that hin/her are representations of a directional little p_ head . This analysis helps capture many of the observat ions made by Zeller (2001) but does so in a DM compliant manner. It captures the fact that prepositions in German have lexical like semantic properties and the fact that postpositions also have a clearly lexical component. It reduces the need to postulate an empty suffix for both prepositions

PAGE 226

226 and postpositions. 12 Instead evidence for p_ is found to be associated with a feature difference. This same feature difference can be used to explain the systematic different meanings that postpositions have in comparis on to their prepositional counterparts. It is noted that further analysis and data needs to be considered to see if this analysis is viable. 6 .3 .2 Evidence from Dutch Much like English, many Dutch adpositional forms are morphologically complex (cf. Zwarts 1995, 1997) , containing a functional prefix followed by in many case a clear lexical item . As mentioned in Chapter 1, Muysken concluded that adpositions in compl ex This section will briefly intro duce some of these forms to show how a little p_ head is also necessary in Dutch and how Dutch provides more phonological evidence for p_ . As was seen in English, there exist a set of forms that are appear to b e more funct i onal than other prepositions (23 ). (23) Some Dutch Functional Prepositions (Muysken 2008:68) Dative Genetive Subject Marker of Embedded Clause Purposive The forms in (2 3 ) are not morphologically complex, do not appear in traditional lexical positions, and are associated with the listed grammatical roles. In addition to the forms in (2 3 ) , there are , however, many adpositio nal forms in Dutch that are morphologically 12 Zeller does not have to worry about the suffix with the preposition because he is working under the assumption that the preposition as a lexical item comes with its own category and projections.

PAGE 227

227 complex, cons isting of a lexical ( like) item and possible category changing morphemes . First take a look at the following items from Dutch in Table 6 3 . Table 6 3. Dutch b ase s Use Meanings 1. uit Preposition Out 2. in Preposition In 3. over Preposition Over 4. Halve/ halver noun Half 5. houd ? Keep 6. naar bound GAP down/under 7. nev bound GAP by to the side 8. tref noun Luck 9. noord noun North 10. zuid bound South 11. oost bound East 12. west bound West 13. zijde noun Face/side C olumn one of Table (6 3) l ists thirteen items, column two lists their common syntactic usage, and column three gives the specified meaning of each item as pertaining to the particular use listed in column two . It should be observed that the first three items can already be used as pr epositions . T hey are therefore analogous to out, in and over in English. It should also be noticed that the rest of the se items are not used as prepositions . 13 The precise reason for these gaps and others in the Dutch data is not explored in this work , but it is surmised that they are accidental as defined in Chapter 3. Nevertheless, while there are a few derivational gaps, most of these forms may undergo categorical derivations by being combined wi th the suffix ( es ) / en/. 14 Derivations with / en/ are g iven in Table (6 4). 13 The use of houd 14 As seen with Table (6 4), / en/ appears to cause Root items to become both verbs and nouns and thus this may be a case of homophonous suffixes.

PAGE 228

228 Table 6 4. Derivation with / en/ Form + / en/ suffix(es) Use Meanings uiten verb innen verb oven GAP GAP halveren verb To halve houden verb To hold Neden/naaren GAP GAP neven GAP GAP treffen Noun/ve rb a battle/ to meet noorden noun North zuiden noun South oosten noun East westen noun West zijden GAP GAP The items in Table (6 4) are organized as in Table (6 3 ). In some cases, the s are used as nouns and in other cases as verbs, indicating that either / en/ is representing two different categorical heads or something else. This is not what is important here. The important thing to note is that this appears to be the intermediate stage for a further adpositiona l derivation, seen with Table (6 5 ). Table 6 5 . Derivation with /b(e)/ Form + /b(e) / prefix use Meaning buiten preposition outside 15 binnen preposition Inside/until/on boven preposition behalve preposition behoudens preposition beneden prep osition benevens preposition betreffen preposition benoorden preposition bezuiden preposition beoosten preposition bewesten GAP GAP bezijden prepositon 15 This one has several meanings.

PAGE 229

229 The data s een with Table 6 4 is very similar to what was seen with the proposed prepositional heads /a / and /be / in English except that this process seems to be more productive in Dutch. One can see relational words whose base form is never beoosten with the addition of head /b(e) /. Likewise this head changes verbalized prepositional forms back into prepositions, suggesting the presence of stacke d categorial heads on a item. This phenomenon in Dutch thus strongly supports the proposal here that we include a little p_ head as a categorizer of items. In addition to the fact that there appears to be a produc tive process in Dutch to deriv e new adpositional forms, there are some unexplainable gaps such as the non usage of westen s that meaning wise could be used as prepositions but for some reason have not been used as such (i.e. s like north/south/east/west; bottom ; right/left) . Thus i t seems perfectly possible for the following to have been in the language or t o become part of the language ( 24 ). (24) Gaps in the English system a. I travelled asouth the Mason Dixon Line . b. The y found a shoe abottom the lake. c. Your book is aright of you. d. I travelled besouth the Mason Dixon Line . e. John found the treasure beleft the rocks . The forms in (24 ) are seemingly not a part of the English language (some being perhaps more possible others) but they represent possible forms in what has been a productive process, categorizing informational in adpositional domains. 6 .3 .3 Evidence from Persian

PAGE 230

230 In Persian there is a small set of clearly functional spatial elements (Samiian 1994; Pantcheva 2007a; 2008 ) According to Pantcheva these such as gen eral location, goal Path, and As seen with English, Mandarin , and Dutch , Persian also has a larger set of /lexical like items that can be used to convey more complex spatial relations. Both types are given with Table (6 6 ) Tabl e 6 6. Persian p repositions 16 Persian FVI Ps Gloss Features Persian LVI Ps Linker æz from [source] tu ye in be to [goal] ru ye on ta up to [dir, goal] jelow ye In front of b ær on [location, contact] pæhlu ye beside d ær in,at [location] daxel e inside ba with [comitative] tæh e On the bottom bi without [ comitative] posht e behind næzdik e near paiin e below xarej e outside birun e outside mian e between dour e around Pantcheva ( 2007a , 2008 ) calls the items from column 1 in Table (6 6) Spatial Class 1 Ps , citing Mace (2003) and Mahootian (1997) in asserting that these items are purely functional items. On the other hand, Pantcheva ( 2007a, 2008 ) calls the items in Column 3, Spatial Class 2 Ps because the items from column 3 ar e suffixed with either / e/ or / ye/ when used as a preposition while the Class 1 Ps are not . Class 2 Ps also appear to be lexical (i.e. they are often considered to be nouns because they may also operate in nominal positions and host pronominal clitics (P antcheva 2008 ) . 17 Therefore 16 Table 6 6 is adapted from Pantcheva (2007a, 2008). 17 Pantcheva (2007) classifies those that take the / ye/ marker as Class 2A and those that take the / e/ as Class 2B.

PAGE 231

231 they are often also classified or treated as nouns (Samiian 1994). Moreover, in Sven onius (2006) and Pantcheva (200 8), they are argued to function as Axial P art prepositions (cf. Chapter 1 Sections 1.5 1.7 ). As for the suffixes / e/ and / ye/, they are generally referred to as Ezafe markers (Samiian 1994; Larson and Yamakido 2005 ; Svenonius 2006; Pantcheva 2007a, 2007b, 2008 ). Describing this marker, however, in a unifying manner has been difficult because the same form (whether or not all instances of this form correspond to the same morpheme is another question) appears in many places. and presently appears with several different categories of post nominal modifiers : d escri ptive n ouns, attributive a djectives, and d escriptively used Class 2 Ps 18 (Larson and Yam a kido 2005; Pantcheva 2007b). It also appears between heads and complements of different categories: adjectives, nouns, and partitives (Larson and Yamakido 2005) , between two nouns in a possessive relationship, between an infinitival verb and complement noun (Pantcheva 2007a), and between a Class 2 P and its DP Ground (Pantcheva 2007a) . Given this distribution, the marker is often called a linking element. More spe cifically Samiian (1994) proposes that Ezafe is a case marker /assigner (which does not seem outrageous given the etymology of the form) . Lik e wise Larson and Yamakido (2005) and Svenonius (2006) treat it as a c ase assigner . I n Svenonius (2006) it is analyz ed as a realization of c ase in a KP at the base of the extended projection of the category P. Here the Class 2 Ps (or AxPart items ) will be analyzed as nodes categorized by a type of p_ head because of their otherwise lexical distribution. As seen 18 This is less of a problem in DM because we can view these items as being a categorical if th ey are

PAGE 232

232 w ith English, Mandarin and Dutch, the Class 1 Ps in Persian will be analyzed as FVIs inserted under p_ . Moreover, it will be proposed that the Ezafe marker that occurs with Class 2 Ps be treated as the spell out of the proposed little p_ head categorizing t hese elements as prepositions. Furthermore, i n a model like DM, the syncretism of / e/ and / ye/ can be at least partially explained by underspecification (i.e. if Ezafe is only associated with a common feature it might win i n several different feature set s if there is no better suited FVI). Continuing with the previ ous work, the feature could be c ase ( little p_ arguably would be a case assigner ) or a general [ LINK ] feature . That is despite the fact that Larson and Yam a kido argue that Ezafe is a c ase marker , they ask a very good case (2005:1). The answer here is that perhaps they do not, suggesting that the Ezafe form can r epresent terminal nodes with a c ase feature and ones without a c ase features. It is thus proposed tha t not only does Persian provide further evidence for the proposal here , but that the DM model and proposal here may aid understanding the large distribution of the Ezafe marker. The remainder of this section will show how Class 2 Ps are different than noun s and show how the Ezafe marker can be viewed as a phonological representation of p_. As mentioned, there exist lexical forms in Persian that can be used as prepositions (Class 2 Ps). I n (25 ), one of these forms is given in a no n prepositional domain and in (26 ) it is seen operating as a preposition. (25) Nominal Use (Larson and Yamakido 2005: 4) a. in ru this top b. un zir a that under pl

PAGE 233

233 (26) Prepositional Use ( Svenonius 2006 : ). a. Ræft posht e xane. went back EZ house b. Sib oftad ru( ye) miz. apple fell face EZ table In (25 ) it is clear that ru zir are being used as nouns as they are both preceded by deictic determiners and zir is marked as plural in (25 b). On the other hand, in (25 b) it can be seen that two items from the same class of items , posht and ru , are being used as prepositions, ostensibly linking DP figures to DP Grounds. As previously mentioned, in this environment these items consist of a base form (i. e. the lexical item) and the Ezafe marker , a phonological representation of p_ . Now if this was the only place whe re Ezafe occurred, the truth of the proposal here would be as obvious as it was in Dutch. In order to begin to accept the possibility that / e/ and / ye/ are phonological representations of p_ , it must be sho wn that the Class 2 Ps as in (26 ) are truly dif ferent than nouns (even though the same form can also appear as nouns) to show that these forms are being derived as prepositions . First compare (27) and (28 ). (27) kif e ch æ rm bag EZ leather (28) zire e der æ xt under EZ tree In (27 ) and (28 ), one may propose that the ident ities of these forms (i.e. the structure aroun d them) is the same, glossing (27 ) as the bag of leather and (28 ) the under area of a tree . I t can, however, easily be seen that the relationship between the form s zire

PAGE 234

234 under and deræxt tree in (28 ) is different than kif bag and chærm leather in (27 ). Class 2 Ps suffixed with / e/ and / ye/ cannot be modified by an adjective (29 a) or be qu antified (29 b) ( Pantcheva 2008 ) , while form s in a nominal position can (30 ) and (31 ) . (29) Inability to be Modified (Pantcheva 2008 : 5) a. Reza ræft be tu ye *tarike e gænje Reza went to inside EZ dark EZ closet b. * hær zir e miz every under EZ table (30) Nominal Linking (Pantcheva 2007b: 2) ketab e xub i book EZ good indef (31) hær kif e ch æ rm every bag EZ leather A noun suffixed with the Ezafe marker can be modified but this cannot happen if the form is operating as it does in (29 ). Thus it is reasonable to label this item a s a preposition as done in (26 ) . The question then is if Ezafe is really marking a categorial p_ head or so mething else as the forms in (30) and (31 ) are not prepositions. As mentioned at the beginning , t he claim here is that the Ezafe marker is not exclusi vely used to spell out p_ because it appears elsewhere. It is also likely not representing a little n_ in (30) and (31 ) because when Ezafe appears between two nouns , they do not have a meaning typical of compounded forms (32 ) . (32) Nominal Linking (Pantcheva 2007a:3) Kelid e d æ r i Key EZ door indef

PAGE 235

235 Moreover, a s can be seen with (3 3 ) , the Ezafe marker is missing in certain post nominal modification arrangements . (33) (Pantcheva 2007a:2) ketab i xub book indef good Example (33) shows that if the indefinite suffix is attached to the noun , it will not take the Ezafe marker , also suggesting that t his is not an FVI exponent for l ittle n_ . From the discussion given in this section, it is proposed that the Ezafe marker is not solely used for framing Root items in adpositional domains. Likewise in English , /be . T his could be accidental (i.e. like / / in happy+er vs. think+er) or be caused by an FVI being underspecified (i.e . / / in I/we/y ou/you(all)/they walk / / vs. / z/ he/she/it walk). Other features may of course bundle with this feature, allowing the same FVI to represent a variety of linking grammatical relations between different categories. In the case of p_ , it is the [ F LINK ] feat ure which may also bundle with a Case feature. Thus as seen in Pantche va (2008), the Ezafe maker remains in stacked constructions despite the fact that a Class 1 preposition would also be a c ase assigner as seen in (34) . (34) (Pantcheva 2008:10 ) a. (d æ r) posht e xane at behind house b. (be) posht e xane to behind house c. *( æ z) posht e xane from behind house

PAGE 236

236 As can be seen with (34 ), Class 1 Ps may co occur with Class 2 Ps. More over, while the Class 1 P in (34a) and (34 b) can be omitted , the Ezafe marker is always present, suggesting that it does not have to correspond with a C ase feature because if the Class 1 P is present, the Eza fe marker would then become unnecessary . Moreover , in (34 c) the Class 1 preposition / æ z/ cannot be omitted for some reason . The reason for this is left for future research as this description is merely beginning to account for the behavior of the Ezafe mar ker in accordance with the ideas presented in this work. Nevertheless f rom the data in ( Samiian 1994; Larson and Yamakido 2005; Svenonius 2006; Pantcheva 2007a, 2007b, 2008) it seems that this form can be explained if one proposes that it is associated w ith a [ LINK ] feature. The purpose of this feature is simply to connect concepts that have not been integrated via other grammatical features. 6.4 Concluding Remarks Chapter 6 reiterated the need for a little p_ head in the DM model as it presented more examples of lexical items being used as prepositions in English (/a / and /be / +lexical item) and showed that in Dutch and Persian adpositions are created from other lexical items . In doing so, it presented strong phonological evidence for p_ in English and Dutch and suggested that hin and her be looked at as possible evi dence for p_ in German. Likewise it was argued that the Ezafe marker is a phonological representation of p_ when lexical items are used as prepositions in Persian. It also showed that Dutch and Persian have a smaller class of strictly functional adposition al items, as was seen with English and Mandarin. The findings here thus suggest that items are often framed in prepositional domains and that the features responsible for the framing often get phonologically realized. It also further supports the ide a that in languages where

PAGE 237

237 items are framed in adpositional domains there exists a class of similar adpositional features (not used to frame items) that are only realized by an FVI adpositional form.

PAGE 238

238 CHAPTER 7 LITTLE P_ IN A WORLD OF N_,V_, AND A_ 7 .1 Introduction In accordance with the discussion of Chapter 3, Chapter 4 showed that many adpositional forms have the distri bution of a node. Chapter 3 also showed that adpositional items often appear to have unpredictable meanings in differ ent categorial environments, suggesting the presence of a bare and categorial head. Chapter 5 therefore proposed a little p_ categorial head, a label for a variety of features unified by the [ F LINK ] feature proposed in Chapter 1, as the categorial h ead responsible for framing items in adpositional domains. In order to control for insertion, s inserted into adpositional domai ns are argued to be listed with little p_ C ontextual F ea tures. Chapter 5 also argued that those adpositional forms la cking a distribution are FVIs that directly represent little p_. In cases where a item from Chapter 3 was categorized as an adposition, it was argued that a null FVI represents p_. In def ense of this proposal, Chapter 6 showed that separate pho nological evidence can be found for p_ , proposing that the /a / and /be / prefixes in adpositions such as along and beside are phonological realizations of p_ in English and that /b(e) / in beoosten ruye ian are also phonological realizations of p_ . This c hapter further justifies the proposed head p_ by comparing its properties and behavior to the other types of categor ial heads . Differences and similarities between variations of little p_ and those of li ttle v_, n_ , and sometimes a_ will be discussed in order to both support the proposal for little p_ as well as draw attention to possible issues. The similarities between p_ and v_ and n_ include function, derivation,

PAGE 239

239 realization, and some featur al makeup. The differences include distribution, featural makeup, a set of functional projections that merge with the categorial head, and perhaps universality. Nonetheless, the point that should be clear is that little p_ will share properties with both n_ and v_ , as all the categorizing functional heads could be viewed as a subtype of little x_ 19 , which can simply be thought of as a set of grammatical features that enable purely conceptual information to be realized in language. Viewing things in this manner will n ot alienate p_ , as it will become clear that not all languages utilize the same variations of little x_ to frame concepts . This Chapter is organized as follows: Section 7.2 discusses the issue of there being universal lexical categories (as examined in Cha pter 1) from a DM perspective. Section 7.3 discusses the general behavior of the categorial heads as discussed in the literature in comparison to little p_ and discusses how the similarities between p_ and n_, v_ and a_ outwe igh the differences. Section 7. 4 then goes on to show that p_ can be the base for additional derivation al operations and that it is associated with a set of functional features just as n_,v_ and a _ are associated with sets of features . 7.2 The Universality of Lexical Categories In Chap ter 4, a little p_ head was proposed to be the categorial head responsible for framing content in an adpositional domain. In contrast to the other categorial heads, one might argue that such a proposal is problematic because not every language has ad positional elements as was suggested by Croft (199 1) and argued by Baker (2003) and discussed in Chapter 1 . If this is true and it is true n_ , v _ and a_ exist in all languages, this would make little p_ distinctly different. One goal of this section, 19

PAGE 240

240 howev er, is to show that excluding little p_ because traditional classification methodology suggests that it is not as common as n_, v_ and a_ is a poor argument because the idea that any of these general labels are universal is controversial and meaningfully unsubstantiated. Another goal is to show that DM provides no explicit argument for limiting categorizing head types . Likewise if the isolation of n_,v_ and a_ is ba sed upon empirical findings then those findings are now insufficient as demonstrated by Chap ters 4 , 5 , and 6 . This section will discuss this issue by first examining the positions of linguists in the general field and comparing that to how the notion of lexic al categories is treated in DM. The proposal here is that the framework of DM encourages discreteness and that the LDH opens the door for different feature values to merge with nodes. 7 .2.1 Lexical Categorie s and Universality As mentioned in Chapter 1, one of the two major problems regarding lexical categories is the inability of lingu ists to more or less uniformly interpret the data to 185). 20 crosslinguistic evidence is not converging on a smallish set of universal (po ssibly innate) cate almost every newly described new category that hardly fits heme of the parts of speech their number, nature, and necessary confines 20 This of course seems to be corollary to the first problem (i.e. the inability to more or less uniformly determine categorical identity).

PAGE 241

241 language has i e n 1965: 91). If this is true, it would seem that cross linguisti c syntactic generalizations could never be based upon lexical distinctions and that the common labels used are merely the result of pedantic pedagogy rather than accurate/meaningful scientific descriptions. In contrast to the view above, is the claim that linguistic items are being organized by means of UG. If this is true, UG must be limiting the grammar with a number of discrete choices (otherwise it could not be universal). 21 There are thus claims that all languages at least have the categories Noun, Ver b, and Adjective (Baker 2003; Dixon 2004). Such claims are of course done under the premise that these labels are in syntax (1991: 2). If this is true, it suggests that these distinctions can form the basis for accurate cross linguistic generalizati ons. 22 Given these two positions, linguists appear to be at a crossroads, where they can either abandon a linguistic explanation for why so many languages have items that linguistic di fferences mentioned by Sapir and Haspelmath. It is argued here, however, that both positions d o not have to be exclusive. 21 It may be true that as UG is simplified it looks more and more like UG is primarily created via Merge as Chomsky muses. H o wever, if the items that can undergo either internal or external merge can have any iden tity then we would expect language to be very different than what is seen. 22 T hese general izations , however, may show exceptions if these putative universal categories are just labels that are more or less appropriate for describi ng items in a specific l anguage .

PAGE 242

242 This controversy can be reframed with the following possibility: perhaps all word classes are potentially universal in the sense that a new language may select its categories from a definitive list. However, given that word classes may be defined by a composite of features and that these features may be allowed by the grammar to bundle differently, there is a great chance for items to d iffer cross linguistically while being composed of similar (or the same) pieces. This idea will be evaluated with what has been said within Distributed Morphology. 7 .2.2 Universal Lexical Categories and DM As mentioned in Chapter s 1, 2, and 4, a majority of the DM literature lists n_ , v _ and a_ as the categorial heads used to frame s. One may surmise from this that those working within the DM model have not only accepted the typological observations of Baker (2003) and Dixon (2004) that all languages have nouns, verbs, and adjectives but also concluded that universality is a fea ture of the categorial heads. This would explain why Embick says that and not others, this is not the case with v , n , and a In this way, these heads are viewed as being special and different from other terminal nodes. Such a position, however, appears to gloss over many facts and certainly does not take into account the kind of data discussed in this work. It is unclear how Embick is defining n_ , v _ and a _ in or der to substantiate his claim. It is unclear if these categorical heads are being viewed in a mono dimensional manner as given in Baker (2003), where nouns and verbs are created by solitary noun and verb features, or if these categories are built from smaller discrete features. If a mon o dimensional approach is being taken then this does not account for the different

PAGE 243

243 l anguages (Haspelmath 2007 ) and it does not account for why adjectives may be limited or seemi ngly non existent in some languages (cf. Kim 2002; McLaughlin 2004 ). On the other hand, if Embick (2012) is viewing this in more discrete terms then n_ , v _ and a_ are not the same thing as the Nouns, Verbs, and Adjectives discussed in Baker (2003) and Dixo n (2004). Moreover, a proposal where n_ v_ and a_ are actually created by a range of similar features appears to correspond with how these heads are treated elsewhere in the literature. According to Harley and Noyer, the L morpheme Hypothesis , an integra l part of DM as discussed in Chapter 2, elements, suc h as noun, verb, and adjective have no universal significance and are 1999:7). While it is also u nclear what these basic morpheme types exactly are, one can conclude that they are talking about a variety of little v_, n_ and perhaps a_ heads. With this approach, claiming that n_ , v _ and a_ are universal is not the same thing as claiming that N, V, and A are universal in a framework that treats these labels as mapping to one undifferentiated feature. Given this discussion, it is proposed that the categorial heads be viewed as a range of features that merge with and frame nodes. Within this range, one may label the heads as n_ , v _, and a_ with the understanding that the specific feature for that head may be di fferent than another head label ed the same way and that a more discrete analysis may at times be warranted. This view means that the heads n_ , v_, a_ and p_ are not themselves primitives . This attempts to account for the difference seen

PAGE 244

244 perhaps a common feature or a class of similar features will be necessary in every language, making these classes universal. In this way little p_ is to be viewed as just another range of categorial features. 7 .2.3 Universal Lexical Categories , DM and Little p_ There is no theoretical reason to claim that only v_, n _ and a_ varie ties categorize nodes. If it could be clearly shown that v_, n _ and a_ are specific and universal features and that universality was a mandatory prop erty for categorial heads in DM then p_ might be excluded for such reasons. However, the universality of v_, n _ and a _ has not been clearly demonstrated and there is no clear reason for why a categorial head woul d have to be universally used. In fact if the features that compose adpositions are also responsible for creating case marking systems, as the two systems share many similarities (Asbury 2006; Hagege 2010), this class would be just as prevalent if not more as the adjectival class. Furthermore, s ince items can be found with identities that are not the same as those standardly given by v_, n_ and a_ , as seen in Chapter s 4 and 6 , to isolate v_, n_ and a_ as being the only categorial heads is no longer empirically valid. We must therefore conclude t hat there is another type of categorizing head or that the bundle of features that normally const ruct a [+v], [+n], or [+a] domain have been altered to form another subclass (i.e. perhaps someone would want to re label p_ as a type of v_ ) . 7 .3 The Behavio r of the Categorial Heads As discussed, there is no clear theoretical or empirical reason to exclude p_ as a categorial head. Moreover, this section shows how a deeper examination of the features responsible for p_ in comparison to the features proposed fo r the other heads only further aligns this head with the other heads. In other words, one can see that p_ is very

PAGE 245

245 similar in behavior to the other categorial heads in its ability to form the base for additional categorial derivations. 7 .3 .1 L i ttle v_ in R elation to L ittle p_ In Chapter 1 it was proposes that little p_ is derived from an [ F LINK ] feature that relates a DP figure in accordance with the meaning of the adposition. This feature may bundle with other spatio temporal features to derive different types of Path, Directional and Place interpretations. Moreover as a preposition, [ F LINK ] connects a DP Figure with a DP Ground and sometimes a structural c ase feature is also present. These proposed features will now be compared to the identified features of the other two best studied heads: n_ and v_ . Harley states studied head of this type is the verb This section will thus examine why little v_ has been proposed, its pr operties, the feat ures associated with it, and how it is use d to account for different predicate structures so it may be compared to the proposed role of little p_. Historically little v_ comes from the expanded VP hypothesis: a hypothesis that attempts to account for more complex verbal structure (Larson 1988) . 23 One of the primary purposes of little v_ was to be the locus for light verbs, hence the name little v_ (Chomsky 1995). In addition to giving a location to light verbs, little v_ was also used to sp lit the job of introducing arguments, giving the role of introducing the external argument to little v_ and that of the internal argument to the lexical V head. 24 As can be 23 It proposed two layers to the verb phrase in order to allow for a better explanation of ditransitive verbs. 24 The job of assigning an external theta role is still a function of little v_ in DM. Perhaps this is analogous to a function of p_ .

PAGE 246

246 surmised, the job of assigning the internal theta role , however, is also with the v_ head, as the lexical Big V head is no longer present in DM because little v_ gives the its identity (Embick 2004). 25 However, t he reason to have a little v_ head is more than just to give a verbal identity to an uncategorized in order to uphold the categorization assumption of Embick and Noyer (2005). The fact that little v_ is a functional node defined by some grammatical feature or set of features in DM felicitously allows for the possibility of there being more than one type of little v_ . 26 Since there are different classes of predicates and because these classes can be explain ed via the proposal of different verbal like grammatical features, several varieties of v_ have been proposed. There have been proposals for [v cause] [v be], [v have], [v do], and [v become]/[v fient] heads to account for the different verbal classes: cau s atives, statives, possessives, actives , and recipients (Harley 1995a , 2008:5 ; Embick 2004; Folli and Harley 2007). These classes can be defined by the following features or features plus some common verbal feature. If little v_ is created by a [ + CAUSE ] fe ature, it assigns an agentive role to the external argument. An example of this can be seen with (1 ). (1) Acquaviva [v cause clarified] the description of DM. Acquaviva 25 A s for the internal argument being introduced by V and the external argument being introduced by v_, the evidence to suggest that was based upon the fact that agents cannot be a fixed part of an idiom. This provided the separation to explain this as V and i ts complement could only form the idiom. Since Big V is gone in DM, Marantz (1997) explains this phenomenon as being the result of the fact that the of the Agent would be above the v_ head that projects the agentive role. If this v_ head is marked as the boundary for special meaning then no incorporated special meaning could be given s ab ove and below this v_ head. 26 This is because we do not expect languages to always package morpho syntactic features in the same manner, although patterns are not unexpected either.

PAGE 247

247 In (1 ) we can explain the specific s emantics and structure of this verb phrase with a v_ head with a [+cause] feature. A verb with this feature requires a DP that can become an agent and a DP to act as a complement. This has also been represented with a [ + AG ] feature for agent (2 ) . (2) Agent F eature with Little v_ (Embick 2004: 366). In (2 ) this feature would project for the structure that would host a DP that would then receive the Agentive theta role. 27 Moreover, if this causative or agentive feature is replaced with another proposed feature such as [+ PASS ] or [ CAUSE ], the result is a clause where no external thet a role is assigned as se en in (3 ). (3) The description of DM [was v cause clarified] In (3 ) as in (2 ), it should be clear that ture [+pass] can be tied to the features of (Embick 2000: 204). In other words just as there is a v_ related to active, transitive syntax, and with a particular feature content, so there is a v_ associated with structures in which only internal argumen 204). Furthermore, it should be obvious that these two features, [+ CAUSE ] [+ PASS ], would not bundle together. Thus perhaps the verbal structure described above could be explained via the presence or absence of the [ CAUSE ] fea ture alone. 27 Another set of related possible little v_ features are [+/ transitive] features. Many SINK may appear in both transitive and intransitive contexts. This difference might also be explained by the presence of the alr eady mentioned [+/ cause] feature (Embick and Noyer 1999; Harl ey and Noyer 2000; Siddiqi 2005 ), possibly eliminating the need to posit [transitive] features.

PAGE 248

248 Another discussed possible little v_ feature is the [+ BE ] feature, which is found when verbs have a stative interpretation as in (4 ). (4) Students [v be fear] not passing the ir comprehensive exams . In (4 ) the little v_ head does not assign an a gentive theta role and instead would assign an experiencer role to the students . It should be noted here that there is not complete agreement on what subtypes of v_ exist. For example Embick (2004) argues that stative verbs are the result of a verbal struc ture that lacks a v_ head, which is repeated in Travis (2005). In these cases obviously no external theta role would be assigned. Nevertheless, if a [v be] exists it may create at least two different thematic links, an experience r role, and a theme role. Compar e HAVE type in (5 ). (5) John [v have has] a small problem. In (5 ) John gets a possessive role while a small problem is a theme. Verbs that appear in this environment all require a complement. 28 Th e same goes for [v do] verbs (6 ). (6) John [v do dances] for joy. In (6 ) John is not an agent or an experience r but rather a doer or theme for lack of a better word. s in this position with this interpretation can be explained via the presence of the [ + DO ] feature. In addition to this, verb clas ses created by the [ + DO ] feature could be further modified by the presence of a [ + TRANSITIVE ] feature to explain the possibili ty of a transitive version of (6 ) such as John dances a jig . This is comparable to the next type of verbal feature, one that can b e summarized as a [ + BECOME ] feature (7 ). 28 The question remains whether the [ v have ] feature is inherently transitive as the [cause] featu re is argued to be or if we should conclude that it always comes with a [transitive] feature. The issue of a [+/ transitive] features in addition to [ +v ] feature also app ears with [v do] verbs .

PAGE 249

249 (7) John [v become grew] six inches taller. In (7 ) John is the recipient of an internal force. He neither acts nor is acted upon and his role is different than that of an experiencer. These types may also be both tran sitive and intransitive. All in all it was seen that little v_ is a location for functional v erbs, a categorizer , and a the ta role/ c ase assigner . Moreover, the little v _ head can be divided into different descriptive types via the fundamental understandin g that multiple features are really what compose the [+v] feature . 29 All these features are supposed to capture verb classes within a Lexical Decompositional Model. Moreover, these features all share a similar flavor in that they situate the concept provide Root to another. In terms of features, little v_ can be generalized as features that causativity, eventivity/stativity, licensing of external arguments, and certain syntactic hermore, some of these features are not unique to the v_ head, as they can be seen with p_ as well. Little p_ was argued to be a categorizer, a location for functional prepositions, and a relator o f two DPs . Likewise this analogy can be made with the featu res that compose the types of little n_ heads. 7 .3.2 Little n_ in Relation to L ittle p_ We have seen arguments supporting the idea that there is features that are Root but also responsible for determining the overall argument structure of the Verb Phrase. Likewise in cases where Root s appear in nominal environments , there are proposals for different types of little 29 These features are not the only possible little v_ featur es nor is it certain if they themselves are the primitives or if they result from the combination of other features.

PAGE 250

250 n_ heads. This s ection will examine some of these proposals in order to gauge the properties of little n_ in comparison to types of roles performed by little v_ and little p_. Little n_ as a general category inducing functional head is responsible for nominal interpretati Root s. Moreover, beyond merging with a to give a nominal identity, little n_ has also been proposed to project for the structure seen in noun information wi analogy to having verb phrases be projections from little v _, Adger (2003) 30 argues that noun phrases are projections of little n_. Thus little n_ is responsible for the argument stru cture of the noun phrase just as little v_ is responsible for the structure present in the verb phrase, as seen in the previous section (i.e. t he deter miner node in the phrase is analogous to a Tense node in the verb phrase ) . While t he featural consistenc y of little n_ is not terribly clear across the pertinent literature, there are different proposals for little n_ as the locus of several features distinct from little v_ . These heads are most notably responsible for the different types of affixal inflecti ons seen with Root s in nominal environments. In Lowens tamm (2008) and Lampitelli (20 10 ) the argument is made that little n_ is Root s in French and Italian and by extension other languages. These grammatical gender features, which c ould be represented as [n 1, n 2, n i ], are viewed as being responsible for types of n_s . The s e grammatical gender features will simply be listed here as a type of [ CLASS ] feature. 31 30 It should be noted that Adger is not working directly within DM theory. However, some of his observations might prove relevant to the g eneral discussion. 31 This feature is clearly not verbal or adpositional.

PAGE 251

251 In addition to [ CLASS ] features ( recall Alexiadou (2011) from Chapter 3 wher e n_ was described as h osting a lexical plural feature ), t his feature is described as being different than he reason why ASNs [argument both the nominalizing affix and lexical plural realize n , and hence are in complementary n_ while another type of n_ is responsible for ASNs. A regular plural marking would be argued to happen on the Number node in the outer cycle while another type of plural feature can create a functional head that will merg Root , forming the lexical plural. T he literature describes n_ as being responsible for creating the phrase structure of the nP. Furthermore, n_ has been shown to consist of features [ CLASS ] and [ +/ LEX PL ], for lexical plural. In this way n_ can be summarized as representing grammatical meaning into the partitioning or sorting scheme of the language. 7 .3.3 Little p_ in Comparison to L ittle n_ and v_ To recapitulate, w e have seen ar guments for functional heads representing a variety of adpositional, verbal, and nominal like features t hat are capable of licensing a . Root s but also determining the overall argu ment structure of the phrase they head. I t has been shown that the little v_ head is the result of [ CAUSE ], [ BE ], [ HAVE ], [ DO ], and [ BECOME ] features and that these feature types might contain sub variations as well (i.e. [ TELIC ] and [ TRANSITIVE ] features correspond to the v_ head ) . The semantic link is that they are all event oriented. Moreover, little v_ is a c ase assigning head. On the other

PAGE 252

252 hand, the features of little n_ [ CLASS ] and [ +/ LEX PL ed. Like little v_ , little n_ merges with a particular set of functional heads: number, quantity, and definite heads. As for case, little n_ has nothing to do with assigning case . These properties can now be compared to the proposed little p_ head. Recall ing the description of adpositions from Chapter 1, it can be seen that little p_ aligns more with v_ than n_ but also that little p_ differs from both heads . The features for p_ . In addition to this it might be p ertinent to have a [ TRANSITIVE ] feature, which would be absent in the case of particles. However prepositions may also be bundled with [ TELIC ] feature s. This is demonstrated with (8 ). (8) a. John built a house in a week. *b. John built a house for a wee k. In (8 ) built has a telic feature, the action has an endpoint, and thus the preposition in must also share this featur e , which it does. However, in (8 b) the preposition lacks this feature, making it ungrammatical. For instead has a durative meaning. It is the authors judgment that ( 8 b) can be grammatical if built is not thought of as an action with a clear endpoint. This would merely prove that this feature exists outside of the , giving Root both possible interpretations. This can be shown if the subsequent DP object is modified to direct the interpretation (9 ). (9) a.*John built houses in a month. b. John built houses for a month. In (9 ) built has an atelic feature, the action does not have an endpoint and likewise the grammaticality ju dgments with the p repositions are reversed from ( 8 ). This goes to

PAGE 253

253 show that p_s not only can relate one or more DPs as verbs can but that they can have some of the same features: transitivity, and telicity. Thus p_ and v_ at least share the possible featur es [ TRANSITIVE ] and [ TELIC ]. Moreover, the features of p_ require additional structure to be present as the verbal features do. There must be a DP figure if Root item with another, thei Root with (an ) other Root (s) as v_ heads do. Likewise p_ heads can assign case to their object DP. W hile the semantics of the features that make up the core identities of each type of categorial head is different, the ability t Root synthesis (via a giving and receiving role) appears to be a commonality supported in the literature, although not explicitly stated. Given this , the proposal for p_ seems well justified as a close neighbor to v_ . It may though be found wi th further study of postpositions that these variations of p_ might resemble variations of n_ , a s n_ itself needs further study. 7 .4 Structural Similarities This section shows that prepositions may form the base for other categorial derivations, linking t his category with the other categories which also may form the base for other operations. Moreover, recent work on adpositional structure has revealed that this category does often co occur with a set of other functional features, making this head very muc h like the other established categorial heads. 7 .4 .1 Deriving Categories from Categorized s As discussed in Chapter 2, Distributed Morphology does not claim that all categorization occurs from a single categor izing head merging with a bare . It is that determines syntactic category (Marantz 2001:6). Thus

PAGE 254

254 after a obtains a more fixed interpretation by merging with a categor ial head, it may be compositionally reframed by other categorial heads (Arad 2003 , 2005 ) . This is seen with (10 ) (10) In addition to being stackable , as seen with (10 ), 32 each of the categorial heads will have s electional requirements for the type of categorial head it can attach to and this is reflected by how they get spelled out (Marantz 2001). In other words, a little a_ that gets spelled out as / ive / selects for a little v_ head and a little a_ head that gets spelled out as / ous / selects for a little n_ head. 33 There also appears to be hierarchy concerning how high a type of categorizing head can att ach. For example, the suffix / ness / is a high attaching affix, generally forming the outermost layer in these types of derivations . 34 Since many prepositional forms in English are also argued to involve a and categorial head, it is not surprising to also find th e s e forms involved in such derivations. 32 The y show that the exponents / ize/, / able/, / ive/, and / tion/ represent the discussed categor ical s have first been categorized by a categorical head represented by /Ø/. The FVI under each categorical head depends on feature of the head it attached to. 33 A llomorphy may occur between the categorizing hea ( Cf. Marantz 2001). 34 Perhaps it is possible to say something like forgivenessy in something such as He keeps giving all these forgivenessy type sermons .

PAGE 255

255 Little p_ may also be involved in the type of derivation described above, as can be seen in English by above and below taking the suffix / ness/ in (11) and (12 ). (11) a. aboveness and His highness refers to His highness of glory and attributes and greatness . 35 b. My bedroom is above my library. Does the ' aboveness ' (From the Will iam James Reader Vol 1 Page 155) (12) a. Only the Gaffle knows why the Gibble lives deep in the belowness . 36 b. I n the evangile given, something is surely said for belowness . Men are born to it. (James W. Fernandez Persuasions and Performa nces: The Play of Tropes in Culture 1986: 19). In (11) and (12 ) one must consider the forms aboveness and belowness to be composed of at least two categorizing heads: [[above p ]ness n ], [[below p ]ness n] . T he base forms [above] and [below] must be viewed as first being categorized as prepositions because neither base form is use d as a noun, verb, or adjective as can be seen with (13 ) . (13) In (13 ), the Root is first categorized by little p _ and then converted in to a noun by a little n_ head. This makes sense because not only is / ness / understood to be an outer layer affix, it is also an affix that has few selectional requirements. In addition to this , 35 From the blog: wahabi doctri ne of flawed aboveness/ 36 From Bedtime stories: innovation in the spirit of Lewis Carroll

PAGE 256

256 Chapter 4 discussed many examples where an English adposition al form underwent conversion to a noun or adjective. Likewise, there was clear evidence for categorial stacking in Chapter 6. As seen in Chapt er 6 , a form such as bezuiden south of is composed of a zuid that can be used as a noun . Moreover, it can be suffixed with / en/ and also used as a noun . From this form, it may be converted into a preposition by adding the pref ix /b(e) /, as seen with (14). 37 (14) a. b. This is interesting because unlike the English example the prepositional head is clearly selecting an already categorized head as evidenced by / en/. It is unclear whether this prepositional head is select ing a categorized by little n_ as in (14a) or if more structure is contained as in (14b). Either way , p_ is recategorizing an already categorized . Another possible piece of evidence for this type of derivational stacking was seen with forms such as n of in Mandarin in Chapter 3. This analysis though entirely depends on the identities of the suffixes /mian/tou . For sake of simplicity, it will presently be assumed that all three are variations of the same thing. 37 for further study.

PAGE 257

257 Thus the mos t common form will be used to model the possible derivations as seen in (15 ). (15) a. b. c. If is treated as a , then it may be possibl e to treat the derivation is as in (15 a) where two items merge before being categorized as postposition. It may also be that the forms are first categorized as a noun (as nán is also used as a noun) and then a preposition (15b) . If that is the c ase then the categorial heads are stacked and p_ is converting an already categorized into a preposition. Furthermore, if one is to analyze as the FVI representation of an n_ head (15c) then stacking is clearly occurring with Mandarin as well. Thus far we see that the little p_ head actually behaves in a similar fashion to the other more established categorizing heads with regards to stacking. It was shown to appear below a little n _ head , realized by / ness/. In this way the little p_ head woul d be deemed a low affix with selectional requirements that forbid it from attaching above most or all other categorizing heads in English. In Dutch, however, little p_ may be select nouns and verbs with the suffix / en/ and convert them into a prepositions . S ince selection al requirements are necessary to describe the behavior of other functional heads (Marantz 2001), they would likewise be expected with little p_.

PAGE 258

258 7 .4 .2 A Possible Difference This section will discuss a possible difference between p_ and th e other categorial heads with regards to the functional structure that each h ead is associated with. It is here proposed that p_ is commonly found with at least deictic features as discussed in Svenonius (2008) among others . The major lexical categories a re often distinguished from one another by the general types of features they co occur with or project for (i.e. Noun: Definiteness, Number, Class Gender, Case These features tend to co occur wit h each of the major categorial heads, providing an identifying context for s c commanded by them (Harley and Noyer 1999; Embick 2004). 38 Thus t hese features help to further i dentify the type of verbalized present ; however, the category p_ ostensibly does not have such features (16). (16.) Functional Projections Associated w ith the Categorial heads Example (16 ) shows the types of features that commonly co occur with the other major categorial heads, mar king p_ with question marks because this category has not yet been clearly associated with classes of functional features. Thus in comparison to v_ 38 I t is argued that categorizing he ads are fundamentally different than the features under nodes such as Root present. They arguably do not give the type of perspecti ve necessary for Root to be interpreted .

PAGE 259

259 and n_ , a p_ head might appear different as a proposed categorizer as it ostensibly does not have projection As demonstrated in (16), it has been thought the category adposition does not have a set of functional material similar to n_ , v _ and a_. However, if one takes the extended projections of the Category P (Koopm an 2000; den Dikken 2006 an d Svenonius 2008), as discussed in Chapter 1 and 4, postpositions in German, as discussed in Chapter 5, it is more than reasonable to propose that p_ projects or co occurs with type s of degree , directional, and deicti c features as first demonstrated by (17 ). (17.) A few feet right there below the shed. As can be seen with (17) prepositional phrases in English can co occur with degree words such as right and deictic markers such as there . These are argued to also be part of the extended projection of P in Svenonius (2008). Likewise in Chapter 5 it was seen that German postpositions agree with a verb's direction (18). (18.) hin aus awayfrom out her aus ou In Chapter 5, it was suggested that the postpositional markers hin/her were representations of p_ . This may be so but one could instead propose that these features be treated as features that select for p_s just as Tense and Aspect are features that select for types of v_. In German, it may then turn out that hin/her actually are only FVI

PAGE 260

260 representations of the little p_ shell and not p_ itself. The answer to this will be left for further research. 7.5 Conclu ding Remarks The purpose of this chapter has been to show that p_ is not unusual as a categorical head. It may not be universal but universality judgments depend upon absolute categorical definitions for which there is no consensus. The heads n_ , v _ and a_ are more likely to be universal if they are considered a class of feature combinations rather than a single undifferentiated feature. If this is the case, p_ is also an extremely common linguistic class (as it possibly includes the features involved in languages that use case marking systems). V iewed in this manner, it was seen that p_ behaves similarly to v_ with the licensing of thematic roles and assigning structural Case. Also the two heads possibly overlap in ability to bundle with the features [transitive] and [telic]. Furthermore, like n_ , v_ and a_ it takes part in additional based derivations and it appears to have a functional structure in analogy to n_ v_ and a_ . However, this structure is clearly different than n_ , v_ and a_ . Likewise its function differs semantically from that of a verb, noun and adjective as it is p rimarily used for orientation. Thus while the category p_ in every way looks like a categorical head, it also appear s to be a separate class from its closest relative, v_.

PAGE 261

261 CHAPTER 8 THE ORGIN OF LITTLE P_ 8 .1 Introduction In Chapters 5 and 6, s everal a d positions were proposed to consist of a functional categorical head (little p_) and a lexical node in order to account for data from English, Mandarin Chinese, German, Dutch and perhaps Persian. It was also seen that some adpositional items truly appear to be functional in these languages . These were argued to be FVI realizations of p_ . This chapter now explores the ontology of such direct representations. It examine s the properties of grammaticalization, factors contributing to grammaticalization and grammaticalization in accordance with the category adposition. This chapter investigates whet her FVI adpositional forms first began as lexical items and then later became realization of p_ . If so, the subsequent question is if some or all lexically derived p_ s are in the processes of grammaticalization or subject to grammaticalization effects more than other item s? Some diachronic studies suggest that many adpositions are derived from lexical items. Additionally, the grammaticalization cline of Lehmann (1985) can be taken to support the synchronic analysis given in Chapters 5, 6 and 7. Addit ionally, as discussed in Chapter 1, adpositional elements are sometimes argued to be semi lexical it ems. While it was noted that semi lexicalility is an inconsistently defined notion, the notion is nevertheless associated with grammaticalization, the proce ss by which lexical items become grammatical/functional ones. 39 Semi lexical forms can be diachronically viewed as fully lexical items that have 39 As discussed in Section 8.4, grammaticalization is also an unclearly defined notion encompassing several things that may or may not be best viewed as the result of the same process.

PAGE 262

262 become more functional (cf. Hagemeijer 2001). 40 In addition to this, diachronic observations have led many to con clude that once an item begins grammaticalization it continues grammaticalizing until it essentially disappears . This chapter thus intends to discuss these issues in accordance with how they relate to what has already been proposed about adpositions. Ques tions about whether semi lexical items are in fact lexical items becoming functional items will in general remain as the answer not only depends on a more consisten t definition of semi lexicality but also what counts as grammaticalization . I nstead of expli citly trying to answering this, this Chapter will show how observations about grammaticalization and the ontology of adpositions can support the proposal here that adpositions in many languages consist of a functional categorical head (little p_ ) and a lex ical node. 41 The data suggest that certain lexical items were reanalyzed as p_ , signifying that such items performed the function of p_ (i.e. were framed by p_ ) and then were relisted to become the direct representation for a type of p_ . In conclusion the possibility of captu ring grammaticalization within DM is discussed and explored. It is argued here that the DM model can capture an aspect of grammaticalization by explaining it as Contextual Features becoming Associated F eatures. With adpositions , this means going from being listed as a item that can be framed as an adposition to being associated with some grammatical feature or set of features found with p_ . 40 This work also describe s items going form a semi lexical status back to a lexical status. 41 nition would ostensibly mean that lexical. This distinction is therefore not helpful for this Model.

PAGE 263

263 This c hapter is organized as follows : Sec tion 8.2 introduces the idea that grammaticalization provides functional features w ith a phonological representation . Section 8.3 explains the concept of grammaticalization as both a framework and process. Section 8.4 discusses adpositions as a product of grammaticalization and uses English data to show that the claimed FVI adpositions were functional long before modern English and that the proposed LVI forms have never grammaticalized. Section 8.5 th en shows how a specific grammaticalization cline for adpositions partially mirrors the proposal given in Chapters 5, 6 and 7. Accordingly , some changes are examined with the conclusion that the different types of adpositions discussed can be explained as products of a diachronic process. Finally Section 8.6 shows how grammaticalization can be synchronically accounted for in the DM model by pr oposing that grammaticalization be viewed as the changing of Co ntextual Features to Associated Features. 8 .2 Old Language, Language Change and the Emergence of p_ As is with the ontology of any category, there is mystery regarding the origin of the cat egory P (adpositional items and perhaps case markers; cf. Asbury et al. 2006). Some linguists have contemplated the possibility that proto languages consisted of only nominal and verbal items (Heine and Kuteva 2002:394). 42 Many then if not a ll the functiona l elements been derived from a lexical source via grammaticalization . That is , [of grammaticalization] starts with nou ns and verbs, the argument goes, the earliest stages of language might have possessed these elements, but not auxiliaries, pronouns, 42 According to Hopper and Traugott (2003), Humbuldt (1825) proposed that only concrete ideas were expressed in earlier lan guage.

PAGE 264

264 2006:1). 43 If true, this would suggest that the FVI representations of little p_ arose via grammaticalization. A subsequent question is whether functional features such as p_ were suddenly realized (i.e. given a salient phonological exponent) as language evolved or if these functional items pre existed phonological realization. Given the model and per spective of hum an language discussed so far , it seems insightful to suggest that fun ctional items (features composing syntactic structure) pre existed any phonological evidence and that such evidence came about via grammaticalization. 44 This is just a suggestion, however, as there can be no evidence for this. Nevertheless, evidence from more recent history shows that adpositional elements are often derived from lexical items via grammaticalization. This wo uld also suggest that items historically have been available for categorization in such environments. As stated, one can propose that evidence for a class of grammatical p_ features comes from the grammaticalization processes while the ontology of the features themselves putatively will never be definitively known. 43 Newmeyer (2006), h owever , also cautions one from taking a dogmatic view based upon the aforementioned logic because tendencies are not always a trustworthy indicator of the nature of something over a long period of time or of what all Proto Lang uages looked like. The process could very 44 This tendency a ppears to be continually replaying as lexical items become functional items, some thus moving from LVI representations of nodes to FVI representations of p_ .

PAGE 265

265 8 .3 Adpositions as a Product of Grammaticalization The previo us section discussed the likeli hood that adpositional items originate via grammaticalization. In the literature this specific claim is often found as many historical linguists propose that at leas t some adpositional members originated via grammaticalization. content, such as prepositions and tense/aspect morphemes, can also be the product of (2006:1). Moreover, given the variation of the category adposition discussed in Chapter 1, it should not be surprising that the categorical source of adpositions (albeit lexical) also var ies . As seen in Chapter 6, verbs and preposit ions have several si milarities. Accordingly it make s sense that verbal elements would be a source for new prepositions. One example of a verbal element turning into a preposition is the verb /w / Twi becoming reanalyzed as the preposition /w / Campbell 1995: 63). Furthermore , while lly derived from verbs nouns by the process of grammaticalization (Hagege 2010:151). It is accordingly thought that Mandarin adpositional forms were derived from nouns and verbs , with prepositions coming from verbs and postpos itions coming from relational nouns ( cf. Hagege 2010; Djamouri and Paul 2009; Djamouri et al. 2012). One may also find claims that English prepositions have gone from adjunct adverbial like positions to prepositional ones. itions in Present Day . 45 Thus even with a very brief survey 45 s a garb age or catch all category. The makers of the etymological dictionaries are referring to many functional like items as adverbs . Items that appear to function as the Modern English intensifiers too and so receive such a label as well as temporal locative for ms like afterwards .

PAGE 266

266 of languages, it seems as if several different types of lexical items have been the source for adpositional elements. From this short overview, a dpositional elements are claimed to be derived from verbal, nominal, and adverbial categories. What all of these proposals share is that adpositional forms are derived from somet hing that would be considered lexical via grammaticalization. However, to prop erly defend such a proposal , one needs to establish a clear view of what grammaticalization entails because the term encompasses a number of different notions in the literature. 8 .4 Grammaticalization Grammaticalization, also called grammaticisation, is d observable (van Gelderen 2006: 1) 46 and regular, systematic process ( Heine and Kuteva 2002 ; Haspelmath 2004). The term g rammaticalization , as it is more or less conceived of today , was first used by Meillet in 1912 to describe lexical items becoming grammatical items. However, i n addition to referring to a possible specific process th at causes such a change , grammaticalization can also be broadly viewed as the framework that encompasses all linguistic phenomena associated with l anguage change whereby linguistic items obtain grammatical function or become functional (Smith 1996; Lass 1997; Haspelmath1999; Hopper and Traugott 2003; Miller 2010b) and whereby grammaticalized linguistic items often lose autonomy (Lehmann 1985; Hopper and Traugott 2003). Given this, the grammaticalization framework may actually include several different processes that appear to derive a similar effect or similar 46 Grammaticalization has been observed in every language described over time.

PAGE 267

267 effects. It is proposed here that the term grammaticalization is associated with a t least th ree effects: semantic/functional effect (lexical meaning >functional meaning), a lexical category effect (lexical category > functional category > new roles), and a distributional/phonological effect (free morpheme s > bound morphemes); and four processes: analogy, reanalysis, exaptation, and grammation. With this understanding it can be concluded that the implications of grammaticalization claims are limited by how the term is being used and the theoretical assumptions different authors bring. Since s evera l linguists have claimed that adpositions are the result of grammaticalization , t his issue will be discussed to better evaluate such claims . Since grammaticalization theory does not endorse a specific synchronic model, observations flavored by different pe rspective must be translated into one (in this case DM). In an attempt to do this, it is proposed that grammaticalization as a process is change of status from a lexical to functional item via reanalysis or grammation. 8 .4 .1 Grammaticalization Understood via its Effects It has been stated that grammaticalization as broadly understood includes different processes and different effects. Moreover, the description of grammaticalization effects often implies different theo retical positions. These effects and the theoretical positions are the topic of discussion here. The effect of grammaticalization whereby an item goes from lexical to grammatical use is described in a number of ways. From a more traditional lexicalist per spective it can be described as [ing] (Smit h 1996: 142). I t also can be described as process whereby a lexical item obtains a functional use (Hopper and Traugott 2003:1), or as ich lexical

PAGE 268

268 ( Miller 2010b : 67) . These latter two descriptions do not entail category change for the original item . Instead they only describe a new use or a more restricted use for the origina l item. This description would also account for the creation of lexical and functional homonyms. In addition to lexical to functional change, grammaticalization is also often associated phonetic reduction and dependency (Lehmann 1985; Hopper and Traugott 2003; Newmeyer 2006). That is g rammaticalization can be thought of as the historical Newmeyer 2006:1 ), or as when an item becomes less autonomous (Lehmann 1985). This effect, however , can be thought of as a corollary of grammaticalization because meaning/category changes correspond with phonetic reduction. Moreover, dependency changes could be the result of a process that is not causing an item to become functional or more functional. Nevertheless, this effect is heavily associated with grammaticalization as a framework. Meaning, category and dependency changes are all accounted for with Heine and gr ammaticalization steps (7). (7) Grammaticalization Properties (Heine and K uteva 2002:15). a. Extension: the rise of novel grammatical meani ngs when linguistic expressions are extended to new contexts b. Desemanticization/Semantic Bleaching: loss (or generalization) in meaning or content. c. Decategorialization: l oss in morphosyntactic properties characteristic of lexical or other less grammaticalized forms d. Erosion: loss of phonetic substance 47 47 Step (4) will be here considered peripheral to grammaticalization.

PAGE 269

269 The steps given with (7) capture the fact that grammaticalization is often associated with meaning changes ( 7a,b), categorical changes (7c) and dep endency/phonetic changes (7d). These facts are also often described as sequence of small changes across a (Miller 2010b:71) . A common general cline, given in Hopper and Traugott ( 1993: 7; 2 003:4 7) and restated in Miller (2010b:70) 48 , is given with (8). (8) lexical word > functional (grammatical) word > clitic > inflectional affix (> zero) Example (8) shows that at least the category and autonomy of an item is changed as it is grammaticalized ( i.e. lexical items are becoming grammatical items and words are becoming affixes). 49 Additionally the cline demonstrates that these changes are ordered from least to most grammatical ized . Thus while something may seemingly skip a step , it is not expected to revert back to an earlier stage. From the steps of Heine and Kuteva (2002) and the cline of Hopper and Traugott (1993, 2003 ) , it can be seen that the term grammaticalization commonly covers a semantic change, a categorical change and dist ributional/phonol ogical change. Moreover, these descriptions differ slightly in what a lexical to functional move entails (i.e. perspectives appear different regarding the details of lexical to functional change ) . Nevertheless , these changes are understood to proceed in a directional manner, from more to less lexical. 48 The inflectional affix stage has since been changed to just affix. 49 A cline suc h as (8) may also be viewed in terms of a grammaticalization scale or continuum between a lexical and grammaticalized form with the intermediate nodes note necessarily holding any special theoretical status (cf. Lehmann 1985).

PAGE 270

270 8 .4 .2 Grammaticalization and Direction As disc ussed in the previous section, g rammaticalization is often depicted over a cline. This not only says that grammaticalization as a process is directional (lexical > functional) but also that it is unidirectional (*functional > Lexical), meaning that the process is irreversible or is largely irreversible (Haspelmath 1999, 2004 ; Heine and Kuteva 2002 ) . In other words , if an item become semantica lly bleached, it tend s to stay that way. Haspelmath (1999; 2004) for one strongly argues that grammaticalization is overwhelmingly a unidirectional process. Likewise Newmeyer (2006) agrees with Haspelmath that u nidirectionality is the norm and that for items to move in the opp osite direction is quite rare. Given this , unidirectionality constrai nts on possible language change Haspelmath 1999:2). Likewise, Miller (2010) mentions that unidirectionality is one of the few or only principles that the g rammaticalization framework really espouses, suggesting that this principle forms the backbone of the framework. Haspelmath (2004:23) , however, also admits that (depending on the definition of grammaticalization) unidirectionality is not without exception . Possible examples include the fact that the clitic [ ] in English came from an affix (breaking the cline: *affix > clitic) and the word ism came from words containing that suffix (breaking the cline: *suffix > word). Such cases, however, are generally c ontroversial. Miller for one of the arguments for degrammaticalization (or grammaticalization reversal) lack cogency For instance both given examples for degrammaticalization involve uncommon distribution al changes that are no t indicative of a functional to lexical change (i.e. ism does not appear to be a ) . Moreover, since g rammaticalization

PAGE 271

271 theory is not an agreed upon theory (Haspelmath 2004:23), as has been discussed in the previous sections , these may be considered cases of degrammaticalization from a distributional perspective . However, if gramma tical ization and its unidirectional constraint is confined to describing lexical to functional categorical changes rather than distributional changes, these examples would not be counterexamples . Moreover, Haspelmath argues that even if some supposed count erexamples are true cases of degrammaticalization, the asymmetry requires an (1999: 6). It should be clear that grammaticalization as a framework is heavily associated with the idea of unidirectionality. A s a pr ocess that turns lexical items into functional items, it is largely or completely without exception. Counterexamples seem to only break the distributional cline, not the lexical to functional cline . I t is thus claimed here that only a lexical to functional s tatus move really indicate s grammaticalization. When this happens, however, questions regarding whether the nature of the item completely changes, whether the item splits, forming a functional copy, or whether the item becomes semi lexical or semi functi onal remain . W hile there appears to be a correlation between the types of changes discussed, it is unclear if these effects can or should be always attributed to the same cause. 50 8 .4 .3 Grammaticalization Causes It remains debatable though whether both dis tributional/phonological and categorical/meaning changes can accurately be attributed to the same cause and thus these may not be reflective of the same process (Miller 2010b). That is , the move from 50 Kiparsky (2009) proposes a theory where analogy is a part of grammaticalization, contrary to the Meillet defines it in (1912[1958]).

PAGE 272

272 free to bound is not necessarily the same as the move fr om verb to auxiliary verb. T he prosodic structure of a language could have more to say about the status of a morpheme along the spectrum of bound and free than semantic bleaching. For the purposes of this work, such changes will be viewed as possible indic ators of grammaticalization, and thus will not be viewed as the process itself. Otherwise, grammaticalization cannot reasonably be defined as a single process corresponding to absolute effect s . 51 Grammaticalization does not have to change boundedness and mo vement from a clitic to a n affix does not indicate further functionalization (if that definition needs to be held). Moreover , while the causes should render the same/similar effects, evidence for the effect alone does not necessarily indicate the same caus e. Given this, the label grammaticalization and language change can be further broken down into at least four different processes: analogy, reanal ysis, exaptation and grammation. Each of these processes proposes a different cause for what may be considered grammaticalization. Analogy : The process of analogy is simply changing one pattern to mimic another pattern . A simple example is when patterns such as sing, sang, sung, give rise to bring, brang, brung or when help , holp , and holpen became help, helped , and helped . Analogy is certainly a part of language change , but it is unclear if it plays a role in the process of grammaticalization. Analogical based changes are included in some discussions of grammaticalization. A uthors such as Kiparsky (2009) and Fisc her (2008) include types of analogy within the framework of grammaticalization. The argument is that analogical 51 It may also be possible that some writers in the field do not wish to view grammaticalization as a single process.

PAGE 273

273 change can set up grammatical change. each step in a process of ind ependently taken [and] not caused by the grammaticalization process itself but it is the result of a speaker s analogical thinking in combination with the content o his repla cement is based on similarities, both in form and function, between the old structure and other structures that exist for the speaker seems for Fischer (2008) that analogy plays a role along the grammaticalization cline. In this way, analogical change at least could fall within the framework o f grammaticalization. However, many historical linguists consider it a separate source for language change. Most analogical changes, however, do not appear to result in lexical material becoming functional material. As such, a nalogical change such as this can be considered a separate source for language change. Meillet is quoted in Traugott saying In other words, w ith analogy no thing in the grammatical system has to change. 52 Instead a different inflectional pattern is being applied. Accordingly, Meillet separates analogy form grammaticalization because creates new forms and introduces categories which ha d no linguistic expression. It changes the system as a whole" (2004:5). This of course should not be taken to mean that grammaticalization is always radical, but that it is more innovative than analogy. The cause for analogy is different than grammaticaliz ation 52 This looks like differential li sting in DM.

PAGE 274

274 defined by categorical change . 53 In addition to this aspect of change, one can further categorize grammatical like change with the term exaptation. Exaptation: Grammaticalization theory may also be associated/contrasted with exaptation , 54 a term desc ribing the type of inn ovation in language change seen when a form loses its original function but is not lost. I nstead o f being lost from the language , the form remains by gaining a new function (Traugott 2004). This like analogy does not mandate a categor ical change but rather the change of phonological exponents for functional features. In the case of exaptation, however, the change is more innovative because the change is not based upon an established pattern. E involve[ing ] the assignment of new morphosyntactic functions to elements which are already centrally part of the grammar, and typically part of the paradigmatic core of the morphological 438). In such cases, it would appear that those items wher e becoming even more functional (if one views the functional class on a cline) because by taking on new functional roles, the item must become semantically vacuous as to be underspecified for so many functional positions. In this way Exaptation may look mo re like grammaticalization than analogy but ostensibly the process does not take lexical items and reanal yze them as functional items. Reanalysis: In contrast to analogy and exaptation, the process of reanalysis involves categorical change. According to H arris and Campbell , r mechanism which changes the underlying structure of a syntactic pattern (1995: 61) . 53 Another reason to exclude analogy from grammaticalization is that analogy is not unidirectional (Miller 2010b:74). 54 The exact meaning of this term seems inconsistent, sometimes overlapping with reanalysis.

PAGE 275

275 information regarding at least (i) constituency, (ii ) hierarchical structu re, [and] (iii) category labels Campbell 1995: 61). This contrasts with analogy, which as discussed, does not change the underlying structure but rather alters the surface form in imitation of another form pattern. A relevant example of catego ry reanalysis is found with some English prepositions becoming complementizer s from Middle to Modern English as demonstrated by (9a) in contrast to (9b). (9) Constituency Judgment ( from Harris and Campbell 1995:62) a. [It is bet for me] [to sleen my sel f than ben defouled thus] b. [It is better] [[for me to slay myself ] than to be violated thus] It is better for me to slay In (9 a) the form for is a preposition and a part of the senten tial subject while in (9 b) it has been reanalyzed as a complementizer according to Harris and Campbell (1995:62). 55 In other words, t he same form has moved into a new grammatical, syntactic context. Its use has been Extended as described in Heine and Kuteva (2002) and given in (7) . S uch reanalyzes would also be associated with bleaching ( for arguably lo s es its benefactive meaning). Moreover, in such cases there is no claim that this occurred in a single step. T across a cline (Traugott 2008: 225) would suggest otherwise . Finally there is another distinction that should be considered when understanding reanalysis and gram maticalization and that is grammation. 55 However, this change has not r esulted in phonetic reduction. I f we model the preposition as already being an FVI then one is not moving from a lexical category to a functional one. The semantic bleaching would have already happened if the preposition were modeled as an FVI.

PAGE 276

276 Grammation : like the idea of reanalysis is the notion of Grammation . According to is grammation, a change from lexical to grammatical content, or from hybrid/grammatical to other grammatical content, generally via reanalysis of a formative to be directly merged in a functional While reanalysis was also seen to involve functional to functional re categorization (i.e. something simi lar to exaptation), grammation focuses on the reanalysis of lexical or semi lexical items in functional domains. It is thus this aspect of grammaticalization that seems most relevant for a discussion of adpositions and grammaticalization. The section has examined four different causes for what might be called grammaticalization. While included in some discussion of grammaticalization, analogy was deemed to be a separate source for language change because it does not clearly result in a categorical change. Similarly, if grammaticalization has to involve a categorical change, exaptation should be considered a separate process. Moreover, while reanalysis and grammation are all similar in that they describe an item obtaining new uses in a functional domain. Rea nalysis as a process may involve lexical to functional change or function al to functional change as was seen with example (9). However, grammation specifically involves lexical items being merged in functional adpositional forms may further change via exaptation a nd reanalysis, the creation of new adpositional exponents is better understood as a process like Grammation. 8 .5 Adpositions, Grammaticalization, and a General Model As suggested in the previous sectio ns, it is best to view grammaticalization as a lexical to functional change that may or may not be accompanied by phonetic reduction.

PAGE 277

277 Moreover, those processes described as reanalysis and grammation account for such a status change and suggest that these c hanges come from a lexical item becoming heavily associated with a particular grammatical domain, being reanalyzed as an indicator of that domain. This section will examine specific data from English to see if a difference is found between the proposed FVI and LVI adpositional forms . 8 .5 .1 Historical Analysis of Proposed Adpositional FVI Prefixes in English It was argued in Chapter 5 that forms such as beside and aside consisted of a prefix that represented a type of p_ head. These prefix es were ostensibl y not always such as they appear to have undergone grammaticalization in terms of bleaching, reanaly sis, and distribution al change . This section will examine the historical source of these forms as they can be traced back to at least Gothic, 2nd century C. E, Old English, ~700 C.E, and Midd le English, ~1300 C.E, Table (8 1) . Table 8 1. Etymological source and meaning of English adpositional p refixes 56 Eng. Gothic Old Eng . Meaning/F unction : M iddle Eng. Meaning/ Function : a ana a n o adverbs and adjectives afire a hunting b eing used in prepositional like phrases: alive, afloat, asleep 57 be 58 bi be bi Forms: bespatter As seen in Table ( 8 1 ) , the forms a and be did not start off as affixes. In Gothic a is a free . Be on the other hand appears a s a both a prefix While Gothic is the oldest attested Germanic 56 The sou rce for information found in Tables 8 1, 8 2, 8 3, 8 4, 8 5 and 8 6 is Barnhart (1995), Harper (2001), and Slocum and Krause (2013) 57 forms listed in Bar nhart. Moreover, these forms appear to be more as results/statives than prepositions. 58 Be 2001)

PAGE 278

278 language, it is not the mother of English. Thus in other Germanic language, including English, be survived as a free morpheme, suggesting that it started off as one. I t is thus clear that both of these forms underwent dependency changes, suggesting grammaticalization. It is very likely that these forms had ob tained a grammatical identity as there is no indication of lexical use. In addition to this it should be noted that these forms had a range of meanings (/a be / English, as an affix. With a D M perspective, such variation in meaning can be taken as evidence that they were underspecified FVIs. In general, it may be best to argue that they have no meaning at all in present English (only the function of representing functional features) . One could take this to mean that sem antic bleaching occurred in Old English for /bi/ and /an/. In fact , further evidence for this is seen when evaluating the source of combined affix and adpositional forms. 8 .5 .2 Historical Analysis of Adpositional Affixes and Items At the start of the section, the historical source of the proposed adpositional prefixes /a / and /be / were briefly discussed. While it was shown that were derive d from free forms, it could only be suggested that they began with a lexical identity. Nevertheless, a historical analysis of these prefixes joining wi th other lexical items further suggests that the prefixes /a / and /be / underwent gram maticalization . It can be seen that the form / be / is u sed fo r a range of functions, suggesting that it had a similar/connected meaning is expected with i contrast is seen with Table (8 2) and (8 3 ).

PAGE 279

279 Table 8 2. Etymologica l source and meaning of English A derived a dpositions Eng. Time Source: Old English Meaning: Function Middle English Meaning: Function aboard 15c O.Fr. à bord, from à "on" + bord "board," within shippes borde . about 10c. "from on "on" + be "by" + utan "outside," onbutan "on the outside of," Forced out ymbe, the old English word for about. above 896 *ufan: reduction of abufan across 1325 a on preposition 1590s again 1031 agan 16c differentiated from against/ in the south only an adverb, but North dialects kept it as a preposition a gainst 1160 genitive o o the t arrives possibly from its superlative use. a long 887 andlang "entire, continuous; a mid 725 considered an adverb on middan Late O.E used as a preposition a mong 899 preposition before 1121 a round 11c Rare before 1600 aside 59 1300s side a top 1650s Not fully established as one word until late 1800s It should be noticed fro m Table (8 1 ) that the source for the proposed a prefix is not homogenous, except for it always involves a prepositional like element. Moreover th e 59 Asidely, asideward

PAGE 280

280 Root . Thus it may be better to analyze them as single forms. However, with those forms that are clearly partitionable, it looks as if several Place related adpositions leveled to form a . Thi s same pattern can be seen with / be / with Table 8 3 . Table 8 3. Etymological source and m eaning of Engl ish Be derived a dpositions Eng. Time Source: Old English Meaning: Function Middle English Meaning: Function before 725 beforan of, in the presence of, in former times," behind 725 Behindan "behind, after," prepositional sense started here. b elow 1325 be "by, about" + logh, lou, lowe "low" variant of earli er a low, which is English but used adverbally) analogous to a fore and be fore . 1500s used a preposition. beneath 854 beonethan (adverb) (prep) b eside 725 be sidan (two words) "by t be sidan bisidan between 750 b betweonum Bitwene: (prep) and (adverb) beyond 885 b Begeondan: used as a preposition during this time. Unlike the forms in Table (8 3), Table (8 4 ) show s that be derived prepositions occurred from a regular process of prefixing a variation of by to a locational word. Nonetheless, as already mentioned , this prefix obtained other uses. In this way it seems to have lost its own meaning. As this goes it picks up the meaning of the terminal node it is inserted Be can also be privative (cf. behead ), causative, or have just about any sense

PAGE 281

281 I n Old English it appears that /a / and /be / were already grammaticalized because of their meaning and function . From this point we can explain what happened as exaptation. It is interesting to view these forms in comparison to the other proposed FVI prepositional forms in English because while similar meaning changes are seen, the FVIs that ha ve maintained a word status appear to have only more recently grammaticalized. 8 .5 .3 Historical An alysis of Proposed English FVI W ords Chapter 3 proposed that several adpositional forms in English are functional. These forms are traced back as far as can be reasonably reconstructed, Proto Indo European, and Proto Germanic 1st c. C.E, with attested forms coming from Gothic, 2nd century C.E, Old English, ~700 C.E, and Mid dle English, ~1300 C.E, Table (8 4 ). Table 8 4. Etymological source and m eaning of Eng lish f unctional prepositions Eng. PIE Meaning: Proto Germanic Gothic Meaning: (Category) Old English Meaning: (Category) 1. to *do ([to:] Old Saxon) direction of, purpose, furthe r m 2.from *pr *fr forward; by, since, on result, since, forth, forward movement, 3. at *ad *at 4. mid /with *medhyo *medjaz together with, with, together with, 5. with * wi tero , "more apart," * withro opposite; over against, by, near, to, toward, 6. of *apo *af of/ æf

PAGE 282

282 Table ( 8 4 ) gives the historical source of those items that are here argued to be FVIs in Chapter 1 in Modern English. I tems (1 6), al l can be traced back to PIE. Items 4 and 5 are included because item 4, which is lexical in PIE becomes with in Modern English while the original with form only survives in compound forms such as withstand of withdraw . If one follows the directionality h yp othesis, the forms in Table (8 4 ) started off as LVIs. It seems reasonable to conclude that these forms have undergone grammaticalization via semantic bleaching and reanalysis from an LVI to an FVI at some point in their history. Interestingly the forms fr om Table (8 4) appear to be prime targets for many of the Germanic languages (e.g. the ad prefix in Latin comes from the same source as at ). 60 It can also be seen that most of these forms have undergone phonetic reduction (a common by product of grammaticalization as discussed). Moreover, one can see that these forms can be used as the representation of several different functional features. It would thus appear that th ey had already lost an identifiable mean ing in Gothic and Old English. For instance Item 1, at, apparently could be used in several domains, represent ing meanings associated with at least [goal] , [locational], [associative], and [ablative] type feature s . 61 Moreover, t 60 T he function of these words in PIE i s unknown. That is while the comparative method can within reason reconstruct phonological forms and perhaps prototypical meanings, reconstructing syntax in order to associate certain forms with grammatical function seems more than a few steps beyond our ability. 61 Compare with the form, as , which can be used in a variety of linking positions, including a prepositional one in modern English. This meaning differential appears to be a good way to indicate if something is an underspecified FVI.

PAGE 283

283 to at ", while to has been lost and at very much behaving like a n FVI. Likewise in G othic, it can be seen that at least the forms du , fram , and miþ map to several different grammatical type relations. They therefore do not appear to have a clear mea ning of their own and can be taken to rather represent whatever grammatical feature was pre sent at the terminal node . Following this logic if one of these forms were lost, another FVI would be inserted in its place through a process like exaptation. 62 Furthermore, as can be seen in Column 5, the final column, the reported uses for these forms in Old English is largely prepositional (compare with historical LVI uses reported in Section 8.3 ). It may have also been noticed that in Column 5, Forms 2 and 3 are both listed as having prepositional and adverbial uses. T his adverbial listing , however, c ould merely mean that these forms were not necessary as case assigners in Old English. It does not mean that the forms should be analyzed as LVIs at this time. This could also be taken as evidence for the proposal that to was an LVI at one point and at thi s point it was suffixed with ward, as briefly discussed in Chapter 5 . This form was later reanalyzed as a single chunk corresponding to a type of p_ head. Given this we can speculate that before/during Old English , Form 3 was still an LVI that could be cat egorized by p_. It may also be that by then it was already two FVIs (to and toward) and during this time a case of exaptation occurred where morpho syntactic features moved from one FVI to another ([case] becomes marked by to ) . 62 In some contexts in English perhaps the dual use has been maintained such as in The bull ran at the scared man .

PAGE 284

284 This section discussed some of the historical evidence to support the assertion that several of the proposed FVI adpositional forms in English underwent grammaticalization a long time ago. However, it also suggests that these forms had a lexical past, perhaps explaining the creation of the word toward as ward as a suffix would have attached to a lexical item. 8 .5 .4 Historical Analysis of Proposed English Adpositional LVIs In contrast to those forms argued to be FVIs preposition, a historical analysis of the English LVI prepositional forms shows several difference s supporting the notion that these forms come from a lexical source and that they have not yet undergone grammaticalization as defined as: lexical functional change , Table s (8 5) and (8 6) . Table 8 5. Etymological s ource an d meaning of English lexical a dpositions Eng. P I E Meaning: ( Category ) Proto Gmc. Meaning: ( Category ) Gothic Meaning: (Category) Old Eng. Meaning: Category *ambhi be bi (prep/adv) du 2. in *in in (prep); inne (noun) at, among; about, during; within, 3. out *ud [u:t] [u:t] (adv); utian outside; to expel 4. on 63 *ana over, to,upon a an on (prep/adv) 63 an the word had a wider function than it does in modern English and took rnhart (1995:521).

PAGE 285

285 Table 8 6 . Etymological s ource and meaning of English lexical a dpositions 2 Eng. P I E Meaning: ( Category ) Proto Gmc. Meaning: ( Category ) Gothic Meaning: (Category) Old Eng. Meaning: Category 5. off 64 *apo *a f Of offe 6. up iup u:p, uppe (adv): uppian (verb) 7. down (a development of English) ofdune (adverb) 8. over *uper *uberi ofer (prep/adv): uffera (adjective) upon, in, across, 9. under *ndhero lower *under undar Prep/adverb Among between;subordin ate 10.through *tere þurh 11. back *bakam bæc bæc (noun) T he f orms in Ta ble (8 5) and Table (8 6) can all be traced back to either Proto Germanic or Proto Indo European. These forms can be contrasted with the FVI forms because t he meanings of these form s are more consistent , despite their different lexical usage. An other inter esting observation is the greater amount of lexical, categorical flexibility shown with these fo rms than th e FVI forms shown with Tables (8 2 ) , (8 3) and (8 4 ) in the previous section s . In Old English the form in d inn. Likewise a type of adjectiv e was derived from the form over and verbs from up and out . Given that grammaticalization is a slow process, if it is operating on these forms, they s (8 1) and ( 8 4 ) . That is , 64 Originally the same word as of .

PAGE 286

286 these forms show more of a lexical distribution than both the bound FVI forms /a / and /be / and the free FVI forms ( of, at, from , etc.). It can thus be taken that the historical data supports the more general claims of this work as well as the idea that adpositions arise via grammaticalization. 8.6 Modeling Adpositions with Grammaticalization As seen in the previous sections, the idea that adpositions are grammaticalized forms is a common position. Moreover, this position is generally suppo rted by the data given in section 8.4 . However, in order to account for the different details that emerge from the data presented , mo re specific models are needed. These are discussed in this section. 8.6.1 Adpositions and G rammaticaliza tion S tages A mode l for the grammaticalization of adpositional forms is given in Lehmann (1985). According to Lehmann (1985), the grammaticalization cline for adpositional forms is as follows: (10). 65 (10) Relational Noun + Adposition or Case Affix + NP + Adposition or Case affi x > Secondary Adposition + NP + Adposition or Case affix > Primary Adposition + NP +/ Case affix > Agglutinative Case Affix + NP or N > Fusional Case Affix + N Crucially what (10) demonstrates is that relational nouns often get reanalyzed as a dpositions and then these adpositions appear to sometimes get reanalyzed as case affixes. 66 Moreover, interestingly the ste p where a Secondary Adposition becomes a Primary adpositions appears to capture the change from Old English to Modern English. 65 As discussed in Lehmann, the cline is a scale, meaning that other points could be identified. 66 This cline is more controversial as it is also thought that adpositional systems can develop from cas e marking systems (i.e. Latin case system > prepositions in the Romance languages; Germanic case system > prepositions in Germanic language, especially English).

PAGE 287

287 In Old English prepositions, postpositions and case markers were all used to spatially relate a DP. In Old English adpositional items co occurred with case markers, optionally appearing as pre or post positions ( 1 1) . (11) Old English Prepositional Use ( Smith 1996:1 54) a. cwoe oe m mannum He spoke the men to cwoe oem mannum He spoke to the men 154). However, while Smith is familiar with many linguistic concepts, it is hard to know what he means by adverb . T he point has been made tha t it may be best to think of as having a similar meaning to present day too or also (this makes sense: too ) . In Old English to) except where the adverb retained its stre ss (tired and hungry too); there it came to be written with can be omitt ed in both environments and as a postposition occurred for stylistic reasons such those found in poetry. I n fact , it was very common to have no prep osition in Old English, especially in cases with dative indirect objects. 67 Nevertheless, when the case marking system eroded, these elements became essential. Smith says a lexical function, has in Present Day English 67 Many thanks to Dr. Gary Miller, Dr. Sara Pons Sanz, and Dr. Jules Gliesche for grammaticalit y judgments of the cited example and for further information regarding it.

PAGE 288

288 A t some poi nt these prepositional items were reanalyzed as both the case assigner and preposition , removing the need for the case affix . In addition to captu ring co o ccurring functional items in Old English (prepositions and case markers) described in this work regarding how little p_ is realized (i.e. some s were argued to correspond with a salient p_ , others a null representation, and the FVI adpositions were argued to directly represent p_ ). 8 .6 .2 Revisions to the Adpositional C line (1985) cline for adpositions is not only helpful for descri bing prepositional items in Old English, it also roughly captures some of the synchronic variation discussed in this work. However, it fails to note when grammaticalization (defined as lexical to functional item) occurs. This section discusses some possibl e additions and revisions to the adpositional cline so far discussed. While the cline given in Section 8.5 (1 0 ), clearly involves a item eventually becoming a functional item, it also captures the fact that when items get positioned in an adpositional domain, there is often visible morphological complexity. The observation here is that these changes are seen with the synchr onic grammar as well. This can be seen more clearly with the LVI top in (12 ). (12) a. John is on the top of the mountain b. John is on top of the mountain c. John is atop the mountain d. ???John is top the mountain (not yet attested) In (12a) the LVI top is used as a relational noun in a full DP licensed by on . O f can be argued to represent c ase, as in Svenonius (2008), for the subsequent DP. Furthermore,

PAGE 289

289 when top is used in (12b) (recall the Axial Part position proposed by Svenonius 2006), it still functions as a Ground and needs of to license the subsequent DP. From a traditional perspective, the part of speech seems to have taken on more of a grammatical role , so this position can be thought of being somewhere between a relation al noun and secondary adposition. Moreover, when top is categorized as a preposition, as in (12 c) , of is not needed as the preposition atop can now licens e the following DP. I n this position , it is acting as a secondary adposition according the Lehman n (1985) description . Moreover, it does not seem impossible for top to eventually look more like a primary adposition as in (12d), but one needs a finer distinction than the cline given by Lehmann. According to Lehmann , the distinction between primary and secondary adpositions an elementary objective or a grammatical meaning and is morphologically simple, such as of , in 68 Given that in appears to also be a item in English because of its distribution and meaning, it seems better to include another possible intermedia te step, secondary preposition with salient FVI> secondary preposition with null FVI. It is thus possible that in is more likely to join the FVI list than low or top , but if top gets used as in (12 d) , it will be more like in . To account for this possibility, a revision to the cline is offer ed with the proposal that the different exponents for little p_ described in this work reflect stages of grammaticalization along a cl ine . 68 An analysis of during as a possible adposition is not taken here. It is not suggested that ing is an FVI for p_ .

PAGE 290

290 Table 8 7 . Revision of adpositional c line Category Title Consists of Co occur with Lexical/Functional Relational N oun item with spatial/temporal meaning Adposition or Case Affix +DP (+ Adposition or Case affix) Lexical Axial Part item with spatial/temporal meaning Adposition or Case Affix +DP (+ Adposition or Case affix) Lexical Secondary Adposition type 1 item with spatial/temporal meaning Phonological Realization of p_ + DP (+ Adposition or Case affix) Lexical Secondary Adposition type 2 item with spatial/temporal meaning Null Realization of p_ + DP (+ Adposition or Case affix) Lexical Primary Adposition Morpho syntactic features DP (+/ Case affix) Functional Agglutinative Case Affix Morpho syntactic features DP or N Functional Fusional Case Affix Morpho syntactic features + N Functional Table (8 7) captures the different types of adpositions seen in present day English. Crucially, moreover, it indicates where grammaticalization occurs. (i.e. the move from Secondary to Primary adposition). After this change, it is predicted that adpositional forms will not be used in lexical doma ins. 8 .7 Grammaticalization and DM Often diachronic process[es su ch as] relevant for their models because base their grammar on or deduct it from the synchronic outpu t t 2003:448). I t is thus not surprising that grammaticalization is not addressed in Distributed Morphology. I t is , however, proposed here that diachronic and synchronic

PAGE 291

291 models should not be exclusive and that the DM model can in part captur lexical to functional categorical change . I t is proposed that the DM model c an aid in our understanding of g rammaticalization by showing that for the individual speaker g rammaticalization occurs when a lexical item s use is restricted . A lexic al item with a restricted use is more likely to be reanalyzed as a grammatical item . It is proposed that grammaticalization is first captured as the limiting of available insertion contexts by means of Contextual F eatures followed by reanalyzing a contextu al feature as an associative feature. 8 .7 .1 Contextual to Associated Features With a model such as DM, an item (LVI or FVI) cannot be both functional and lexical at the same time as discussed in Chapter 2 and Chapter 5 . Nevertheless, a s discussed in Chap ters 2, 3 and 4 every lexical item is contextually associated with certain functional heads (i.e. n_, v, a _, and now p_ ) because these lexical items must be categorized by these functional heads . In Chapter 3 it was shown that many LVIs have distributional gaps and that at least some of these g aps (i.e. arbitrary ones) are best explained with Contextual F be thought to be unspecified for a context). Moreover, i environment is contextually narrow, this narrow context is proposed to be the trigger for reanalysis. Accordingly, some item s labeled as semi lexical items can be view ed as LVI s that have C o ntextual F eatures that limit their insertion to a few specific environment s. Ove r time this context may be further narrowed to only one context, causing the LVI to

PAGE 292

292 appear as the exponent of a functional head. 69 In DM this can be modeled as a feature changing from a {C on textual} to [A ssociated ] F eature, as seen with Figure (8 1). Figure 8 1. Cont extual and Associative Feature s wap . Figure (8 1 ) demonstrates the theoretical grammaticalization of the form ana from its Proto Germanic form to its modern attestation in English . It shows a bundle of C ontextual F eatures becoming a bundle of Associated Features. Once the LVI has Associated F eatures, it ceases to be an LVI and becomes an FVI. At this point, there is no need to keep an additional set o f duplicate Contextual F eatures so they can be deleted. At the le vel of the individual speaker , this specific environment is argued to be the catalysis for reanalysis or grammation, as such an environment has already been described as the trigger for some types of reanalysis. Moreover, t he change modeled with Figure (8 1 ) may either spread or be rejected by other speakers. If rejected, grammaticalization has not occurred at the E language level. If this occurs, it is conceivable that at the I language level, the new Associated F eature ( s ) may reve rt back to Contextual F ea ture(s) from the pressure at the E language level . 69 les of the Subset Principle. The Revised Subset Principle of De Belder (2011) over generates and is not independently descriptive of better way. Discussed fu rther in a separate Chapter.

PAGE 293

293 8.7.2 The Proposed M odel and Unidirectionality The issue and question here is if l exical material that is reanalyzed as an exponent of n_ , v _ , a _, or p_ (as demonstrated in Figure 8 1) can go back to LVI hood and reconnect with encyclopedic informa tion? It is proposed here that such change is extremely unlikely and that this causes the asymmetry seen with grammaticalization as defined by the changing of the categorical identity of a VI . There is no consis tent or compelling reason for FVIs , which lack an inherent connection with Encyclopedic information , to be chosen as new items because assumingly once an item becomes an FVI it is no longer listed in the Encyclopedia . 70 More discussion and research is needed to answer this topic, but it is thought that the proposal of the previous section offers a starting point for addressing unidirectionality in DM and explaining how and why lexical items become the exponent for little p_ . 8.8 Conclu ding Remarks Thi s chapter discussed the idea that the realization of functional features in human language came about via gramm a ticalization , suggesting that realizations of little p_ also came about via grammaticalization. It was also shown that this claim is made elsewh ere in the literature. However, this claim needed to be clarified since the term grammaticalization was found to correspond with several processes and potentially related effects. From a discussion of these processes and effects, it was determined that a l exical to functional identity change most directly corresponded with the central 70 Nevertheless, it is conceivable that List B could contain homophonous LVI and FVI copies of a common etymological source. There may be instances where the LVI form came from the FVI exponent, but the move from FVI to LVI wo uld entail the gaining of Encyclopedic information and it would be unclear why this knowledge would link to an FVI exponent.

PAGE 294

294 idea of grammaticalization and the observation of unidirectionality. In other words, it appears that the unidirectional effects of grammaticalization can be best defended as a n to this change, it is this particular change /step that is most likely irreversible , accounting for the fact that grammaticalized items rarely regain lexical uses . It c ould be thought that those items that do regain lexical uses really have not grammaticalized as defined by the categorical change process via grammation . Nevertheless, the cause for the changes leading up to reanalysis correlates with reanalysis, meaning t hat restricted use most often results in reanalysis. Accordingly this view accords well with the claim that adpositions are derived from lexical items if it is clear that only some adpositions have been grammaticalized . This accounts for a set of items th at may be used as lexical items and a set of items that ma y not (as shown in Chapter 4). Furthermore, a historical analysis of English prepositions show that those forms that do not show a lexical distribution in modern in English show the effects of gramm aticalization long before modern English. In addition the source for the proposed /a / and /be / heads ap pear to have undergone grammaticalization even early in the evolution of the language. Accordingly those forms that have a lexical use in Modern Engli sh appear lexical or more lexical than the FVIs prepositions within the same time span. This finding was important as it follows what would be expected from the discussion of unidirectionality and true grammaticalization. That is we would not expect to see lexical prepositions coming from older more grammatical looking forms. Taking this information, this Chapter showed that much as ontology might recapitulate phylogeny, the a diachronic analysis of English prepositions mirrors the

PAGE 295

295 synchronic analysis of p repositions in English given in Chapters 4, 5 and 6. However, while we would expect the /a / and /be / heads to eventually erode to / ø /, it is unclear if we should expect the other FVI p_ s to ever appear with a categorized and it is unclear if those items that cause / ø / to be inserted for p_ are more likely to be reanalyzed as a VI for p_ than those items that cause /a / and /be / to be inserted. These questions will be left for further thought and research .

PAGE 296

296 CHAPTER 9 CONCLUSIONS AND IMPLICATIONS 9.1 Summary Chapter 1 of this work discussed how categorical distinctions and the functional/lexical distinction, while seemingly capturing something true about language organization, are not concise outside a framework which may or may not artificially group items together. Moreover, thoughts do not appear to be converging on the identity of universal categories if there be any. This was especially clear with the category adposition which traditionally defies t he lexical/functional distinction. Moreover, it was shown that the category could not be unified via semantic meaning or syntactic function, especially if one desires to account for instances of particle prepositional forms (i.e. P items). Thus it was prop osed that the category P can be unified by the presence of a feature called [ F LINK ] that can bundle with spatial, temporal, and c ase orienting features to create all the attested versions of p_ . Chapter 2 of this work overviewed the DM framework used to analyze adpositional items. It argued that the DM model is superior to traditional lexical positions because it makes fewer assumptions regarding the ontology of word and sentence structure. However, it was pointed out that adpositions as a unified categor y might be difficult for the model because its insertion mechanisms require all insertion points to be distinguished as either l morpheme or f morphemes. As indicated in Chapter 1, despite being unifiable via the [ F LINK ] feature, simply analyzed adpositio nal forms appear in lexical domains but also represent functional features. Thus adpositional forms represented a potential problem for DM. The question posed is whether the class could or should be viewed as originating with a lexical node, a functional n ode or whether it

PAGE 297

297 should be split. It also was crucially pointed out that the DM model views items as being a categorical units than must receive their category from a local feature in the syntax. items in List A. It also discussed a methodology for testing whether or not linguistic i tems originate at a node since it is unclear if items in List A can come with any identifiable features. As discussed in Chapter 2, s were deemed to be a categorical entities that become categorized by grammatical features in the syntax. T his not only explains why the same can surface in a variety of different grammatical contexts in different languages but it also may help to explain the variable (i.e. underived) interpretations that s receive in such contexts. Nevertheless, the possibility of different types of features being present with items in List A was discussed in accordance with different proposal from the literature to explain specific properties (i.e. properties that appear to influence the grammar). Another reason for this was to highlight potential identifying features and to explore possibilities for controlling LVI insertion into particular nodes to explain derivational gaps. However, given the strong theoretical reason s to maintain Fea ture Disjoint ness and the empirical reasons to uphold the Lexical Decomposition Hypothesis, it was determined favorable to keep items in List A featureless and to try to explain gaps another way. Largely the discussion of gaps determined that most we re accidental. Nevertheless it was also recognized that some concepts seem to be more flexible and that others may not be conceivable in certain grammatical frames (i.e. those items that reference an entity or property do not work in an orientation f rame). To account for this, it was

PAGE 298

298 dis cussed that indices along with Contextual F eatures could be used as a theory internal way of modeling the fact that certain lexical concepts ( instantiations of a node) fail for nodes in a given context. Thi s setup the empirical argument of Chapter 4 that many adpositional items begin as items and that gaps should not be unexpected. Chapter 4 showed that in English there exists of a set of seemingly mono morphemic adpositional forms that can be found (n oting gaps with specific forms for certain domains) in transitive and intransitive verbal domains, nominal domains and adjectival domains. Given the identification guidelines discussed in Chapter 2, this distribution strongly suggests that such items originate with a node. It also showed that these forms could not always be explained as prepositions converted into verbs, nouns, or adjectives but rather as bare s because the meaning of the verbal, nominal, or adjectival form was not derived preposition. Moreover, the chapter showed that there was also a set of adpositional forms in English that are never used as nouns, verbs, and adjectives. Accordingly, these forms were determined not to originate as nodes . These findings were likewise found with adpositional items from Mandarin Chinese where all the postpositional forms and some of the prepositional forms were found to have a lexical distribution, suggesting the presence of a node in such cases. As f or those prepositional items that failed to show a lexical distribution, they were determined to be functional adpositional elements. Given the empirical findings of Chapter 4, Chapter 5 analyzed the data within the DM model, proposing that a little p_ he ad (unified by the presence of an [ F LINK ] feature)

PAGE 299

299 categorizes nodes in adpositional domains. The lexical distribution of forms that also appear in adpositional domains is then explained by the same item (i.e. the same LVI in a node) categorized by a type of n_, v_ or a_ . As for the functional adposi tional forms, they were argued to be FVIs inserted directly under the proposed little p_ head. This explains why they do not have a lexical distribution. Before this analysis was given, however, an alternative analysis given by De Belder (2011, 2013) was d iscussed. It was determined that De Belder's analysis was undesirable because it overgenerated FVI to LVI insertion, ignored the observation of unidirectionality in the framework of grammaticalization (as discussed further in Chapter 8), and required major stipulative changes to the Subset principle. This analysis was thus rejected in favor of the little p_ analysis. Nevertheless, with the data given, little p_ was only phonologically realized when it failed to categorized a (instead being merged with a DP). It had to be argued that a null FVI was being inserted under p_ in cases where it did categorize a . The explanation in Chapter 5 that a null FVI was inserted under p_ when it categorized a item would appear to be all too convenient if i t were not for other phonological evidence for a categorial p_ head. Chapter 6 presented such evidence by arguing that /a / and /be / be considered as the FVI representation of p_ in complex adpositional forms such as aside and beside . It was shown that su ch an analysis accounts for the fact that these two prefixes systematically change items into prepositions, particles, or results. It also accounted for the fact that be headed forms cannot operate as particles as defined by the inversion test and fronting tests. Thus /a / and /be / still appear to be morphemes and most reasonably can be assigned to the

PAGE 300

300 class of features labeled p_ . Moreover, in addition to this it was found the Dutch also clearly has phonological manifestations of a p_ head. It was shown that prepositions can be created form other categories in Dutch and when this happens the prefix /b(e) / is inserted. Likewise it was suggested that the Ezafe marker in Persian be considered as a phonological representation of p_ , as this marker also appears with items that have been converted into adpositions. Chapter 7 discussed how DM does not have a theoretical reason to exclude another categorial head type. It also further discussed the idea that these heads correspond to a class of features. It was dete rmined that the behavior and properties of the proposed p_ head resembled the properties of n_,v_, and a_ as much as these other heads resemble one another. It noted that p_ has a set of functional forms (i.e. of, at, to , etc.) just as v_ (auxiliaries) and n_ (pronominals) and perhaps a_ ( the item so as argued in Schütze ( 2001) ) . It has a class of Associated F eatures, as it co occurs with deictic and directional features, just as v_ often co occurs with temporal features, n_ sortal features, and a_ comparat ive features. Moreover, out of the three categorial classes it appears that p_ most resembles v_ . Chapter 8 presented a historical discussion of the category P. It discussed how many have claimed that adpositional elements are derived from lexical items v ia grammaticalization. However, to properly evaluate this claim, it showed that a more concise understanding of what grammaticalization entails needed to be established. It found that as a single process it was most helpful to view grammaticalization as a categorically lexical to functional change caused by grammation (a type of reanalysis). The effects of this process ofte n result in the other phenomena associated with

PAGE 301

301 grammaticaliz ation but these other phenomena may not occur for other reasons. This accou nt could then be applied to the claim that adpositions are derived from lexical items. A historical analysis of English adpositions suggested that those forms elsewhere argued to be FVIs had already categorically moved from lexical to functional items in O ld English while those forms argued to be LVIs appeared to be lexical items in Old English and Gothic. This corresponded with the principle of unidirectionality, meaning that one would not expect to find grammaticalized adpositional items regaining lexical uses. Moreover, the synchronic analysis given in Chapter 4, 5, and were shown to roughly mirror a specific model for the grammaticalization of adpositional items as given in (Lehmann 1985). It was thus suggested that this is further evidence that the pro posal here is correct because we would expect to find layers of the historical process contained in the synchronic process. Thus we see lexical items being categorized a prepositions with overt morphology, lexical items being categorized as prepositions wi th no overt morphology, and grammaticalized prepositions. 9.2 Implications and Future Research The conclusion of this work means that there exist a class features centered on the proposed [ F LINK ] feature that can frame conceptual information in the gramm ar. This occurrence is thus expected to be found in many more languages than this work has discussed. It is also unclear why this featural class does not appear to categorized s in all languages. That is , further research is needed on the relationship between Case markers and the category p_ as described here. If such markers also turn out to be types of p_ then the number of languages that use p_ to frame material grows substantially. However, in most cases it would seem that an analysis of Cas e markers

PAGE 302

302 as a type of p_ would have to involve s becoming nouns and then being framed by prepositional features. It thus may be unclear how such an analysis would be different than p_ relating a full DP to another DP as claimed with the free FVI form s in English and Mandarin. In addition to this issue, there needs to be further research on justifying Contextual F eatures . As argued, they may be used as a theory internal way to exclude certain LVIs from filling nodes in adpositional domains. Howe ver, unlike situations where non usage can be determined to be accidental, prescriptive, or allomorphic, the inability of using the DOG in and adpositional domain is not prescriptive or accidental. It seem that conceptually the sense of the it em cannot be applied in this domain. One might think that a putative index marker, as discussed in Chapter 3, individuates a prior to merge and thus perhaps provides a sneaky way for the Encyclopedia to influence w hat a item may merge with. This, make clear that s are void of both phonetic and semantic features in the derivation. As Embick (2012) conjectures, this pro blem requires more investigation. There thus remains no formal way to account for the effects of semantic categories on the distribution of items into grammatical categories. Finally given the brief discussion of adpositions and grammaticalization, over time one can expect that some LVI adpositional forms will be reanalyzed as direct representations of p_ . It is unclear if null headed LVIs or /a / and /be / headed LVIs are equally liable to be reanalyzed as FVIs. Further researc h is also needed on th is point as the model provided by Lehmann (1985) would suggest that the here argued LVIs out,

PAGE 303

303 on, u p are more likely to become FVIs than the LVIs round , side salient phonological spell out of the features of p_ . 9.3 Distributional Que stions and Curiosities In addition to the issues discussed in the previous section, there remain some specific distri butional issues regarding LVIs that may be used in adpositional domains in adjectival domains . As noted in Chapter 4 , when many P items ar e used in putative adjectival domains , further modification with very is either strange or ungrammatical. The reason suggested here for this is that the item in many cases is actually categorized as a particle, not an adjective (1). (1) Particle Predicate Do main a. We are (??very) off. b. I am (??very) up. c. It is (??very) near. 71 d. I am (??very) out. e. We are (??very) in. f. They are(??very) down. g. I am (??ve ry) down with a cold. 72 h . Th e lexicon era is(??very) over. i . The store is going (??very) under. j. The light is (??very) on. The domain seen with ( 1 ) is not an adjectival domain in the t raditional sense as these form+meanings resist very modification. They, therefore, cannot be equated with forms like tired/beautiful/blue in the predicate positions (it is very tired/beautiful/blue). This suggests that they are functioning as particles ( wh ich was argued in Chapter 1 and 6 to be another lexical posi tion), linking to the subject ( the DP Figure ) as particles have been 71 Not quite like the others because one must be near something. 72 Idiomatic meaning to have or possess a cold.

PAGE 304

304 proposed to do . On another note, the way P items generally inflect for the comparative and superlative is curious . M any of the argued LVI P items appear in adjectival domains as is evidenced by the presence of adjectival comparative and superlative morphology , but these markers generally do not attach to P items in the manner expected for monosyllabic adjectival forms . Instead the analytic forms more and most are generally used ( 2 ) . (2) Attributive Adjectival Domain (Comparative/Superlative) a. You are more off than him. b. Out of that bunch of weirdos, he is the most off individual c. Of the t wo, the upper dryer needs repair. d. The upmost/uppermost dryer needs repair. e. Step back into the outer ring. f. Step back into the outmost/outermost ring. g. Step forward into the inner ring. h. Step forwa rd into the inmost/innermost ring i. Here you can be nearer to nature. j. Where is the nearest store? k. He is downest ( Young Ni *&^% by Tupac (2pac) Shakur) l. She can stretch more under than you. m. Stretch to the undermost position. n. The most on light of the three is the red one. o. Of the two, this one looks more on . As shown in ( 2 ) at least seven of the forms in question can be found in comparative and super lative domains , indicating that these forms are indeed acting as adjectives . (3) OFF UP NEAR OUT IN DOWN OVER , UNDER , ON [ ] / {a_ , } However, how the items are placed in this domai n requires more investigation because even though the se form s are monosyllabic, the superlative is rarely / est / and in most cases there is a locational meaning (suggesting an adposition frame inside an adjectival one) . It is thus possible that these forms are being categorized as an adposition before

PAGE 305

305 being used in the adjectival domain. The question, however, is if there is a s ystematic re ason for this? T h ere is also the issue of a slight overlap with the degree adjectiv al domain ( 4 ). (4) Degree Adjectival Domain/Intensifier Domain a. I nearly missed. The degree domain of (4 ) i s itself a controversial domain as intensifying elements such as too , very , quite and so would be classified as FVIs by distribution and semantic tests. However, the form of many Degree Adjectives or Intensifiers appears in multiple domains: awful ly good , really good , bloody good , pretty good , terribly good etc... In many cases, though, the semantics are not connected, so the meaning of pretty , bloody and awful ly as Degree Adjectives has ostensibly nothing to do with the meaning of pretty , bloody and awful as attributive adjectives. 73 One may then argue that these are FVI homonyms of other LVIs and that the connection between form and meaning of the LVI has been lost with the FVI. 74 With other forms, however, one could see the opposite argument as really as a Degree Adjectives is similar to the meaning of real . In AWE TERR have neither a negative or positive meaning, allowing for terrible versus terrific and awful versus awesome . If that is the case then the meaning of terribly as a Degree Adj. could be derived from a single . No w while this point is not directly crucial to what is being generally discussed here and will not be discussed further, it is 73 Considering the arguments from Chapter 3 for an unspecified root meaning, one may object to this point. 74 It is crucial to recognize that this is not the case for adpositions, as the meaning seen in a lexical position is similarly seen in the adpositional position.

PAGE 306

306 important to realize that this is another context that may contain a node. Since nearly and overly contain a meaning that is related to near and over (i.e. if one accepts the notion of completion within the meaning of over), one may account for these attestations by linking the common forms and meanings to a node filled by the same LVI. Again, though, more research is needed. 9.4 Final Remarks The goals of this project were at least threefold: further our understanding of the adpositional class, describe adpositions in the model of Distributed Morphology, and to focus our attention to the possibility that issue s over universal versus specific language categories can be better approached by viewing things such as v_, n_, a_ and p_ as labels for a class of discrete features of which some may or may not be present in any given language. This work found that Distributed Morphology has the tool necessary to accommodate the traditionally difficult category P. All that needed to be done was to propose a categorial p_ head and exclude certain items from this domain with Contextual F eatures. This was found to be superior for semi lexical items in DM, the Revised Subset Principle) because the approach taken here does require major revisions to DM, accords with grammaticalization da ta, and does not overgenerate ( albeit Contextual Feature s are used to prevent this). Moreover, this work found that P items could be unified with a type of relational feature called [ F LINK ] and that this feature could then be added to other spatial and temporal features to form the individual p_ heads of the c lass P. The more adpositions were analyzed, the more they appeared to be like other lexical categories and the more little p_ looked like v_, n_ and a_ . Thus while a language may for whatever reason not frame a concept

PAGE 307

307 with a particular feature bundle, it is likely to do so with a similar feature bundle , explaining both categorical differences and similarities seen cross linguistically and providing a possible out for the hypothesis of Baker (2003) in contrast to Haspelmath (2007) .

PAGE 308

308 LIST OF REFERENCES Abne y , Steven P. 1987 . The English noun phrase in its sentential aspect . Ph.D. dissertation, Massac husetts Institute of Technology, MA . Acquaviva, Paolo. 2008. s and lexicality in Distributed Morphology. Ms., University College Dublin and Universität Kons tanz. Acquaviva, P aolo , and Phoevos Panagiotidis. 2012 . Lexical decomposition meets conceptual atomism. Lingue e linguaggio 11 (2) : 165 180. Acquaviva, Paolo. 2014 . s, concepts, and word structure. In Morphology and Meaning: Selected papers from the 15th International Morphology Meeting, Vienna, February 2012 , 327 :49 . John Benjamins Publishing Company. Adger, David. 2003 . Core syntax: A minimalist approach . 33. Oxford: Oxford University Press. Alexiadou, Artemis. 2011. Plural mass nouns and the mo rphosyntax of Number. In Proceedings of the 28th West Coast Conference on Formal Linguistics , ed. by Mary Byram Washburn et al. , 33 41, Somerville, MA: Cascadilla Proceedings Project. Arad, M aya . 2003. Locality constraints on the interpretation of s. Natural Language a nd Linguistic Theory 21: 736 778. Arad, M aya . 2005 . s and patt erns: Hebrew morpho syntax. Berlin: Springer . Asbury, Anna B. Gehrke, and V. 2006 . One size fits all: prefixes, particles, adpositions and cases as me mbers of the category P. UiL OTS yearbook 1 17. Baker, Mark C. 2003. Lexical categories : Verbs , nouns and adjectives . Cambridge , MA : Cambridge University Press. Barnhart, Robert K . (ed.). 1995. The Barhart Concise Dictionary of Etymology: The origins of American English Words . NewYork, NY: Harper Collins Publishers. Bhattacharya, Tanmoy . 2001 . Numeral/quantifier classifier as a complex head. In Semi lexical Categories , ed. by N. Corver and H.C. van Riemsdijk , 191 222 . Walter de Gruyter . Boertien, H a rmon . 1997. Left Headed Compound Prepositions. Linguistic Inquiry 4: 689 697. Cambridge , MA : MIT Press. Bolinger, Dwight Le Merton . 1971 . The phrasal verb in English . 27 . Cambridge, MA: Harvard University Press.

PAGE 309

309 Borer, H agit . 2005 . The Exo Skeletal Tril ogy, Book III. Ms., University of Southern California Borer, H agit . 2009. s and categories. Talk given at the 19th Colloquium on Generative Grammar, University of the Basque Country, Vitoria Gasteiz. Botwinik Rotem, Irena. 2004. The Category P. Features, Projections, Interpretation . Ph.D. dissertation, University of Tel Av iv. Brennan, Jonathan . 2007 . Irish Prepositions: Agreement and Impoverishment. In Proceedings WCCFL 26 . Butt, Miriam, and Wilhelm Geuder. 2001 . On the (semi) lexical status of light verbs. In Semi lexical Categories , ed. by N. Corver and H.C. van Riemsd ijk , 323 370 . Walter de Gruyter . Cardinaletti, Anna, and Giuliana Giusti . 2001 and Germanic. In Semi lexical Categories , ed. by N. Corver and H.C. van Riemsdijk , 371 430 . Walter de Gruyter . Celce Murcia, Mariann e, and Diane Larsen Freeman . 1999. The Grammar Book: An course 2 nd ed. Boston: Heinle and Heinle. Cheung , Hung nin S. 1994 . A Practical Chinese Grammar . Hong Kong : The Chinese University Press. Chomsky, Noam. 1970. Remarks on nominaliz ation. In Readings in English transformational grammar , ed. by R. Jacobs and P. Rosenbaum Waltham, 184 221. MA: Ginn. Chomsky, N oam. 1981 . Lectures on Government and Binding: Foris, Dordrecht. Chomsky, N oam. 1986 . Knowledge of Language: Its Nature, Origi n and Use . Westport, CT: Praeger. Chomsky, N oam . 1995. The minimalist program. Cambridge, MA: MIT Press. Croft , W illiam . 1991. Syntactic Categories and Grammatical Relations. Chicago: The University of Chicago Press. Corver, Norbert, and Henk v an Riems dijk . (e ds.). 2001 . Semi lexical categories: the function of content words and the content of function words . Walter de Gruyter. Deacon , Robert J. 2011 . P forms in Distributed Morphology: accounting for a type of semilexical form. MA Thesis , University of Florida .

PAGE 310

310 DeLancey, S cott . 2005 . Adpositions as a non universal category. Linguistic diversity and language theories 185 202. Dehé, N icole. (e d.). 2002 . Verb particle explorations . 1 . Mouton de Gruyter. den Dikken, Marcel. 1995. Particles: On the Syn tax of Verb Particle, Triadic and Causative Constructions . New York , NY: Oxford U niversity Press . den Dikken, M arcel. 2006 . On the functional structure of locative and directional PPs. Ms., CUNY. De Belder, M arijke . 2011 . s and affixes: eliminating lexical categories from syntax. LOT 282 . De Belder, Marijke , and Jeroen van Craenenbroeck . 2011 . How to merge a . Ms., HUBrussel and Utrecht University. De Belder, Marijke , and Jeroen van Craenenbroeck . 2013 . On voc abulary insertion . M s. ,HUBrussel. Di Sci ullo, Anna M aria , and Edwin Williams. 1987. On the Definition of Word. Cambridge , MA : MIT Press. Dj amouri, Redouane, and Waltraud Paul. 2009. Verb to pr eposition reanalysis in Chinese. I n Historical syntax and linguistic theory , ed. by Paola Crism a and Giuseppe Longobardi, 194 211 . Oxfo rd: Oxford University Press . Djamouri, Redouane , Waltraud Paul , and John Whitman. 2012 . Postpositions vs. prepositions in Mandarin Chinese: The articulation of disharmony. In The oretical approaches to disharmonic word orders , ed by T. Biberauer and M. Sheehan. Oxford: Oxford University Press. Embick, David. 2000. Features, syntax, and categories in the Latin perfect. Linguistic Inquiry 31( 2 ) : 185 230. Embick, D avid. 2004. On the Structure of Res ultative Participles in English. Linguistic Inquiry 35(3). 355 92. Embick, David . 2012 . s and features (an acategorial postscript)*. Theoretical Linguistics 2012 38(1 2): 73 89 . Embick, David , and Alec Marantz. 2008 . Architecture and blocking. Linguistic Inquiry 39 (1): 1 53 .

PAGE 311

311 Embick, David and Rolf Noyer . 2005 . Distrib uted Morphology an d the Syntax Morphology Interface. I n The Oxford Handbook of Linguistic Interfaces , ed. by Gillian Ramchand and Charles Reiss , 1 28. Oxford: O xford University Press . Emonds, J oseph . 1972 . Evidence that indirect object movement is a structure preserving rule. Foundations of Language 546 561. Emonds, J oseph . 1985 . A unified theory of syntactic categories . 19 . Walter de Gruyter. Evans, Nicholas , and Toshiki Osada. 2005 . Mundari: The myth of a language without word classes. Linguistic Typology 9 (3): 351 3 90 . Fischer, Olga . 2003 . Principles of Grammaticalization and Linguistic Reality. In Determinants of Grammatical Variation in English , ed. by G. Rohdenburg and B. Mondorf, 445 478 . Berlin and New York: Mouton de Gruyter. Fischer, O lga . 2008 . On analogy as the motivation for grammaticalization. Studies in language 32 (2) . Folli, Raffae lla, and Heidi Harley. 2007. Causation, Obligation, and Argument Structure: On the Nature of Little v. Linguistic Inquiry 38:197 238. Gelderen, Elly van . 2006 . Econ omy of Merge and Grammaticalization: Two Steps in the Evolution of Language. M s. Grimshaw, J ane . 1991 . Extended projec tion. Ms., Brandeis University. 1997. Projection, heads, and optimality. Linguistic Inquiry 28 : 373 422 . Grimshaw, J ane . 20 05 . Words and stru cture . Center for the Study of Language and Information. Stanford University. Haider, H ubert . 2001 . Heads and selection. Semi lexical Categories 67 96 . New Y ork: Mouton de Gruyter . Hagege, Claude . 2010 . Adpositions . Oxford: Oxford University Press. Ha gemeijer, T jerk . 2001 . Underspecification in serial verb constructions. In Semi lexical Categories , ed. by N. Corver and H.C. van Riemsdijk , 415 51 . Walter de Gruyter . Halle, Morris , and Alec Marantz . 1993 . Distributed Morphology a nd the pieces of infle ction. In The view from building 20 , ed. by Kenneth Hale and Samuel Jay Keyser, 111 176 . Cambridge: MIT Press . H alle, Morris , and Alec Marantz. 1994 . Some Key Features of Distributed Morphology. MIT Working Papers in Linguistics 21 : 275 288 .

PAGE 312

312 Halle, Morris , and J. R . Vergnaud. 1987 . Stress and the cycle. Linguistic Inquiry 18 (1): 45 84. Harley, H eidi , and R olf Noyer. 1999. State of the Article: Distributed Morphology . Glot International 4(4): 3 9. Harley, Heidi , and Rolf Noyer. 2000 . Formal versus Encyclo pedic Properties of V ocabulary: Evidence from nominalizations. In The lexicon encyclopedia interface , ed. by B. Peeters . Elsevier: Amsterdam. Harley, H eidi. 1995a . Subjects, Events and Licensing, Ph.D. dissertation , Massachusetts Institute of Technology , MA . Harley, H eidi. 1995b . Sase bizarre: The st ructure of Japanese causatives. In Proceedings of the Canadian Linguistic Society Meeting, ed. by P. Koskinen. Harley, Heidi. 2005a One replacement, unaccusativity, acategorial s and Bare Phrase Structur e. Harvard Working Papers in Linguistics , ed. by Slava Gorbachov and Andrew Nevins. Harley, H eidi . 2005b. On the Causative Construction, M s., University of Arizona. Harley, H eidi . 2006. The morphology of nominalizations and the syntax of vP. Ms. , Univers ity of Arizona . Harley, H eidi . 2008. Compounding in Distributed Morphology. In The Oxford Handbook of Compounding , ed. by R. Lieber and P. Stekauer . Harley, H eidi. 2009 . The morphology of nominalizations and the syntax of vP. Harley, H eidi. 2011 . On th e identity of s. lingbuzz/001527 . Harper , Douglas ( ed. ). 2001 . Online Etymological Dictionary . Retrieved February 15, 2013, from Haspelmath, M artin . 1999 . Why is g rammaticalization irreversible? Linguistics 37(6): 1043 10 68. Haspelmath , M artin . (e d.). 2001 . Language typology and language universals: an international handbook . 20 . Walter de Gruyter. Haspelmath, M artin . 2004 . On directionality in language change with particular reference to grammaticalization. Typologica l Studies in Language 59 : 17 44 . Haspelmath, M artin . 2007 . Pre established categories don't exist: Consequences for language description and typology. Linguistic Typology 11 (1): 119 132 .

PAGE 313

313 Haugen, Jason D., and Daniel Siddiqi. 2013 . s and the derivation. Linguistic inquiry 44 (3): 493 517 . Hayes, Bruce . 1990 . Precompiled Phrasal Phonology. I n The Phonology Syntax Connection , ed. by S. Inkeles and D. Zec, 85 108 . Chicago: The University of Chicago Press . Heine, Bernd, and Tani a Kuteva. 2002. On the evolution of grammatical forms. In The transition to language , 376 397. ed. by Alison Wray. Oxford: Oxford University Press. Her, One Soon . 2006 . Justifying part of speech assignments for Mandarin< i> gei. Lingua 116 (8): 1274 1302 . Hill, Kenneth , and Mary E. Black. 1998. A sketch of Hopi grammar. In Hopi dictionary / Hopìikwa Lavàytutuveni: A Hopi English dictionary of the third Mesa dialect , ed. by the Hopi Dictionary Project, 861 900. Tucson: University of Arizona Press. Hi r st, G. 2009 . Ontology and the Lexicon. In Handbook on ontologies 269 292 . Berlin : Springer. Hopper, Paul J., and Sandra A. Thompson . 1984 . The discourse basis for lexical categories in universal grammar. Language 703 752. Hopper, Paul J., and E. C. Trau gott. 1993 . Grammaticalization. In Cambridge textbooks in linguistics . Cambridge: Cambridge Univeristy Press . Hopper, Paul J., and Elizabeth Closs Traugott . 2003 . Grammaticalization . Cambridge: Cambridge University Press. Hudson, R ichard . 2000. Grammar without Functional Categories. In The Nature and Function of Syntactic Categories ed. by R. Borsley , 7 35 . Academic Press. Hurskainen, A rvi . 1992 . A two level computer formalism for the analysis of Bantu morphology. Nordic journal of African studies 1 (1) : 87 119 . Jackendoff, R ay. 1973 . The base rules for prepositional phrases. A festschrift for Morris Halle 345 356. Jackendoff, Ray. 1977 . Syntax: A Study of Phrase Structure. Cambridge, MA: MIT Press. Jespersen, O tto. 1965 . The philosophy of gramma r . University of Chicago Press.

PAGE 314

314 Johnson, Kyle. 1991. Object positions. Natural Language and Linguistic Theory 9: 577 636. Kim, Min Joo . 2002. Does Korean have adjectives. MIT Working Papers in Linguistics 43 : 71 89 . Kiparsky, P aul . 1983 . Word formation and the lexicon. In Proceedings of the 1982 mid America linguistics conference 3:22 . Lawrence, Kansas: Department of Linguistics, University of Kansas. Kiparsky, P aul . 1997 . Remarks on denominal verbs. In Complex predicates , ed. by A. Alsina, J. Bresnan a nd P. Sells, 473 499 . Stanford, CA: CSLI. Kiparsky, Paul . (2009) . ion.pdf Koopman, Hilda. 2000. Prepositions, postpositions, circumpositions and particles: The structure of Dutch PPs. In The syntax of specifiers and heads , ed. by Hilda Koopm an , 204 60 . London: Routledge. Kortmann, Bernd , and Ekkehard König. 1992 . Cat egorial reanalysis: the case of deverbal prepositions. Linguistics 30: 671 697. Kracht, Marcus. 2002. On the semantics of locatives. Linguistics and Philosophy 25: 157 232. Lampitelli, N icola . 2010 . Nounness, gender, class and syntactic structures in Ita lian nouns. Romance Languages and Linguistic Theory 2008. Selected papers from'Going Romance'Groningen 195 214. Langacker, Ronald W . 1987 . Foundations of cognitive grammar: Theoretical prerequisites . 1 . Stanford University P ress. Larson, Richard, and H iroko Yamakido . 2005 . Ezafe and the deep position of nominal modifiers. In Barcelona Wor kshop on Adjectives and Adverbs. Barcelona . Lass, Roger. 1997. Historical Linguistics and Language Change . Cambridge, UK: Cambridge University Press. Legate, Julie Anne . 1999 . The morphosyntax of Irish agreement. In Papers on Morphology and Syntax, Cycle One : MIT Working Papers in Linguistics , ed. by K. Arregi, B. Bruening, C. Krause, and V. Lin, 33: 219 240 . Li, Yen hui Audrey 1990 . Order and constituency in Manda rin Chinese. Dordrecht: Kluwer.

PAGE 315

315 Lieber, Rochelle . 1992. Deconstructing Morphology. Chicago: University of Chicago Press. Lieber, Rochelle . 2006 . The category of s and the s of categories: what we learn from selection in derivation. Morpholog y 16 (2): 247 272. Lowenstamm, J. 2008 . The Sounds of Silence: Empty Elements . I n Syntax and Phonology , ed. by J. Hartmann , V. Hegedus and H. van Riemsdj ik Amsterdam: Elsevier. Lü, Shuxiang et al. (eds.) . 2000 . Xiandai hanyu babaici . Beij ing: Shangwu chubanshe. Mardale, A. 2011 . Prepositions as a semilexical category. Bucharest Working Papers in Linguistics 13 (2): 35 50. Marantz, Alec. 1997 Privacy of Your Own Lexicon. UPenn Working Papers in Linguistics 4: 201 225. Marantz, A lec. 2001 . Words and things. Ms. , MIT . McCawley, James D. 1992. Justifying part of speech as signments in Mandarin Chinese , Journal of Chinese Linguistics 20( 2 ): 211 245. McLaughlin, Fiona. 2004 . Is t here an adjective class in Wolof? In Adjective classes: a cross linguistic typology , ed. by R.M.W. D ixon and Alexandra Y. Aikhenvald, 242 262 . Oxford: Oxford University Press. Miller, Gary D . 2006. P Words in Theory, Typology, and Old English . Ms. , Uni versity of Florida. Miller , Gary D . 2010a. Language Change and Linguistics Theory I: Approaches, Methodology, and Sound Changes . New York: Oxford University Press Miller , Gary D . 2010b. Language Change and Linguistics Theory II: Morphological, Syntac tic, and Typological Change . New York: Oxford University Press Muysken, Pieter . 2008 . Functional categories . Cambridge: Cambridge University Press. Newmeyer, Frederick , J. 2006 . What can grammaticalizati on tell us about the origins of language. The evol ution of language 434 435. Norman, Jerry . 1988 . Chinese. Cambridge: Cambridge University Press. Prepositions and Particles in English: A Discourse Functional Account . New York: Oxford University Press.

PAGE 316

316 Ouhalla, Jamal. 1991 . F unctional categories and parametric variation . London: Routledge. Packard, Jerome L. 2000 . The Morphology of Chinese . Cambridge: Cambridge University Press Padrosa Trias, S. 2009 . Compounds in Distributed Morphology: Pros and Cons. In Talk given in th e LyCC Colloquium Series, CCHS CSIC Sem inar on Theoretical Linguistics. Madrid . Pantcheva, Marina. (2007 a Ms., University of Tromsø . Pantcheva, M arina . (2007 b ). Persian preposition classes. Nordlyd 33 (1). Pantchev a, M arina . (2008). The place of PLACE in Persian. Syntax and semantics of spatial P 120 : 305 . Pfau, Roland . 2000 . Features and categories in language production . Ph. D. dissertation , Johann Wolfgang Goet he Universität, Frankfurt/Main, Germany . Pfau, Ro land . 2009 . Grammar as processor: A Distributed Morphology account of spontaneous speech errors . 137 . John Benjamins Publishing. Robins, Robert H . 1952 . Noun and verb in universal grammar. Language 28 (3): 289 298. Ramchand, Gillian , and P eter Svenonius . 2002 . The lexical syntax and lexical semantics of the verb particle construction. In Proceedings of WCCFL 21: 387 400 . Riemsdijk, Henk . 1990 . Functional Prepositions Unity and Diversity: In Papers Presented to Simon C. Dik on his 50th Birthday ed. by H. Pinkster and I. Genée, 229 241. Foris: Dordrecht. Riemsdijk, Henk . 1998 . Categorial feature magnetism: The endocentricity and distribution of projections. The Journal of Comparative Germanic Linguistics 2 (1): 1 48 . Rooryck, Johan . 1996 . Prepositions and minimalist case marking. Studies in comparative Germanic syntax 2 : 226 256 . Ross, John R. 1972 . The category squish: endstation hauptwort. In Chicago Linguistic Society , 8: 316 328 . Oxford University Press.

PAGE 317

317 Samiian, Vida . 1994 . The Ezafe constructio n: some implications for the theory of X bar syntax. Persian Studies in North America 17 41. Sapir, E dward. 1921 . An introduction to the study of speech. Language . Schü tze, Carson . 2001 . Semantically empty lexical heads as last resorts. In Semi lexical C ategories , ed. by N. Co r ver and H.C. van Riemsdijk , 127 187. Walter de Gruyter . Selkirk, E lisabeth. 1995 . The prosodic structure of function words. In Signal to syntax , ed. by K. Demuth and J. Morgan . Hillsdale, NJ: Erlbaum. Siegel, Dorothy . 1974 . Topic s in English morphology. Ph . D . dissertation, MIT. Published 1979, New York: Garland. Siddiqi , Daniel. 2005. Distributed Morphology Without Secondary Exponence: a Local Account of Licensing Thematic Licensing of Vocabulary Items and Strong Verb Alternation s . Coyote Papers 14 : 108 143 . Siddiqi, Daniel. 2009. Syntax within the word: Economy, allomorphy, and argument selection in Distributed Morphology . John Benjamins. Siddiqi, Daniel. 2010. Distributed Morphology. Language and Linguistics Compass 4( 7 ) : 524 542 Slocum, Jonathan and Todd Krause. (ed . ). 2013. Gothic Online Base Form Dictionary . University of Austin Texas Linguistic Research Center: Retrieved February 17, 2013 from BF X.html Smith, Jeremy . 1 996 . An Historical Study of English: Function, Form and Change . London: Routledge. Suelmann, Thomas . 1999 . L'afrikaans de poche . Assimil. Svenonius, P eter . 2004. Adpositions, particles, and the arguments they introduce. Ms. , University of Tromsø. Svenon ius, P eter . 2006 . The Emergence of Axial Parts. CASTL, University of Tromsø . Svenonius, Peter. 2007 . Adpositions, particles and the arguments they introduce . In Argument Structure , ed. by E. Reuland, T. Bhattacharya and G. Spathas , 63 103 . Amsterdam: Benj amins . Svenonius, Peter. 2008. Spatial P in English . CASTL, University of Tromsø.

PAGE 318

318 Taylor, J ohn . 1993. Prepositions: Patterns or Polysemization and Strategies of Disambiguation. In Natural Language Processing: The semantics of Prepositions , ed. by C. Zeli nsky Wibbelt, 151 175. The Hague: Mouton de Gruyter. Traugott, Elizabeth C. 2004 . Exaptation and grammaticalization. Linguistic studies based on corpora 133 156. Traugott, E lizabeth C. 2008 . Grammaticalization, constructions and the incremental develop ment of language: Suggestions from the development of degree modifiers in English. Trends in Linguistics Studies and Monographs 197 : 219 . Travis, L isa . 2005 . Passives, states, and s and Malagasy. UCLA working papers in Linguistics 12 . Vincent, N. 1995 . Exaptation and grammaticalization. Amsterdam Studies in the Theory and History of Linguistic Science Series 4: 433 433. Zaenen, Annie, Joan Maling, and Höskuldur Thráinsson. 1985 . Case and gramm atical functions: The Icelandic passive. Natural Langu age and Linguistic Theory 3(4): 441 483. Zeller, J ochen . 2001 . Lexical particles, semi lexical postpositions. In Semi lexical Categories , ed. by N. Corver and H.C. van Riemsdijk , 505 49 . Walter de Gruyter . Zwarts, J ochen . 1995 . Lexical and functional direction in Dutch. Linguistics in the Netherlands 12 (1): 227 238 . Zwarts, J ochen . 1997 . Lexical and functional properties of prepositions. In Grammatische Eigensc haften präpositionaler Elemente, ed. by Haugmann Dagmar and Stefan J. Schierholz , 371: 1 1 8 . Tübingen: Linguistische Arbeiten . Zwart, Jan Wouter . 2005 . A note on functional adpositions. In Organizing Grammar: 689 695 . Berlin: Mouton de Gruyter .

PAGE 319

319 BIOGRAPHICAL SKETCH One often has to sit back an d ask of himsel f what he did with all the time, to ponder the ebb and flow of the restlessness, the stillness, and the time forgotten. The reason is a mirage , as it is the path and not the p lace that is sought, reckoning this earth in frenzy an d sl oth, in sickness and health, and in solitude and as companions. Friends for a time shared bread, told stories and drank wine. Their eyes salivated over th e murmuring stars. All was wished and well was all . Tears of joy are unexpected . Ch ildren remind yo u of everything, of every hole left with every passing foot . And for t hes e little goals, b irth and death, everyone makes a print, tosses a coin into the recesses . Robert J Deacon was born and raised in Florida. He received a degree from the Univ ersity of North Florida, major ing in English and minoring in p sychology. Subsequently, he received a degree in l inguistics from the University of Florida in 2011 and a Ph.D. in 2014 . In the meantime , life completely changed. He was married to Yuka ri Nakamura, and they are expecting their first child . With these changes, t he real challenges and joys of life bega n.


$UFKLWHFWXUHDQG%ORFNLQJ $XWKRUVf'DYLG(PELFNDQG$OHF0DUDQW] 6RXUFH/LQJXLVWLF,QTXLU\9RO1R:LQWHUfSS 3XEOLVKHGE\7KH0,73UHVV 6WDEOH85/ . $FFHVVHG Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact . The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


Architecture and Blocking David Embick Alec Marantz We discuss theoretical approaches to blocking effects, with particular emphasis on cases in which words appear to block phrases (and perhaps vice versa). These approaches share at least one intuition: that syntactic and semantic features create possible "cells" or slots in which particular items can appear, and that blocking occurs when one such cell is occupied by one form as opposed to another. Accounts of blocking differ along two primary dimensions: the size of the objects that compete with one another (morphemes, words, phrases, sentences); and whether or not ungrammatical forms are taken into consideration in determining the correct output (relatedly, whether otherwise wellformed objects are marked ungrammatical by competition). We argue that blocking in the sense of competition for the expression of syntactic or semantic features is limited to insertion of the phonological exponents of such features (the Vocabulary items of Distributed Morphology) at terminal nodes from the syntax. There is thus no blocking at the word level or above, and no competition between grammatical and ungrammatical structures. The architectural significance of these points is emphasized throughout the discussion. Keywords: architecture, blocking, competition, Distributed Morphology 1 Introduction Intuitions about blocking are driven by certain canonical cases emphasized in the linguistic literature. For example, a prevailing intuition is that the "irregular" form gave blocks the otherwise expected "regular" *gived as the past tense of give. This intuitive notion of blocking evokes the cells of a paradigm structure: in a list or table of inflected forms of the verb whose stem is give, there exists a slot for the past tense of the verb. If this slot is filled by a memorized form, then the word formed by the regular past tense "rule" of adding -(e)d is blocked from filling this slot by the memorized form. The intuitive notion of blocking is illustrated in (1), which shows paradigmatic cells or slots and their contents. (1) Slots Lexemes Present 3sg Past WALK walk walks walked GIVE give gives gave The authors are indebted to Morris Halle. Linguistic Inquiry, Volume 39, Number 1, Winter 2008 1-53 © 2008 by the Massachusetts Institute of Technology 1 This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


2 DAVID EMBICK AND ALEC MARANTZ In this article, we will support the view that blocking as captured by the intuitive description that gave blocks *gived is not part of the grammatical system of language. Crucially, the intuitive notion of blocking, the basis for a set of formal proposals in the literature that have become standard, relies on two assumptions that we examine in detail here. First, the standard approach assumes that blocking involves competition at the level of the word, or perhaps at the level of larger objects as well (phrases, sentences). Second, it assumes that blocking involves consideration of * 'ill-formed' ' words like *gived (or ill-formed phrases/clauses, in some theories). These assumptions define a range of competition-based approaches to blocking effects that hold that the existence of some irregular or memorized forms renders certain other forms deviant, even though the forms in question are not problematic as far as independent principles of the grammar are concerned. The losers of the competition are marked deviant solely because some ''listed" or "better" way of expressing that meaning is found in the language. We argue for a very different perspective on these matters. Generalizing, the standard approach to blocking effects centers on two questions that motivate our investigation here: Question 1 (Locality of competition question) Is the computation of morphophonology local to the terminal nodes from the syntax, or more global, at the level of the phrase or sentence? Question 2 (Grammaticality question) Does the grammar involve comparison of two or more otherwise grammatical expressions (i.e., expressions that would be fine connections between sound and meaning, if they did not lose the competition to a "better" expression)? We answer these questions as follows: 1. Competition in the relevant sense is limited to the level of the morpheme: in the model that we elaborate below, this amounts to competition between Vocabulary items, for insertion into terminal nodes in a syntactic structure. In the example of give and gave, the zero allomorph of the past tense node T[past] competes with the regular -(e)d allomorph for realization of the past tense terminal node from the syntax and wins in the context of give (but not in the context of, say, walk); a morphophonological readjustment rule changes the phonology of give to gave in the context of the past tense morpheme. No larger objects words, phrases, clauses enter into the competition. 2. The change in perspective involved in this approach to blocking has consequences for the treatment of ungrammatical forms. From the point of view of the grammar, *gived is ill formed because it will not be part of the phonological form of a grammatical sentence; that is, it is not generated by the grammar. Thus, what is called "blocking" in the literature does not involve consideration of forms of a word rendered ungrammatical via competition (as it would if *gived were marked ungrammatical by virtue of losing the competition with gave). Rather, blocking involves the interaction of stored information about morThis content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 3 phemes and the syntactic and phonological systems that build forms (and do not, crucially, yield *gived). Much of the discussion in this article centers on the analysis of blocking effects within Distributed Morphology, and on the examination of alternative approaches to blocking that make different predictions about the range of possible crosslinguistic variation. As indicated in the questions posed above, accounts of blocking differ crucially with respect to the locality of interactions between morphology and syntax. The centrality of questions of locality in this domain has a history that connects these questions with questions of architectural significance. Aronoff s (1976) groundbreaking treatment of blocking is based on the idea that blocking effects involve listedness or irregularity: "lexical" properties. In line with lexicalist assumptions about the division of labor in the grammar, this means that blocking effects are expected to be found in the lexicon, that is, in the domain of (certain types of) words. Poser (1992) directed the attention of the field to cases of what looks like blocking in which a phrase and a single word express the same meaning. To cite one of Poser's examples, if a phrase like *more smart is blocked by a word like smarter in the same way that gave blocks *gived, then it is necessary to extend the competitors relevant to blocking "out of the lexicon" (this is in fact the point that Poser makes). For a theory like Distributed Morphology, in which all word formation is syntactic, the interaction of word formation and phrasal syntax is in some sense unsurprising. A particular syntactic configuration might yield either a phrasal or a single-word expression, depending on conditions governing the particular rules of syntactic affixation that might apply. For example, a condition limiting the merger of the comparative head to adjectives of a particular phonological shape would yield a single word like smarter or taller when the rule applies, and a phrasal expression like more intelligent when the rule does not apply. Cases of word/phrase alternation are simply a subcase of syntactic affixation in which the rule that affixes one piece to another is "partial" in some sense. As far as the grammar is concerned, a single set of mechanisms responsible for affixation in syntactic structures is all that is required; there is no need to implement blocking or competition. For theories that adopt blocking competition between expressions as a mechanism in grammar, on the other hand, cases of word/phrase "Poser blocking" require an extension of blocking beyond word-word comparisons and also beyond any notion that only "listed" words block words or phrases that are produced by regular processes (since, for example, comparatives like smarter are regular and need not be listed, but nevertheless appear to block phrases like *more smart). In asking whether there is competition in word-phrase interactions, we also consider whether there is evidence for word/word blocking in the first place. On our view, there is no fundamental architectural distinction between word-word and word-phrase (or phrase-phrase, etc.) interactions. Thus, rather than "extending the lexicon" in the manner associated with Poser blocking, our approach holds that there are explanations for all the relevant interactions that do not require blocking, and that, moreover, the explanations are syntactic in nature. Thus, part of the shift in perspective that comes from the theory we propose here involves the kinds of objects that have This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


4 DAVID EMBICK AND ALEC MARANTZ to be considered in accounting for blocking effects. We demonstrate below that even "canonical" cases of blocking those that appear to involve only words within a lexicon implicate comparisons between phrasal and singleword expressions. Our initial comments above illustrate intuitions about blocking with reference to the relation between gave and *gived for the past tense of give. Consider now that did . . . give, with tense and the verb expressed in two separate words, wins over both gave and *gived as the past tense of give in the presence of negation {didn't give, *gaven't) and in questions or emphatic contexts (You did give him the paper, didn't you?). This is an obvious and extensively discussed facet of English clausal syntax, but the importance of such observations for architectural matters has not been fully appreciated in the literature on blocking (but see the discussion of Andrews 1990 below). There is no escaping syntax in the discussion of blocking, and considerations of economy of expression should be as relevant at the level of the phrase as at the level of the word if these considerations drive blocking. The importance of clausal syntax for blocking can be acknowledged both in local (nodeonly) theories of competition like ours, and in theories that extend competition to larger objects. The question then is which approach to grammar correctly predicts the empirical facts of blocking. The phenomena usually described as involving blocking, particularly Poser blocking, provide an empirical base for deciding these issues. When the approaches to blocking involving comparison between otherwise grammatical expressions are made explicit, they all necessarily involve global competitions and they appear to make the wrong predictions about the range of blocking phenomena observed. An alternative architecture of the grammar, that of the Minimalist Program as instantiated in Distributed Morphology, fares better in these cases precisely because it limits competition to allomorphy and limits morphophonological computation to the individual terminal node in its syntactic environment. In section 2, we outline Distributed Morphology and what it has to say about blocking effects, as a preliminary to the discussion of "standard blocking" in section 3. Addressing some of the standard cases of blocking from the literature such as the relationship between glory and *gloriosity, we show that there is no clear evidence for word/word blocking. In section 4, we discuss word-phrase interactions of the type associated with Poser blocking. We show that a theory with global (clause-clause) competition along the lines proposed by Bresnan (2001) makes incorrect predictions about the relevant interactions. Similarly, theories that implement Poser blocking with words winning over phrases under particular circumstances turn out to be inadequate. Instead, the generative treatment within Distributed Morphology makes the correct predictions. We discuss some implications of our results in section 5. 2 Distributed Morphology 2.1 Basics of Distributed Morphology Distributed Morphology is a syntactic, piece-based, realizational approach to morphology in which there is at least some late insertion of phonological material into terminal nodes. The nodes are the primitives of syntactic derivations; many "morphological" operations are part of the PF component of the grammar (shown schematically in (2)). This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 5 (2) The grammar Syntactic derivation v (Spell-Out) Morphology / \. PF LF The nodes that are manipulated in syntactic derivations are of two types, functional morphemes and Roots. (3) Terminals a. Functional morphemes are composed exclusively of nonphonetic features, such as [past], [pi], or the feature (or features) that make up the determiner node D of the English definite article the. b. Roots make up the open-class or "lexical" vocabulary. They include items such as VCat, Vox, and VSrr. The functional morphemes are functional categories in the sense familiar from syntactic theory. In the PF component of the grammar, these morphemes receive phonological representations in the process of Vocabulary Insertion. This process involves Vocabulary items like those in (4), which spell out the past tense node T[past] in English; these items compete according to specificity, so that the most highly specified wins. When two Vocabulary items tie on measures of specificity, as is the case for the irregular items with ~t and -0'm (4), either there is no ordering, a possibility discussed in section 3.1, or the items are extrinsically ordered. (4) Vocabulary items for past tense (T[past]) T[past] <-t/{\/LEAVE, \/Bend, . . . } nttt T[past] <-0/{VHit, VQuit, . . . } nttt T[past] <-ed In displaying the effects of Vocabulary Insertion, we represent the node in question with its features/label, as well as the phonological exponent associated with the Vocabulary item. The verb kicked, for example, is represented as in (5). (5) Structure for kicked v v T[past,-ed] VKick [v,0] This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


6 DAVID EMBICK AND ALEC MARANTZ As noted above, we take Roots to be category-neutral. The members of the typical openclass categories nouns, verbs, and so on are Roots combined with a category-defining functional head n, v, and so on (e.g., Marantz 1997, 2001, Arad 2005). Thus, for example, the noun cat is complex, consisting of a Root and a category-defining n; the latter has the phonological form -j0for this particular Root. (6) The Root \/Cat as "noun" n \/Cat [n,0] We assume that every Root must combine with a category-defining functional head. (7) Categorization assumption Roots cannot appear (cannot be pronounced or interpreted) without being categorized; they are categorized by merging syntactically with category-defining functional heads. If all category-defining heads are phase heads in Chomsky's (2001) sense that is, if they are heads that initiate spell-out the categorization assumption would follow from the general architecture of the grammar (see Marantz 2007). Concerning the functional heads themselves, we assume that there exist different types of n, v, and so on, distinguished by virtue of their feature content (although we will not provide a theory of such features here).1 As we discuss extensively below, competition in this approach is restricted to Vocabulary Insertion, which targets individual terminal nodes in the structure. 2.2 "Wordhood" and Structure Much of the discussion below concentrates on cases of so-called Poser blocking, in which there are apparent interactions between ' 'words" (e.g., smarter, *intelligenter) and phrases (*more smart, more intelligent). In traditional terms, these are alternations between synthetic ' 'one-word" forms and analytic (or periphrastic) "twoword" forms. The theory of Distributed Morphology does not have a primitive notion of "word" directly relevant to cases of Poser blocking (the phonological word may or may not play a role in the operation of phonological rules or constraints). Instead, the notions relevant to analytic versus synthetic expression are structural, and involve how the heads in a syntactic structure are packaged 1 Some literature exists on different v heads and their feature content; see, for example, Harley 1995 and subsequent work in this vein. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 7 for phonological interpretation. We assume that the theory of constituent structure makes clear the notions "head," "complex head," and so on. According to the structural view envisioned here, multiple terminal nodes that are packaged as one complex head by the syntax or PF are 4 'one word' ' in an informal way of speaking, whereas terminal nodes realized as separate heads are, in the same informal way of speaking, "two words." This is illustrated for two heads X, Y in (8) (analytic) and (9) (synthetic), where we take (9) to be the output of head movement. (8) Analytic, "two words" (9) Synthetic, "one word" XP XP X YP X YP Y ... Y X • There are other ways that a complex head could be formed in addition to head movement; these are discussed below. The important point is that the difference between the "oneword" and "twoword" types of expression has to do with the syntactic structure and, in particular, how the heads in the structure are packaged. 2.3 Notions Relevant for Blocking Effects Many of the phenomena that fall under the heading of blocking effects are accounted for by mechanisms included in Distributed Morphology, although in ways that do not necessarily implicate competition-based blocking. As a way of introducing some of the basic points to be advanced later in this article, we present an overview of the relevant mechanisms and their effects here. The process of Vocabulary Insertion assigns phonological content to syntactic nodes. We assume that each node receives a single phonological exponent in this process. (10) Single-Vocabulary-Insertion assumption One exponent per terminal node; that is, Vocabulary Insertion applies only once to a terminal node. Vocabulary items like those in (4) are thus competing with one another, and when one wins this competition, it prevents others from doing so. For example, when -t appears as T[past] in the context of the Root \/Bend, it is at the expense of the default case, which has the exponent -ed. It might be said in this case that -t blocks -ed (more precisely, that the Vocabulary item with -t as its exponent blocks the Vocabulary item with -ed as its exponent). It is crucial to note here, however, that the blocking effect is limited to the phonology of a single node. It is not the case that one word blocks another word: bent does not block *bended. The ungrammatical forms in this example, those with "incorrect" allomorphs of T[past] like *bend-ed are never generated or considered in the derivation of bent. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


8 DAVID EMBICK AND ALEC MARANTZ Beyond the effects of Vocabulary Insertion at a single node, there are other ways in which conceivable forms are excluded from the language on this approach. That is, in considering the operation of the grammar as a whole, we as linguists are able to distinguish different ways in which an ungrammatical form fails to occur, of which allomorphic ''blocking" of the type just described is one possibility. Importantly, our metalinguistic analysis of the ungrammatical forms does not imply that these forms play any role in the speaker's competence. Including allomorphy of a single node, some of the relevant ways that forms can fail to exist, chosen with reference to our discussion below, are these: (11) a. Allomorphically: The structure is well formed syntacticosemantically, but the morphophonology of the language simply does not produce the pronunciation under consideration. Example: [bend T[past]] is well formed, but is pronounced bent, not *bended. b. Syntactically/ Semantically: There are two subcases: one in which a structure is never possible, and one in which the combination of Root and functional structure is restricted but possible in limited cases. i. Example 1: The "potential" (adjectival) head a pronounced -able attaches outside v but not outside, say, n. Thus, ^atrocity able is ungrammatical because the structure Root merged with n, then [Root n] merged with potential head a cannot be derived in the first place. ii. Example 2: Some functional heads have a restricted distribution and only go with a limited set of Roots. For example, while feminine forms exist for certain nouns, such as actress and lioness next to actor and lion, they exist for only a handful of nouns (of the appropriate type: animate and so on). Thus, for any given noun, even of the appropriate semantic class like jaguar, one does not expect that the syntax of "feminine noun" will necessarily be available for that noun. Nothing rules out the structure that would underlie *jaguaress as a whole (since forms like lioness and tigress are possible); but at the same time there are substantial restrictions on the distribution of the relevant functional heads.2 c. Combinatorially (in terms of complex head formation): Some process that creates a complex head may apply under restricted circumstances. When the process does not apply, thfere is no single "word" to consider. Example: There is a rule that combines the degree element Deg and the adjective smart into a single complex head that is pronounced smart-er. The rule that affixes 2 Some clarifications are in order about the "Root-specific" (selectional) effects identified in this classification. By assuming an approach in which Roots are category-neutral, we are not predicting that every Root should appear felicitously in every possible environment, for example, in every different "lexical category" (cf. Borer 2004 for another conception). While a theory of Root-functional head combinations must be part of a comprehensive theory of competence, we cannot provide such a theory here. Any theory able to account for so-called conversion the appearance of the same Root in multiple categories faces the same issues as the present approach. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 9 Deg to adjectives does not apply in the case of the adjective intelligent. Thus, the synthetic form *intelligent-er is not created by the morphophonology; the only way of realizing the syntactic structure "comparative of intelligent' " is with the analytic form more intelligent.3 These, then, are various ways an ungrammatical form can fail to be generated by the grammar. Crucially, for our purposes, none of them involves competition at the level of "words" or "phrases." Instead, the grammatical forms are derived, and conceivable ungrammatical forms are not. We take it that (11) clarifies what it means to say that the only type of competition that takes place in this approach is at the level of the morpheme, where one Vocabulary item can win over another (as in (lla)), and that there are many ways that forms can fail to occur that do not result from competition. 3 Blocking In the previous section, we outlined several ways in which the basic architecture of Distributed Morphology derives effects sometimes associated with blocking. In this section, we examine blocking directly, with reference to specific proposals concerning such phenomena. It is convenient to refer to theories according to certain very general positions they endorse. We take the "standard" approach to blocking to be as follows: (12) Standard blocking Some forms are ungrammatical only because other forms happen to exist and win over them; competition takes place at the level of the word, phrase, or sentence. Whenever we consider approaches in which one form is said to block another, the approaches are of this standard type. As a cover term, we refer to theories where competition takes place at the level of the word or above as competition-based. (13) Competition-based theories In competition-based theories, the forms that compete for expression of meaning are words, phrases, or sentences; that is, these theories endorse standard blocking. We reserve the term Poser blocking for approaches that implement word-phrase interactions in terms of standard blocking. (14) Poser blocking Words win over phrases by standard blocking. 3 Note that every theory of word formation has to say that there is a general rule of comparative formation, and that this rule does not apply to intelligent, because it only applies to words of a particular phonological shape. Thus, in some sense every approach in which words are preferred to phrases must hold that there is no word intelligenter to consider here, as we discuss further with reference to Kiparsky 2005 in section 4.5. Where our approach differs from others is in holding that there is no more smart to consider (and block) either. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


10 DAVID EMBICK AND ALEC MARANTZ Our arguments about blocking and Poser blocking in this section and section 4 proceed through a few steps. These are the major points of the discussion to come: • Following a brief sketch of our take on derivational morphology, we show in section 3.1 that there is no evidence in favor of competition-based theories even at the level of the 4 'word" (the traditional domain of blocking). After paving the way with a discussion of "synonymy blocking" (section 3.2), we examine standard cases of what is supposed to be word/word blocking, like glory I *gloriosity, and show how the relevant patterns can be analyzed without such competition (sections 3.3, 3.4). This of course leaves open the possibility that word-phrase interactions (Poser blocking) are the only instances of standard blocking, motivating the discussion of section 4. • In section 4.1, we outline a generative approach to analytic/synthetic alternations (and "affixation" generally), centering on the primary assumption of such theories: namely, that rules apply when their structural descriptions are met. This paves the way for a comparison with competition-based theories. • In section 4.2, we demonstrate that theories that implement competition at the sentence level to accommodate blocking effects, such as Bresnan's (2001) (and perhaps Kiparsky's (2005)), make incorrect predictions; thus, there is no evidence for competition at this level. • In section 4.3, we show that theories that implement Poser blocking with a principle that words win over phrases under specific structural conditions make precise predictions about the configurations in which analytic/synthetic alternations could occur. • In section 4.4, we show with reference to specific examples that the Poser blocking approaches cannot predict where words and phrases interact in the relevant sense. The generative approach, with syntactic structures and syntactic (or postsyntactic) movement processes, makes the correct predictions. • In section 4.5, we show in addition that there are interesting ways in which the principle "Prefer words over phrases" is problematic in the first place. This article contrasts the approaches to blocking in (12)(14) with the largely syntactic account of blocking phenomena made available within Distributed Morphology. 3.1 Derivational Morphology As noted in the preceding section, we assume a syntactic approach to derivational morphology in which category-neutral Roots combine with functional heads n, v, a (see Marantz 2007 for recent discussion). With x ranging over these heads, we use the term Root x to refer to a structure in which x is the element that categorizes the Root. Thus, our example cat in (6) is a root nominalization (or root nominal)', red is a root adjective (Root combined with adjectival head a); and so on. In addition to attaching at the Root level, it is of course possible for the x heads to attach outside other x heads. These are cases of true "category-changing" morphology. So, for example, This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 11 we assume that the root adjective vaporous can be nominalized by n to yield vaporousness, as shown in (15) and (16). (15) Structure for vaporous (16) Structure for vaporousness a n V Vapor [a,ous] a [n,ness] V Vapor [a,ous] Schematically, we refer to Root-attached x heads as being in the inner domain; outside other x heads, category-defining heads are in the outer domain. A significant set of generalizations that are captured in this approach hinge on differences between the inner and outer domains. These generalizations, which we assume in our analysis of blocking effects below, are summarized in (17). (17) Generalizations a. Allomorphy: For Root-attached x, special allomorphy for x may be determined by properties of the Root. A head x in the outer domain is not in a local relationship with the Root and thus cannot have its allomorphy determined by the Root.4 b. Interpretation: The combination of Root-attached x and the Root might yield a special interpretation. When attached in the outer domain, the x heads yield predictable interpretations. The workings of these assumptions can be illustrated with reference to the forms in (18), which also play a role in our analysis of blocking effects below. (18) curious, curiosity generous, generosity verbose, verbosity The first two are examples of what appear to be -ous adjectives being nominalized with -ity; the third is an adjective ending in -ose. In these cases, we propose an analysis in which -ity is in the inner domain, as in (19). (19) Structure for curiosity n VCurious [n,ity] 4 Some cases of interest, which appear to be sensitive to the identity of the Root for insertion in an "outer" domain, are discussed in Embick 2003. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


12 DAVID EMBICK AND ALEC MARANTZ According to this analysis, the Root is VCurious; our position is that this treatment accounts for significant generalizations that are lost in alternative analyses. In particular, it might seem possible in principle to treat curious as containing the piece -ous, with the Root \/Cury as a kind of cran-morph, as in (20). (20) Structure for curious? a VCury [a,ous] If (20) were the structure for curious, then curiosity would be a deadjectival noun, as in (21). (21) Structure for curiosity? n a [n,ity] \/Cury [a,ous] Given our assumptions in (17), (21) cannot be the structure for curiosity. There are two reasons for this. The first is that, as we discuss below, n is pronounced -ity only when it appears in the context of particular elements, and these elements have to be listed. Since [a,ous], which is visible to n in (21), is not on the list for -ity (see below), the n head could be pronounced -ity only if the Root VCury were in a local relationship with n. This is disallowed by (17). Similarly, curiosity has an idiosyncratic interpretation (a curiosity can be an artifact of specific interest/ strangeness). By (17), then, (21) is not a possibility. Importantly, as we discuss in section 3.4.1, there are other reasons to think that the analysis that takes curiosity to be a root nominalization makes the correct predictions; see in particular the discussion of putative doublets, where we show that the Root in curious in a structure like (21) yields curiousness. Notice that it is also possible, given the pattern in (18) and our assumptions in (17), that there is a monomorphemic -osity that is Root-attached in the nominal forms, as in (22). (22) Alternative structure for curiosity n VCury [n,osity] According to this view, the adjectival forms would involve -ous or -ose, attached to bound Roots. The analysis in (22) would be compatible with (17); the question is whether there is evidence in favor of this alternative as opposed to the one in (19). It might be possible to argue that there is a generalization within English according to which Roots that take an adjectival a head in -ous This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 13 or -ose also form a nominal in -osity, and that the analysis in (22) states this more directly than the one in (19).5 While this is a possibility to be explored, we will assume (19) here, since the further differences are not essential to our discussion of blocking. Before we address blocking, we wish to illustrate another point concerning Vocabulary Insertion in derivational morphology. As noted above, we assume that the n, a, v heads can be distinguished further in terms of their feature content. It seems that this is a necessary component of any theory with the general properties of the one supported here. However, it is also important to note that in some cases it is not clear that differences in feature content are responsible for differences in form and interpretation. In one set of cases that highlight these issues, the same Root surfaces in more than one root nominalization. So, for example, we assume that the single Root VCover forms two different root nominalizations, cover (23) and coverage (24). (23) Structure for cover (24) Structure for coverage n n /\ x\ VCover [n,0] VCover [n,age] At this point, the central question concerns the status of the n heads in these two trees: are they the same, or are they different? If we assume the latter (i.e., that there is some head n! in cover, and some head n2 in coverage), then the effects on allomorphy and interpretation reduce to this difference. This could very well be the correct analysis in this particular case. Or we could assume the former (i.e., that there is one n in these two structures). For this analysis to work, we have to configure Vocabulary Insertion so that the same Root may appear on more than one list, as in (25). (25) Vocabulary items, n inner domain n <-0/LISTj nttt LIST! = VCat\ \/Dog, VCover, . . . n <-age/LIST2 nttt LIST2 = VMarry, V^ond, VCover, . . . In (25), the Root VCovER appears on more than one list. If we treat Vocabulary Insertion in such a way that two Vocabulary items that are not related to one another by inclusion are not ordered, and therefore one cannot always take precedence over the other, then either could be inserted. Thus, in a grammar containing (25) and the "nonordering" assumption just mentioned, both cover and coverage could be derived.6 In this scenario, there is part of the grammar in which cover and coverage are identical. The interpretive difference between these forms arises from the 5 Note that while specious, precious, and impecunious seem to support the generalization that adjectives with -ous take root nominalizations in -osity, other adjectives, such as various, tenacious, and pious, are counterexamples (see (34)). 6 We put aside the question of what would be involved in ' choosing the correct outcome in any particular instance of use. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


14 DAVID EMBICK AND ALEC MARANTZ fact that they are used to refer to different parts of semantic space, perhaps in the way that sofa and couch or (as we discuss below) thief and stealer are in competition at the level of use (not grammaticality). While we will not investigate these different accounts further here, the connection with what it means to have competition for use provides a natural transition to our main topic. Whatever the status of these different approaches to cover and coverage may be, this discussion leads up to an issue that is central to blocking: the difference between competition for grammaticality and competition for use. This is seen clearly in a standard question for blocking theories, the question of lexical relatedness. 3.2 Competition, Lexical Relatedness, and Synonymy Blocking One question about blocking that must be answered at the outset concerns the scope of putative competitions. In particular, what can potentially compete with what? Returning for expository purposes to our example of blocked *gived, note that certain conditions must be present for a blocking relation to obtain between *gived and gave. All accounts of blocking assume that the competing forms must "mean the same thing" in some sense. For most accounts, this entails that the competing forms must be seen as forms of the same Root, the property of lexical relatedness. (26) Lexical relatedness in blocking Competing forms are forms of the same Root. This of course raises issues for apparent cases of blocking brought up elsewhere in the literature, where assumption (26) appears to be denied. For example, can thief be said to block stealer, on the assumption that the latter is unacceptable in some sense? Since thief and stealer do not seem to share a Root, this blocking would seem to require equating the meaning of thief with that of "one who steals" certainly a bleaching of what one would generally call the meaning of the word.7 In any case, there are treatments of blocking effects that extend the notion of standard blocking to the thief/ stealer relationship (Giegerich 2001). Theories of this type warrant further discussion as a means of elaborating our stance. For convenience, we call approaches with competition between lexically unrelated items in this sense theories with synonymy blocking. (27) Synonymy blocking Competing forms simply have to "mean the same thing"; they do not have to share a Root. There are some prima facie difficulties with this position. Given the various specialized uses of thief (see footnote 7), one would imagine that a more general stealer would have room to acquire non-thief meanings, so it is not entirely clear how the blocking relationship would actually work. In general, a notion of blocking based on "meaning" independent of "lexical relatedness" (built on the same Root) could only possibly account for the (nonoverlapping) sharing of semantic 7 That is, thief in the normal sense means 'one who steals professionally or habitually', for example. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 15 space the possible meanings of a form like stealer not the ungrammaticality or nonexistence of a form.8 Marantz (2003) discusses some instances of putative synonymy blocking from the literature and points out that, in cases like stealer, the feeling of ill-formedness that speakers have is arguably independent of the existence of single words like thief. Parallel forms, such as breaker, for someone who breaks things, have a similar status (i.e., are putatively deviant), although there is no "listed" form for someone who breaks things (something like clumsy oaf, klutz, or butterfingers). It seems clear that there is no obvious sense in which breaker is deviant because some other existing word blocks it. While it is thus unclear how a theory with synonymy blocking could connect the effects found with stealer and breaker, these nominals exhibit a pattern that has nothing to do with blocking: transitive verbs of the semantic class including steal and break generally require direct objects, even in their agentive nominal form. So base-stealer is fine, and Web research suggests that password-stealer, girlfriendstealer, scenestealer, and a host of others are in common use. The same is true of breaker, unsurprisingly {rule-breaker, heart-breaker, etc.). These considerations, which implicate the type of Root involved as well as the structure(s) it appears in, require us to look at different aspects of the derivation of stealer. In particular, which structure or structures is this form realized in? For thief, we assume that the structure is that of a "noun"; that is, it is a Root combined with an n, the latter with a null phonological exponent, as in (28). (28) Structure for thief n /\ VThief [n,0] Given the theory of category-determining morphology presented in Marantz 2001, at least two analyses of stealer are possible. One is identical to the structure in (28), where a nominalizing "little n" creates a noun from the Root VSteal; the other is a deverbal agentive -er nominal, in which a Root and a verbalizing head combine and are subsequently nominalized by an n. These two structures are shown in (29) and (30). (29) Root nominalization (30) Deverbal agentive nominal n n /\ /\ VSteal [n,er] v [n,er] VSteal [v,0] 8 Compare "pragmatic" theories of blocking, already critiqued effectively in Poser 1992. Pragmatic approaches to blocking continue to be advanced (e.g., Williams 2007) without addressing Poser's original arguments. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


16 DAVID EMBICK AND ALEC MARANTZ One may ask whether the existence of thief might lead speakers away from the analysis of stealer as a root nominalization (29). If the need arises for a word to describe someone who steals for a living, against the law, the existence of thief might make the creation of a novel root nominalization pronounced stealer unnecessary (i.e., superfluous given that the semantic space for which stealer would be used is already filled). That is, the kind of principle invoked by language acquisition specialists to explain why children spread the words they hear around semantic space, instead of assuming that every word spoken around a dog, say, means dog (cutie, shaggy, tiny) a "uniqueness" principle for sharing semantic space might work against the root nominalization analysis of stealer. See, for example, the Principle of Contrast, discussed by Clark (1993) and others. It might be that what stealer in (29) could be used for depends to some extent on what other words happen to exist in the language. If stealer is used when thief exists, there might be strong motivation ("Contrast") for the hearer to assume that stealer does not mean what thief means. In the absence of a clear use for the root nominalization in (29), the hearer might conclude that stealer is not a root nominalization.9 Thus, (a) the pressure exerted by the existence of thief might make the analysis of stealer as a root nominalization unlikely, given its relation to the semantic space for which it could be used at the same time (also see footnote 10 for other problems); and (b) the deverbal agentive nominal analysis in (30) has problems as well, since the requirement that an object be present is not met. As a result, there is something odd about stealer in some contexts. Crucially, while there is potentially some sort of interaction between thief and the hypothetical root nominalization stealer, it is not competition for grammaticality. Rather, the effect has to do with what a root nominalization stealer could be used for (in terms of semantic space) given that thief exists. Thus, there is no blocking effect that determines what is grammatical and what is not; any effect of thief on stealer has to do with how objects that are generated by the grammar might be employed, not with whether the object in question can be generated in the first place.10 What we find with thief and stealer contrasts with cases involving lexical relatedness. In the case of someone who cooks, for example, the root nominalization cook already exists, beside the more professional chef In the theory under discussion here, cook could "block" cooker, but this is only a manner of speaking. In our view of competition, all of the action occurs at the level of the morpheme, not the word. In this case, this means that the zero nominalizing suffix would win the competition for insertion over -er in the environment of the Root VCook. What is really at issue, then, is the phonological form taken by the n head in (3 1 ), with reference to the Vocabulary items in (32). n 9 Although compare robber, burglar, and so on. 10 As far as the root nominalization analysis goes, our considerations elsewhere might raise the question whether the grammar leads us to expect that a root nominalization with the allomorph -er of n should be possible with V Steal in the first place; see (1 lbii). The answer is probably negative, but the possibility has to be considered nonetheless. 1 ' It could be that VCook is on both lists in (32), in which case further reference to structures and features would be required in order to determine the pronunciation and meaning of root nominalizations formed from this root (consider, This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 17 (31) cook-0 vs. cook-er n /\ VCook n (32) Vocabulary items n <-> -er/Y nttt n <-> -0/X nttt X = VCook... Reiterating our point from above, the same issues do not arise when there is no shared Root; thus, there is no grammatical sense in which chef could block cooker/cook, nor thief stealer. In the literature on blocking, the existence of an ill-formed word and a word that intuitively would "mean the same thing" is often taken as sufficient evidence to substantiate a blocking relation. The cases outlined in section 2.3 illustrate several ways in which some form could fail to exist for reasons that have nothing to do with competition and blocking, as we illustrated for stealer above. Other cases of putative synonymy blocking from the literature can be analyzed in these terms as well. For example, Giegerich (2001) gives pairs like *horsess/mare as cases of blocking. However, there is no reason to expect horse ss for 'female horse' independent of the existence or nonexistence of mare: forms like *turtless are deviant independent of any words we might know to express 'female turtle' (and sticking to the semantic domain of lioness, consider leopardess and tigress but *jaguaress, *pantheress; leopardess exists and *jaguaress apparently does not, but not because there is some other word for 'female jaguar'). There are thus no strong arguments in the literature that a grammatical blocking relation holds between words that "mean the same thing" but do not share a lexical Root.12 (33) There is no synonymy blocking. In the next section, we argue that there is also no word/word blocking relation between words that do share a Root, a point that arises in the analysis of certain facts that are central to all discussions of blocking effects. for instance, pressure-cooker and other examples where cooker is fine). One possibility is that there is no structural/ featural difference at play here, as discussed at the end of section 3.1. If this is the case, then the Vocabulary items in (32) are such that neither wins the competition on the basis of specificity alone, so that either could appear for a Root that appears on both lists (recall section 3.1). 12 Putting aside the question of suppletion: there is an important parallel between went/*goed and gave/*gived that can only be captured by recognizing wen(or went) as an allomorph of the head that is pronounced go. If thief were analyzed as an allomorph of steal, a blocking/competition analysis of a connection between thief and the root nominalization stealer (along the lines of cook-0 versus cook-er above) might be attempted. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


18 DAVID EMBICK AND ALEC MARANTZ 3.3 Blocking in Aronoff 1976 While synonymy blocking can be dismissed in the manner outlined above, there remain cases in the literature involving what is claimed to be (a) blocking between whole words that mean the same thing and are built from the same Roots that (b) do not yield to an analysis involving competition among allomorphs for realization of a functional head. Perhaps the most famous case, found in Aronoff s work, involves the relation between "bare" (i.e., not overtly affixed) nouns, adjectives formed from these nouns with the suffix -ous, and the possibility of further nominalizing the adjectives with -ity and -ness. The existence of forms such as curiosity and viscosity suggests that -ity can attach to adjectives in -ous {curious, viscous) and create abstract nominals meaning, among other things, possession of the quality named by the adjective. A question then arises about -ous adjectives for which this relation apparently does not hold: for instance, if curious goes to curiosity, why does glorious not go to * gloriosityl Why, moreover, is the form gloriousness allowed, unlike * gloriosity! As Aronoff (1976) explains in the first detailed discussion of these types of facts, no simple notion of blocking between words based on meaning will be able to account for *gloriosity, since consideration of the meanings of -ity and -ness would predict that gloriousness would mean what gloriosity would mean; thus, gloriousness should have the same "blocked" status as * gloriosity, contrary to fact. The analysis that Aronoff develops in response to this and some related observations involves components that figure in all subsequent work on blocking effects, as we discuss below. Roughly speaking, Aronoff s generalization about such examples says that if the -ous adjectival form decomposes into an independent noun (e.g., glory) and the affix -ous, then the -ity form is blocked. If, on the other hand, the -ous adjective cannot be decomposed in this way, then the osity form is (potentially) grammatical. Thus, the absence of cury, as a word, allows curiosity. (34) Nominals from adjectives in -ous (Aronoff 1976:44) Xous Nominal -\-ity #ness various * variety variousness curious * curiosity curiousness glorious glory *gloriosity gloriousness furious fury *furiosity furiousness specious * speciosity speciousness precious * preciosity preciousness gracious grace *graciosity graciousness spacious space *spaciosity spaciousness tenacious * tenacity tenaciousness fallacious fallacy *fallacity fallaciousness acrimonious acrimony *acrimoniosity acrimoniousness impecunious * impecuniosity impecuniousness laborious labor *laboriosity laboriousness bilious bile *biliosity biliousness pious * piety piousness This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 19 Aronoff attributes the difference between -ity and -ness to the regularity of -ness compared with -ity. Since -ity forms are not entirely predictable, they must be listed when they do occur.13 What "listing" means is that a slot represented as "Nominal" for the Root in question in (34) contains the output of the word formation rule that assigns -ity. Because their existence is not predictable, "simple" nouns like glory must be listed as well. Blocking then can occur between listed forms, so that, for example, glory blocks *gloriosity; technically, what this means is that glory occupies a slot associated with this Root, so that slot cannot be occupied by an -/ry-affixed form. From the point of view of listed lexical items, the "Nominal" and " + ity" columns of (34) constitute a single lexical slot, for which only one form should exist. At the same time that this blocking relationship obtains between listed forms, productive formations like "Nominal in #ness" in (34) will not necessarily be blocked by words that mean the same thing, at least for this sort of derivational morphology. In terms that we used above in explaining the intuition behind blocking, it is important to note that this approach makes crucial reference to paradigmatic notions: glory occupies the listed spot that *gloriosity would occupy if the latter were formed. Similarly, curiosity would block *cury as a backformation, since curiosity is itself listed and occupies the slot. Four components of this treatment of blocking effects have been highly influential in subsequent analyses: (35) Components of Aronoff 's blocking a. Paradigmaticity: The blocking effect arises because each "lexical item" has associated with it a set of cells expressing different meanings for that lexical item. Each cell may be occupied by (at most) one phonological form. b. Lexical relatedness: The competition that results in blocking is between words that share the same Root. c. Irregularity: Irregularity is crucial to blocking. Only elements that are irregular in some respect must be listed in the lexicon that is, must be recorded in the "paradigm slots" ("The words which must be listed are blocked, and those which must not be listed are not blocked" (1976:45)). Therefore, blocking effects may obtain only between formations each of which is "irregular" or "unproductive." d. Wordhood: The objects that are entered into paradigm slots and thus compete with each other by virtue of blocking each other are words. Our reexamination of word/word blocking centers on these points, and on how each point or the effects it is meant to cover are stated in a theory different from Aronoff s. 3.4 A Reexamination of Word/Word Blocking Recall from our discussion of synonymy blocking that the absence of certain forms in what appear to be blocking situations might actually be independent of any putative competition between the 13 Aronoff relates the irregularity of -ity to the application or nonapplication of Truncation (e.g., atrocious/atrocity but not curious/* curiety). Similar points concerning the interpretation of -/fy-affixed words arise in his discussion as well. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


20 DAVID EMBICK AND ALEC MARANTZ ungrammatical form and the grammatical form. In the case at hand, a similar approach must be examined as well. In line with the general architectural principles that underlie our approach, is there any reason to expect, say, gloriosity in the first place? That is, would the form that has to be blocked in theories with word/word blocking ever be derived by the grammar? More specifically, in our framework the question is whether either of the objects in (36) and (37) is found in English. (36) Structure 1 (37) Structure 2 n n VGlory [n,osity] a [n,ity] \/Glory [a,ous] We address points related to each of these structures in the discussion below. We argue that there is no reason to expect gloriosity as the phonological form for either of them. The relationship between glory and gloriosity is thus not an argument for blocking at the word level, since the properties of gloriosity can be accounted for independently. Generalizing, our conclusion is as follows: (38) There is no word/word blocking between lexically related words that "mean the same thing." As discussed in section 2, Distributed Morphology allows competition for the phonological form of individual nodes. Competition among larger objects for example, word-word competition of the type proposed by Aronoff (1976) and others following him cannot be formulated in the theory. Taking up points from section 2 and earlier parts of this section, we demonstrate first that there is no need to prevent the * gloriosity in (37) via competition, given the correct analysis of English derivational morphology (section 3.4.1). When the structures and their allomorphic properties are understood, * gloriosity is not derived. A second question, considered (and answered negatively) in section 3.4.2, is whether or not some evidence for word/word blocking can be gleaned from occasional attestations of gloriosity or a hypothetical root nominalization glori-osity (as in (36)). 3.4.1 Accounting for Aronoff s Observations without Word/Word Blocking Beyond the (negative) conclusion in (38), there are stronger and more interesting things to say about glory/ gloriosity and related cases that support the notion that grammatical competition is waged at the morpheme level, between Vocabulary items competing for insertion once we recognize the full force of Aronoff s observations. Our approach is based on identifying possible structures (combinations of Roots, n heads, a heads, etc.) and, further, the phonological exponents of the functional heads. Two aspects of This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 21 the distribution of the -ity allomorphy of n must be taken into account. The first is that in a particular domain, where an n is attached to a Root, -ity appears sporadically that is, with a certain set of Roots that simply must be listed. Thus, we find root nominalizations like atroc-ity and curios-ity, but there are also many, many other root nominalizations without -ity. This is only part of the picture, however. As explored further in Baayen and Renouf 1996, for example, while the distribution of -ity is highly restricted in what for us is the root nominalization structure, -ity is in fact productive in another context, attaching to adjectival-forming a heads with the exponents -able and -al. Since -ity has the listed property that it attaches to these suffixes, it will generally win over -ness as the realization of the relevant nominalizing node, yielding a preference for, say, categorizability over llcategorizableness. The Vocabulary item with the exponent -ness functions as a kind of default for n, as shown in (39). I4 (39) Vocabulary items n «-» -ity/X nttt X = Roots (\/Atroc, VCurious . . .); [a,able], [a,al] n <-ness As outlined above, for gloriosity there are two different structures to consider: one in which the nominalizing head n is attached to the Root, as in (40), and one in which it is attached to an adjective (Root combined with adjective head a), as in (41). (40) Root attachment (41) Outer domain attachment n n VRoot n an VRoot a As shown in (39), one of the possible allomorphs of n in (40) is -ity; this is a listed property, which is correlated with the presence of certain Roots, but not others (there are a number of other allomorphs of n in this context: compare cat-0, marri-age, act-ion, etc.). The Root VGlory is not on the list in question. Thus, there is nothing in the grammar of English that leads one to expect *gloriosity (or for that matter *glority) with the structure in (40); see also section 3.4.2. In any structure like (41) in which the n attaches outside a, some aspects of what happens at n are determined by the properties of a. With some a heads (e.g., the one with the -j^allomorph 14 Further specification might be required here, referring to structural properties of the n in question, and in particular to whether it is attached directly to a Root or to another category-defining head. For instance, -ness is a kind of default for n only outside other category-defining heads, not in the Root-attached domains. See Embick 2003 for discussion of how these conditions are relevant to the specification of Vocabulary items. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


22 DAVID EMBICK AND ALEC MARANTZ in curious-0, or the one with the -ous allomorph in glori-ous), the n head defaults to the phonology -ness. In the case of the ' 'potential" a head with the exponent -able, the situation is different; the -ity allomorph of n is strongly preferred. In other words, the grammar of English provides no reason for a speaker to suppose that gloriosity from glorious [y Glory [a,ous]] exists, because there is no reason to expect to find -ity in this case. The problem here is not with the syntactic structure. The grammar generates a structure in which VGlory combines with an a head, this structure then being nominalized by n, as in (42). (42) Structure for the deadjectival nominal of \/Glory n /\ a n /\ VGlory a The point is about allomorphy. For gloriosity, the question here amounts to what phonological form is assigned to a and n. For a, the answer is clearly that in (42) this head is pronounced -ous. The real question regarding gloriosity is about n, and the answer is that in (42) it is pronounced -ness. Putting these observations together, the structures for the cases under discussion, with exponents of category-defining heads included, are as shown in (43)-(45). (43) Structure for (44) Structure for (45) Structure for glory glorious gloriousness n a n \/Glory [n,0] \/Glory [a,ous] a [n,ness] /\ VGlory [a,ous] The word gloriosity with the structure in (45) is not blocked by glory; it is simply not derived. In some sense, then, the ill-formedness of ^gloriosity is similar to that of *jaguaress. Neither of these forms is marked ungrammatical by virtue of being blocked by something else. Instead, each is deviant for independent reasons. However, the precise sense in which the two cases are deviant is not the same, in terms of a distinction made in section 2. In the case of ^gloriosity, although the hypothesized structure that in (42) is well formed, what is found in that structure is a different allomorph of n. In this sense, ^gloriosity is like *gived. In the case of *jaguaress, on the other hand, it seems that in general there is no productive way of expanding the paradigm space so that any noun (from Roots of the relevant semantic classes) can be affixed to form a This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 23 "feminine noun." The latter type of fact is independent of allomorphy, in the sense that the "paradigm slot" understood here as a particular structure produced by the syntax is not created in the first place. As an interim conclusion, then, we have an account of why *gloriosity does not exist, one that makes no reference to the existence of glory. With this analysis in hand, we now have a nice account of "doublets," cases in which there is evidently no strong preference for either -ity or -ness: curiosity! curiousness, ferocity/ ferociousness, verbosity I verbousness, and so on. No competition is relevant in apparent doublets like these because the -ity form is built on a category-neutral Root, as shown in (46), while the -ous form is built from the adjective (Root combined with adjectival head a), as shown in (47)-(48). (46) Structure for (47) Structure for (48) Structure for curiosity curious curiousness n a n /\ /\ /\ VCurious [n,ity] Various [a,0] a [n,ness] /\ X/Curious [a,0] This is most apparent in cases like feroc-ity/feroc-ious-ness cases of so-called Truncation where -ousl-os is not part of the Root and one can see -ity attaching overtly outside an object that is phonologically different from what -ness attaches to. This is illustrated in (49)-(51). (49) Structure for (50) Structure for (51) Structure for ferocity ferocious ferociousness n a n VFeroc [n,ity] VFeroc [a,ious] a [n,ness] \/Feroc [a,ious] But the generalization covers cases like curios-ity I curious-ness as well, where -ousl-os is included in the stem form. The doublets stand beside pairs like *gloriosity/gloriousness, with preference for the -ness forms, and pairs like categorizabilitylV.categorizableness, with a strong preference for the -ity forms. Where there is competition for realization of the nominalizing head in a single structure, a winner emerges (-ity outside -al and -able and a set of listed Roots; -ness as default, as in (39)). The relevant pairs behind the doublets do not involve competition, because the two members of the pair have different syntactic structures: one in which the n head is attached to This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


24 DAVID EMBICK AND ALEC MARANTZ the Root, and one in which it is attached to an a head.15 Different Vocabulary items win the competitions in these distinct structures. Note that the account proposed here is still consistent with Aronoff s original observations, since -ity outside -al and -able functions as -ness does that is, as "productive" or predicted morphology. It should be exempt from the Aronoff-style blocking to the extent that -ness is. However, as implemented in our theory, local competition for insertion goes beyond Aronoff s blocking between listed words, as it includes cases of "productive" affixes ordered by the specificity of their insertion environments. The productive insertion of -ity in the environment of -able and -al, preferentially over -ness in these cases, illustrates that local competition produces apparent "blocking" effects even in the case where neither competing word needs to be listed.16 3.4.2 Additional Considerations We have provided an account in which the structures where gloriosity might conceivably appear do not receive that particular phonological form. There are two further aspects of gloriosity to consider. The first is that instances of gloriosity are indeed attested. The second returns to questions about word-word competition: to the extent that something like gloriosity exists, what does this imply for word/word blocking theories? We show that the possible structures for gloriosity are quite restricted; for the speakers who have this form, it appears to be a kind of emphatic of glory (following in some sense the general principle that emphatics are larger than their nonemphatic counterparts). Moreover, word/word blocking accounts, to the extent that they predict anything, seem to predict that glory should block this emphatic gloriosity more strongly than a deadjectival gloriosity. This prediction is not supported by the facts. Some of the groundwork for the first part of this discussion is found above in section 3.1. If the grammatical system is explicitly comparing glory-0 and glori-osity, this comparison can be reduced to a comparison similar to the one discussed earlier with reference to cook-0 and cook-er. The zero form of n exists for glory and cook, and thus the motivation for a new root formation, gloriosity or cooker, is not present.17 That is, given (52), there is pressure (in terms of "semantic space") for gloriosity to be analyzed as something other than (53); if it is analyzed with that structure, it should have to be different from glory in some salient semantic sense. 15 If there is any competition when the same structure is at issue, then it is along the lines sketched at the end of section 3.1 with reference to cover and coverage (recall also footnote 11). 16 "Blocking effects" between two entirely predictable affixes are found in other domains as well. Take passive prefixes in the Hokan language Seri (Marlett and Stemberger 1983, Carstairs 1990). This prefix has the allomorph -pwhen it occurs before a vowel-initial Root, and -a:?elsewhere. (i) -p-esi 'be defeated' -a:?-kasni 'be bitten' The form of the passive prefix is entirely predictable throughout this system; there is no need for either type of competitor to be listed (i.e., we expect passive prefixes on nonce verbs to be treated in an entirely productive fashion). Nevertheless, only one of these prefixes appears with any given verb, "blocking" the other despite the absence of listedness or unpredictability. 17 There is a third option parallel to part of what was considered for curiosity above, one that would also require the Root to be analyzed so that it has an allomorph glorious all in one piece. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 25 (52) Structure for glory (53) Structure for gloriosity (root nominalization) n n /\ /\ VGlory [n,0] VGlory [n,osity] The alternative in (53) relies on the hypothesis that in some varieties of English -osity is a single affix (i.e., a single piece). Informal Web research reveals that -osity might be a productive affix in at least some dialects of English (see, e.g., the Bangles' Babe-Osity album).18 Gloriosity is used not infrequently; its primary meaning is religious, referring to the glory of a deity. Note that the primary use of glorious is connected not to the type of glory that accrues via actions or deeds but to beauty and splendor, as in glorious morning. Glory is most often attributed to people, while glorious is associated with, say, weather, or with physical beauty in people. In its most frequent uses (as far as Web searches and intuitions reveal), gloriosity relates to glory and not glorious, suggesting that it involves an -osity suffix. One can think of this glori-osity as a kind of ' 'emphatic" version of glory. This means that to the extent that gloriosity exists, it does not have a structure in which a Root combines with an a head and then an n head; rather, it is a Root combined with an n (perhaps an "emphatic" n, in the structure in (53)). For the purposes of how blocking works, simple word/word blocking theories might suggest stronger blocking of gloriosity by glory on this reading than of gloriosity by glory if the former were analyzed as being created via affixation of -ity to the adjective glorious, with the meaning 'the state of being glorious, as of a morning or a beautiful person'. As noted above, glory does not occupy the semantic space of a gloriosity formed from the most common meaning of glorious, which would be the abstract property related to the basic meanings of glorious. On the other hand, glory does seem to occupy the semantic space of the actual uses of gloriosity (or is closer to it). For a theory with word/word blocking, then, the relationship between glory and gloriosity does not go as predicted. If anything, such theories predict that an existing gloriosity with a meaning "close" to that of the noun glory should be blocked by glory, more than a gloriosity related to glorious would be. This prediction is not borne out. Thus, even when we look further into gloriosity, it does not seem to provide an argument for word/word blocking; if anything, the patterns go in the opposite direction. Close examination of the uses and meanings of glory, glorious, and gloriosity provides no support for a word/word blocking approach to the apparent ill-formedness of gloriosity. 3.5 Synopsis The main line of the argument above is that there is no evidence for word/word blocking. The change in perspective motivated by our reanalysis puts a new light on each of the components of the standard treatment of blocking described in (35). 18 In our dialect(s), uses of -osity in the relevant sense are only tongue in cheek. It is quite possible that some attested "productive" uses of -osity have this status as well, to judge from initial Web searches. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


26 DAVID EMBICK AND ALEC MARANTZ 3.5.1 Paradigmaticity All reasonable accounts of blocking are paradigmatic in this sense: they assume that sets of syntactic or semantic features (of the sort usually assumed to be expressed by affixal morphology and/or function words) establish a space of discrete cells or slots for various forms of lexical stems or Roots. In theories such as the Distributed Morphology framework adopted here, the paradigm space generated by grammatical features is a virtual one: it derives from the inventory of functional heads and the generative process. The cells and the shapes of paradigms play no direct role in the grammar. For blocking, local competition for insertion of Vocabulary items into terminal nodes explains why a given paradigmatic "cell" lacks multiple residents, without direct reference to the paradigm. The syntax generates a structure, and morphophonology provides it with a phonological form. 3.5.2 Lexical Relatedness To the extent that there is something to say about competition effects, it is in the case of two forms that share the same Root, which can under specific circumstances be seen as competing for the same "slot." In English, the fact that verbs have past tense sets up a potential slot for the past tense of any verbal stem. The notion of two forms "meaning the same thing" reduces to the notion that two forms built on the same Root or stem potentially express the same set of features that is, fill the same cell or slot in a paradigmatic space generated by the possible meaning distinctions expressible morphologically or syntactically in a language. In our theory, in which paradigms are not reified as objects of the grammar, "filling the same slot" amounts to appearing in the same syntactic structure. As we showed with pairs like *gloriosity/ glory, there is no strong evidence that a competition exists between a potential word, gloriosity, and the existing word, glory, in a way that must be invoked to account for the apparent ill-formedness of the former. Rather, the relative ill-formedness of *gloriosity can be explained in terms of generalizations about the distribution of the n head with the exponent -ity, without invoking any sort of competition with glory. To use these generalizations about -ity in the case of *gloriosity, it is necessary to examine the nominalization of a complex form, glori-ous, that contains glory. The competition between -ity and -ness is won hands down by -ness when the nominalizing suffix attaches outside the adjective-forming -ous. Thus, lexical relatedness, internal structure, and "selectional" information about Vocabulary items are all relevant to determining when competition between Vocabulary items yields a preference for one form over another. However, the explanations here do not require or suggest blocking relations between whole words that "mean the same thing." 3.5.3 Listedness Does blocking require "listed" forms taking precedence? Evidently not, as the discussion of -ity and -ness outside of -able and the like shows. What is crucial for the blocking effect is competition for insertion at a single node, not whether or not the output of some process is "listed."19 19 The same considerations arise in the case of Poser blocking as well; both the phrasal more Adj and the wordlevel Adj-er are entirely regular, modulo the phonological restriction on adjectives with -er. Speakers tested on novel adjectives like wug judge that if John is wug, then Bill is wugger than John, not Bill is more wug than John. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 27 3.5.4 Wordhood The one aspect of the standard approach to blocking that remains to be examined in detail is the "wordhood" assumption in particular, the question of whether blocking effects exist between objects larger than the word and therefore outside the lexicon. Following Poser's (1992) lead, most approaches that consider analytic/synthetic alternations try to make such interactions look like "standard" blocking, in which a single word takes precedence over an analytic expression. In the next section, we examine such approaches and the alternative provided by Distributed Morphology. 4 Word-Phrase Interactions and Blocking In the previous section, we explained why there are no strong arguments for standard blocking at the word level. Nevertheless, data like those discussed by Poser (1992) and others apparently involving word-phrase competition might still require standard blocking, perhaps with an additional principle that smaller objects block larger ones. In this section, we argue against approaches that extend Poser's line of reasoning about words competing with phrases, and we show that a syntactic approach (Distributed Morphology) without competition among larger objects makes the correct predictions. 4.1 No Competition: Rules Apply Our approach to word-phrase interactions is a "traditional" generative one, one that became unpopular when theories began to exploit an assumed split between lexical and syntactic derivations. If the same syntactic structure is found in both phrasal and single-word expressions, the apparent preference for the single-word expression, where possible, falls under the generalization that Rules Apply. (54) Rules Apply Perform a computation when the structural description of the rule is met. If there is a rule of affixation a rule that adjoins one head to another that applies in a particular structural context, it will create something that is pronounced as "one word." However, the operative principle is not "Words are better than phrases"; it is (54). For instance, Lowering of the Tense node (T) to the verb in English is a process that applies when T is in the appropriate syntactic configuration with respect to the verb. It is "obligatory" in the sense that phonological and morphophonological rules are obligatory. Further aspects of the past tense cast light on other facets of this type of approach. At least since Chomsky 1957, a standard analysis of tense in English puts tense features in a functional head higher than the phrase (VP) in which a main verb might be generated. This positioning of tense features captures (a) the fact that tense has sentential scope, not just scope over the verb to which it attaches, and (b) the facts concerning the distribution of tense morphology in questions, negative sentences, and so on. Abstracting away from various details in competing generative analyses, the relevant syntactic tree for a sentence like John walked to the store is something like (55), where some details (e.g., subject copy) have been ignored. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


28 DAVID EMBICK AND ALEC MARANTZ (55) Structure for John walked to the store TP DP T A ^\ John T[past] vP v VP VWalk v ... to the store Following the standard analysis, main verbs in English do not raise to T; instead, T must be lowered. (56) T-Lowering T lowers to v. This rule a kind of merger process applies when its structural condition is met. Following earlier work on merger operations, Embick and Noyer (2001) hypothesize that a kind of merger called Lowering, an operation that relates a head to the head of its complement, applies in English as in (56). As Embick and Noyer discuss, syntactic affixation (adjunction; ' 'morphological merger") can be of different sorts, governed by different structural conditions, depending on the stage in a derivation at which the affixation occurs. The "rule" is an instantiation of the most general recursive operation of the grammar, merger, as realized within the parameters dictated by the properties of the structures in which it applies. In the case of T-Lowering, the target for T is a verb (i.e., v) and the locality conditions for merger are met if no head intervenes between T and the verb; when a head X intervenes, as in (57), T-Lowering does not apply. (57) No T-Lowering T T XP X vP v y/? A This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 29 So, for example, when a negation head Neg or an emphatic node 2 (found in / DID (too) walk to the store) intervenes between T and v, merger does not apply, because T and v are not in the configuration specified in the rule. In such cases, the T head is pronounced as an affix on a dummy v, the light verb do (see Embick and Noyer 2001 for proposals concerning the relationship between dtf-support and Lowering). It is crucial to note that this type of analysis involves no duplication of the lexical and the structural. Every theory requires some account of past tense formation. Standard lexicalist accounts rely either on a lexical rule of tense affixation or on a general notion of affixation or merger in the lexicon that allows compatible morphological pieces to get together and percolate their features to the word level. On the Distributed Morphology account, whatever one says about the structural details that define the affixation of T, it is always mediated by a structure that is essentially syntactic. This is a simple consequence of the architecture of the theory, which has no lexicon where complex objects can be built in the first place. There is therefore no sense in which we ask, faced with a complex word apparently consisting of a stem and an affix, whether the affixation was the result of a syntactic "merger" or of lexical affixation. All affixation results in a syntactic structure, one that displays the arrangement of syntactic features necessary for the derivation of the syntax, semantics, and phonology of the relevant word. The differences between "standard" head movement and the different "lowering" operations discussed in Embick and Noyer 2001 (see below) crucially involve the sensitivity of word formation to the syntactic environment of the words. The same factors must be taken into account on any approach. Thus, on some lexical treatment of word formation, the difference between how T and the verb relate to one another in, say, English versus French might be recast in a different way, but some difference has to be acknowledged. Thus, the Distributed Morphology approach is not under any special burden of proof because it posits different movement operations for affixation. Now consider an approach that sets the phrasal did walk in competition with the lexical walked to compute the proper realization of the sentence John walked to the store. As far as we can tell, only a system like the one made explicit in Bresnan 2001 would be able to implement this sort of competition formally (see section 4.2.2). The reason is that the putative competition between lexical and phrasal expressions implicates the presence or absence of negation, emphasis, and/or interrogative structure, such that it is difficult to compute the notion of the availability of the phrasal or lexical expression of tense without a more global consideration of the expression of negation and interrogative force, as Bresnan makes clear. What is needed, then, is some set of constraints on the possible realization of tense, negation, and so forth, within the sentential domain. One can imagine a variety of constraints on the expression of negation that would cause John didn 't/did not walk to the store to win over *John not walked to the store and *John walked not to the store. The important point here is that something like Bresnan' s system seems to be required on this type of competition theory, and the details of the analysis of negation, emphasis, and interrogative force must be made explicit and defended against the treatment in the standard generative approaches; it is this challenge that Bresnan' s work rises to. The generative position here claims that what one needs to say anyway in an attempt to generate the possible competitors for a global competition analysis already yields just the grammatical structures. That is: This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


30 DAVID EMBICK AND ALEC MARANTZ (58) Generative position The space of competitors ' 'meaning space," "paradigm space" must be generated in syntactic derivations. These potential competitors are a. grammatical, and b. operated on phonologically to yield "pronunciations," and at LF to yield "meaning." For the purposes of word-phrase interactions, the relevant points center on how syntactic affixation occurs in a single structure that could in principle yield one word or a phrase (more than one word). As noted earlier, our version of the generative approach recognizes different means by which one terminal can be affixed to another (derived from work on "(morphological) merger"; Marantz 1984, 1988). Beyond head movement, there are two other operations that can package terminals into a complex head. The example of packaging that we employed above, the much-discussed case of tense in English, is an instance of Lowering in the terminology of Embick and Noyer (2001). Formally, a head lowers to the head of its complement. This operation is the "downward" version of another complex-head-creating operation, "standard" head movement.20 A third type of operation is defined in terms of linear adjacency rather than hierarchical notions of headedness. This operation is referred to as Local Dislocation in Embick and Noyer 2001: it affixes one element to another when they are linearly adjacent. We summarize these operations in (59) (with linear order inside complex heads irrelevant; the strikethrough in (59b) is for expository purposes). (59) a. Head movement (Head X raises and adjoins to head Y) YP Y XP X Y X ... b. Lowering (Head X lowers and adjoins to head Y) XP X YP /\ Y Y X 20 Questions have been raised about the status of head movement; see Chomsky 2001. Matushansky (2006) treats the process as movement to a specifier position that feeds a lowering operation. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 31 c. Local Dislocation (X adjoins to Y under linear adjacency) X~Y -> [[Y]X] Each of these processes crucially involves syntactic notions of locality and intervention, as shown in our initial discussion of T-Lowering above (in the case of Local Dislocation, the relevant linear notions are derived from a syntactic structure). We discuss the implications of these locality effects in greater detail below. Our central claim is that a theory with syntactic affixation correctly predicts the possible range of word-phrase interactions (i.e., that they occur in the structural configurations implicated by the movement operations above), whereas competition-based theories of different types do not. 4.2 Competition Theories Two types of competition theories will be considered here. One type implements competition among multiword expressions (phrases, clauses, etc.), as mentioned with reference to Bresnan 2001 above. These we refer to as sentential competition or global competition approaches. The second type of competition theory, with Poser blocking, implements Poser's (1992) original insight: that there are specific nodes in a structure at which comparisons between words and phrases can be made, with a word winning over a phrase when the two ''mean the same thing." We outline some key differences between these types of competition theories before comparing them with the generative approach. Within competition theories, one important question is whether or not the grammar contains a general principle regulating how words and phrases interact with one another. One global competition theory, proposed by Kiparsky (2005), develops the intuition that blocking is governed by two competing constraints on grammatical expression: they should realize as much of the information to be conveyed as possible, and they should be as economical as possible in expressing information. These constraints favor expressions with fewer morphemes and, perhaps, singleword over phrasal expressions, if the notion of economy can be tweaked in the appropriate way. These considerations by themselves do not favor stored, irregular forms over generated, regular forms when these are equally complex. The Poser blocking theory of competition is similar to the global competition theory except that it is more explicit about how information is preserved in competition and about how economy of expression is defined. On this view, the lexical always takes precedence over the phrasal. Putting pieces together according to the general constraints of a language (say, the phrase structure rules) generates feature structures via conventions of feature percolation and combination. Two generated trees can be compared at any node. If a node in one tree has the same syntactic features as a node in the comparison tree, one can then ask about the number of morphemes dominated by the node used to create this feature structure. If one tree has a single word expressing the feature structure at the given node but the other tree has a complex structure of words expressing the same feature structure, the single-word expression blocks the multiple-word expression, rendering it ungrammatical. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


32 DAVID EMBICK AND ALEC MARANTZ Because the theories we discuss have different takes on the matter, we must ask whether the principle that words are better than phrases (all other things being equal) is an important part of the grammar. We refer to this hypothetical principle as Lexical Preference. (60) Lexical Preference Use a word instead of a phrase when they both express the same features. (That is, use a phrase exhausted by one lexical item over a phrase that contains more than one lexical item.) The status of this principle figures prominently in the discussion below. For present purposes, one major difference between the two types of competition theories involves situations where it seems that a less expressive structure wins over a more expressive structure that is, where considerations of blocking seem to yield a sentence that says less than it should. A crucial example here is Aren't I lucky?, expressing what the otherwise ungrammatical Amn't I lucky? would express. Are is underspecified for person and number compared with am, but it seems as though the unavailability of the form amn 't causes the less expressive structure to be chosen here. Optimality Theory (OT) machinery can be employed, as in Bresnan 2001, to compute the competition among various structures expressing a set of syntactic features; OT assumptions allow for the possibility that some features will not be expressed in the winning candidate. This type of consideration weighing expressiveness and factors surrounding lexical gaps can be treated under the Kiparsky-style blocking but falls outside the scope of Poser blocking approaches, since no straight competition between the lexical and the phrasal is involved. The second important difference between the two competition approaches to blocking is that the Poser blocking theory limits blocking locally to a node in a syntactic tree; in fact, competition must be assessed at every node in a tree, if the theory is to be fully generalized.21 A Poser blocking approach thus makes precise predictions about the locality of competition effects, predictions that differ crucially from the ones our theory makes. While Poser blocking theories tightly circumscribe the locus of possible effects, this is not the case in a sentential competition theory weighing general constraints about the economic expression of features, where the domain of competition depends on the domains in which sets of features may be expressed, as well as on possible interactions among the ways that features are expressed. Bresnan (2001) implements these sorts of ideas via competition at the sentential level, where the competition is not between a singleword expression of the sentence and multiword expressions but among various multiword expressions. In sections 4.2.1-4.2.3, we exemplify global competition theories and show that they make incorrect predictions. 4.2.1 Kiparsky 2005 Kiparsky (2005) hypothesizes that two constraints are operative in the domain of blocking phenomena. These constraints are supposed to exhaust what there is to say about blocking effects and are evidently operative in all cases to which the term blocking has been applied. 21 In Poser's original conception, competition is constrained to nodes dominating heads. Full comparison of competition and noncompetition approaches requires generalizing this notion. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 33 (61) Kiparsky's (2005) constraints a. Economy: Avoid complexity. b. Expressiveness: Express meaning. The first constraint requires that the "simplest" form be chosen; the second, that meaning be expressed in the appropriate way (i.e., that the expression appear in the correct paradigmatic slot, to use our language from above). To operationalize the notion "simplest" relevant for the first constraint, Kiparsky suggests a procedure that counts morphemes, noting that other metrics could be conceived of. For our purposes, we will simply assume that there is a way of formulating Lexical Preference along these lines, since what is at issue is how words and phrases interact with one another. Kiparsky's theory should prefer, say, smarter over more smart by Economy (i.e., by that theory's version of something like Lexical Preference). This is something that blocking theories of many types could accomplish. What happens in other cases is more revealing. Part of what Kiparsky has in mind is a conceptual preference for a "uniform" conception of blocking effects, based entirely on competition. He considers the view of blocking expressed here incorrect on the grounds that Distributed] M[orphology] in effect stipulates blocking twice: once by positing that merger processes are obligatory an undesirable stipulation in itself and secondly as the Subset Principle. (2005:1 18) (The Subset Principle is one way of formulating the idea that Vocabulary items compete for insertion at a single node.) The claim here is that Distributed Morphology is missing a generalization by saying that some forms fail to occur because of allomorphic competition (e.g., *giv-ed), while other forms fail to occur because of the nature of packaging processes like merger (e.g., *more smart). The nature of the objection is conceptual, in the sense that the problem is supposed to be that a "uniform" notion of blocking is preferable. In this light, it is instructive to consider that Kiparsky's treatment accounts only for one case under consideration: blocking of *more smart by smarter. Because the grammar generates the synthetic form here, there exists a word that means the same thing as the phrase; and because the former is shorter in the relevant sense, it wins over its analytic competitor by Economy. The same cannot be said about the relationship between more intelligent and * intelligenter, however. Kiparsky does not consider this case. If the system generates * intelligenter, or if it exists as a possible output form, then it must, by Economy, block more intelligent it is smaller in the correct fashion, by virtue of being a word, and thus should block the phrase. To avoid this result, it must be the case that intelligenter is impossible in Kiparsky's system for independent reasons that is, reasons other than the interplay of Expressiveness and Economy. These principles might be stated as they are in other accounts; specifically, intelligenter is underivable because it violates the phonological conditions that govern the acceptability of synthetic forms. Whatever the relevant independent conditions are, the point is the same: Kiparsky's account "stipulates" blocking effects twice; as far as this sort of arithmetic is concerned, his proposal and ours are similar. One type of blocking arises when the two constraints interact, and another type because certain forms are ill formed or underivable for other reasons. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


34 DAVID EMBICK AND ALEC MARANTZ What is at issue, then, is not a conceptual preference for one versus two principles ("stipulations"), but which approach makes the correct predictions about the range of analytic/synthetic alternations: the competition-based theory or the generative one. In the earlier discussions of English comparatives and the English tense/verb system, we showed that a competition-based blocking approach to word-phrase interactions must make explicit assumptions about the possible phrasal and lexical expression of comparative, tense, negation, and question features, and the like the whole range of features that play a role in apparent cases of Poser blocking. Since Kiparsky is vague on the general assumptions about syntax that lie behind his approach, we need to turn to a more fully realized system, that proposed by Bresnan (2001), to compare the generative and competition-based hypotheses. 4.2.2 Bresnan 2001 The insight behind Kiparsky 's approach to the contrast between smarter/ *more smart and *intelligenterlmore intelligent is twofold: a single-word expression is preferred over a phrasal expression, but when the single-word expression is unavailable, the phrasal expression emerges. Within the generative approach, the same insight is captured without referring to a preference for single-word expressions (Lexical Preference). Rather, the syntactic structure that underlies both single-word and phrasal expressions yields the single-word expression when a rule's conditions are met, and a phrase when the conditions are not met. One question that ultimately must be addressed is whether there are clear arguments for adopting Lexical Preference or not. We will examine the exact status of that principle later, with respect to more intricate examples. For the moment, we will continue our examination of generative versus competition-based theories by asking whether the global competition approach and the generative approach can be shown to differ in their predictions. Bresnan (2001) points out a strong prediction of the competition-style approach that, if proved accurate, would separate it from the generative approach. Bresnan argues for her global competition analysis by claiming that it can explain how an apparent gap in the lexicon can cause an otherwise blocked phrasal expression to emerge. Generative theories do not make this prediction. In a generative approach, a structure has whatever status it has independent of any other structure that the grammar may happen to derive. In a competitionbased framework, on the other hand, the deviance of some structure could result in another popping out, that is, becoming more grammatical than it would be otherwise. In an example that Bresnan examines in detail, she argues that the absence of amn 't allows the possibility of Am I not (verb) ... ? in some dialects of English, as shown in tableau (62). (62) Analytic expression emerging Input: interrogative wide negative lsg This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 35 Note that the input in this tableau refers to the person and number of the subject, as well as the scope (sentential, not constituent) of negation. The ranking of the various *Neg constraints encodes the preference for lexical expression of negation (Neg-I) over phrasal, and a preference for low expression (Neg-VP) over high (Neg-C). The constraint Lex enforces language-specific lexical gaps in the case at hand, the fact that amn't is disesteemed (i.e., does not exist). If the input were 3sg rather than lsg, a candidate in which negation and the auxiliary are expressed together would not violate Lex (since isn't is all right), and thus *Neg-VP would rule out Is he [not working]? in favor of Isn't he working?. In the lsg case, on the other hand, anything with amn't violates Lex; the result in terms of the ranking in (62) is that analytic Am I not ... is optimal. The amn 't lexical gap thus allows the emergence of an otherwise dispreferred Neg-VP expression of sentential negation. On a generative approach, a structure's grammaticality is not determined by considering alternative realizations of the same set of features. Therefore, the status of Am I not working? must be determinable in a way that is independent of the lexical gap that results from the illformedness of amn 't. Here, the generative approach apparently makes the correct prediction, while Bresnan's theory produces the wrong results. The competition-based approach with popping out predicts that if a dialect of English allows Am I not working? only because amn't does not exist, then in this dialect Am I not working? should be better than Is he not working?, since the latter should be blocked by Isn't he working? given the constraint ranking shown in (62). Bresnan provides no evidence for dialects in which Am I not ... is preferred to Is he not . . . in this way, and our own research fails to find such a preference among speakers who allow Am I not ... or prefer it to Aren 't I . . .?. The general point here concerns the popping out effect; it does not seem to happen the way theories like Bresnan's predict that it should. For past tense, it is easy to demonstrate that Bresnan's predictions are not borne out. It is not the case that the existence of a gap renders some other (otherwise deviant) form grammatical. For instance, the lexical gap in the past tense for forgo (i.e., the deviance of "forwent) does not make "did forgo grammatical, as Bresnan's account predicts. For us, "forwent is ill formed. At the same time, */ did forgo dinner is not improved (or better than */ did consume dinner), despite the ill-formedness of "I forwent dinner. Nor is the absence of stride's participle "stridden associated with a grammatical analytical form that is otherwise impossible (e.g., has done stride, or whatever else it might be). In general, there is no evidence that an apparent gap of this sort in the lexicon improves a conceivable phrasal alternative. On the generative account, this is because the structural constraints on lowering force the operation of merger, so that the phrasal "alternative" is never generated in the first place (i.e., there is no phrasal alternative, because the grammar is simply not structured to generate one). The apparent ill-formedness of the past tense of forgo cannot interact with the structural constraints on merger. As we argue later, it is a mistake to group forms like "intelligenter with forms like "forwent and "forgoed. In the case of "forwent, speakers are confronted by the apparent degraded status of a word generated by the grammar.22 In such cases, the grammar does 22 We are aware of no account of the degraded status of these past tense forms that is incompatible with our approach. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


36 DAVID EMBICK AND ALEC MARANTZ not automatically yield an alternative that then emerges as grammatical in some sort of OT computation. In the case of *intelligenter, the word is not generated by the grammar. Finally, it should be noted that our argument here goes well beyond the familiar difficulty that OT systems have with ineff ability. One could integrate some attempt to deal with that problem into Bresnan's system for example, the null parse (McCarthy and Wolf 2005) and the primary prediction of the global competition approach would still not be verified. All that the null parse (or its equivalent) would do is reduce the number of instances in which effects of the strong prediction of Bresnan's approach popping out could be detected.23 4.2.3 Blocking and Sentential Competition Bresnan (2001) acknowledges that a blocking account of word-phrase interactions requires comparisons between constructions at the phrase or clause level, and not just competitions between words (i.e., it does not suffice to simply have ate block eat). An examination of Andrews 1990, which Bresnan cites as precursor of her approach, reveals why. Andrews proposes a Morphological Blocking Principle that can block a structure S containing a particular word Lj by consulting the lexicon for lexical entries related to the word (in practice, for lexical entries containing the same Root, and thus the same Root meaning, as the word in question). If there exists another lexical entry L2, such that Lj has a subset of the features of L2 and L2 has a subset of the features expressed in S, then Lj is blocked (in S). For example, at the sentence level for the sentence John did eat the beans (with unstressed did), eat has the feature [pred] and did the feature [past], as shown in (63). Assume further that the lexical items in (64) are under consideration for appearing in this structure. (63) Structure for John did eat the beans TP NP T A /^\ John T[past] VP did V[pred] NP eat the beans (64) Lexicon eat, [pred] ate, [pred,past] 23 Put slightly differently, it would be noteworthy if all cases in which the predictions of the competition-based theory could be tested were cases in which it is the null parse that pops out. The competition-style theory advocated by Bresnan predicts that there should be instances in which what pops out is in fact something other than the null parse. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 37 The features of eat are a proper subset of the features of ate, and ate determines a subset of the features of John did eat the beans, which thus blocks eat from the structure John did eat the beans and renders the sentence ungrammatical. Andrews acknowledges a deep problem for this account of blocking, which essentially involves a comparison between lexical entries but in a particular sentential context. Continuing with the example John did eat the beans, the problem is that the existence of ate should block eat not only in this sentence, but also in its negative and interrogative versions, where the auxiliary did is required. (65) a. Did John eat the beans? b. John didn't eat the beans. All that the sentences in (65) add to the structure in (63) are an interrogative feature and a negative feature, respectively. Given that ate blocked the structure with did in (63), it is predicted that its existence should block the sentences in (65) as well, but it does not. Andrews describes two solutions to this problem that will allow the use of eat in interrogative and negative sentences, but both involve having the auxiliary add features to the sentence that would not be present without the auxiliary. For example, if did adds a [ + aux] feature to any sentence in which it occurs, the features of the whole [ + aux], [past], and so on will not be a subset of the features of any lexical verb in the past tense, with the result that the existence of such a verb (e.g., ate) will not block a sentence containing did plus the verb (e.g., did eat). The consequence is that the Morphological Blocking Principle will no longer cause ate to block John did eat the beans. That is, in order to allow eat in sentences like (65) with past tense features, the Morphological Blocking Principle cannot block eat in any sentence with the auxiliary did. In general, the Morphological Blocking Principle is not able to account for the distribution of phrasal versus single-word expressions of inflectional meanings when structural properties of sentences determine where the single-word expression is allowed (e.g., when there is necessarily some local structural relation between the verb stem and the functional head containing the inflectional features). To have blocking account for the ungrammaticality of *John did eat the beans, one must compare this sentence with the alternative John ate the beans', it does not suffice to simply notice the existence of ate. Thus, a lexicalist account of word-phrase interactions must either adopt Bresnan's whole-sentence-comparison theory, problematic for reasons already discussed, or something like the implementation of Poser blocking found in Poser 1992 and discussed more recently by Hankamer and Mikkelsen (2002, 2005), to which we now turn. 4.3 Competition and Poser Blocking In neither the generative approach nor Bresnan's sentence-level competition-based approach to blocking does a general preference for words over phrases do any real work. On the other hand, for Poser's (1992) initial idea that blocking ' 'extends the boundaries of the lexicon," this principle appears to bear a great deal of the explanatory burden. Whether any explanatory burden should be placed on Lexical Preference is an important question. A kind of conceptual argument could be made against the generative approach based on whether "Merge when the conditions are met" This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


38 DAVID EMBICK AND ALEC MARANTZ is a hidden way of saying ' 'Merge (i.e., prefer a single-word expression) whenever you can." If so, then the generative approach would be stipulating on a case-by-case basis something that should actually be treated as a universal principle if words are "better" than phrases. Part of what is at issue is illustrated in (66). (66) *I did walk to the park, (unstressed do) The generative approach says that the ungrammaticality of (66) falls under the generalization Rules Apply; things that are not generated by the grammar, like (66), are ungrammatical. An intuition behind blocking, on the other hand, says that the ungrammaticality of (66) is the result of the preference for the lexical over the phrasal; walked wins over did walk. In theories that accept some kind of blocking, reactions to Poser's (1992) proposals diverge in terms of how they relate to Lexical Preference. One line of research building on Poser's observations, the line that extends through Andrews 1990 and Bresnan 2001, implements a general comparison between sentences along a number of dimensions. In such approaches, Lexical Preference ends up being a consequence of a constraint ranking rather than an overarching principle. Because lexical gaps do not in general license the emergence of phrasal expressions, we noted above that the intuition behind Lexical Preference does not work out correctly in a system like Bresnan' s. In this section, we turn to another line of research that extends Poser's proposals by trying to make precise the structural conditions under which words can block phrases. Poser's (1992) original formulation of word/phrase blocking attempts to restrict the competition between words and phrases to particular structures: "small categories," like a node dominating two heads. The idea also discussed by Hankamer and Mikkelsen (2002, 2005) is as follows. Whenever a "small category" is built in the phrasal syntax, the lexicon is scanned; if a word expressing the same features is found in the lexicon, then that word blocks the phrase. As noted earlier, this implementation of Poser blocking can be generalized so that at each node in the structure, the lexicon must be scanned for an appropriate word. For the purposes of comparing approaches, we consider the generalized version, stated in (67). (67) Generalized Poser blocking For each node in the syntactic structure, scan the lexicon for a word that expresses the same features. If such a word exists, use the word in place of the phrase. The condition in (67) is a more precise instantiation of Lexical Preference that specifies how a word can compete with and in some cases win out over a phrasal expression. For many of the cases we are concerned with, what (67) might mean in structural terms is that a word can block a phrase when the features that are expressed by the word are provided by elements in a sisterhood relationship. (68) Blocking possible X/Y[a,p] X[a] Y[0] This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 39 (69) Lexical items WORDjKp] WORD2[a] WORD3[p] The idea here is that the lexicon of a language contains words that express a, (3 separately, along with a lexical entry that expresses both. Given generalized Poser blocking, the structure in (68) is one in which WORDj must be pronounced. We discuss how this might work in section 4.4. Theories that implement Poser blocking make a very clear prediction about when a word can block a phrase: namely, only when the word expresses features of a node. Thus, a word may block a phrase when the node in question dominates sisters whose features are expressed. However, other configurations, like those in (70)-(72), are predicted not to show blocking of phrases by words. (70) No blocking (71) No blocking (72) No blocking XX X X[a] YP X[a] YP X[a] YP /\ /\ /\ Y[p] ZP ZP YP YP ZP Y[p] ... Y[(3] Because X[a] and Y[(3] are not in the relevant structural configuration in (70)-(72), the word WORDi expressing [a,p] cannot be used. Rather, what is expected in this case is an analytic expression, with the words WORD2 and W0RD3 appearing as separate words within a phrase. On the other hand, an account where X/Y forms a "word" that is based on complex head formation in terms of syntactic and postsyntactic locality makes different predictions. Depending on the type of operation that applies that is, depending on whether it is hierarchically defined, like head movement (and Lowering), or sensitive to linear order (Local Dislocation) it would be possible to form a complex head X/Y in these cases and have a single word express [a,p]. In this way, the syntactic approach is less restrictive than the lexicalist alternative with generalized Poser blocking, in that it allows synthetic expression of syntactically distributed features in a greater range of cases.24 4.4 Case Studies The real issue is whether the generalized Poser blocking theory or the generative theory makes the correct predictions. We argue that the generative approach is correct and that it generalizes 24 On the other hand, generalized Poser blocking could allow an entire XP head, complement, and specifier to be expressed as a single word, with no predictions about that word's internal structure (if any). The Distributed Morphology approach does not have this property. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


40 DAVID EMBICK AND ALEC MARANTZ in a way that the Poser blocking theory cannot. We illustrate this point with Danish definite formation and English comparative and superlative formation, two cases that figure prominently in the literature. 4.4.1 Danish Definites Hankamer and Mikkelsen (2002, 2005) provide an explicit argument for Poser blocking in their analysis of definite nouns in Danish.25 Nouns in this language are suffixed with a definite element under certain conditions, as in (73a), such that analytic expression of D and N is ungrammatical, as in (73b). (73) a. hest-en horse-DEF 'the horse' b. *den hest the horse 'the horse' The affixed nominal is not found with a prenominal adjective. (74) a. *gamle hest-en old horse-DEF 'the old horse' b. *den gamle hest-en the old horse-DEF 'the old horse' c. den gamle hest the old horse 'the old horse' Given this alternation between analytic and synthetic expression of D and N, one option that Hankamer and Mikkelsen consider involves a lexically created "definite noun" like hesten blocking the phrasally constructed den hest. They implement this in terms outlined more or less abstractly in our preceding discussion. One step involves a lexical rule of the language, Rule D, which, for at least some nouns, creates an affixed "definite noun" of category D. The rule is stated as follows: (75) Rule D (Hankamer and Mikkelsen 2002:155) The second step is to state how the outputs of Rule D interact with phrasally constructed DPs. The interaction is like the one schematized above in our introduction to generalized Poser 25 The lexicalist treatment with Poser blocking is presented in Hankamer and Mikkelsen 2002. In Hankamer and Mikkelsen 2005, the authors consider a nonlexicalist alternative as well. For present purposes, what is important is the claim that this is a test case for something like generalized Poser blocking. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 41 blocking. The syntax creates structure (76). The lexicon is checked at each node to see if there is a lexical expression that contains all the features included in that tree structure. In the case at hand, the lexicon contains derived hesten, a definite noun of category D, by virtue of Rule D. Given Lexical Preference as instantiated in generalized Poser blocking, the phrasal expression (76) is blocked by the lexical item hesten. This single word is employed and projects a DP, as shown in (77). (76) DP/NP (77) Definite "noun" DP DP D' D' D NP D den N' hesten N hest In the cases with prenominal adjectives, it is clear from the structure in (78) that blocking of this type cannot occur. (78) Prenominal adjective DP D' D NP den AP NP A' N' A N gamle hest This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


42 DAVID EMBICK AND ALEC MARANTZ There is no node that expresses just the features of N and D, as there is in the case of (77). Therefore, the phrase (78) is grammatical, with no affixation of the definite element to N.26 As far as this case goes, then, the Poser blocking approach is consistent with the facts. While Poser blocking makes the correct prediction for prenominal adjectives, it does not predict any kind of left-right asymmetry in the prevention of synthetic definite noun formation. Hankamer and Mikkelsen (2005) discuss this component of Poser blocking with reference to postnominal PPs. On the standard assumption that these are attached to an NP as in (79), the Poser blocking account predicts that no synthetic expression should arise in such cases; the analytic form should surface. (79) Postnominal PP (ungrammatical) DP D' D NP den NP PP N' med bla pletter N gris Structurally, a theory in which words express nodes predicts that cases like this should behave just like the above cases with prenominal adjectives. However, this prediction does not hold; instead, a postnominal PP does not prevent the use of the affixed N. (80) a. gris-en med bla pletter pig-DEF with blue spots 'the pig with blue spots' 26 This raises the question of why an AP can never be associated with a DP that is headed by the output of Rule D. Hankamer and Mikkelsen (2005) state that it is because Rule D's output heads a DP, and APs attach to NPs. It should be noted that their approach requires additional assumptions to rule out (i) with the structure (ii) This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 43 b. *den gris med bla pletter the pig with blue spots 'the pig with blue spots' In examples of this type, the synthetic form appears even though D and N do not form a constituent. Hankamer and Mikkelsen discuss several possible treatments for patterns of this type, noting that they are forced to adopt a treatment in which all post-head material in NPs must be regarded as adjoined to DP. Unlike the Poser blocking approach, an analysis based on syntax and morphological merger makes the correct predictions about locality in this and other cases. For the case at hand, it appears that D affixes to N under linear adjacency.27 (81) D-sujfixation D[defTN -+ [[N]D[def]] (i) *hest-en gamla horse-DEF old 'the old horse' (ii) Structure for *hest-en gamla DP D' D NP I /\ hesten AP NP A' N' A N gamla 0 or, for that matter, a case in which the output of Rule D occupies D and an overt N occurs in NP. In general, Ds created by Rule D must, unlike normal Ds, be prevented from combining with NPs. A reviewer notes that an account might be formed along semantic lines, whereby the output of Rule D is of the wrong type to combine with nouns; we defer discussion of such a possibility in the absence of a fleshed-out proposal. 27 Hankamer and Mikkelsen (2005) note the existence of various Ns that do not allow affixation with D, something that is not unexpected under the Local Dislocation approach (cf. the discussion of comparative/superlative formation in the next section). The rule in (81) must be assumed to have additional specifications in its structural description that account for the exceptional nouns; that is, it must be made Vocabulary-sensitive in the required way. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


44 DAVID EMBICK AND ALEC MARANTZ This rule says that when D[def] is concatenated with N (i.e., with an n-headed element), D is adjoined to N, where it is realized "affixally." This rule directly accounts for the basic facts of Danish definite DPs.28 Importantly, adjacency-based merger predicts the kind of left-right asymmetry found in Danish and in other cases of affixation under adjacency. In cases with prenominal adjectives, (81) cannot apply because D and N are not linearly adjacent. When, on the other hand, the NP contains post-N material, such as a PP, nothing prevents the rule from applying, as its structural description is met. For these cases, nothing further need be said, as the adjacency-based account already contains the empirically correct locality conditions.29 4.4.2 English Comparatives and Superlatives In his original discussion of word-phrase interactions, Poser (1992) analyzes a well-known effect in the formation of English comparatives and superlatives: an alternation between synthetic and analytic forms that depends on the properties of the adjective.30 (82) a. more/most intelligent b. smarter/smartest In light of the previous sections, the question of course is how the forms in (82) relate to those in (83) and, in particular, whether the former are blocking the latter. (83) a. *intelligent-er/*intelligent-est b. *more smart/*most smart Note that with shorter adjectives those normally taking synthetic forms like the ones in (82b) analytic forms are impossible with regular comparative interpretation. However, they are possible with a "metalinguistic" type of reading, as has been noted in the literature. For present purposes, since we are concentrating on normal comparatives and the processes responsible for the analytic/synthetic alternation, we mark such examples with an asterisk to indicate that *more smart is not a grammatical pronunciation of the normal comparative of smart; we put aside discussion of the morphosyntax of metalinguistic comparison here (see Embick 2007a and references cited there). The patterns in (82)-(83) provide another important case study for the present discussion. According to Poser (1992) and others who have followed him, such as Kiparsky (2005), smarter does in fact block *more smart. We claim, though, that these patterns can be properly analyzed within the generative approach and that the Poser blocking alternative makes the wrong predictions. 28 Embick and Noyer's (2001) discussion of the interaction of determiner heads and definiteness agreement in Swedish wrongly assumes that D and N combine via head movement, an assumption that is drawn from elsewhere in the literature. 9 Other cases, such as restrictive relative clauses, are special for independent reasons. See Hankamer and Mikkelsen 2005 for discussion of the facts and some pertinent considerations. Di Sciullo and Williams (1987) also note with reference to English comparative/superlative formation and the formation of Latin passive perfects that blocking effects extend beyond the domain of the "word." For an analysis of the Latin case that does not invoke blocking, see Embick 2000. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 45 Any account of the morphophonology of comparatives and superlatives must refer to the syntax of such constructions. We assume, with many authors who concentrate on the syntax and semantics of comparatives, that the structure of the comparative is (84) (see Bhatt and Pancheva 2004 for a recent discussion); in this structure, the Root moves to a by head movement. (84) Structure of comparative aP DegP a(P) Deg (XP) a \/p (than-c\ause?) \/Root a \/RQQT • • • We indicate with parentheses in the DegP the fact that we are not taking a stance on certain aspects of comparative syntax, in particular, whether or not the than-clause is generated as a sister of Deg inside DegP and then extraposed to yield surface orders like Mary is more intelligent than John.31 For the purposes of blocking, what is important about (84) is how this structure relates to synthetic forms like smarter and analytic forms like more intelligent. For our approach, this is the question of the conditions under which Deg combines with an adjective ( = [\/Root a]). This question is addressed in Embick and Noyer 2001 and examined in greater detail with respect to issues of blocking in Embick 2007a. In terms of the head-packaging operations discussed earlier, the evidence suggests that when Deg is moved, it is combined with its host by a process that operates under conditions of linear adjacency that is, by Local Dislocation. The rule in 31 The structure in (84) represents one approach; for present purposes, the structure could also be (i). These structures are equivalent for the point we wish to make about blocking, although of course they are not syntactically. (i) Alternative structure for comparative/superlative DegP Deg XP Deg aP r/wrt-clause . . . adjective . . . In either case, an adjacency-based treatment like the one discussed in the text makes the correct predictions, whereas the blocking alternative does not. Other effects for example, whether or not unpronounced copies are visible for adjacencybased affixation are implicated in the choice of comparative structures; see Embick 2007a for some discussion. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


46 DAVID EMBICK AND ALEC MARANTZ question affixes Deg to the adjective when (a) they are linearly adjacent, with the further condition that (b) the adjective have the correct phonological properties. This is stated schematically in (85). (85) Local Dislocation for comparatives Deg Adjective > [[Adjective] Deg] where Adjective has the relevant phonological properties This is a rule of the PF component of the grammar, one that creates complex heads. As far as the syntax is concerned, the structure of all comparatives and superlatives is that shown in (84). In the case of shorter adjectives like smart, the rule in (85) applies. The effect is to place Deg inside the same complex head as the adjective that is, to affix it. Inside a complex head, Deg is pronounced -er, and we have a synthetic comparative, smarter. In the case of phonologically heavier adjectives like intelligent, the rule in (85) does not apply. Deg and the adjective thus remain distinct heads in the representation. In this case, Deg is pronounced more, so that we have an analytic comparative, more intelligent. By providing a blocking alternative to this generative account, we can formulate a Poser blocking account with Lexical Preference quite straightforwardly along the lines sketched above. This treatment assumes that the lexicon is capable of generating some synthetic comparative forms (i.e., comparative adjectives like smarter), which then function as Deg elements in the syntax (analogous to Hankamer and Mikkelsen's treatment of definite nouns in Danish). The syntax creates a structure in which Deg combines with an adjective. (86) Structure (87) Employed item AP AP DegP A A' A I I Deg smart A smarter The lexicon is then scanned, and the action takes place at the AP node. In the case of (86), an object in the lexicon, smarter, expresses all of the features that appear in this AP (i.e., the features of smart and the comparative features). Therefore, the two objects (86) and (87) are in competition with each other, and by Lexical Preference the single-word expression wins, thus blocking the phrase. The syntax then employs a "comparative AP" headed by smarter (87), parallel to the discussion of Danish definites above. This account with Poser blocking requires that the derived word in effect substitute for a node (technically, that an AP headed by the single lexical item smarter be used in preference to the AP that contains additional internal structure, more smart). Recall that Danish raises a problem This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 47 for the Poser blocking analysis in that postnominal material does not force the appearance of an analytic DP. Along these lines, parallel examples can be formed for the English comparative, showing that there are clear cases of adjectives with AP-internal material that nevertheless allow synthetic comparative forms, contrary to the prediction of Poser blocking. For postadjectival material, there are a number of different configurations in which an AP that appears in a comparative has internal structure. The prediction of Poser blocking is clear: such cases should uniformly disallow the creation of a synthetic comparative. Yet they do not. (88) Adjectives with complements a. Raising: John is [likeli-er [Jeh» to win the race]] . . . b. Control: Mary is [quick-er [PRO to spot counterexamples]] . . . c. Transitive adjectives: Bill is [proud-er [of his accomplishments]] . . . d. 7bwg/z-construction: Susan is [easi-er [to understand]] . . . For instance, we take it that the structure of (88a) is as in (89), where the raising infinitival clause is the complement of the adjective (which we represent as A and not [ vRoot a] to facilitate exposition). (89) Substructure for (88a) AP DegP AP A ^\ Deg A TP likely John to win the race There is no node that could be pronounced as the features of Deg and the adjective alone. Thus, for Poser blocking likelier should be impossible. The adjacency-based treatment outlined above does not suffer from this problem. In the cases in (88), Deg and the adjective are linearly adjacent, and thus the rule that affixes Deg under adjacency can apply. This approach can account for the facts, whereas the alternative cannot.32 32 We note in addition that material to the left of the adjective appears to prevent the formation of synthetic comparatives. The facts are somewhat complex (see Embick 2007a for some discussion), but it can be shown that in cases like the following, the comparative Deg takes scope over an adverbially modified AP, and, as predicted, no synthetic form is possible: (i) a. Mary is [more [amazingly smart]] than Bill. amazingly smarter, on the relevant bracketing b. John is [more [ploddingly slow]] than Susan. ploddingly slower Much care must be taken in cases of this type to ensure that there is a "true" comparative interpretation (as opposed to This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


48 DAVID EMBICK AND ALEC MARANTZ It seems that both Poser blocking and the generative approach make clear predictions about locality and analytic/synthetic alternations. The predictions made by the generative approach are correct for the cases we examined above, and generalize to others as well; we are in fact aware of no cases with the properties that are expected if Poser blocking is part of the grammar (section 4.6). 4.5 Remarks on Lexical Preference To this point, we have established that the Distributed Morphology approach makes the correct predictions about word-phrase interactions, while generalized Poser blocking does not. As noted at various points, these two theories differ in terms of whether they accept the intuition that words are better than phrases. In the generative view, this is not a principle, whereas in the Poser blocking approach and other theories, it is supposed to be doing a lot of the relevant work. Here, we present some additional remarks on Lexical Preference. Assuming that the features that are being expressed are held constant, theories with lexicalist Poser blocking express the intuition that we have encoded as Lexical Preference. It is important to note that Lexical Preference cannot by itself account for what is found in, say, comparative formation; it has to be augmented by additional principles or constraints. In a competition-based view, such as Kiparsky's (2005), one candidate for expressing the meaning 'comparative of intelligent' is the synthetic form intelligenter. If Lexical Preference (in Kiparsky's terms, Economy) were the primary factor in deciding the competition that selects the grammatical form, then intelligenter should be preferred over more intelligent, contrary to fact. Some other constraint in the grammar must ensure that intelligenter is disallowed for morphophonological reasons; that is, there must be some constraint ranked higher than Lexical Preference that is doing most of the work here. The only alternative would be to hold that more intelligent and intelligenter are not actually competing with one another to express the same meaning (= ' 'paradigmatic slot," in our metaphor). However, this move is problematic. Why smarter and more smart would express the same features, while intelligenter and more intelligent would not, is unclear. Technically, this could take the form of the proposal that -er introduces/expresses features when it occurs with intelligent that are not introduced/expressed when it is affixed to smart, but this seems to be missing the point. Thus, even in theories that want to elevate Lexical Preference to an important grammatical principle, it is not doing all of the relevant work. As noted at several points above, a more pressing problem with Lexical Preference is that in many if not all of the cases where it could potentially do some work, comparisons require considering alternatives at the phrase level and not simply at the level of words. The examples a metalinguistic comparative interpretation) and that the bracketing is [Deg [Adv Adj]], not [[Deg Adv] Adj]. When these factors are accounted for, it appears that the adjacency-based Local Dislocation of Deg is prevented, as predicted. Along related lines, Bresnan (2001) notes problems with Poser blocking and the comparative, with reference to examples like [exactly three times more] expensive. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 49 discussed above show one type of interaction, in which intelligenter cannot be preferred to more intelligent for morphophonological reasons. But there is a stronger point here as well. In the analysis of analytic/synthetic alternations, it quickly becomes clear that simply noticing that an analytic form (e.g., ate) exists does not suffice to account for where it occurs at the expense of the corresponding synthetic form (e.g., did eat), as we noted with respect to Andrews 1990. Instead, as we noted in connection with Bresnan 2001, the relevant comparisons involve phrasal objects, in which tense, negation, and so on, are expressed. From this perspective, the operative principle is not that words are better than phrases; rather, it is that certain types of phrases are better than certain other types of phrases, with Lexical Preference a possible emergent side effect of the overall constraint ranking. The general question is whether a theory that does not encode Lexical Preference, either one like ours or one like Bresnan' s, is missing a crucial generalization. One way to approach the question head-on is simply to ask whether the grammar seems to function as if it contains Lexical Preference as an inviolable principle in the first place. There are some cases that make it look as if this principle can be overridden. To take one often-discussed example, consider prepositions and determiners in French. Certain combinations require a ' 'fused" form, as in (90a), whereas other combinations do not, as in (90b). (90) a. "Fusion" i. du chat 'of. the cat' (*de le chat 'of the cat') ii. aux enfants 'to.the children' (*a les enfants 'to the children') b. No "fusion" i. de la mere 'of the mother' ii. a la femme 'to the woman' Note that our discussion here does not rely on whether there is one Vocabulary item {du) or two ( d and u) in a branching head. The point is that for the purposes of morphophonology, the special P/D forms are realized in a single complex head. The cases in (90a) look like prime examples of the operation of Lexical Preference: du and aux exist, and they must appear at the expense of phrasal two-word alternatives. However, it is significant that the "fused" P/D elements cannot occur when the element following the D is vowel-initial. (91) a. de l'arbre 'of the tree' b. *du arbre 'of.the tree' Thus, whatever Lexical Preference might encode, it is not inviolable, and in general it is not the sole factor in determining the winner between competing forms that (by hypothesis) "mean the same thing." The only competition-based means of accounting for these patterns would evidently require an OT computation in which Lexical Preference can be outranked by other constraints in the case at hand, perhaps something like *Hiatus. The interaction between P, D, and whatever follows D poses no prima facie problems for a generative approach, although this phenomenon does raise some interesting questions (it might, This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


50 DAVID EMBICK AND ALEC MARANTZ for example, illustrate something about cyclicity; see Embick 2007b for an analysis that takes cyclicity into account).33 Overall, it does not appear that theories without Lexical Preference are missing significant generalizations. 4.6 Synopsis The kind of constituency-based word-phrase interactions required on the formalization of Poser blocking discussed above make very clear predictions about when words should be able to block phrases. As detailed above and elsewhere, these predictions are not borne out. Abstracting, what is not attested is a case in which, for X and Y that potentially form "one word," both pre-Y Z(P) and post-Y Z(P) whether complements to Y or specifiers/adjuncts in the phrase headed by Y prevent a lexical form from occurring. Another way of putting this is that the Poser blocking approach predicts no blocking of phrases by words in either of the configurations (92) and (93) (showing complement and noncomplement status for ZP), where linear order of ZP in particular is irrelevant. (92) Configuration 1 (93) Configuration 2 X X X Y X Y /\ /\ Y ZP Y ZP Y (...) It is not difficult to describe what such a language would look like. One case would be a language with an interaction between T and the verb like that found in English, but in which transitive verbs required analytic forms, unlike intransitives. (94) Hypothetical English' a. John laugh-ed. b. John did eat the apple. Naturally, this hypothetical English' relies on a number of assumptions about constituent structure that are subject to question. The general point is that in this domain and others that have been 33 As an aside, we note that both de Varbre and *du arbre can be seen as containing two "words," if French is assumed to have something like Hankamer and Mikkelsen's Rule D for vowel-initial nouns. One might therefore consider a theory in which it is, as far as some global metric is concerned, the overall number of nodes (lexical items) that is optimized. In such an account, both competing outputs in (91) contain the same number of words that is, could be seen as equally "economical" given some general idea that the optimal case involves the fewest words. Our point still holds, in that Lexical Preference by itself cannot explain why (91a) is grammatical and (91b) is not. This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 51 examined, cases with the property schematized in (92)-(93) are not found. In general, we are not aware of any cases that pattern in the manner predicted by Poser blocking, that is, cases where any material that interrupts "node sharing" prevents the creation of a synthetic form. On the other hand, every case that has been studied systematically shows locality properties definable in terms of a syntactically derived structure and operations on it that is, in terms of putting heads together in a way that respects locality in hierarchical syntactic structure, or respecting locality of the linear relations derived from the hierarchical structure. The generative approach that makes these predictions is completely general, in the sense that it is not a specific theory of word-phrase interactions. This approach offers a theory of syntax, and a theory of how the heads in syntactic structures are packaged, and this covers affixation in general. Word-phrase interactions are a subcase of this general theory of syntactic affixation. We conclude from these considerations that the Poser blocking approach fails, not for conceptual reasons, but because it makes the incorrect empirical predictions. Poser blocking cannot account for the range of cases in which word-phrase interactions (or affixation more generally) occur. Other competition-based theories, like Bresnan's (2001), do not overly restrict the size of objects in which competition takes place (Bresnan (2001:16) in fact critiques this aspect of Poser blocking, as noted above). However, Bresnan's theory of competition at the sentence level makes incorrect predictions as well. The generative approach within Distributed Morphology, on the other hand, is able to explain the attested patterns. 5 Conclusions An analysis of blocking effects requires specific assumptions about the architecture of grammar along numerous dimensions that define a space of competing theoretical approaches. Cases of apparent competition between single-word and phrasal expressions, in what has been termed Poser blocking, highlight the need for any theory of grammar to explain the connection between affixation and the sentential distribution of information carried by closed-class items (e.g., tense). Although it is true that a theory in which all affixation is syntactic, like Distributed Morphology, leads one to expect the sorts of interactions exemplified by cases of Poser blocking, our primary argument here has been in service of a stronger conclusion: there are clear empirical domains in which this grammatical architecture makes the correct predictions, whereas others do not. Looking primarily at (a) the locality of competition effects and (b) whether otherwise wellformed structures are marked ungrammatical as a result of competition, we showed that the generative approach to grammar as formalized within Distributed Morphology forces an analysis that explains the facts. Other alternatives, based on competition between larger objects (words, phrases, and sentences as opposed to Vocabulary items) and different notions of what it means to be (un)grammatical, do not make the correct predictions. Lexicalist approaches to blocking, as inspired by Poser's (1992) work and as formalized by Andrews (1990) and by Hankamer and Mikkelsen (2005), stumble because single words are (sometimes) not constituents from the point of view of the functional structure of sentences, arising instead from the syntactic manipulation of heads (via head raising, merger, and Local Dislocation). Recognizing this problem with the This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


52 DAVID EMBICK AND ALEC MARANTZ narrow lexicalist account of blocking, Bresnan (2001) has proposed that blocking involves an OT competition at the sentential level. However, global approaches of this type fail because they wrongly predict phrasal expressions to emerge as grammatical when single-word expressions are ill formed or unavailable. There is thus no evidence for blocking at this global level, where a blocked but otherwise well-formed expression becomes grammatical when an otherwise more harmonic expression is removed from the competition. Moreover, the generative approach makes the right predictions about directionality effects on syntactic affixation: for example, Distributed Morphology provides an account of why material to the left of a head in a head-initial structure might result in a phrasal expression where a singleword expression would otherwise be available, while material to the right would not. As a general point, the lexicalist approach explicitly predicts symmetrical effects on word-phrase interactions, and these are never exemplified. We take these results to constitute a strong argument for generative approaches to grammar in general, and for our version of such an approach in particular. References Andrews, Avery. 1990. Unification and morphological blocking. Natural Language and Linguistic Theory 8:507-557. Arad, Maya. 2005. Roots and patterns: Hebrew morphosyntax. Berlin: Springer. Aronoff, Mark. 1976. Word formation in generative grammar. Cambridge, MA: MIT Press. Baayen, R. Harald, and Antoinette Renouf. 1996. Chronicling the Times: Productive lexical innovations in an English newspaper. Language 72:69-96. Bhatt, Rajesh, and Roumyana Pancheva. 2004. Late merger of degree clauses. Linguistic Inquiry 35:1-45. Borer, Hagit. 2004. Structuring sense. Oxford: Oxford University Press. Bresnan, Joan. 2001. Explaining morphosyntactic competition. In Handbook of contemporary syntactic theory, ed. by Mark Baltin and Chris Collins, 1-44. Oxford: Blackwell. Carstairs, Andrew. 1990. Phonologically conditioned suppletion. In Selected papers from the Third International Morphology Meeting, ed. by Wolfgang Dressier, Hans Luschutzky, Oskar Pfeiffer, and John Rennison, 17-23. Berlin: Mouton de Gruyter. Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton. Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A life in language, ed. by Michael Kenstowicz, 1-52. Cambridge, MA: MIT Press. Clark, Eve. 1993. The lexicon in acquisition. Cambridge: Cambridge University Press. Di Sciullo, Anna Maria, and Edwin Williams. 1987. On the definition of word. Cambridge, MA: MIT Press. Embick, David. 2000. Features, syntax, and categories in the Latin perfect. Linguistic Inquiry 31:185-230. Embick, David. 2003. Locality, listedness, and morphological information. Studia Linguistica 57:143-169. Embick, David. 2007a. Blocking effects and analytic/synthetic alternations. Natural Language and Linguistic Theory 25:1-37. Embick, David. 2007b. Linearization and Local Dislocation: Derivational mechanics and interactions. Linguistic Analysis 33(3-4):2-35. Embick, David, and Rolf Noyer. 2001. Movement operations after syntax. Linguistic Inquiry 32:555-595. Giegerich, Heinz. 2001. Synonymy blocking and the Elsewhere Condition: Lexical morphology and the speaker. Transactions of the Philological Society 99:65-98. Hankamer, Jorge, and Line Mikkelsen. 2002. A morphological analysis of definite nouns in Danish. Journal of Germanic Linguistics 1 4: 1 371 75 . This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions


ARCHITECTURE AND BLOCKING 53 Hankamer, Jorge, and Line Mikkelsen. 2005. When movement must be blocked: A response to Embick and Noyer. Linguistic Inquiry 36:85-125. Harley, Heidi. 1995. Subjects, events, and licensing. Doctoral dissertation, MIT, Cambridge, MA. Kiparsky, Paul. 2005. Blocking and periphrasis in inflectional paradigms. Yearbook of Morphology 2004, 113-135. Marantz, Alec. 1984. On the nature of grammatical relations. Cambridge, MA: MIT Press. Marantz, Alec. 1988. Clitics, morphological merger, and the mapping to phonological structure. In Theoretical morphology, ed. by Michael Hammond and Michael Noonan, 253-270. San Diego, CA: Academic Press. Marantz, Alec. 1997. No escape from syntax: Don't try morphological analysis in the privacy of your own lexicon. In Proceedings of the 21st Annual Penn Linguistics Colloquium, ed. by Alexis Dimitriadis, Laura Siegel, Clarissa Surek-Clark, and Alexander Williams, 201-225. Pennsylvania Working Papers in Linguistics 4.2. Philadelphia: University of Pennsylvania, Penn Linguistics Club. Marantz, Alec. 2001. Words and things. Handout of a talk, MIT, Cambridge, MA. Marantz, Alec. 2003. Blocking. Handout of a talk, MIT, Cambridge, MA. Marantz, Alec. 2007. Phases and words. In Phases in the theory of grammar, ed. by Sook-Hee Choe, 191-222. Seoul: Dong In. Marlett, Stephen, and Joseph Stemberger. 1983. Empty consonants in Seri. Linguistic Inquiry 14:617-639. Matushansky, Ora. 2006. Head movement in linguistic theory. Linguistic Inquiry 37:69-109. McCarthy, John, and Matthew Wolf. 2005. Less than zero: Correspondence and the null output. Ms., University of Massachusetts, Amherst. Poser, William J. 1992. Blocking of phrasal constructions by lexical items. In Lexical matters, ed. by Ivan Sag and Anna Szabolcsi, 111-130. Stanford, CA: CSLI Publications. Williams, Edwin. 2007. Dumping lexicalism. In Oxford handbook of linguistic interfaces, ed. by Gillian Ramchand and Charles Reiss, 353-382. Oxford: Oxford University Press. (Embick) Department of Linguistics 619 Williams Hall University of Pennsylvania Philadelphia, PA 19104-6305 embick @ ling, upenn. edu (Marantz) Department of Linguistics New York University 726 Broadway, 7th Floor New York, NY 10003 This content downloaded from on Tue, 27 May 2014 12:29:46 PMAll use subject to JSTOR Terms and Conditions