AUTHOR ""
TITLE "Functional Structure in Nominals: Nominalization and ergativity"
SUBJECT "Linguistik Aktuell/Linguistics Today, Volume 42"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Functional Structure in Nominals
Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.
Series Editor Werner Abraham University of Groningen, Netherlands University of California, Berkeley University of Vienna
Advisory Editorial Board Guglielmo Cinque, University of Venice Günther Grewendorf, J.W. Goethe-University, Frankfurt Liliane Haegeman, University of Lille, France Hubert Haider, University of Salzburg Christer Platzack, University of Lund Ian Roberts, University of Stuttgart Ken Safir, Rutgers University, New Brunswick NJ Lisa deMena Travis, McGill University Sten Vikner, University of Stuttgart C. Jan-Wouter Zwart, University of Groningen
Volume 42 Functional Structure in Nominals: Nominalization and ergativity by Artemis Alexiadou
Functional Structure in Nominals Nominalization and ergativity
Artemis Alexiadou University of Potsdam
John Benjamins Publishing Company Amsterdam/Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data Alexiadou, Artemis Functional Structure in Nominals : Nominalization and ergativity / Artemis Alexiadou. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166–0829 ; v. 42) Includes bibliographical references and index. 1. Grammar, Comparative and general--Nominals. 2. Grammar, Comparative and general--Ergative constructions. 3. Grammar, Comparative and general--Word formation. 4. Functionalism (Linguistics) I. Title. II. Linguistik aktuell ; Bd. 42. P271 .A44 2001 415--dc21 isbn 90 272 2763 2 (Eur.) / 1 58811 055 9 (US) (Hb; alk. paper)
2001025181
© 2001 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O.Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O.Box 27519 · Philadelphia pa 19118-0519 · usa
Table of contents Acknowledgements
ix
Chapter 1 Introduction . Some theoretical and methodological preliminaries . Verbs and deverbal nominals in English . Derived nominals and gerunds . Word formation and the lexicon . Word and category formation in DM . An ambiguity in the nominal system . Process versus result nominals: Grimshaw’s diagnostics . The role of the event argument . The VP analysis . The fine structure of process nominals: a first encounter . Nominalization patterns across languages . Outline Chapter 2 The Functional Architecture of Nominalizations . Introduction . NPs are dominated by D . AgrP, NumberP and GenderP . On the A vs. A¹ status of Spec,DP . Some notes on the Greek DP . Greek nominal formation . General remarks . Nominals derived from transitive predicates . Nominals derived from intransitive predicates . Nominals derived from ditransitive predicates . Nominals derived from psychological predicates . The verbal properties of process nominals . DP-internal adverbs . Morphological reflexes
vi
Table of Contents
.
. . .
. Aspectual distinctions . The structural differences between process and result nominals Process nominals lack tense . Absence of phenomena related to T: EPP, ECM and raising . Evidence for nominal tense On the obligatory licensing of arguments within process nominals Processes in morphological structure Conclusion
Chapter 3 Intransitivity in Nominalization . Event nominals are ergative constructions . Greek event nominals . Event nominals in English and other Germanic languages . Event nominals in Romance . Slavic event nominals . Semitic event nominals . Hungarian event nominals . On the properties of passive nominals . Background . Affectedness and aspectual properties . Referential adjectives . No process of passivization . The structure of process nominals revisited . By-phrases and more on the lack of external arguments Chapter 4 Variation in Functional Structure . Variation in the set of verbal projections . Presence vs. absence of C . Presence vs. absence of Aspect . Variation depending on the type of v/Voice (transitive vs. intransitive) . Variation in the set of nominal projections . Presence vs. absence of number: licensing of adjectival modification . Variation depending on the type of D . Summary
Table of Contents
Chapter 5 Nominalization and Ergativity . The Case patterns in nominalization and ergative languages . Unaccusativity in ergative languages . Ergative as a lexical case . v is deficient in ergative languages . Remarks on Case within DPs . Genitive is a structural case . Case as a morphological property . Locus of genitive . Agents, ergativity and the perfect . Background . The crosslinguistic distribution of the possessive agent . Possessor subjects in the perfect Auxiliary selection . HAVE-BE and the perfect . Auxiliary selection languages and only HAVE languages . Conclusion Chapter 6 Conclusions References Index
vii
Acknowledgements
This book is based on my Habilitationsschrift ‘Post-doctoral thesis’ On the Syntax of Nominalization and Possession: remarks on patterns of ergativity, submitted to the Philosophical Faculty II of the University of Potsdam in May 1999. It is concerned with an investigation of the phrasal architecture of nominalizations and their relationship to other linguistic phenomena. The book differs from the thesis both in its organization and in its content. In particular, chapter 4 is to a great extent novel, and certain sections of the thesis are not included here for clarity of exposition. While working on the material in the book I have had fruitful discussions with several people who offered helpful comments, raised interesting questions, provided intriguing data. For all this I would like to thank Hagit Borer, Henry Davis, Ulrike Demske, David Embick, Miriam Engelhardt, Martha McGinnis, Eric Haeberli, Liliane Haegeman, Heidi Harley, Norbert Hornstein, Sabine Iatridou, Graham Katz, Winfried Lechner, Renate Musan, Rolf Noyer, David Pesetsky, Irene Rapp, Tom Roeper, Arnim von Stechow, Wolfgang Sternefeld, Friderikos Valetopoulos, and Heike Winhardt. Very special thanks go to Alec Marantz, Gisbert Fanselow, Melita Stavrou, and Elena Anagnostopoulou who also provided detailed written comments. Thanks are also due to Gereon Müller and Ursula Kleinhenz. Parts of the material included in this study were presented at the 29th NELS conference held at the University of Delaware in October 1998, at the workshop on adverbs and adjunction held at the University of Tromsoe in April 1999, at the 23rd GLOW Colloquium in Vitoria-Gasteiz in Spain in April 2000 and at various seminars at the Universities of Thessaloniki, Tübingen, Pennsylvania and Potsdam. I thank all these audiences for their comments. Many thanks to Werner Abraham and Kees Vaes for all their help and guidance with the manuscript. The DFG grant AL554/1-1 is hereby gratefully acknowledged. Potsdam, Germany Artemis Alexiadou September 2000
Chapter 1
Introduction
The goal of this work is to study certain aspects of the internal structure of nominal/determiner phrases (DPs). The body of the book is composed of Chapters 2 to 4. These chapters contain an in depth investigation of the phrasal architecture of nominalizations across languages and define criteria on the basis of which the organization of their functional structure can be determined. Chapter 5 examines the interaction between nominalization and patterns of ergativity, the aim being to explore the hypothesis that nominal clausal structure is analogous to ergative clausal structure.∀ This chapter offers an overview of the phenomena to be discussed, presents my theoretical assumptions and sketches an outline of the approach taken in this book.
.
Some theoretical and methodological preliminaries
. Verbs and deverbal nominals in English At a superficial investigation English has different types of nominals with similar meaning and distribution (see Chomsky 1970, Lees 1960, Fraser 1970, and also Marantz 1997, Harley and Noyer 1998b, Hoekstra 1999, Zucchi 1993 among others for a recent discussion). Some of these nominal constructions are exemplified in (1–3): (1) John’s criticizing the book. (2) The barbarian’s destruction of the city (3) Belushi’s mixing of drugs led to his demise
Gerund Derived Nominal Mixed Nominalization
Native speakers of English perceive the nominals in these examples as being related to some sentential construction. Thus, for instance, the noun destruction in (2) appears to bear the same semantic relation to the noun phrases the barbarians and the city, as the verb destroy does in (4).
Functional structure in nominals
(4) The barbarians destroyed the city.
A more accurate inspection reveals, however, that the nominals in (1–3) differ among themselves both in their distribution and in the range of interpretations they allow. I turn to a brief description of these differences in the next subsection. . Derived nominals and gerunds In English only gerund formation as in (1) is productive. According to Lees (1960), all verbs have a corresponding gerund, but not all of them give the corresponding derived nominal. Moreover, (1) differs from (2) in that only in the gerundive context is the presence of an auxiliary licit: (5) John’s having criticized the book.
Furthermore, morphological -ing affixation is transparent only in gerunds. Finally, the theme argument of derived nominals is introduced by the preposition ‘of ’. There is also an interpretive difference between the two types of nominal constructions. As Vendler (1967) observed, while nominals of type (1) like that clauses and fact-NPs cannot occur with predicates such as is slow, is sudden, takes a long time, this is not the case with nominals of type (2). This is shown in the following examples: (6) a. John’s performance of the song was slow. b. John’s performance of the song was sudden. c. John’s performance of the song took a long time. (7) a. *John’s performing the song was slow. b. *John’s performing the song was sudden. c. *John’s performing the song took a long time.
Vendler claims that nominals of type (2) may denote eventualities, while nominals of type (1) denote propositional entities. In the literature a certain amount of consensus has emerged that gerunds (a) are NPs and (b) directly dominate a VP node (see Abney 1987), although the details of their syntactic analysis vary from author to author. Evidence that English gerunds are NPs comes from the fact that they are able to appear in a number of positions typical for NPs. For instance, they may appear as objects of prepositions, or subjects of sentences:
Introduction
(8) a. I learned about John’s smoking stogies b. Would John’s smoking stogies bother you?
However, it remains a matter of controversy whether the structure of derived nominals includes a VP node or not (see Section 2). ‘Mixed nominalizations’ show properties of both groups. While their syntactic behavior is parallel to that of ‘derived’ nominals, in the sense that their direct objects are inserted by the preposition of, they are formed by -ing affixation (see Chomsky 1970 and Harley and Noyer 1998b for a recent discussion). Parsons (1990) refers to the gerunds of type (1) as verbal gerunds, while to the ones of type (3) as nominal gerunds. Central to the study of nominalization is the definition of the syntactic category and the projection of the derived nominal (see for instance the overview in Lefebvre and Muysken 1988, and Siloni 1997). In general the syntactic nature of structures showing a categorially ambiguous behavior, such as nominalization, has not been easy to determine within most approaches formulated in the spirit of Government and Binding (GB) theory, as the distinctions between syntactic phenomena that merely echo lexical information and those that result from an actual syntactic operation is sometimes blurred due to miscellaneous factors. In fact, the tension between the lexicon and the syntactic component is natural under any approach assuming a separate generative component such as the/a lexicon. In the development of GB the tension between the lexicon and the syntax has had as a result the constant re-examination of the division of labor between the two components. The background idea has been the assumption that given that the grammar is a tightly organized system, enriching one component involves widespread modifications in the other components. In the next section I turn to a more detailed discussion of these issues. . Word formation and the lexicon Much of the work done on word formation during the early 80’s was based on the assumption that there is an independent word formation component, the lexicon, but its interaction with the syntax is severely restricted by some version of Lapointe’s (1979) Lexical Integrity Hypothesis (LIH). A recent formulation of this principle is given in di Sciullo and Williams (1987): (9) The Atomicity Thesis Words are atomic at the level of phrase syntax and phrasal semantics. The words have features or properties, but these features have no
Functional structure in nominals
structure and the relation of these features to the internal composition of the word cannot be relevant in syntax. Standardly, the way in which the LIH is enforced in many of the models is by assuming that the word formation component is ordered prior to D-structure, that is prior to the availability of any syntactic operations. The word formation component and the syntax interact only in one fixed point: the output of the former is the input to the latter (cf. 10):
(10)
Lexicon Syntax postlexical phonology
LF
Such views clearly state that the lexicon generates words, while the syntax generates structures in which words are combined into phrases. The generative aspect of the lexicon is recognized also in approaches that allow for the burden of word formation to be divided between the lexicon and the syntactic component. A clear example of such an approach is the discussion between verbal (11) and adjectival (12) passives in Wasow (1977). (11) Three poems were written by me (12) Three poems in this book are well written
Wasow proposed that verbal passives are formed in the syntax, while adjectival passives are formed in the lexicon. According to him, this fact correlates with a number of differences between the two types of passives. First of all, the lexical formation of adjectival passives is associated with idiosyncrasy in form and meaning, e.g. the hung jury, the shaven man. Second, lexical formation cannot interact with syntactic rules, thus no raising to object followed by passive can take place for the adjectival passive construction, e.g. John was believed to be sick vs. *John remained believed to be sick. Third, lexical word formation is associated with change in lexical category (verb to adjective). Hence the adjectival passive can be modified with modifiers that standardly modify adjectives, e.g. a very driven worker. This does not hold for the verbal passive. A similar approach has been adopted to account for the differences between the nominals in (1) and (2), exemplified in Section 1.1. In order to explain the more verbal nature of gerunds, a number of researchers sided with the view that derived nominals are the output of an operation done in the
Introduction
lexicon, while gerunds are the output of a syntactic nominalization (see Siloni 1997 for a recent discussion). On this view, the discrepancies between the two nominal types might follow. However, matters are not that simple. If one adopts the view that deverbal nominals are constructed in the lexicon, then the following questions become acute: what are the means for signaling the relations between action nominals and their arguments as compared to the situation in verbal clauses? How are certain verbal properties of nominalizations best to be captured, e.g. the fact that they seem to take complements and subjects? In recent years much work (notably inspired by Baker 1988, and Pollock 1989) can be characterized as an attempt to deny the word formation component its status as a generative component. On such views, the morphemes that form words are scattered across the clausal architecture and are brought together via movement or lowering operations. The basic objective is to show that word-formation phenomena adhere to syntactic constraints and interact with syntactic rules, and hence are best characterized as syntactic phenomena. As Baker (1988: 71) notes, the relationship between morphology and syntax is such that it allows principles that are fundamentally morphological to determine syntactic structure in various ways. Consider (13), containing two Chichewa sentences (from Baker 1988: 69): (13) a.
Msangalatsi a-ku-yend-a ndi ndodo. entertainer sp:pres-walk:asp with stick ‘The entertainer walked with a stick.’ b. Msangalatis a-ku-yend-er-a ndodo. entertainer sp:pres-walk-appl:asp stick ‘The entertainer walked with a stick.’
The highlighted applicative morpheme in the (b) sentence is associated with assigning a semantically transparent instrumental thematic role to the postverbal NP. The same role is assigned to NP by a preposition in the (a) sentence. The Uniformity of Theta-assignment Hypothesis implies that this morpheme is an independent constituent in D-structure. Hence the (b) sentence is derived by preposition incorporation. In other words, the verb and the affix come together in the syntax. In fact the most direct effect morphology has on syntax is to state which types of incorporation are impossible. In the next section I take a closer look at the proposal that all word formation is syntactic. For the purposes of my discussion I concentrate on the framework proposed in Marantz (1999) within the theoretical model referred
Functional structure in nominals
to as Distributed Morphology (DM), but see e.g. Schoorlemmer (1995), van Hout and Roeper (1998), Borer (1999) for related views. . Word and category formation in DM In the grammar architecture assumed in DM the interaction between syntax and morphology is direct. Vocabulary items are inserted under terminal nodes into positions in syntactic structures. Operations such as Merger, Fusion, Fission may apply to adjust the structure, but what the morphological spell-out rules see is effectively the output of the syntactic derivation.
(14)
Syntax Morphology (addition of morphemes, Merger, etc.) Vocabulary insertion Phonological rules PF
LF
Following Chomsky (1998), it is assumed that Universal Grammar (UG) provides a set of features. A subset of these is chosen by a language and packaged into the terminal nodes in the syntax. According to Embick (1997), the types of features that are present in the syntactic terminals are those that are relevant for semantic interpretation at LF. There is a second type of features which are only relevant for morphological well-formedness, and these are introduced at the morphological structure (MS), i.e. at the PF branch (cf. 15). These latter features are completely irrelevant syntactically and semantically. Both types, however, affect morphological realization. (15) Feature Types Syntactic-Semantic Tense Aspect
Morphological Theme vowel Gender
Within DM several arguments have been constructed showing that a generative lexicon of the familiar type is not justified. As Marantz (1997, 1999) discusses in detail, the modularity arguments (see the discussion on Wasow in
Introduction
Section 1.3) break down on both the phonological and the semantic branches of the grammar. On the phonological side, the breakdown of the modularity argument is found in cases in which objects that have to be assembled syntactically nevertheless show lexical phonology. Semantically, special meanings are not found with simplex lexical items alone, but must be associated with objects created in the syntax. Thus syntax is the only generative system available in the grammar. Specifically, Marantz outlines an approach to word formation according to which lexical elements, unspecified for syntactic category are introduced into variable syntactic environments. Depending on the functional layers that dominate the unspecified item, this is spelled-out as a verb or a noun or an adjective. In other words, the underlying parts of speech like verb destroy, noun destruction, are abstract roots, which are unspecified for syntactic category. Category labels such as N, V, etc are irrelevant. These abstract roots enter into relations with higher functional heads such as n, v, a, and give verbs, nouns, and adjectives respectively. For example, in (16), when √DESTROY is placed in a verbal environment, it becomes a verb (16b); when it is placed in a nominal environment (16c), the result is a nominalization (see also Aoun 1981, van Riemsdijk 1983 and Picallo 1991 for related ideas).
(16) a. √=√DESTROY b.
c.
VP
D°
agent v¹ v
D¹
√
nP n
√
On such an approach, functional heads fully determine the category of a lexical head (see also Alexiadou 1999b, Embick 2000a for further discussion). This view crucially departs from Grimshaw’s (1991) proposal that the category of lexical heads determines that of the functional category, thereby eliminating the redundancy between lexical and functional categories. The roots that appear in multiple syntactic environments are taken to belong to certain semantic classes of the type discussed in the work of Levin and Rappaport (1995) and Levin (1993). Thus the root in (16a) denotes a change of state, which is not internally caused and implies an external causer or an agent. On the other hand, a root like √GROW denotes a change of state that is internally caused, i.e. does not imply the presence of an agent.
Functional structure in nominals
According to Marantz, correlations among phonological, syntactic and semantic properties of word formation are accounted for the same way such correlations among properties of phrasal formation are accounted for, i.e. in terms of locality domains defined by syntactic heads. In fact Marantz attempts to reconstruct the ‘two places’ word formation of the type developed in Wasow’s work without assuming a lexicon. The two places emerge from the operation of the syntax, both structurally and derivationally. The uniformity of morpho-phonology follows from having the syntax perform all the merger operations, including those between morphemes within a word. According to this proposal, one place to build words is in the domain of the root. This involves attachment of the morpheme to the root before attaching the functional head that determines the syntactic category. A second place to build words is outside the domain of a functional head that determines syntactic category, the little v’s, n’s and a’s. Take a concrete example, the nominal dancer. This nominal receives an eventive/agentive interpretation, namely it refers to a person that performs the activity of dancing. In this case, the root √DANCE first combines with v, the head that contains agentivity/eventivity features (see Section 3) and then with n. My aim here is to further consider the functional determination hypothesis, although the basic ideas and generalizations to be expressed are theory neutral. This choice is due to the following reasons. First, a framework where all word formation takes place in syntax seems superior to models adhering to an omnipotent lexicon (see also Borer 1999). Second, the redundancy between lexical and functional categories is eliminated, i.e. categories need not carry double specification. All category/word formation is syntactic and functional. Moreover, the intuition in any research on nominalization has been that nouns share some of the semantic properties of the related verb. This is readily captured in structural representations such as the ones offered in (16) above. Here the differences between nouns and verbs are located in the functional layers of the structure, e.g. presence vs. absence of Tense. That the source of these differences might be related to the presence vs. absence of Tense has been extensively discussed in the typological literature (see for instance Croft 1991) and in semantic approaches (see for instance Pustejovsky 1995). This can be straightforwardly expressed within the approach to word formation developed here. In general such a model facilitates a better understanding of the deficient character that nouns are felt to have as opposed to verbs. Furthermore, given that in several languages the distinction between verbs and nouns is not so clear-cut, e.g. Eskimo (see Wunderlich 1994), adopting the view that such distinctions
Introduction
exist only at a superficial level enables us to readily capture this property. Finally, the differences among the various types of nominals will be derived from variation in their respective functional structure, a result consistent with the spirit of much recent syntactic research. Having introduced my basic assumptions, in the next section I turn to an ambiguity in the interpretation of nominals of kind (2) that has preoccupied the recent literature and illustrate how this can be reformulated in the framework outlined above. A detailed discussion on other types of nominal constructions is given in Chapter 4.
. An ambiguity in the nominal system Although nouns and verbs are semantically related, the issue of whether nouns, like verbs, license argument structure has provoked a great deal of controversy. For instance, Anderson (1983), Higginbotham (1983), and Dowty (1989) among others have argued that nouns crucially differ from verbs in that the former, unlike the latter, do not take arguments. Their reasoning runs as follows. Given that nouns fundamentally differ from verbs in that they only optionally take arguments, e.g. the race lasted an hour vs. Sam’s race lasted an hour, nouns must lack argument structure altogether. However, in recent years, there is a certain amount of consensus that nouns do not behave uniformly with respect to the licensing of argument structure. Some nouns are systematically like verbs in their argument taking capacities, while others are quite different and in fact take no arguments at all. This distinction correlates with an interpretational distinction between nouns expressing an event (a process) and nouns simply naming an entity in the world or the output of an event (see Grimshaw 1990, Lebeaux 1986, Roeper 1987, Zubizarreta 1987 and references therein). The former take arguments, while the latter and do not do so. Specifically, Grimshaw claims that nouns denoting complex events such as the examination of the students, like verbs, also have an argument structure, since they denote events breaking into aspectual subparts. On the other hand, nouns that denote simple events such as trip and race and result nouns such as exam do not have an argument structure. Result nominals can be both derived or non-derived. In fact for Grimshaw nouns such as examination are in fact ambiguous between two interpretations: a complex event interpretation, in which the noun takes arguments, and a result interpretation in which it does not take arguments. In the latter reading the noun would be similar to nouns
Functional structure in nominals
(cat, dog, house) which are not derived from verbs and which also do not have arguments. The term complex event then is interpreted as signaling that the nominal, much like its verbal counterpart, has a complex event structure that can be broken down into various aspectual subparts. Since argument structure is composed from the aspectual and thematic analyses of a predicate, in the sense that event participants are projected as syntactic arguments, any predicate that lacks an aspectual analysis also lacks an argument structure. In this sense, one could understand the term process nominals as referring solely to argument supporting nominals. The distinction between complex event and other nominals corresponds roughly to the distinction between process and result nominals (see also Borer 1993, Zubizarreta 1987). Henceforth, I use the term result nouns to refer to what other researchers call non-argument supporting nouns, and the terms process and event nouns interchangeably to refer to Grimshaw’s complex event class, though semantically events are distinguished from processes in that the former are terminative, while the latter are durative (see Verkuyl 1993 for a recent discussion). This semantic distinction does not affect the argumenttaking properties of nouns. It will become relevant for some of the facts discussed in Chapters 2 and 3. As will become clear in the course of the discussion, here I am not concerned with the argument supporting property in the way Grimshaw and others following Grimshaw view it, since in principle all nouns can have complements in my system. Rather I pay attention to how process/event/result readings are expressed configurationally and how these interact with the licensing of argument structure. I argue that the difference between process and result nominals is that the former include a set of functional categories standardly associated with verbal clauses that bring about the process/event reading, while the latter lack such projections. That is, event properties are associated with specific functional nodes inside clauses and not with specific categories/words (see Borer 1999, van Hout and Roeper 1998 and references therein for related views). In the next subsection, I turn to some of the criteria that distinguish between the two noun types, established in Grimshaw (1990), so that I can refer back to them. . Process versus result nominals: Grimshaw’s diagnostics According to Grimshaw, the most salient differences, which can be explained in terms of absence vs. presence of argument structure, between the two types of nouns are:
Introduction
(1) Process nominals denote an event; result nominals denote the output of an event. In (17) below the nouns examination and exam refer to an event, which can be located in time and to an entity in the world, respectively: (17) a. the examination of the students at noon b. the exam
(2) Process nouns take internal arguments obligatorily, while result nominals never do. Thus the former can function as theta assigners (although the thetamarking is mediated by the preposition of in English), while the latter cannot. (18) a. The examination of the papers b. *the exam of the papers
Process Result
(3) Process nominals can take agent-oriented modifiers, while this is not possible with result nouns. (19) a. the instructor’s intentional examination of the student b. *the intentional exam is desirable
(4) Result nouns do not permit aspectual modifiers, while these are allowed with process nominals. In fact process nominals admit the same aspectual modifiers as their verbal counterparts, while result nouns do not permit such modifiers (see Vendler 1967): (20) a. the examination of the papers in three hours b. *the exam for three hours c. The teacher examined the papers in only two days
Process Result
(5) Result nouns may be modified by a, one while this is not possible with process nominals. These can only be definite. (21) a. an exam b. *one examination of the papers
(6) Result nouns are count nouns, and they may pluralize, while event nouns are mass nouns. (22) a. one exam, two exams b. the examination of the papers c. *the examinations of the papers
(7) Modifiers like frequent may appear with plural result nouns, and are ungrammatical within singular result nouns. On the other hand, such modifiers may only appear with singular process nominals, and never with plural process nominals, since these cannot pluralize in the first place.
Functional structure in nominals
(23) a. b. c. d.
the frequent examination of the papers *the frequent examinations of the papers the frequent exams *the frequent exam
(8) The prenominal genitives that appear with result nouns are possessives, while the prenominal genitives with process nominals are agents. Note, however, that the possessive reading does not exclude an interpretation in which the possessor was the instigator: (24) a. (*)The instructor’s examination b. The instructor’s examination of the papers
(24a) is grammatical if the NP is interpreted as a possessive modifier. But if it is construed like an agentive subject, the sentence is ungrammatical. The instructor cannot be the agent of the examination. In (24b), however, the instructor is necessarily understood as the instigator of the examination. (9) By phrases appear in non-argumental contexts with result nouns, but denote arguments with process nominals. This is illustrated in (25): (25) a. a picture by a painter b. the destruction of the city by the enemy
(10) Implicit argument control is possible with process nominals, but not with result nominals. (26) a. the assignment of easy problems in order to pass all the students b. *the exam in order to pass all the students
(11) Result nouns may appear as predicates, while this is not possible for process nominals. (27) a. this is an exam b. *this is an examination of the students
In the next subsection I briefly review previous analyses of the differences discussed here. . The role of the event argument According to Grimshaw (1990), the differences between process nominals and result nominals are derived from the fact that the former have an associated event structure, which breaks events into aspectual subparts. As a result, they
Introduction
also have an argument structure, which must be satisfied, hence the obligatoriness of their arguments. On the other hand, result nominals lack an aspectual analysis, and as a result they lack an argument structure analysis. On this view, result nominals differ from event nominals in that the latter select the event argument (Ev), while the former select the referential argument (R), as an external argument (in the sense of Williams 1981). The selection of R leads to a referential reading. The selection of Ev leads to an event reading. A noun gets Ev as its external argument only if it has an event structure. No noun with R as its external argument can ever have an event structure associated with it. In Grimshaw’s system referential elements contrast with eventive elements in that they may not have a theta-grid. In other words, it is the presence of Ev that explains the presence of an argument structure and the aspectual properties associated with process nouns. The availability of an event argument is associated with the individual affix. For instance, affixes such as -ation and -ment in English are ambiguous between a result and an event reading. The affix -ing in nominals has an event reading only. Zero derived nouns have the result reading only. Grimshaw’s account crucially attributes the verbal character of these constructions to the lexical properties of the nominals (stem-affix)., For Grimshaw then, argument structure is a property linked to both verbs and nouns. Both result and process nominals, however, are pure nominal extended projections in her system. Here I point out some problems with this account. First of all, in Grimshaw’s approach result nominals can never surface together with complements. However, this is contrary to fact. Consider the following Catalan examples (from Picallo 1991): (28) a.
la discussió de les dades va durar tot el dia the discussion of the data lasted all day b. la discussio de les dades es va publicar a la revista the discussion of the data was published in the journal
Process Result
(28a) clearly involves a process nominal, as expressions denoting events or process can be placed in time; in (28b), on the other hand, the nominal receives the result interpretation, as only the outcome of an event can be published. But, the PP ‘of the data’ bears the same thematic relation to both instances of the noun in (28). A similar contrast is illustrated in (29) and (30) (French data from Siloni 1997: 98, English from Grimshaw 1990: 99).
Functional structure in nominals
(29) a.
la présentation de livres de ce journaliste est toujours raffinée the presentation of books of this journalist is always refined b. *la présentation frequente de livres de ce journaliste.. the presentation frequent of books of this journalist
(30) a. Their conclusion that there is no relevant data. b. The decision to leave at six.
In the French example in (29a) the noun presentation is thematically related to the PPs of books and of this journalist. However, it can only receive the result interpretation in this context, as is shown by the ungrammaticality of adjectival modification that could bring about the event reading. Other nouns that behave similarly are ‘translation’, ‘review’ and so on. In (30) the nouns conclusion and decision do not bear eventive readings, and yet the clauses that follow them can only be interpreted as complements of the respective nouns. Finally, as will be shown in the next section, the fact that certain adverbs, which are standardly taken to modify VPs, can appear within process nominals remains unaccounted for in Grimshaw’s system. . The VP analysis According to some other researchers, the verb-like properties of event/process nominals, i.e. the presence of aspectual modifiers, the obligatory presence of internal arguments etc, are attributed to the presence of a VP node within these nominals. For instance, Lebeaux (1986) posits a VP present at LF as a result of affix raising at LF. Hazout (1995), Borer (1993), and Fu (1994) among others argue for an underlying VP which surfaces as the complement of a nominal head. For these linguists, it is only the category verb that is associated with the elementary notional type ‘event’/‘process’. Thus, the properties observed by Grimshaw necessarily make reference to a VP node. More specifically, Borer (1993) argues that the appearance of an argument structure associated with event nominals results from the presence of a full VP in these structures. It is the head of this VP that assigns thematic roles to the arguments, not the derived nominal. Result nominals, on the other hand, do not include a VP, and hence lack argument structure. Similarly, simple event nominals, are not associated with a syntactic VP, and they do not have an argument structure. The fact that they refer to an event has no interesting syntactic consequences. Thus, the proponents of the VP analysis distinguish between a verbal
Introduction
domain, which licenses the presence of arguments, and a nominal domain that renders the verbal head into a nominal. A version of this proposal is depicted in (31):
(31)
DP Spec D
D¹ NP N¹ N°
VP
a³x V NP N° hosts the nominalization affix. The verb gets associated with it by means of head-movement that takes place in the syntax. Result nominals differ from process nominals in that they are directly inserted under N°. The most convincing argument in favor of the view that a VP is present inside nominalizations would be the existence of cases showing that crucial properties standardly associated with VPs also show up with event nominals. Two obvious such cases are: (i) adverbial modification and (ii) assignment of accusative case. However, no adverbs and no accusative case marked objects occur in English nominalizations (see Borer 1993): (32) a. I made Pat collect mushrooms secretly b. Pat’s collection *(of) mushrooms (*secretly) went on all afternoon
There are languages, however, in which modification of a nominal by adverbials is possible. For instance, Hebrew and Greek permit adverbs in process nominals, but disallow them in result nominals. Consider the following sentences (Hebrew data taken from Hazout 1995, for Greek see also Alexiadou 1997, Alexiadou and Stavrou 1998a): (33) Harisat ha-cava et ha-kfar be-axzariyut destruction the-army acc the-village cruelly ‘The army’s destroying the village cruelly’ (34) a.
i katastrofi ton egrafon prosektika/me prosohi the destruction the documents-gen carefully/with care b. *i katastrofi prosektika the destruction carefully
Functional structure in nominals
In both the Hebrew and the Greek example, an adverb/adverbial phrase is present together with a process nominal. As the ungrammaticality of (34b) shows, when the nominal is used under its result interpretation, signaled by the absence of the DP complement, the presence of the adverb leads to ungrammaticality. In other words, the occurrence of adverbs in these constructions is syntactically conditioned, since the adverb cannot appear unless the rest of the complement system accompanies it. Under the standard assumption that adverbs modify VPs and not NPs (see Jackendoff 1977), the presence of adverbs in process nominals is unexpected. (33) further shows that Hebrew process nominals permit objects bearing accusative case. Greek, like English, differs from Hebrew in that it does not allow accusative case marking on the DP argument. At first sight then, the data discussed here seem to provide a clear argument for the presence of at least a VP inside process nominals. In the next section I briefly outline my proposal for the internal structure of process and result nominals respectively.
. The fine structure of process nominals: a first encounter The proposal I make here (Chapters 2 to 4) with respect to the internal structure of (process) nominals is as follows: process nominals include nominal as well as projections that we standardly associate with verbal clauses. Specifically, event nominals include Voice/v and Aspect, projections linked to the expression of eventivity and Viewpoint Aspect. Result nominals, on the other hand, lack such verbal projections (see also Schoorlemmer 1995, Borer 1999, van Hout and Roeper 1998 for related views). As a consequence, externally, process nominals behave like nouns. Internally, however, they display verbal properties. My proposal then takes the notional type event/process to be associated with functional layers and not with a specific lexical category, in the spirit of much recent work, and it is embedded within the framework of DM outlined in Section 1.4. Hence my structural account of the differences between process/event and result nominals capitalizes on the possibility of embedding lexical roots under a different set of functional projections in each case. My task then is to (a) determine the number and the type of verbal and nominal projections found within process nominals and (b) determine their function in accounting for the differences between the two nominal types. As will be shown in detail in Chapter 4, variation in the number and the type of the projections
Introduction
included in the nominal structure is responsible for the various types of nominals found within a language and across languages. Specifically, (35) below represents the ‘minimal’ structure which surfaces in various ‘eventive’ environments (Alexiadou 1999a,b for nominalization, Embick 2000a,b for participles):
(35)
AspectP Aspect¹ Aspect°
vP v
LP L° √DESTROY
Comp (=theme) the city
I take the following properties to be associated with the functional heads in (35). Following Kratzer (1994a, b) and Chomsky (1995), and see also Harley (1995), Marantz (1997), Arad (1999), v (i)
is the locus of agentivity, i.e. it contains features relevant to the licensing and interpretation of external arguments. (ii) contains features related to eventivity. (iii) bears the Case features for the object. (iv) comes in two types: one that introduces an external argument, and one that does not. On the other hand, the verbal functional head Aspect contains features that relate to the semantic properties of the event denoted by the verb: for instance perfective for a completed event, imperfective for an ongoing event. The feature specification of Aspect and of v, plus their presence vs. absence from the internal architecture of nominals, will be argued to be the source of much of the variation found in nominalizations across languages and within a language. A few notes on the dual nature of v are in order here. Harley (1995) argues that light v, which she labels EventP, can be of two different types: Cause and Happen/Become. Only the first one can have a specifier, which will end up being an external argument with an agent/causer thematic role. If the head is of the type Happen/Become, it will not merge with an external argument. It thus appears with unaccusatives. EventP contributes to the event interpretation
Functional structure in nominals
while the lower VP rootP contributes information not related to eventiveness of the verb (e.g. Aktionsart). Collins (1997) also proposes that for unaccusative verbs v, is present, but is somehow deficient: it does not check accusative case and it does not assign an external theta-role. In other words, properties (i–ii) create the semantic content of little v. On the other hand, properties (iii–iv) are the transitivity property of little v (Burzio’s generalization). The two sets of properties may be dissociated from each other: there are verbal heads that share the semantic content of v but not its transitivity property. A case in point is passive formation, i.e. the sentence the book was read does have an eventive interpretation and implies the presence of an agent, although it lacks the transitivity property. What is crucial here is that both transitive and non-transitive strings denote events. The difference is that the former denote an event that is brought about by an external causer. The properties of v will become crucial in the discussion in Chapter 3 concerning the (in-)transitivity of process nominals. Nominalizations, in spite of their similarity to verbal clauses, crucially differ from them in that no accusative case is assigned to their DP argument, and that no agent is syntactically projected in Spec,vP. If both these attributes are associated with v, then the conclusion is that nominalizations either lack such a head or only include the type of v found with unaccusative predicates. Since, however, the eventive readings of these nominals are linked with v, I will conclude that in such constructions, the type of v included is the deficient one, i.e. the one that does not assign an external argument, and does not check accusative case. Thus, nominalizations constitute a reflex of Burzio’s generalization. This result brings the proposed structure for nominalized clauses close to analyses of the nature of ergative languages. According to some authors, ergative languages are reflexes of a passive/unaccusative system (see Hale 1970, Nash 1995). In fact my investigation of nominalization patterns attempts to substantiate the hypothesis that aspects of the syntax of DPs of nominative– accusative (NA) languages are strikingly similar to aspects of the syntax of ergative (E) languages. Specifically, I show that constructions such as the destruction of the city by the enemy can be argued to have a direct reflex in the syntax of E languages. In this sense, certain E patterns receive an analysis analogous to the one proposed for these nominalizations. I argue that in both contexts, the by-phrase and the ergative phrase respectively can be analyzed on a par as instantiations of deep-causers/instrumental phrases. The unifying point between the two will be the proposal that both nominalizations in N/A languages and E languages have deficient vs (Chapter 5). On this view, in both
Introduction
nominalizations and patterns of ergativity there is only one syntactic argument checking structural case, and only one functional head bearing Case features in the respective structure. Coming back to the structure of nominals, my proposal is that process nominals include (35) (Alexiadou 1999a,b) in addition to nominal functional projections such as Number or Agr (36a). On the other hand, result ones do not do so and thus have the structure in (36b).
(36) a.
DP D°
FP (Numb/AgrP)
AP
FP F°
AspectP Aspect¹ Aspect°
vP v
LP L°
Comp (=theme)
DP
b. D
FP F°
LP
The structures in (36) differ from the structure in Marantz (1999) in a crucial way: in (36) no category changing position, i.e. n°, is included. In principle Marantz’s system could explain the eventive nature of certain nominalizations by allowing affixation to take place above v. However, the intuition behind Marantz’s proposal is that there is a notion of a nominalizer/adjectivizer, even if this notion is in a sense functional. This view introduces a theory internal inconsistency, and hence will not be adopted here. Rather, the structure in (35) is ‘spelled-out’ as verb or noun, depending on the general environment it is inserted: noun in the domain of D/Number, verb in the domain of Tense. Identical or distinct nominal affixes attach to positions that differ in height, i.e. below or above Aspect, below or above v, at MS. Thus I derive the ambiguity
Functional structure in nominals
observed with certain nominalization affixes, i.e. the fact that they do not discriminate between a process and a result reading, from a difference in ‘attachment height’. Note also that this proposal for the structure of nominals differs from Embick’s treatment of participles, which derives all their properties from features located in functional heads, in that it does not recognize the need to include an aspectual type of head specified [RESULT] within result nominals. I come back to these issues in Chapter 2. On my account, the difference between result and process nouns relates to the presence vs. absence of certain verbal-like functional projections, much like the VP analysis of process nominals. However, it differs from this approach in that it allows for result nominals to take complements as well, since both types are derived from unspecified roots that can take internal arguments. Note that in order to account for the fact that result nominals of the type the announcement that John resigned also select for complements in Grimshaw’s system, one has to draw a distinction between complements present in lexical conceptual structure, possible with both types of nominals, and arguments projected in argument structure, possible only with process nominals. Here this discrepancy does not arise. Event readings are associated with functional layers, not with the presence of argument structure.
. Nominalization patterns across languages According to my proposal, words and categories are the result of combining abstract roots with functional projections. Evidently, such a formation owes a lot to the presence vs. absence of certain functional layers, to their feature specification, and to their selection requirements. This way of viewing the internal structure of DPs offers a new way of interpreting the nominalization patterns found across languages and within a language. As I point out in detail in Chapter 4, the various types of nominals reflect differences in the number and the type of functional projections. That is the agentive nominal dancer does not contain the same set of nominal or verbal functional projections as the noun destruction. In order to briefly outline how this idea works consider (37), an instance of a transitive nominalization: (37) John’s destruction of the city
Introduction
While (37) is grammatical in English, it is ungrammatical in Greek. In Chapter 4 I argue that the crosslinguistic availability of constructions of this type is related to the availability of Spec,DP to function as a thematic position. Hence such constructions are possible in English, where Spec,DP is a thematic position, but they are not possible in Greek, where Spec,DP functions as an A¹-position (see Chapters 2 and 4). Interestingly, the investigation of nominalizations of this type will bring me again to a discussion of ergativity patterns from a different perspective. In many discussions of the origin of ergativity, nominalization is also advanced as one possible source for it (see Comrie 1978). On these views, ergative marking is considered to be either (a) the result of the reanalysis of nominalizations in which the agent is expressed by means of a possessive phrase as in the example John’s destruction of the city above (see Johns 1992, Siewierska 1998 for a recent discussion and references), or (b) to be related to constructions such as the destruction of the city by the enemy (see Comrie 1978 and the discussion in Section 3). Though the view that aspects of N/A syntax resemble domains of E syntax has been formulated in the form of hypotheses in other places as well (see Bok-Bennema 1991, Postal 1977, Williams 1987, among others), to the best of my knowledge no detailed account seems to exist. Here, by investigating both these constructions in detail within the nominal syntax of N/A languages, I revisit typological hypotheses on the source of ergative marking and I attempt to ultimately identify the constructions in which these two different markings, i.e. the ‘passive’ and possessive are encountered in different languages and constructions. Crucially, the view on the internal structure of DPs generalizes: variation in functional structure gives different types of categories, clauses and languages. In certain parts of this book, I pay special attention to empirical data taken from Greek. However, this study adopts a strong comparative methodology. It compares characteristics of Greek grammar to properties shown by other grammars, drawing conclusions of general interest.
. Outline The book is structured as follows. Chapter 2 contains a brief overview of the literature on the structure of DPs; it further presents certain aspects of nominalization in Greek and my proposal for the internal structure of process nominals in detail. Chapter 3 provides empirical support for the claim that
Functional structure in nominals
process nominals are actually unaccusative/ergative constructions and discusses passive nominals in some detail. Chapter 4 discusses the variation in nominalization types found across languages and within a language. Chapter 5 turns to an investigation of the similarities between nominalization and ergative languages by first examining the respective Case patterns. Second, given that in the literature ergative case is either seen as a prepositional or as a possessor case, the chapter looks at the relation between these two patterns and (transitive) nominalizations. In addition, the formation of the (possessive) perfect across languages is dealt with, since such constructions tend to appear in the perfect. Chapter 6 summarizes the conclusions drawn in this work.
Notes . It is assumed that the gerunds in (1) and (3) have two different uses: a propositional and an eventive one. What is the relation between the two syntactic forms, nominal and verbal gerund, and the two semantic forms, propositional and eventive (see also Zucchi 1993, Katz 1999a, Hamm 1999 for a recent discussion)? Vendler (1967) points out that the question whether we are dealing with a propositional or an eventive use of the gerund has a great deal more to do with the context in which it occurs (Vendler’s container) than the syntactic form of the gerund itself. Although the verbal gerund is almost always usable in contexts in which a propositional reading is natural, and the nominal gerund is always usable in contexts in which an eventive reading is natural, the forms mix quite freely in many contexts. Fraser (1970), however, notes that these two -ing forms clearly receive a different semantic interpretation, namely (1) is interpreted as an assertion of a fact, while (3) is interpreted as an action, an activity, an event. Derived nominals can almost always replace nominal gerunds. If Fraser’s view is on the right track, then only (2) and (3) constitute true action nominalizations. . The following discussion draws from Borer (1993), Carstairs-MacCarthy (1992), Marantz (1999), and Embick (2000c) among others. It should be noted that what follows is not meant as an exhaustive overview of the field. For approaches to word formation in the early days of Generative Grammar see Lees (1960) and Chomsky (1965). Lees took the position that nearly all (if not all) nominalizations, i.e. noun phrases between whose parts there is a syntactic relationship should be derived from sentences through the use of syntactic transformations. See Wunderlich (1994) for a revision of the categories put forth in Chomsky (1970). . A number of operations take place at MS. Merger involves structurally adjacent nodes, whereby the heads of two independent phrases are brought together but nevertheless remain separate morphemes in the newly formed item. Fusion brings together sister terminal nodes under a single terminal node, as is the case of the single affix for Number and Case in Greek (and many other Indo-European languages). Finally, fission results in the split of features carried by one node into a sequence of nodes. Fission and Fusion are
Introduction
the two main morphological processes that immediately disturb the isomorphism between syntactic and phonological features. The application of the operations that modify the syntactic tree is completed before vocabulary insertion at MS. At the same time, the addition of terminal nodes at MS, changes the number of terminal elements that might be phonological realized and thus contributes to the noted lack of isomorphism between PF and SS (see Halle and Marantz 1993: 115). . Arad (1999) points out that in Romance and Semitic the same root can form different types of predicates. This is exemplified below for Italian. This fact supports the view of the lexicon put forth within the DM model. (i) a. b.
√romp + ere (break, causative) √romp + ersi (break, inchoative)
. However, Davis and Matthewson (2000) point out that the group of Salish languages, which has been traditionally analyzed as lacking the V-N distinction provides evidence that the roots carry lexical specification. . In principle, the generalizations to be made could be expressed if one were to propose that process nominals include ‘passive’ VPs (see for instance van Hout and Roeper 1998 for a recent discussion). However, as it will be shown in detail, process nominals do not behave like passive VPs. . A similar view to the one outlined above is presented in Williams (1987) who proposed that there is no distinction between nominalizations with and without argument structure. Rather, all nouns are equipped with an argument structure. In fact in Williams (1981) it is proposed that nominalization, which he regards as a morphological rule, applies to verbs to form a new lexical item, affects the argument structure of these verbs, in that it forces internalization of the external argument. In a sense, the main idea of my proposal eventually supports the view that all nouns have an argument structure, and that agents are internalized within process nominals, but it substantiates the hypothesis that result nominals differ from process nominals in certain functional aspects of their internal structure. This distinction is not drawn in Williams’s work. . However, it has been observed that in several languages, e.g. Hungarian (see Szabolcsi 1994) and Greek (see Markantonatou 1992), process nominals can also be plural, a property which correlates with the aspectual class the nominal belongs to, i.e. achievement vs. process (see next chapter). . A similar contrast is found in German (from Audring 1999): (i) Jans Prüfung John’s Exam (ii) *Jans Prüfung der Tatsachen John’s examination of the facts The presence of a possessor clearly brings about the result interpretation. . Naturally, a number of semantic accounts exist as well; see in particular Ehrich (1991 and subsequent work), Bierwisch (1989), Zucchi (1993), Pustejovsky (1995), Hamm (1999) and references therein. Reference to these is made when necessary.
Functional structure in nominals
. Szabolcsi’s (1994) analysis is similar to Grimshaw’s in that both account for the eventive reading of nominals in terms of the inclusion of an (abstract) event argument in the relevant structures. However, she differs from Grimshaw in that she argues against the suppression of the external argument in passive nominals and in favor of its construal as PRO (controlled or arbitrary). . While affixes such as -tion, -ing and -er seem to be able to co-occur with the complements of their parental verb, it has also been observed that many deverbalizing affixes prohibit the presence of the underlying verb’s arguments (cf. Roeper 1987). (i) a. *an employee by Mary b. *a trainee with a great effort(cf. van Hout and Roeper 1998) c. the trainer of dogs d. the saver of lives It has been pointed out in the literature that affix -ee binds a patient argument, while -er binds an external argument. On the other hand, -ion binds something like a theme. An analysis that capitalizes on the thematic binding properties of affixes crucially believes that stems and affixes contribute to the interpretation of the lexical item, and that affixes play a role in licensing arguments in the lexical conceptual structure (see also di Sciullo and Williams 1987, and references in Grimshaw 1990). Marantz (1999) distinguishes between -ee and -er nominalizations in suggesting that the former include a functional nominal head that embeds an unspecified lexical phrase, while the latter include a functional nominal head that embeds a verbal projection (see the discussion in Section 1.4). A similar view is defended in van Hout and Roeper (1998). . However, Fu et al. (1998) have claimed that English nominalizations also permit adverbs. Not all native speakers agree with these judgements. (i) a. b. c.
??the protection of the children completely ??the removal of evidence purposefully ??the explanation of the problem immediately
. Siloni (1997) observes that there is an interesting fact concerning the distribution of adverbs in Hebrew process nominals: the adverbs one finds with process nominals always have a PP structure, such as the one in (33) which is preceded by the preposition be. Adverbs lacking a PP structure are not found. This does not hold for Greek. Here simple adverbs are found as well as PPs. . In fact affixes such English: -ize ‘nationalize’, -en ‘deep-en’ or Greek, -pio: kako-pio ‘harm’, -ono: veve-ono ‘certify’ can be seen as overt realizations of the functional head v. . Alexiadou and Stavrou (1998a) argue for the presence of a single functional projection bearing the event property that selects for a VP inside nominalizations. . Note that there are constructions where accusative is assigned in the absence of an external argument. For instance, Northern Russian passives discussed in Timberlake (1976), and the Japanese passives discussed in Harley (1995) and references therein. . From the point of view of a N/A language, the E system is most reminiscent of passive
Introduction
constructions in a nominative/accusative language. Thus, passive constructions were initially advanced as a possible source for ergativity (see Hale 1970). . Van Hout and Roeper (1998) assume an N° head containing a nominalization affix. . Note that the hypothesis to be entertained here makes an interesting prediction with respect to the ‘lexical’ categories one can find across languages. For instance, if a language lacks Tense as a separate category, then this language is expected to neutralize the difference between verbs and nouns. As will be shown in the next chapter, this prediction appears to be borne out.
Chapter 2
The Functional Architecture of Nominalizations
In this chapter I substantiate the proposal that the functional structure of event nominals includes an Aspect Phrase (AspP) and a light v/Voice Phrase (vP). The chapter is organized as follows. In Section 1 I first give an overview of the literature on the DP. Large part of the discussion in the literature on nominalizations has been devoted to the question of their internal structure. Thus, reviewing the literature on the status and the nature of DP internal functional projections serves as a general background for my purposes. Section 1.4 is devoted to a presentation of the general properties of the Greek DP. In Section 2 I discuss the various verbal classes, i.e. transitive, intransitive and di-transitive, and the types of nominals derived from them in terms of the process vs. result distinction. This section offers a detailed study of derived nominals in Greek and compares Greek nominalization patterns with the betterknown cases from English, Romance, Slavic and Hebrew/Arabic. In Section 3 I concentrate on the verbal properties of event nominals and I present my account of the structure of process nominals, which crucially relies on the presence of AspectP and vP within their functional architecture. I also briefly discuss the internal structure of result nominals. In Section 4 I present a number of arguments against the presence of TenseP inside process nominals. In Section 5 I turn to the conditions enforcing the licensing of argument structure within process nominals. In Section 6, building on Embick (1998), I give an outline of the realization of the morphological object noun in a system that denies the existence of lexical categories as primitives, such as the one adopted here.
.
Introduction
. NPs are dominated by D A more recent issue in the investigation of noun phrases concerns the functional structure they contain. Traditionally noun phrases were taken to be the
Functional structure in nominals
maximal projections of nouns, as depicted in (1) (see e.g. Jackendoff 1977):
(1)
NP Determiner
N¹ N
The theoretical developments in the eighties have called this structure into question. More specifically, the extension of the X-bar schema to the sentential functional elements (Chomsky 1986) and the increasing work on headmovement (Baker 1988) have led researchers to elaborate a more articulated syntactic representation for the noun phrase. Arguing that the functional nominal material should fit into the X-bar schema, Abney (1987), Fukui and Speas (1986), Horrocks and Stavrou (1987), and Szabolcsi (1983) among others hypothesized that noun phrases, like clauses, are headed by a functional element. Horrocks and Stavrou (1987) label this functional head Art(icle), while Abney (1987) calls it D(eterminer), and proposes that noun phrases are the maximal projections of D, the base position of articles. Abney’s proposal capitalizes on the symmetry between NPs and clausal projections. Clauses are actually VPs dominated by a functional projection, IP. Abney proposes that in the same way the noun phrase should be seen as a projection of N dominated by a functional projection, as shown in (2):
(2)
DP D¹ D
NP
The proposal for the introduction of a functional category in the NP structure has received strong empirical support by a series of studies (cf. Ritter 1987, Delsing 1988, Ouhalla 1988, Taraldsen 1990, Haider 1988 among many others). The introduction of the functional category D has been linked with the presence of overt noun movement in a number of languages, e.g. Romanian and Scandinavian. In fact the syntactic behavior of articles in these languages, i.e. their affixal character, has led linguists to identify the landing site of the raised noun as D, the base position of the article (see Cornilescu 1993 for Romanian and Delsing 1993, Taraldsen 1990 for Scandinavian among others). Alongside this research, several studies have developed a principled explanation of why the nominal expression is the maximal projection of D (cf.
The functional architecture of nominalizations
Szabolcsi 1987, Stowell 1989, Longobardi 1994). The basic insight is that D is the element that converts the nominal expression into a referential phrase that consequently can be used as an argument. In this respect, D can be argued to parallel the complementizer of sentential complements: each turns its complement (NP and IP respectively) into an expression that can appear in an argument position. As will be shown in 1.3, this property correlates with the syntactic status of D in certain languages. . AgrP, NumberP and GenderP Subsequent studies of Germanic, Semitic and Romance languages have suggested that the structure of noun phrases is indeed more articulated and includes additional inflectional structure between DP and NP (see e.g. Ritter 1991, Valois 1991, Cinque 1993, Bernstein 1993, Fassi-Fehri 1993 to mention a few). For instance, Szabolcsi, on the basis of Hungarian data such as the ones in (3) below, proposes that AGR is present within the DP: (3) a.
az en kalap-om the I-nom hat-1sg b. a te kalap-od the you-nom hat-2sg
Kalap ‘hat’ is a noun that agrees with its possessor, marking its person and number. The possessor phrase bears nominative case, as would the subject of a sentence. If one maintains the idea that AGR assigns nominative, then we conclude that NPs in (3a) and (3b) contain an AGR as well. Ritter (1991) provides evidence for the existence of a different type of functional projection between D and N in Hebrew. The projection is labeled Number Phrase and is taken to be the locus of plural affixes. According to Ritter, the affixation of plural marking on nouns is similar to the affixation of tense and agreement affixes on the verb. In the same vein, Picallo (1991) observes that all Romance nouns express Gender and Number distinctions (cf. 4a). Picallo claims that the order in which the suffixes appear at S-structure, reflects that successive cyclic movement of the head N up to Num° has applied (cf. 4b). (4) a.
el gat, la gata, els gats, les gates the cat(s) M/S, F/S, M/P, F/P b. Les [NumP noveles d’[GenP en Pere de [NP Nabokov the novels of Pere of Nabokov
Functional structure in nominals
In related work, which, however, crucially capitalizes on the relative position of the head noun with respect to the adjectives modifying it, Cinque (1993) and Crisma (1993), (1995) present arguments for the presence of additional functional projections. For instance, in (5) the noun ‘destruction’ must appear to the left of the adjective ‘Roman’. Assuming that the referential adjective is located in a specifier position in the extended projection of the noun, the patterns below show that overt N-movement to an intermediate projection applies in Italian. Actually, the noun in Italian is claimed to move to a position higher than NP and the category immediately dominating it, since it leaves behind the thematic adjective romana, as well as classifying adjectives like ‘bronze’ (from Zamparelli 1994): (5) a.
La distruzione romana di Cartagine the destruction Roman of Carthago b. *La romana distruzione di Cartagine c. Le ultime tre famose sculture bronze della moglie the final three famous sculptors bronze of the woman
. On the A vs. A¹ status of Spec,DP Apart from determining the functional architecture of the DP, another issue that has preoccupied the syntactic literature has been the status (A vs. A¹) of Spec,DP. Under Abney’s (1987) account, Spec,DP is a thematic-position in English. Evidence for this comes from the fact that no expletives can be licensed within English DPs, and no raising takes place either (cf. 6; but see Section 4.1 for a different account). (6) a. *there’s arrival b. *Mary’s appearance to be sick
On the assumption that expletives are licensed in non-thematic positions and that raising involves movement from a theta-position to a non-theta position, the ungrammaticality of the examples in (6) is straightforwardly accounted for, if Spec,DP is a theta-position. Horrocks and Stavrou (1987) argue that a crucial difference between English and Greek is that in English Spec,DP corresponds to Spec,IP, thus being an A-position, while in Greek it corresponds to Spec,CP, thus being an A¹-position. Since Spec, DP is not an argument position in Greek, whmovement naturally takes place within this domain. Their argument is based on examples such as the following:
The functional architecture of nominalizations
(7) a. a.¹ b. b.¹ c. c.¹
to vivlio tu Chomsky the book the-gen Chomsky tu Chomsky to vivlio i kritiki tu vivliu the review the-gen book-gen tu vivliu i kritiki to endhiaferon ja to arthro afto the interest about the article this ja to arthro afto to endhiaferon
In each of the primed examples in (7) the constituent that follows the head N in the non-primed examples is fronted to a pre-N position. The interpretive effect of such fronting is one of focalizing. According to Horrocks and Stavrou (1987:86), all this is obviously reminiscent of the fronting of constituents that takes place in sentences for the purpose of bringing a particular constituent into prominence. Sentential focusing in Greek is illustrated in (8): (8) a.
edhose to vravio tis Afrodhitis gave-3sg the prize-acc the-gen Aphrodite-gen ‘he gave the prize to Aphrodite’ b. tis Afrodhitis edhose to vravio c. to vravio edhose tis Afrodhitis
As shown in the examples above, the indirect object tis Afrodhitis and the direct object to vravio may be fronted for focalizing effects. Horrocks and Stavrou (1987) show that focalization in the clause has the properties of A¹-movement (see also Tsimpli 1995). In (8) focalization takes place within the nominal projection. If the fronted constituents in the examples in (8) occupy [Spec,DP], this means that [Spec,DP] is similar to [Spec,CP]. The parallelism between interrogative clauses and interrogative DPs in Greek strengthens the assumption that DP is to NP what CP is to VP. (9a) is an echo question: the wh-constituent ti (‘what’) does not move to the sentenceinitial position. In (9b) the wh-constituent is fronted (all examples from Horrocks and Stavrou 1987). (9) a. ekane ti did-3sg what ‘He did what?’ b. ti ekane what did-3sg
Functional structure in nominals
The same pattern is found DP-internally: (10) a.
to vivlio tinos the book who-gen ‘whose book’ b. tinos to vivlio
Example (11), Horrocks and Stavrou’s 1987:89 (14), illustrates the interaction between wh-movement at the clausal level and DP-internal wh-movement. (11) a.
b. c. d. e.
mu ipes oti diavases [to vivlio tinos] me told-2sg that read-2sg the book whose ‘You told me you read whose book?’ mu ipes oti diavases [tinosi [to vivlio ti] tinosi mu ipes oti diavases to vivlio ti [tinosj [[to vivlio tj]]i mu ipes oti diavases ti [to vivlio tinos]i mu ipes oti diavases ti
All the above examples are fully grammatical. Example (11a) is taken to represent the structure prior to movements. In (11b), the genitive phrase has moved to Spec, DP. From this position subsequent movement to Spec, CP as in (11c) is possible. (11d) and (11e) involve fronting of the whole DP to Spec,CP. In the former case wh-movement has not applied DP internally, in the latter it has. Hungarian offers further confirmation for aligning the D node with C, as shown by Szabolcsi (1994). The Hungarian pre-nominal possessor may occupy two positions. It may follow the determiner, in which case it has nominative case (12a), or it may precede the determiner, in which case it has dative case, as shown in (12b). (12) a.
a Mari kalap-ja the Mari-nom hat-3sg b. Mari-nak a kalap-ja Mari-dat the hat-3sg
It seems reasonable to propose that Marinak in (12b) occupies the specifier position of DP, as shown in (13).
The functional architecture of nominalizations
(13)
DP Spec Marinak
D¹ D a
NP kalapja
Topicalization and wh-movement in Hungarian suggest that [Spec,DP] is also an escape hatch for A¹-movement. In (14), the dative possessor Marinak is topicalized independently from the noun with which it is construed. Such topicalization of the nominative possessor is not possible (cf. 15). In (16), similarly, the interrogative dative possessor has undergone wh-movement independently of the noun with which it is associated: (14) [CP [TopP Marinaki [FocP Peter làtta [IP [DP ti a kalapja]]]]]. Mari-dat Peter saw the hat ‘Peter saw Mary’s hat’ (15) *[CP [TopP Marii [FocP Peter làtta [IP [DP a ti kalapja]]]]. (16) [CP [FocP Kineki làtta [IP Kati [DP ti a kalapja]]]? whose-dat saw Kati the hat ‘Whose hat did Kati see?’
Nevertheless, in studying the parallelisms between clausal and nominal structure, researchers have often noticed the defective character of nominals in their functional structure and the fact that such projections are radically different from the ones encountered in the extended projection of the verbs. Thus, functional projections inside the DP make reference to morphological properties of nouns and thus are related to number and gender. On the other hand, the presence of a Voice Phrase, an Aspect Phrase, a Tense Phrase or even a Mood Phrase inside the DP is a bit more difficult to substantiate. In Section 3 evidence for the presence of certain of these categories inside process nominals will be presented. In the next section I turn to a brief examination of the Greek DP. . Some notes on the Greek DP The nominal inflectional system of Greek, as manifested both in deverbal and in common regular nouns is rather rich (see Alexiadou and Stravrou 1998b for a detailed discussion). It manifests morphological distinctions for three distinct
Functional structure in nominals
grammatical genders (masculine, feminine, neuter) and two numbers (singular and plural) on both the noun and the determiner, as well as any adjectives modifying the head noun (see 17). In this respect Greek is similar to Romance, although it displays one more grammatical gender (neuter). However, Greek is unlike Romance in that it further displays both a full case system on the noun (like Latin) and the determiner. (17) a.
ti megal-i katastrof-i the-fem:sg:acc big-fem:sg:acc destruction-fem:sg:acc b. ton megal-on katastrof-on the-fem:pl:gen big-fem:pl:gen destruction-fem:pl:gen
There are four distinct morphological cases in the singular and four in the plural, nominative, accusative, genitive and vocative (vocative is in most of the cases non-distinct from nominative; genitive is morphologically distinct from the nominative in most declensional classes). Case, number and gender appear in the form of inflectional affixes on the noun. However, there is a high degree of syncretism in the language, whereby different features are realized by the same affix. Unlike Romance, number and gender are never affixed to the stem by distinct affixes (cf. the contrast between the Spanish nominals in (18) and the Greek ones in (19); see Ralli 1994; 1998, for a detailed description of the Greek nominal inflection system): (18) a.
l-a pelicul-a (sg.) the film
b.
l-a-s pelicul-a-s (pl) the films
(19) a.
o anthrop-os (sg.) ‘the human’
b.
i anthrop-i (pl) ‘the humans’
But neither is case signaled by a distinct affix from gender and/or number (anthrop-os masc:sg:nom, anthrop-u masc:sg:gen). In Greek, as has been shown convincingly by Ralli (1994, 1997), gender as opposed to number, is not a feature of the inflectional affix but it is an inherent feature of the noun stem. This in turn means that gender specification will show up under the terminal node N°. It is reasonable then to assume that this intermediate functional projection relates to the features of number, features that are syntactically relevant (i.e. determined by syntax, see Harris 1991 for Spanish). In other words, NumberP is one of the functional projections that provide instructions for LF (see also de Wit 1997, Willim 1995), and thus is present for independent reasons in the structure (see the discussion in Chapter 1).
The functional architecture of nominalizations
(20)
DP D
FP F°
NP
Number
N
In (20), one could argue that it is actually Number that is primarily responsible for nominalizing unspecified roots. A further characteristic of Greek is that it does not show any effects of noun displacement within DP, at least with respect to the criteria that have been crosslinguistically established, i.e. affixal articles and adjective placement. As mentioned in the previous section, Cinque (1993) and related literature have suggested that the relative order between adjectives and nouns is a criterion for assuming noun-movement in a language or not, under the assumption that adjectives attach to the same sites crosslinguistically. On this view, languages in which nouns precede adjectives are languages with noun movement. In Greek, all adjectives always precede the noun, as is illustrated in the examples in (21) and (22): (21) *to spiti megalo the house big (22) *i katastrofi italiki the destruction Italian
Moreover, given that the definite article is not an affix in Greek, no argument in favor of overt N-to-D movement can be formulated either. With this general background, let us have a closer look at Greek nominal ‘formation’ primarily.
. Greek nominal formation . General remarks Much like their counterparts in English, Semitic and so on, Greek deverbal nouns pattern externally like common nominals, i.e. they appear in positions restricted to nouns, e.g. complement positions of verbs and prepositions.
Functional structure in nominals
(23) a.
ida tin katastrofi tis polis saw-1sg the destruction the city-gen ‘I saw the destruction of the city’ b. lipithike me to kapsimo ton vivlion felt-sorry-3sg with the burning the books-gen ‘He felt sorry about the burning of the books’
As with common object nouns, derived nominals follow the determiner, and the adjectival modifiers. DPs in the genitive follow the noun as in the case of possessives in non-derived nominals (cf. (24a) and (24b)). Note that both in derived and non-derived nominals, the genitive DP can appear in initial position, that is before the definite article (cf. (25)). As mentioned in Section 1.3, these patterns have been analyzed in Horrocks and Stavrou (1987) as involving focalization movement to Spec,DP from the complement of N-position, an instance of A¹-movement: (24) a.
i katastrofi tis polis the destruction the city-gen ‘The destruction of the city’ b. to vivlio tu Petru the book the Peter-gen ‘Peter’s book’ (25) a. tis polis i katastrofi the city-gen the destruction ‘The city’s destruction’ b. tu Petru to vivlio the Peter-gen the book
There are a number of nominalizing affixes in the language. Some of the nominals consist of a stem (very similar to that of their related verb) and they are followed by an affix that carries gender and number information (cf. 26), as discussed above (see also Valetopoulos 2000). In (26) the affix -i is added to the verb stem, which undergoes morpho-phonological conditioning, to give a nominal. (26) i katastrof-i the destruction-fem:sg
V=katastref-o destroy-1sg
Other nominal endings are -ma/(i)mo(s), -(s)i, -ito, -ura, -ala, -idi, -tis. In some cases, the infixes m and s appear in combination. The vowel -i- that precedes
The functional architecture of nominalizations
the consonants in each case does not contribute to the meaning of the strings: (27) a.
to lis-i-m-o the untying-neut:sg b. to raps-i-m-o the sewing-neut:sg c. ksefon-i-t-o scream-ing-neut:sg d. kend-i-d-i embroidery-neut:sg e. horef-tis danc-er
lino untie-1sg ravo sew-1sg ksefon-o scream-1sg kend-o do embroidery horevo dance-1sg
As Valetopoulos (2000) points out, the affixes ‘idi’, ‘ito’, ‘ura’ and ‘ala’ do not seem to support argument structure or to have event readings for that matter. Hence I assume that these morphemes attach to the root, before this combines with any higher functional layers (see Chapter 1) and I will not be discussing examples containing them any further. The affix -tis is used for the formation of agent nominalizations, as it refers to the person performing an action. As was the case with English, most of the affixes do not disambiguate between an event and a result reading. This is possible only when the nominal is put in a context that brings about one or the other interpretation, for instance aspectual modification or predication. Consider the following examples. (28) a. *i paratirisi tu Jani epi dio ores itan anakrivis the observation the John-gen for two hours was inexact ‘the observation of John for two hours was inexact’ b. i paratirisi tu Jani itan anakrivis the observation the John-gen was inexact ‘John’s observation was inexact’ c. i paratirisis tu Jani the observations the John-gen ‘John’s observations’
The noun paratirisi ‘observation’ is ambiguous between the event and the result reading. As Picallo (1991) and others observe, only results obtaining from certain activities can have the properties of being inconsistent/inexact. Events or processes do not have such possibilities. The examples above show precisely this contrast. Event nominals are not grammatical with predicates that denote
Functional structure in nominals
the state of being inconsistent. Hence (28a) is ungrammatical, but (28b) is fine. Note that the interpretation of the genitive argument is different in these two examples: in (28b) we interpret John as the person to whom the observation belongs. Similarly in (28c), where the noun is in the plural and thus it clearly expresses a result on the basis of Grimshaw’s criteria, John is interpreted as the possessor. The one interpretation that could be possible for (28a) is one where John is the person being observed. According to the main studies on nominalizations in Greek (cf. Markantonatou 1992, Kolliakou 1995, Alexiadou and Stavrou 1998b), the language shows deverbal formations derived from agentive transitive verbs (e.g. katastrofi ‘destruction’), and nominals derived from intransitive verbs, exemplified in (29a) and (29b) below. The latter distinguish between nominals derived from unaccusative verbs (e.g. afksisi ‘raise’), and nominals derived from unergative predicates (e.g. kolimpi ’swim’). (29) a.
katastrofi perigrafi metafrasi b. afiksi kolimpi skarfaloma
‘destruction’ ‘description’ ‘translation’ ‘arrival’ ‘swim’ ‘climbing’
In the next section I consider the various classes of nominals found in Greek classifying them on the basis of the type of verbs these are related to, i.e. transitive or not. For ease of exposition I use the standard terminology, but it should be kept in mind that I take transitivity and verb/category formation to be a derived property (see the discussion in Chapter 1). . Nominals derived from transitive predicates The first observation to be made about nominals of the type in (29a) is that this group splits into two classes. In the process reading both surface with internal arguments. The difference between the two classes is that nominals such as destruction always receive a process reading, hence the ungrammaticality of (30b). These nominals are derived from roots that denote a resultant state and have an implicit agent/causer component. Nominals such as description, on the other hand, are ambiguous between a process and a result reading. These seem to differ from destruction type nouns in that their complement is not ‘affected’
The functional architecture of nominalizations
and that their meaning alternates between a telic and an atelic one (see the discussion in Chapter 3): (30) a.
i katastrofi tis polis se pende lepta the destruction the city-gen in five minutes ‘The destruction of the city in five minutes’ b. *i katastrofi tis polis (se 5 lepta) itan anakrivis the destruction the city-gen (in 5 minutes) was inexact
(31) a.
i perigrafi tu Jani ja 1 ora the description the John-gen for an hour ‘The description of John for an hour’ b. *i perigrafi tu Jani ja mia ora itan anakrivis the description the John-gen for an hour was inexact c. i perigrafi tu Jani itan anakrivis the description the John-gen was inexact ‘John’s description was inexact’
In (30b) the predicate be inexact creates an environment which brings about the result reading. The example is ungrammatical even if the aspectual modifier is removed. This is not so with nouns such as description. Once the process interpretation is not available any more, i.e. when the aspectual modifier is not present, the example is grammatical (cf. 31c). On the process reading (31a), the genitive argument is always interpreted as the object/entity affected by the predicate, i.e. being destroyed or being observed. On the other hand, as we saw above with (28), in a context that brings about the result interpretation, the genitive argument cannot be interpreted as a theme; it is necessarily interpreted as a possessor (see the discussion in Chapter 1). Example (31) crucially contrasts with (30) where the city is only interpreted as the entity that got destroyed. Nominals such as those in (30) are referred to in the literature as passive (see Cinque 1980). There is a further difference between these two groups. A characteristic property of Greek nominals derived from agentive transitive predicates such as destroy is that no more than one single argument can be present bearing genitive case (see Horrocks and Stavrou 1987, Markantonatou 1992, Alexiadou and Stavrou 1998a). On the other hand, the presence of two genitives is possible with nouns such as translation, description etc (see e.g. Cinque 1980, Zubizarreta 1987 for discussion). Consider (32)–(33):
Functional structure in nominals
(32) a.
b. (33)
i katastrofi tis polis apo tus Italus the destruction the city-gen by the Italians ‘The destruction of the city by the Italians’ *i katastrofi tis polis ton Italon the destruction the city-gen the Italians-gen i metafrasi tis tragodias tu Kakridi the translation the tragedy-gen the Kakridis-gen ‘Kakridis’ translation of the tragedy’
(32b) shows that co-occurrence of two genitive phrases is impossible within destruction type nouns. The agent argument can appear only when it is introduced by a preposition (cf. (32a) and also Horrocks and Stavrou 1987). On the other hand, in (33) the presence of the two genitives is licit. I come back to this difference in Chapters 3 and 4 arguing in the latter that nominals that permit the two genitives are actually result nominals. That is, while the by-phrase necessarily denotes an agent in (32a), the second genitive denotes a possessor in (33). . Nominals derived from intransitive predicates Though Greek is a language lacking most of the standard tests distinguishing between the two classes of intransitive predicates, i.e. unaccusatives and unergatives, Markantonatou (1992), Alexiadou and Anagnostopoulou (1999) among others present certain criteria that can diagnose unacussativity in Greek. Hence in what follows, I assume that the unaccusative vs. unergative distinction is also observed in Greek. What is interesting for the discussion here is that unaccusative and unergative nominals do not behave alike with respect to the criteria distinguishing event from result nominals. As pointed out by Markantonatou (1992) and is illustrated in (34), aspectual modification is one domain where the behavior of these two types of nominals radically differs. (34) a.
i eksetasi ton fititon epi dio ores the examination the students-gen for two hours ‘The examination of the students for two hours’ b. i ptosi ton timon epi tris vdomades the fall the prices-gen for three weeks ‘The fall of the prices for three weeks’
The functional architecture of nominalizations
c. *to kolimpi tu Jani epi mia ora the swimming the John-gen for one hour
While nominals derived from unaccusative predicates happily co-occur with aspectual modifiers, this is not the case with nominals derived from unergative predicates, as the ungrammaticality of (34c) indicates. Thus, unergative nominals qualify as result nominals with respect to the criterion established by Lebaux (1986), Grimshaw (1990) and others. Moreover, unaccusative nominals can be modified by aspectual adjectives (cf. 35a), while unergative nominals are not compatible with such modifiers. And, as (35b) further shows, they are not compatible with agentive modifiers either: (35) a.
i sihni ptosi ton timon the frequent fall the prices-gen ‘The frequent fall of the prices’ b. *i siniditi/sihni ergasia tu ipalilu the conscious/frequent work the employee-gen
The genitive that appears within unaccusative nominals bears the thematic role of theme, much like the single DP argument found with their corresponding verbs. On the other hand, the single genitive that appears within unergative nominals receives a possessor interpretation (see also Chapter 3). Interestingly, Greek nominals derived from unaccusative predicates seem to challenge the view that process nominals are singular nouns (similar observations hold also for other languages). As Markantonatou observes, in certain instances unaccusative nominals receive the event reading only when the nominal is inflected for plural number. This is illustrated in (36) below: (36) i afiksi/√i afiksis turiston oli ti nihta the arrival/the arrivals tourists-gen all the night ‘*The arrival of tourists during the whole night’ √‘The arrivals of tourists during the whole night’
As Markantonatou points out, plural nominals in Greek qualify as event nominals, if they correspond to predicates that have no internal temporal structure and if their aspectual interpretation changes from perfective to imperfective, when the reference of their theme argument changes from a quantized to a cumulative one. As known, unaccusative predicates, and as a consequence, nominals derived from such predicates have necessarily telic readings. In fact
Functional structure in nominals
if the mass/count distinction in the individual domain corresponds to the process/state vs. achievement/accomplishment distinction in the event domain, telic nominals qualify as count events and should therefore license pluralization (see the discussion in Section 3.3). However, as noted by Markantonatou, a durative interpretation is also possible when the reference of the theme argument of unaccusative nominals is not quantized, as in (36). When the theme argument is quantized the example becomes ungrammatical, as shown in (37): (37) i afiksi/*i afiksis tu Gani oli ti nihta the arrival/the arrivals the John-gen all the night
These facts support Tenny’s (1994) Aspectual Hypothesis, according to which aspectual roles are mapped on internal argument positions, and these affect the aspectual interpretation of the whole predicate. Given that unaccusative predicates only have internal arguments, this state of affairs is expected. Several verbs that belong to the group of alternating unaccusative verbs give two different types of derived nominals. Compare the nominals in (38) below derived from the verb alazo ‘change’, one of the unaccusative predicates that enter the causative alternation (see also the discussion on ‘passive nominals’ in the next chapter): (38) a.
i alagi tis katastasis the change the situation-gen b. to alag-ma ton ruhon the change the clothes-gen
The interpretation in the two examples is strikingly different. In (38b) we understand that the clothes have been changed by someone, while in (38a) it is the ‘situation’ that has changed. Several similar examples can be found. There is a certain tendency in interpreting nominals that contain the infix -m- as ‘passive’, however this is not entirely systematic (see 3.3.3 for discussion on the aspectual properties of this infix). To sum up, here I discussed the behavior of intransitive verbs with respect to nominalization. I pointed out that process nominals can be derived from predicates classified as unaccusatives in Greek, but not from unergative predicates. . Nominals derived from ditransitive predicates Greek has ditransitive constructions (39a), corresponding to the English one in (39b):
The functional architecture of nominalizations
(39) a.
edosa to vivlio sto Jani gave-1sg the book-acc to-the John-acc ‘I gave the book to John’ b. I gave the book to John
Like English, Greek has a dative alternation, which, as Anagnostopoulou (to appear) convincingly argues, is like the double object construction in English. In this construction the goal argument bears morphological genitive case: (40) a.
edosa tu Jani to vivlio gave-1sg the John-gen the book-acc ‘I gave John the book’ b. I gave John the book
Anagnostopoulou further shows that Greek has two classes of double object verbs. It distinguishes between the two classes through the case morphology of the two arguments. In the second class both arguments bear accusative. However, in the first class the goal qualifies as a double object, while in the second class the goal behaves like a direct object (cf. 41b): (41) a.
edose tu Jani to vivlio gave-3sg the John-gen the book-acc ‘He/she gave John the book’ b. didaski ta pedia ti gramatiki teach-3sg the children-acc the grammar-acc ton arheon elinikon the Ancient Greek-gen ‘He teaches the grammar of Ancient Greek to the children’
In fact, Anagnostopoulou points out that some Greek verbs permit the double accusative construction in addition to the PP and the double object construction, while other verbs belonging to the same semantic class only permit the double accusative construction. For example, among the verbs that inherently signify ‘giving’, serviro ’serve’ licenses all three constructions, while taizo ‘feed’ only the double accusative construction: (42) a.
Servira to fagito s-ton Petro served-1sg the food-acc to-the Peter ‘I served the food to Peter’ b. Servira tu Petru to fagito served-1sg the Peter-gen the food-acc ‘I served Peter the food’
Functional structure in nominals
c.
Servira ton Petro to fagito served-1sg the Peter-acc the food-acc ‘I served Peter the food’
What is of interest here is that in Greek, nominalizations related to the verb form with to are possible, while nominalizations related to the double object construction are not, like in English. Consider the examples in (43) and (44): (43) a. *To harisma tis Marias enos vivliu (apo ton Petro) the gift the Mary-gen a book-gen (by the Peter) ‘*Peter’s gift of Mary of a book’ b. *To dosimo tis Marias ton ruhon (apo ton Petro) the giving the Mary-gen the clothes-gen (by the Peter) (44) a.
To harisma enos vivliu s-tin Maria (apo ton Petro) The gift a book-gen to-the Mary (by the Peter) ‘Peter’s gift of a book to Mary’ b. To dosimo ton ruhon s-tin Maria (apo ton Petro) the giving the clothes-gen to-the Mary (by the Peter)
Anagnostopoulou argues that the ungrammaticality of the examples in (43) is an effect relating to the presence of a zero head introducing the goal argument, like English nominalizations related to the double object construction. I come back to this point in Chapters 3 and 4. Verbs licensing the double accusative construction permit process nominalizations in which the goal surfaces as the complement of the noun. The fact that aspectual modification, both in form of an adverbial in (45a) and in form of an adjective in (45b) is licit suggests that such nominals bear the process reading: (45) a.
i didaskalia ton pedion epi ores the teaching the children-gen for hours b. i taktiki pliromi ton ipalilon the regular payment the employees-gen
It is possible to form nominalizations where the theme is expressed, but then the goal remains unexpressed: (46) a.
didaskalia ton arheon se takta hronika diastimata the teaching the Ancient Greek-gen in regular time intervals b. *i didaskalia ton arheon ton pedion
The functional architecture of nominalizations
c.
i taktiki pliromi ton logariasmon the regular payment the bills-gen d. *i pliromi ton logariasmon tu ipalilu the payment the bills-gen the employee-gen
Anagnostopoulou (to appear) further points out that nominalizations based on the genitive construction and nominalizations based on the double accusative construction are not identical from the point of view of their event structure. In particular, the former are ‘less verbal’ than the latter. While both appear to support argument structure, nominalizations from give verbs do not permit adverbial modification. (47) *To harisma enos vivliu s-tin Maria ja ores the gift a book-gen to-the Mary for hours ‘the gift of a book to Mary for hours’
I discuss this further in Chapter 4. . Nominals derived from psychological predicates Like other languages, Greek has three classes of psychological predicates (cf. Anagnostopoulou 1999). Class 1 includes psychological predicates like agapo ‘love’, miso ‘hate’. Class 2 experiencer predicates e.g. anisiho (worry), provlimatizo (puzzle), enohlo (bother), diaskedazo (amuse), fovizo (frighten), endiafero (interest) qualify as transitives in that they are causatives, and show evidence of assignment of accusative case. Class 3 experiencer predicates e.g. aresi (like), ftei (bothers/matters) qualify as unaccusatives (see Alexiadou and Anagnostopoulou 1999). In Greek Class 1 predicates, the experiencer has nominative and the theme accusative case (cf. Anagnostopoulou 1999): (48) o Janis agapai ti Maria the John-nom loves the Mary-acc ‘John loves Mary’
In Class 2, the experiencer has morphological accusative case; the theme/causer has nominative case and agrees with the verb: (49) i kategida fovizi to Jani the storm-nom frightens the John-acc ‘The storm frightens John’
Functional structure in nominals
The non-causative counterpart of this construction involves a theme being introduced by a preposition and an experiencer bearing nominative: (50) o Petros fovate me tin katastasi the Peter-nom fears with the situation
Class 3 selects for a dative experiencer (PP as in (51a) or morphological genitive as in (51b)) and a nominative agreeing theme: (51) a.
To krasi aresi ston Petro the wine-nom like-3sg to-the Peter ‘Peter likes the wine’ b. To krasi tu aresi tu Petru the wine-nom cl-gen like-3sg the Peter-gen ‘Peter likes the wine’
Consider now the nominals from these classes of verbs. It has been observed that nominalizations can only be related to the non-causative forms of the predicates that distinguish between causative and non-causative readings, i.e. Class 2 predicates (cf. Chomsky 1970, Grimshaw 1990, Pesetsky 1995:72, Markantonatou 1992: 90f.). These seem to be able to be modified by aspectual modifiers: (52) a.
to endiaferon tu Jani ja ta fita epi dekaeties the interest the John-gen for the plants for decades b. o fovos tu Jani ja tis kategides epi tosa hronia the fear the John-gen for storms for many years c. *o fovos tis kategidas epi tris ores itan periergo pragma the fear the storm-gen for three hours was strange thing
In the above examples the genitive DP is interpreted as the experiencer. Similar facts are observed for Catalan (cf. Picallo 1991). At least certain of Class 3 predicates seem to be able to from process nominalizations. A large number of verbs of this class, however, does not have a corresponding nominal: (53) to sinehes fteksimo tu Jani the constant blaming the John-gen
The discussion so far has established the following: Greek shows a preliminary distinction between event vs. result nominals identified by the criteria of manner and aspectual modification.
The functional architecture of nominalizations
In the next section I turn to the verbal properties of process nominals.
. The verbal properties of process nominals In this section, I discuss in detail the verbal properties of process nominals that I link to the presence of VoiceP and AspectP. Evidence for the presence of these functional layers comes from two main areas: (i) the classes of adverbs that can be present within process nominals and (ii) the overt morphological reflexes of the presence of Voice and Aspect that can be observed in several languages. . DP-internal adverbs In the recent literature a number of significant facts about the distribution of adverbs inside nominalizations have been brought into light, an account of which necessitates certain radical assumptions concerning the internal structure of these constructions. As noted in Alexiadou (1997), adverbs can appear in Greek nominalizations (cf. 54): (54) i katastrofi ton stihion olosheros (mas kateplikse) the destruction the evidence-gen completely us shocked
Under the standard assumption that adverbs modify VPs and not NPs (cf. Jackendoff 1977), the presence of adverbs inside nominals is problematic. Adverbial modification is consistent only with a VP structure. If nominalizations are not syntactically derived from a VP, the presence of adverbs inside these constructions cannot be accounted for. As mentioned in Chapter 1, the occurrence of adverbs inside nominals is not licensed simply by the action or process interpretation associated with them, but rather it is syntactically conditioned. As the contrast in (55) illustrates, the adverb cannot appear unless the rest of the complement system accompanies it. In other words only in (55b), where the eventive nominal is used, as signaled by the presence of the internal argument, is the adverb acceptable. (55) a. *i katastrofi olosheros the destruction completely b. i katastrofi tis polis olosheros the destruction the city-gen completely ‘The destruction of the city completely’
Functional structure in nominals
Besides this general observation, Borer (1993), Hazout (1995) for Hebrew, Fu (1994) for Chinese, Fu et al. (1998) for English, and Alexiadou (1997, 1999a,b), Alexiadou and Stavrou (1998a) for Greek point out that not all types of adverbs are acceptable in nominalizations. As shown in (56)–(57) with Hebrew and Greek data, modal adverbs of the type probably and speaker-oriented adverbs of the type fortunately are not permitted. (56) *ktivat Dan et ha-avoda lelo safek writing Dan acc the work doubtlessly (57) *i katastrofi ton stihion pithanos/ilikrina the destruction the evidence-gen probably/frankly
In both Greek and Hebrew, the type of adverbs that are acceptable within process nominals are (cf. 58–59): manner, aspectual (frequency/interval denoting) and certain temporal ones. (58) a.
i katastrofi ton egrafon toso prosektika the destruction the documents-gen that carefully ‘the destruction of the documents that carefully’ b. i katastrofi ton egrafon kathimerina the destruction the documents-gen daily ‘the destruction of the documents daily’ c. i katastrofi ton egrafon ktes the destruction the documents-gen yesterday ‘the destruction of the documents yesterday’
(59) a.
ktivat Dan et ha-avoda bi-mehirut writing Dan acc the-work quickly ‘Dan’s writing of the work quickly’ b. axilat Dan et ha-uga be-minus eating Dan acc the-cake politely ‘Dan’s eating the cake politely’
The same type of adverbs appear in nominals derived from unaccusative predicates: (60) i ptosi ton timon stadiaka the fall the prices-gen gradually
In contrast, no such adverbs can appear with nominals derived from unergative predicates:
The functional architecture of nominalizations
(61) a. *to kolimpi tu Petru prosektika the swim the Peter-gen carefully b. *to skarfaloma tu Jani prosektika the climbing the John-gen carefully
Finally, recall that adverbial modification is possible with nominalizations derived from double accusative verbs. (62) a b.
to servirisma tu pelati me evgenia the serving the customer-gen with politeness to servirisma tu kafe me evgenia the serving the coffee-gen with politeness
The admissibility of certain adverbs in process nominals is not a fact that simply has to do with some kind of semantic compatibility (given Grimshaw’s reasoning, a priori the semantics of a process nominal should not be different from the semantics of a verb). Rather, it is a syntactic fact that has to do with the principles that determine which elements can be attached at which positions in the tree structure. This does not mean that the admissibility of adverbs has nothing to do with the event interpretation associated with process nominals. In fact, Grimshaw’s account could capture the correlation between the admissibility of adverbs and the event interpretation as well, by assuming that adverbs modify eventive predicates, both nominal and verbal. But the interpretation of such nominals as denoting events is not sufficient as an explanation to the restrictions on the distribution of adverbs. But why is adverb distribution significant and why does it show us something about the internal structure of process nominals? Recent work on the syntax of adverbs has put forth the hypothesis that adverbial phrases are related to distinct functional projections (cf. Alexiadou 1997, Cinque 1999). This hypothesis runs in parallel with research on the type and content of functional projections that constitute the clausal architecture. It crucially builds on the transparent semantic relation that can be observed crosslinguistically between projections and adverbs. According to this hypothesis, the presence of aspectual adverbs is linked to an Aspect phrase, while manner adverbs arguably bear a tight relation to Voice phrase. On this view, the presence of such adverbs signals the presence of certain ‘low’ verbal projections within process nominals. Thus on the basis of this reasoning, I conclude that process nominals contain a domain that shows clear verbal properties.
Functional structure in nominals
. Morphological reflexes Interestingly, nominals in certain languages have a morphological reflex of the inclusion of Voice and Aspect. As far as Voice is concerned, observe that certain Greek process nominals tend to include the infix -m-, which is related to non-active voice morphology in Greek, as can be seen in the passive perfect participle (63a): (63) a.
diavas-men-os read b. diavas-m-a reading
This suggests that at least a number of process nominals in Greek might be shown to inflect for voice. However, this is not very systematic. A more systematic connection between voice morphology and nominals comes from Turkish. In Turkish, as reported in Comrie (1976a), action nominals may show voice in the same way as a finite verb does. Compare the Turkish sentence in (64) to the action nominal in (65). Both contain the passive morpheme -il-: (64) Mektub yaz -il -di letter write pass past ‘The letter was written’ (65) mektub-un yaz -il -ma-si letter-gen write pass vn-its ‘the writing of the letter’
Similar patterns are observed in Korean, West Greenlandic, Amharic, Bantu languages, and Maori. With respect to Aspect, in Slavic languages the perfective vs. non-perfective opposition is also observed in process nominals. Consider the following examples containing verbal nouns in Polish (from Schoorlemmer 1995). As (66) shows, the noun ‘evaluation’ inflects for perfective and imperfective aspect much like the corresponding verb does in the language: (66) a.
ocenienie studentów przez nauczycieli nast˛apiło szybko evaluation-pf the students-gen by teachers occurred quickly b. ocenianie studentów przez nauczycieli ciagn˛eło si˛e evaluation-imp students-gen by teachers lasted refl przez cały tydzien´ through the whole week
The functional architecture of nominalizations
Similar patterns are found in Archi, Inuit, Buryat, Mongolian, Turkish, Tuva, and Tagalog. . Aspectual distinctions In the previous sections I pointed out that the presence of aspectual modification and of aspectual morphology constitutes an argument in favor of the presence of AspectP within nominals. In this section I show that there is something more to this. Specifically, different types of nominals receive distinct aspectual interpretations. The readings discussed here cannot be related to lexical aspect only but seem to suggest a combination of telicity/–telicity and perfectivity/ imperfectivity. In the recent literature telic/–telic readings in combination with the distinction perfective vs. imperfective have been linked to the presence of an Aspect Phrase (see Iatridou et al. 2000 for a recent discussion and the discussion in Chapter 1). .. -ing vs. non -ing nominals in English and -ung nominals in German It has been noted that -ing nominalizations differ in their aspectual properties from -ion nominalizations, in the sense that -ing nominalizations entail an imperfective event. This property led e.g. Pustejovsky (1995) and Siegel (1997) to analyze -ing as a progressive marker. In particular, Pustejovsky points out that there is no interpretation of -ing nominalizations as the result of an event, as there is with -ion and other nominalizations (67b vs. 67c). Moreover, achievements are much less acceptable as -ing nominalizations than are processes (67d vs. 67e). (67) a. b. c. d. e.
destroying the memo took an hour *the destroying was widespread the destruction was widespread *the arriving of John the arrival of John
A further argument for assuming that telicity is encoded in the structure comes from examining the aspectual properties of -ung nominalizations in German. These are sensitive to the type of object (quantized/non-quantized) present. As was the case with unaccusative nominals in Greek, non-quantized objects yield non-telic readings with -ung nominals: (68) a. *die Ausgrabung antiker Schätze in einem Jahr the excavation antique treasures-gen in one year
Functional structure in nominals
b. die jahrelange Ausgrabung antiker Schätze c. die Ausgrabung eines antiken Schatzes in einem Jahr d. *die jahrelange Ausgrabung eines antiken Schatzes
.. Passive nominals in English Tenny (1994) and Snyder (1998) have pointed out that there is a difference in aspectual interpretation between passive nominals, i.e. nominals which involve fronting of the complement to initial position, and other nominal types. Specifically, passive ones only denote delimited events (cf. 69). (69) contrasts with (70) where the nominal can receive both a telic and an atelic interpretation (see also the discussion in Chapter 3): (69) a. the city’s destruction in five minutes b. *the city’s destruction for just five minutes (70) a. the destruction of the city in five minutes b. the destruction of the city for five minutes
.. ma/mo vs. non ma/mo nouns in Greek According to Kolliakou, certain nominalization affixes in Greek show selectional restrictions that are sensitive to the aspectual type of the respective nominal. For instance (cf. Kolliakou 1995: 211f.), prototypical state and telic event predicates do not give grammatical nominalizations when they combine with the affix -ma/mo (cf. 71): (71) a. *agapima loving b. *katastrema destroying c. *dolofinima assassinating
agapi love katastrofi destruction dolofonia assassination
Moreover, as the contrast in (72a-b) shows, while nominals derived from predicates denoting accomplishments do not tolerate adjectival aspectual modification denoting repetition, this is possible with ma/-mo. English and Hebrew accomplishment nominals also resist modification by frequent (cf. Borer 1993). Kolliakou concludes from this that -ma/-mo nouns, if they denote events, they denote processes.
The functional architecture of nominalizations
(72) a.
to sihno plisimo ton piaton the frequent washing the dishes b. *i sihni dolofonia ton ipefthinon the frequent assassination the responsible-gen
Further support for this conclusion comes from the observation that in the cases where two forms of the nominal exist, as is the case with ‘correction’ below, only the ma/mo form is compatible with modifiers such as for an hour. (73) a.
i diorthosi ton grapton se pende lepta/*ja mia ora the correction the papers-gen in five minutes/for an hour b. to diorthoma ton ruhon se pende lepta/ja mia ora the correction the clothes-gen in five minutes/for an hour
Given that -ma/mo nouns are compatible with such modifiers, one can conclude that these are ambiguous between a process (durative) and an event (telic) reading in the sense discussed in the introduction. On the other hand, non -ma/mo nouns tend to refer to telic events only. Note that stative nominals are also incompatible with adjectival modification signaling repetition. (74) *i sihni agapi tu Jani the frequent love the John-gen
In fact in the above examples both stative and achievement nominals pattern like Grimshaw’s result nominals with respect to modification by adjectives such as frequent. According to Grimshaw, such modifiers may appear with plural result nouns, but only with singular process nominals. This contrast is repeated in (75a and b). In Greek, as (76a and b) show, both statives and achievement nominals are compatible with such modifiers, when they appear in plural form: (75) a. *the frequent examinations/√examination of the papers b. the frequent exams/*exam (76) a.
i sihnes dolofonies politikon the frequent assassinations politicians-gen b. i sihnes agapes the frequent loves
Thus (76a and b) seem to constitute evidence that these nominals belong to the
Functional structure in nominals
class of result nouns. However, if, as mentioned earlier on, one assumes that the mass/count distinction in the individual domain corresponds to the process/state vs. accomplishment/achievement distinction in the event domain, telic event nominalizations should license pluralization and disallow repetition modification in their singular forms., In other words, the fact that Grimshaw’s tests do not work as expected with achievement nominals suggests that these qualify as event, i.e. telic, and not as process, i.e. durative, nominals. Coming back to stative nominals, some of the tests used in Grimshaw to diagnose argument structure actually only diagnose the presence of an agent and hence yield deviant results with such predicates. For instance, while by-phrases can occur within achievement nominals, they are not possible within stative nominals. (77) a.
i dolofonia tu Athanasiadi apo ti 17 Noemvri the assassination the Athanasiadis-gen by the 17th of November ‘The assassination of Athanasiadis by the 17th of November’ a.¹ i dolofonies politikon apo ti 17 Noemvri the assassinations politicians-gen by the 17th of November ‘The assassination of politicians by the 17th of November’ b. *i agapi tis Marias apo to Jani the love the Mary-gen by the John b¹. i agapi tis Marias ja to Jani the love the Mary-gen for the John ‘Mary’s love for John’
The ungrammaticality of (77b) is related to the fact that the by-phrase receives a restrictive interpretation within argument supporting nominals, as will be discussed in Chapter 3. It can only denote a causer, and never an experiencer. The presence of a by-phrase is licit even when the noun in (77a) is in the plural (cf. 77a¹). As (77b¹) shows, when the theme argument is present within the stative nominal, it is preceded by a preposition. This pattern is clearly different from the nominalization pattern of non-stative transitive predicates. The genitive argument in (77b¹) can be only interpreted as an experiencer. Moreover, while modification by agent-oriented adjectives is possible with achievement nominals, the presence of such modifiers within stative nominals is ungrammatical: (78) a.
i eskemeni dolofonia tu Athanasiadi the intentional assassination the Athanasiadis-gen
The functional architecture of nominalizations
b. *i eskemeni agapi tis Marias the intentional love the Mary-gen
The ungrammaticality of (78b) is again due to the fact that there is no agent involved in states in general. Finally, while it is possible to modify achievement nouns with manner adverbials, it is ungrammatical with stative nouns. If the presence of a manner component is somehow related to the presence of agentive features (Hale and Keyser 1993, Levin and Rappaport 1995), then again the ungrammaticality of (79b) is expected. However, stative nouns are compatible with aspectual modifiers of the type for decades: (79) a.
i dolofonia tu Athanasiadi me frikto tropo the assassination the Athanasiadi-gen with horrible manner ‘The assassination of Athanasiadis with a horrible manner’ b. *i agapi tis Marias ja to Jani me trelo pathos the love the Mary-gen for the John with crazy passion c. i agapi tis Marias ja to Jani epi dekaeties the love the Mary-gen for the John for decades ‘Mary’s love for John for decades’
If stative nominals can never denote events, then (79b) can be taken as evidence that these lack a functional layer bringing about an event interpretation, since manner modification goes hand in hand with such a layer. In fact, as Katz (1999b) points out, manner modification is generally very restricted with stative predicates. However, these facts say nothing about the availability of a stative nominal to take arguments. As to why modification by aspectual modifiers of the type for decades is licit, this could be related to the lexical semantics of the root. Note that the facts presented here pose certain problems for Grimshaw. According to Grimshaw, only nominals that have a complex event structure, i.e. an event structure that can be broken down into aspectual subparts, can give argument-supporting nominalizations. This view is embedded within the assumption that nominalization necessarily involves suppression of an external
Functional structure in nominals
argument. The data examined so far show that this mapping does not correspond to the actual nominalization possibilities. As we have seen above, unergative verbs that have activity readings do not give argument-supporting nominalizations, although arguably they possess an external argument. On the other hand, unaccusative predicates that have achievement readings give argument-supporting nominals, although they only correspond to the second subpart of the aspectual decomposition of a predicate. Moreover, one would expect that stative predicates, which cannot be argued to involve aspectual subparts, would also not be able to give argument-supporting nominalizations. However, examples such as (79c) above are perfectly grammatical. To conclude, nominals show aspectual oppositions, but AspectP should be present in all the types of nominals discussed in this section (though see Chapter 4 for certain refinements). It receives a different specification in each case. . The structural differences between process and result nominals Thus far, the following general properties of process nominals have been identified: • •
event nominals can refer to different types of events. In certain occasions such nominals bear morphological aspectual marking. event nominals have a number of verbal properties, in the sense that they permit adverbial modification. More importantly, there are severe restrictions on adverbial modification, in the sense that ‘low’ adverbs are licit within such nominals, but ‘high’ adverbs are disallowed.
To capture these properties structurally, at least some verbal functional layers must be present within process nominals. Hence I propose that vP and AspectP are present within process nominals only. On this view, result nominals and process nominals are similar in the sense that both include a root as their basic component, but only the latter include verbal, i.e. eventive, functional layers. The structure assumed for event nominals is shown in (80).
The functional architecture of nominalizations
(80)
DP FP (NumbP/AgrP)
D° the AP
FP F° AspectP Aspect¹
Aspect°
vP v
LP
L° Comp (=theme) √DESTROY the city AspP and a vP count as ‘low’ functional projections responsible for the licensing of adverbs. In fact vP and AspP are functional projections which can appear both under T and D/Number. Embedding under D°, and perhaps further nominal layers, results in a morphological noun, as illustrated in (80) above. On the other hand, embedding under T° and further verbal functional layers results in a morphological verb. In a sense, this analysis makes derived nominals look very similar to untensed clauses (see de Hackbeil 1984). On the other hand, the structure depicted in (81) must be assumed for result nominals. (81) differs from (80) in that no Aspect and no v are included, namely the two projections that contribute to event interpretation:
(81)
DP D
FP F
LP L
DP
One question that may arise here is whether (81) suffices to capture the properties and meaning of result nouns. In the introduction I assumed that the semantic of roots denote a resultant state, which is actually the meaning associated with result nominals. Thus the presence of an additional functional category that bears such interpretation is not necessarily needed. However, one
Functional structure in nominals
could argue that what we understand as a result is derived from the presence of an aspectual head specified [RESULT] which dominates the root, as has been argued to be the case for stative participles by Embick (2000b). On this view, the ‘result’ meaning is derived through a combination of Aspect and the root. Clearly, the issue is related to the general discussion about whether roots bear some component of meaning or not. For the purposes of the discussion here, I take roots to roughly bear the semantics of a resultant state. States are taken to be primitives in the spirit of Dowty (1979) and others building on Dowty. Thus, the presence of an aspectual projection specified [RESULT] is not necessary. Note that this seems a necessary step in view of the restrictions noted in Section 3.3. That is though the feature specification of the functional layers and the selectional requirements among these contribute a great part of the meaning, some part of it seems to be ‘constant’ (see also the related discussion in Section 5). (81) makes explicit what I have already suggested: the difference between result and process nominals is not one of argument structure, since both nominal types can have complements (see 82 which was discussed also in the introductory chapter), but rather it relates to the presence vs. absence of functional layers that bring about process/event readings. This is consistent with the fact that result nominals disallow modification by manner adverbs and aspectual modifiers. On the other hand, a nominal FP is present in both types of nominals, as discussed in Section 1. (82) a.
la demonstracio d’en Joan del teorema de Pitagores the proof of Joan of the theorem of Pythagoras b. la demonstracio d’en Joan del teorema de Pitagores the proof of Joan of the theorem of Pythagoras es inconsistent is inconsistent
Several issues arise. Most importantly, however, if constructions like (82) are generally possible, how can we explain the fact that the presence of complements is obligatory with process nominals only? Recall that result nominals can, but need not, have complements. I come back to this in Section 5. The fact, however, that result nominals can have complements is unexpected both under an approach to nominalizations that assigns a VP structure only to process nominals but not to result nominals, as well as under the lexicalist approach which associates the presence of arguments with the presence of an event structure. First of all, under an analysis of process nominals in which a VP is
The functional architecture of nominalizations
included in their internal structure, it is expected that these and only these will have complements, as the basic claim of this approach is that process nominals have the same internal arguments as their verbal counterparts because both contain a VP. Since a VP is not present within result nominals, it is not immediately clear why complements are present within such nominals. To deal with this, Picallo (1991) suggests that result nominals, enter D-structure categorially marked as NPs like intransitives, non-deverbal, and active transitives. The arguments of lexical nominals do not assign structural case to their arguments. Rather they enter D-structure specified as assigning inherent case (genitive or oblique) to their arguments. Grimshaw (1990) and to some extent Zubizarreta (1987), on the other hand, make a distinction between the notion of complement, i.e. an expression that corresponds to a position in the lexical conceptual structure of the head, and the notion of argument which is licensed by argument structure. While complements can be present within result nominals, only arguments can be present within process nominals. Having shown how the properties of process nominals are accounted for in the system proposed in the introduction, in Sections 4 and 5 I address the following issues in turn: (i) are there arguments for the presence of higher than Aspect functional projections inside process nominals? (ii) what determines the presence of argument structure?
. Process nominals lack tense . Absence of phenomena related to T: EPP, ECM and raising In my discussion of the morphological properties of nouns, I pointed out that these suggest that the inflectional projections of Voice and Aspect are present within nominals. On the other hand, certain other verbal inflectional categories, such as Tense and Mood seem to be totally absent. In general, nominalizations seem not to be able to express tense and modality in a manner identical to their related verbs (cf. Croft 1991). Koptjevskaja-Tamm (1993) investigated nominalizations in nearly 70 languages and from her sample it is also evident that nominals tend to inflect at least for Aspect and Voice, if at all. In this section I discuss certain phenomena that are straightforwardly accounted for, if we assume that nominals lack T.
Functional structure in nominals
There are several properties of nouns, widely discussed in the literature, which distinguish them from verbs (see e.g. Stowell 1981, Kayne 1984, Grimshaw 1990, Giorgi and Longobardi 1991, Drijkoningen 1993, Schoorlemmer 1995, de Wit 1997). These are summarized below: (a) There are no expletives in nominal constructions: (83) a. there arrived a man b. *there’s arrival c. *its rain
(b) Nouns do not allow raising (84) Mary appears to have left (85) *Mary’s appearance to have left
(c) Nouns do not allow passivization across sentence boundaries (86) The baby is estimated to weigh 8 pounds by the doctor (87) *the baby’s estimation to weigh 8 pounds by the doctor
(d) Exceptional case-marking does not occur with nominals: (88) I believe Mickey to be a genius (89) *my belief of Mickey to be genius (90) Jans horen van Marie liedjes zingen Jan’s hear of Mary songs sing
The absence of these phenomena within nouns was taken to be due to the fact that Spec,DP is a thematic position (cf. Abney 1987), unlike Spec,TP that is a checking and not a thematic position. In what follows I account for the absence of these phenomena, by showing that these are linked to the presence of T in a given clause. Given that Tense is absent from nominals, we do not expect to find Raising or expletives within them either. In Chomsky (1995), Collins (1997), Alexiadou and Anagnostopoulou (to appear) among others, it has been proposed that there are two features associated with T: the Extended Projection Principle (EPP) feature and the [assign nominative] feature. The first obvious effect of the absence of T in nominalizations is the fact that the single DP argument does not appear with nominative case. Second, under the assumption that EPP-related phenomena (expletive merge and/or argument movement to Spec,TP) take place in order to satisfy a universal property of T (Chomsky 1998), then it is expected that we
The functional architecture of nominalizations
never find those within DPs, if those lack T (see Abney 1987, Doron and Rappaport-Hovav 1991, Drijkoningen 1993, Alexiadou and Stavrou 1998a, Schoorlemmer and de Wit 1996 among many others for claims that EPP is not respected within nominals). Hence, this explains the fact that expletives cannot be merged inside DPs. Moreover, arguments of nouns do not exhibit obligatory movement to a designated functional position in order to satisfy the requirements of this projection the way subjects move to Spec,TP or to T°., Kayne (1984), and Grimshaw (1990) among others have argued that the absence of ECM and Raising in nominal structures is due to the fact that N is syntactically impaired with respect to V. I believe that this is correct. Nominal clauses differ from verbal clauses in that they lack T. More specifically, the lack of these types of constructions from DPs results from the absence of a landing site in the higher clause. On the basis of the assumption that raising involves movement of the argument from the lower Spec,TP to the higher one, Raising is expected to be illicit within DPs, if these lack TP. Within the minimalist framework, ECM structures have been used as a strong argument against Greed because the XP raises overtly to its EPP position, namely Spec,TP, which is not identifiable with its Case position, namely Spec,AgrOP (cf. Lasnik 1993). From a different perspective, it has been argued that further raising to AgrOP of the higher clause, which is the Case position of the embedded subject, is made possible by forming a tense-chain between the two tenses, i.e. the tense in the lower clause and the matrix tense (cf. Bennis and Hoekstra 1989). Building on this specific proposal, de Wit 1997 proposes that ECM constructions in nominals are impossible as the lower T cannot be part of a tense-chain, since the higher clause lacks Tense. Obviously, the structure proposed here which excludes Tense permits a natural implementation of this analysis. In other words, no tense chain can be formed, as the higher clause lacks tense, in fact it lacks any tense specification. Further evidence for the lack of higher than Aspect functional projections inside process nominals comes from the distribution of adverbs presented in the Section 3. As discussed there, only lower adverbs can be tolerated within derived nominals, i.e. adverbs of the type that do not match with T or other higher functional projections. Moreover, as discussed in Siloni (1997), the presence of T itself crucially affects the presence of D or C, assuming that D is the counterpart of C. Siloni (1997), in a rather different context from the one of the present discussion, investigates in some detail the restrictions on what can appear as a complement of D. Elaborating on Stowell’s (1982) correlation between the existence of a CP
Functional structure in nominals
level and the presence of a tense operator in the clause, she suggests that what determines the choice between CP and DP is the presence vs. absence of a tense operator. Clauses that have no internally determined tense are headed by D. In other words, while C is associated with tensed clauses, D is associated with tenseless clauses. In principle this establishes a correlation that certain verbal projections situated lower than tense can be included within DPs. As soon as tense is active, the projection turns into a verbal clause (see Chapter 4 for further discussion). A similar point is made in Pustejovsky (1995: 157f.). Pustejovksy notes that the way in which a verbal sentence denotes an event is different from the way DPs denote events. Although both strings Mary arrived at 3 pm and Mary’s arrival at 3 pm may refer to the same event, they do not express the same content about the event. The former is an assertion of an event having occurred while the latter denotes an event without an assertion that it has occurred. The difference between the two is that in the former the event is tensed and thereby interpreted as a proposition. Only through tensing of an event secting predicate can the event in the latter string contribute propositional information as in Mary’s arrival was at 3 pm. That is both sentences and NPs denote events, but the manner with which they denote is quite different distinguished by their types. A proposition is seen as the result of applying tense to an event description. To this end, tense acts as a generalized quantifier over event descriptions. D takes this function in the nominal context. This view makes the prediction those languages that lack Tense as a separate category should neutralize the distinction between nouns and verbs. This prediction seems to be borne out. The group of Salish languages lacks a unitary category T (Matthewson 1996, Davis and Matthewson 1997). These authors argue that in this language group nominal and verbal projections are identical. First, the elements that introduce nominal and verbal projections are phonologically identical. Second, they have the same syntactic distribution, as illustrated in (91)–(92): ‘complement’ of intransitive verb: (91) qaním=łkan [kw=šnlíq=i] hear-intr-1sg.subj arrive-3pl.poss ‘I heard that they came’ (92) kiˇc-xal kwu=pəták ʔi=ʔuxwalmixw=a lie down ku=potato pl.det=person=exist ‘The people plant potatoes’
verbal projection
nominal projection
The functional architecture of nominalizations
Third, the elements that introduce verbal and nominal elements also encode the same semantic distinctions. To conclude, in this section I have argued that process nominals are roots appearing beneath the functional head Aspect but do not combine with Tense. While realization of the morphological object verb amounts to combining with Tense (either via head-movement or Merger), this combination would fail in nominals, as the functional head D (or a nominal functional category e.g. Number following Ritter (1991)) dominates the root. Crucially, AspP can combine with either D or T, but not both. In the next section I turn to some arguments in favor of the presence of Tense inside nominals and show how they can be re-interpreted in a different way fitting the proposal made here. . Evidence for nominal tense It has been pointed out in the literature that at the semantic level the absence of tense in the DP can be challenged. Consider (93): (93) Every fugitive is now in jail.
This sentence is an assertion about past fugitives who are in jail at present. The temporal interval of the referents of the DP being fugitives crucially does not coincide with the interval of them being in jail. This suggests that DPs must be able to be given a temporal reading independently of that of the clause in which they appear (Enç 1987). The independent temporal interpretation of DPs is constrained by syntactic factors, though, as shown by Musan (1995). In (94) the NP students can only be assigned a dependent reading, i.e. one in which the interval of the referents’ being students coincides with that of their being sick. (94) There were three students sick.
With respect to temporal readings, DPs seem to behave like embedded clauses in that they may be dependent (95a) or independent (95b) from the tense of the higher clause. Temporal dependence of clauses is usually referred to as the ‘sequence of tense’ (see e.g. Enç 1987, also Haegeman and Guéron 1999). (95) a. b.
Mary said she would come Mary said she will come.
If the temporal interpretation of a clause is encoded in a specialized projection, TP, then of course by analogy one might well wish to postulate that the
Functional structure in nominals
temporal interpretation of the nominal also be related to TP. In the more familiar languages such as English, French, Dutch, German, Italian, Spanish, Greek, etc. this TP would not have any morphological reflex. In some languages, nominal tense does have an overt morphological reflex. Burton (1997) shows that Halkomelem, a Salishan language spoken on the Northwest Coast of North America, has overt past tense marking on nouns. The tense marker on nouns is the same as that on verbs. With verbs the past tense marker occurs on a pre-verbal auxiliary, as illustrated in (96a). The same past tense marker lh is also found on Ns as illustrated in (96b–c): (96) a.
i-lh ímex tel sí:le aux-past walk my grandfather ‘My grandfather walked’ b. tel sí:le tel sí:lalh my grandfather my grandfather-past ‘my grandfather’ ‘my late grandfather’ c. tel xeltel tel xeltel-elh te thá my pencil my pencil-past ‘my pencil’ ‘the pencil which was mine or ‘my broken pencil’
As shown by the translations, the nouns with past tense marking mean one of three things: (i) death (96b), (ii) loss of possession (96c), or (iii) destruction (96c). However, the above data can receive an alternative interpretation along the lines of Davis (2000). Davis points out that the Tense marker attached to nouns applies to the referent of the noun, that is the entity that it denotes, or, in the case of (96c) the referent of the NP including the possessor. Crucially it does not apply to some state of ‘being a pencil’ or ‘being a grandfather’. According to Davis, this is expected if nouns, as the ones in the examples above, are aspectless. That is tense distinctions bear on aspectual ones. As a result Tense within these nouns does not have the same function as the tense markers do with ‘verbs’. The same kind of inflectional endings are signaled by Hockett (1958: 238) for Potawatomi, for the Salishan languages St’at’imcets (see Demirdache 1996, Davis and Matthweson 1997, Lecarme 1998), and for Somali (Lecarme 1996, 1998). Specifically, in Somali, a Cushitic (Afroasiatic) language the tense morpheme attach to definite determiners, providing yet again morphological evidence for a DP-internal TP. According to Lecarme, nominal tense is a property of any DP (common noun phrase). Tense morphology affixes to D, i.e. the
The functional architecture of nominalizations
definite article -k (masculine)/-t (feminine), encoding a [±past] opposition. Near/far demonstrative enclitics are in complementary distribution with tense morphemes. The DP-tense is independent of the predicate-tense: nominal tense and verbal tense may coincide or differ: (97) a.
dhibaatá-dii Khalíij-ku wáy dhammaatay problem-DetF [+past] Gulf-DetM[+nom] F+3s ended [+past] b. ardáy-da baan kasin students-DetF[–past] F+neg understood [+past] ‘The students (who are present, who I am telling you about) did not understand your question’ c. ardáy-dii wáy joogaan students-DetF[+past] F+3P are-present-[−past] ‘The students (the students I told you about) are present’
As was the case in Halkomelem, the past tense marking on the Somali DP may indicate that the referent no longer exists. In (98) the past tense indicates that the speaker believes the exhibition is closed, the non-past that he believes it is still running. (98) bandhíg-ga/-gii máad daawatay? Exhibition-DetM[−past]/det M[+past] Q+2S saw [+past]
Again such data seem to challenge the view that Tense is not present within nominals. However, in most of the cases discussed, it is not immediately transparent whether the presence of temporal morphemes establishes reference, or temporal location. In fact the relation between D and T is not very clear. Lecarme (1996) suggests that the tense morpheme is attached outside the Determiner morpheme. For instance in (97a) she takes d to be the reflex of D and -ii to be the reflex of tense. If the linear order of the morphemes correlates to the syntactic hierarchy of the heads (Baker 1988, the Mirror Principle), one would expect TP to dominate DP. Lecarme proposes, though, that DP dominates TP and that the overt Tense morpheme is actually a specifier of TP which then cliticizes to D. In later work (1998) Lecarme takes the [±past ] morpheme to be syncretic with the definite determiner morpheme. While both D and T specifications are normally expressed by a single, syncretic head and thus project only a single maximal projection, DP is obligatorily split when enclitic possessive pronouns or other material are realized. This view, i.e. the view that T specification actually expresses reference is adopted in Alexiadou (1997).
Functional structure in nominals
Nominal T does not locate events/individuals in time. In fact nominal T could not even locate events in time, if one of the semantic requirements of T on its complement is that this is finite, which nominal phrases definitely are not (see Davis and Matthewson 1997 for discussion). Hence T, even if present inside nominals, does not function like verbal T. Having addressed the functional structure of process nominals, I turn to the issue of the obligatory presence of arguments within them.
. On the obligatory licensing of arguments within process nominals In the approach to nominalization followed here, the question of licensing of arguments must be seen under a different light. As mentioned several times, on this approach it is expected that all types of nominals, result and event ones, in principle to be capable of licensing arguments. However, though it is true that result nominal can license internal arguments, it is also true that the presence of arguments is obligatory only in the case of event nominals. Why is this so? Following much recent work on event structure (van Hout 1996, Borer 1999, Levin 1999 and references therein), I assume that event structure plays a crucial role in licensing the obligatory presence of arguments. As known, events must involve a minimum number of participants, i.e. an event cannot take place in the absence of participants. On the view expressed here, roots become eventive in the presence of functional structure (Aspect and v). This amounts in suggesting that the presence of functional structure is responsible for the obligatory presence of arguments. However, this view is not without problems. In a system such as the one proposed in Borer (1999), where arguments are directly projected in the specifier position of various aspectual layers, this intuition is captured straightforwardly. Arguments are obligatorily present, as only event nominals contain aspectual projections. It is not immediately transparent how the advantages of her system can be translated in mine, though Borer cannot capture the fact that arguments can indeed be present within result nominals as well. To deal with this issue, I build on recent work by Beth Levin. Levin (1999) argues that verb-meanings have two components: one component which is provided by its event structure, and one component provided by the core meaning, i.e. the part of the meaning which is idiosyncratic to a specific verb. This second type is referred to as the ‘constant’. Constants are then integrated into event structure templates to yield the various verb meanings. Let me
The functional architecture of nominalizations
equate what Levin refers to as constant with an unspecified root. If this is so, then generally the semantics of the root are such that they can license an internal argument, as roots are taken to have the semantics of a state that involves an argument. Thus in principle, presence of arguments is guaranteed independently of the eventive character of the outcome of word-formation. As soon as such constants enter into a relation with event related projections their presence is obligatory, i.e. they become structure participants in Levin’s terms, since they are required by the event structure, although their presence is also required by the constant. The consequence of this for stative nominals is that, if such nominals lack v and any kind of Aspect, then their arguments are licensed from the constant, as in the case of result nouns. In the final section of this chapter I turn to a few remarks on the actual realization of the morphological object noun works. The observations and suggestions made there hold for process and result nominals alike.
. Processes in morphological structure So far I have been assuming that there are no categories such as verbs and nouns as such. Nevertheless, it is the case that destruction is a morphological noun, while destroy is a morphological verb. Moreover, it is the case that nouns receive adjectival modification and agree with adjectives in all phi-features. Furthermore, we have seen that process nominals can be modified by adverbials much like VPs. While a treatment of these phenomena by making reference to two domains, one VP domain and one DP domain, the latter embedding the former, is straightforward, in a system where there no such reference is possible, it is not immediately clear how one could deal with these. In the introduction it was assumed that addition of morphemes and of vocabulary items takes place after Syntax. Following Halle and Marantz (1993) and Harley and Noyer (1998a,b), I assume that destroy is changed to destruction in the nominal context in a post-syntactic Readjustment component. (99) katastrefo → katastrofi/ when governed by D destroy → destruction
The Readjustment Component performs a variety of functions including the partial modification of the phonological forms of stems, as well as the insertion of morphemes that are not present in the syntax, e.g. Gender (see the discussion in the introduction). For L° to get associated with D it needs to
Functional structure in nominals
reach F°, either via head to head movement (see Ritter 1991) or via Merger. Crucial to the question of how roots are provided with inflection is the question of agreement. After all, the agreement patterns of verbs and nouns are not identical. Moreover, while adjectives show agreement, adverbs do not. I assume that the presence of an agreement desisence is conditioned by syntactic factors. That is the presence of agreement is not statable as a property of the roots, but as a property of roots in syntactic environments. Embick (1998) provides an explicit account of the post-syntactic processes that apply to yield inflected, agreeing forms in Latin. In what follows I give an outline of Embick’s proposal, which I adopt here. In a system such as DM, agreement is treated as a property of the Morphology purely, a process involving the assignment of AGR nodes and the copying of information from structurally defined DPs onto these nodes. Two types of AGR nodes are distinguished: (100) a. AGR1: Person/Number b. AGR2: Number/Gender
The pattern Embick proposes is simple: when an element combines with Tense, it will show AGR1 (which is in fact directly associated with Tense). When an element cannot combine with Tense, it will agree in the pattern of Agr2. The categories that receive AGR nodes are the following: (101) AGR node Tense Determiners Nouns Adjectives
No AGR Verbs Prepositions Adverbs Particles
Functional heads such as Tense will be uniquely identifiable in terms of their abstract feature content. AGR assignment crucially makes reference to the specific head involved. AGR1 is adjoined to finite Tense AGR2 otherwise. The restriction to finite Tense excludes infinitives from the AGR-set. (102) X → X-AGR,
X = T or D
Only roots capable of combining with Tense in the syntax, i.e. verbs will show AGR1. The second component in the system of agreement concerns the conditions under which properties of a DP are copied onto an AGR node. There are two copying processes. The first type is for agreement of the subject-predicate type:
The functional architecture of nominalizations
(103) AGR → AGR ["Person, $Number, (Gender]/_gov by DP ["P, $N, (G]
This will cover both agreement between a finite verb and the surface subject and agreement of a predicate adjective with the DP of which it is predicated. Copying of the second type must copy information from a root onto an attributive adjective: (104) AGR ~ N["P, $N, (G] → AGR ["P, $N, (G]
(104) is what accounts for the agreement facts within DPs, i.e. the fact that prenominal adjectives agree with the head noun. The presence of adverbs inside nominals and the lack of agreement with these can be accounted for if we assume that, since adverbs are spelled-out within the verbal layers of the clause, or rather within the layers that are not unambiguously identified as verbal or nominal, this local relation is critical. As a result, adverbs will not show any agreement features. The rules outlined merely provide information to AGR nodes, which are present in accordance with the rules discussed above. They do not have influence in determining the type of agreement to be realized.
.
Conclusion
In this chapter I argued that process nominals are like verbs in the sense that their internal structure involves an abstract root embedded under vP and AspP. Like verbal clauses they allow adverbial modification. However, unlike verbal clauses, the type of modification they allow is restricted, and they lack a number of phenomena associated with Tense. Given that there is no lexical difference between verbs and process nouns, and between result and process nouns, apart from the functional domain, all can take arguments.
Notes . Longobardi (1994) has proposed a source for N-movement related to the semantic properties of NPs. Longobardi builds his arguments on data like the ones in (i), where, as he argues, the proper name ‘Gianni’ raises to D to check its strong referential feature. This process takes place in the overt syntax in Romance, but covertly in English and German:
Functional structure in nominals
(i) Gianni mio Gianni my . To host the adjectives, a number of functional projections are postulated, as illustrated in (i), whose content and nature, however, remain unspecified: (i) [DP D [FP AP F [FP AP F …[NP N … ]]]] . In recent work Rizzi (1997) and others have argued that the left periphery of the clause, i.e. the C domain, consists of a number of functional projections. In other words, CP is actually composed of a number of functional layers, FocusP, TopicP, ForceP and so on. Giusti (1996) claims that there is some fine structure in the DP as well, but this is defective, both from the crosslinguistic and from the language internal point of view. That is, not all languages have a fine DP-structure, and furthermore, each single position is not found in all languages. Giusti identifies two such layers DP-internally, FocusP and TopicP, which are situated either immediately above DP or immediately below DP. These are not obligatorily present in all languages. I will not consider this issue here. . According to Ritter (1993: 409), there appears to be some crosslinguistic variation in the base positions of gender markers. In certain languages e.g. Spanish (contrary to Picallo 1991) gender is specified on the functional head that bears number specification. . It could be argued that this functional node conflates case and number. Halle and Marantz (1993) also maintain that the single affix morpheme simultaneously denoting number and case in many Indo-European languages, in contrast with many agglutinative languages, is the result of fusion of two nodes. . There is one more feature interfering with gender/number and case, what is usually called inflection class (marker) (Ralli 1994; to appear). Inflection class is essentially a classificatory feature shared by both the nominal stem and the ending, which indicates the exact declensional class of a particular noun. The specific value (see Ralli op.cit.) of this feature triggers a particular phonologically expressed marker (roughly corresponding to Bernstein’s (1993) Word Marker and Picallo’s (1991) Gender). Inflection marker, being a purely morphological feature, simply assigns nouns to declensional classes. . Recall that in Greek Spec,DP is a non-thematic position (see Section 1.3. for details). . Markantonatou’s work is within the LFG model, while Kolliakou’s is mainly an HPSG approach. Alexiadou and Stavrou discuss a limited set of the data to be presented here within the principles and parameters model. . Alec Marantz and Friderikos Valetopoulos point out to me that sequences that do not involve an internal argument receive process readings. (i) to diavasma ja pende ores me eksondose the reading for five hours me exhausted Note that such examples are similar to the English example in (ii): (ii) the race lasted an hour
The functional architecture of nominalizations
According to Grimshaw, such nouns denote simple events. As such they can appear in contexts that bring about the event interpretation, but cannot license argument structure. Another possibility that comes to mind is to analyze such examples as involving an empty pronominal object. Such an object pro has been argued to be present with certain verbs in Italian (see Rizzi 1986). Giorgi and Longobardi (1991) show that the same situation is found with nominals: (iii) la tecnica della sue riconciliazioni con se stessi è veramente perfetta the technique of his reconciliations with oneself is really perfect In (iii) the anaphor se stessi is understood as referentially dependent on an arbitrary antecedent corresponding to the genitive internal argument. . This is pointed out also by Markantonatou (1992) and constitutes the basis of the major distinction between possessives and pseudo possessives in Kolliakou (1995). . Many thanks to Elena Anagnostopoulou for pointing this out to me. . This subsection relies on Anagnostopoulou’s (to appear) research on the double object construction in Greek. . Temporal adverbs are included in the set of complement-type of adverbs, i.e. to the type of adverbs that can appear within a VP. Note that the presence of temporal adverbs here is not regarded as evidence for the presence of a Tense Phrase within nominals. On the contrary, it is assumed that temporal adverbs are consistent with an event predicate, i.e. they relate to the event properties denoted by the head nominal, not to morphological tense properties. Gisbert Fanselow observes that in German the following string is grammatical: (i)
der Mann gestern the man yesterday
(i) receives an interpretation, according to which there is a specific man who was prominent in a given context yesterday. If the possibility of allowing adverbial modification were a signal for the presence of functional layers within nouns, then (i) would suggest that such layers should be included within the structure of the NP ‘man’, which is simply a referential NP. (i) could be analyzed as involving a reduced relative clause, i.e. functional layers not related to the internal structure of the NP. If this is the correct analysis for such patterns, then the presence of adverbs in such contexts says nothing about the internal structure of the nominal in terms of the process vs. result distinction. . It is assumed that that the possibility of a predicate to license a manner adverb is related to its voice features (see Travis 1988, Alexiadou 1997, Cinque 1999 for discussion and references). . Croft (1991: 83) notes that crosslinguistically many nonfinite forms and nominalizations do inflect for various categories, particularly Voice and Aspect. Incidentally, Picallo (1991) observes that -cio is the nominalization affix that characterizes Catalan passive process nominals. She argues that it could be analyzed as the head of a functional category that is responsible for the passive character of the DP. If
Functional structure in nominals
this interpretation is on the right track, then the nominal -s- could be argued to be located in a functional head, denoting event properties (i.e. light v, see discussion in the next section). . At first sight, Greek nominals seem to be related to the perfective stem (i)
a. b.
to diava-s-ma tu vivliu me prosohi the reading the book-gen carefully diava-s-a read-perf-1sg
Stavrou and Horrocks (to appear) point out that the ‘s’ found in a number of Greek derived nominals could be argued to be related to the Indo-European -t-, which became -s in later stages of Greek, and thus homophonous to the Greek perfective aspect morpheme. It seems to be the case that historically at some stage -t-/-s- this morpheme got reanalyzed as being part of the stem, and as constituting no longer part of the affix. Note that the observations made here concerning the functional morphemes present within Greek nominals hold also for the result counterparts of these nominals, in the sense that most of the forms are ambiguous between the two readings. This is not, however, used as an argument that the nominals include Voice and Aspect in their result interpretation as well. Rather, the presence of Voice and Aspect has a clear semantic reflex in process nominals, and the absence of these categories from the internal structure of result nominals also has a semantic reflex. . Bierwisch (1989) observed that pluralization of nominals in German does not always bring about the result interpretation. Consider (i): (i)
die Besteigungen der beiden Gipfel dauerten 6 Wochen the climbings the two tops lasted 6 weeks
In (i) the noun clearly has an event interpretation. One could argue that in this case the nominal has denotes what Grimshaw calls a simple event, and thus pluralization is licit. But whether is intuitive is a different matter. . As also pointed out in Kolliakou (1995), the fact that adjectival modifiers such as frequent can co-occur with plural achievement nouns does not necessarily imply that these are result nominals. Given that these modifiers signal repetition, they are licit with plural forms of telic event nouns. On their singular form, these nouns clearly denote a culminated event, which of course cannot be modified by frequent. . Thanks to Winfried Lechner for pointing this out to me. . It was brought to my attention that sequences such as i sihnes katastrofes tis Trias ‘the frequent destructions of Troy’ and i sihnes katalipsis sholion ‘the frequent occupations of schools’ are grammatical. However, note that both strings involve actually result nominals. A test that suggests that this is the case involves the use of possessive clitic instead of NP. (i)
a.
*i sihnes katastrofes tis the frequent destruction hers
The functional architecture of nominalizations
b. (ii) a. b.
i sihni katastrofi tis the frequent destruction hers *i sihnes katalipsis tus the frequent occupations theirs i sihni katalipsi tus the frequent occupation theirs
As Kolliakou (1995) points out the clitic, when present within the DP, is always referential, i.e. an extensional, never an intensional or ‘kind’ modifier of the noun. The fact that the clitic is not licit in the plural nominal construction suggests that this involves a kind-modifier, i.e. a type of occupation. . A similar in spirit proposal, which does discriminate between the two types of nominals, is found in Picallo (1991). According to Picallo, the event/process interpretation for nominals is available when one analyzes them as being derived from a category neutral stem, common to both nouns and verbs (see also van Riemsdijk 1983, Pesetsky 1995 among others for similar ideas). Picallo differs from Marantz (1997) in that she analyses the nominalization affix as an inflectional element. In such a structure the nominalization affix heads an NP, which is a functional category. This head takes as a complement a category neutral-lexical projection (LP) headed by a stem, identical to that of the corresponding verb. The stem L° becomes a noun at the syntactic component, by head raising which makes event nominals necessarily cases of syntactic nominalization. Picallo, however, proposes that result nominals have the structure illustrated below, i.e. no category neutral base is included in those. For Picallo, the Projection Principle applies for nominals only when the lexical projection of a DP is category neutral. In other words, category neutral lexical projections enter the D-structure of syntactic nominals (event denoting DPs), and that of clausal structures, both in active and passive clauses. While in nominalizations, categorial specification of the verbal element obtains via head-adjunction to the head hosting the passive morpheme, in the active sentences categorial specification obtains by head-adjunction to a functional category VP, selecting the category-neutral LP. This functional VP is morphologically overt in Catalan. It is headed by the thematic vowel characteristic of the Romance verbs, as in (ib): (i)
a. b.
En Guillem menjava les pomes Guillem ate the apples [VP Guillem [V’ -a- [LP[L’ menja- [les pommes]]]]]
Adjunction of the verbal stem to the word marker allows the verb to assign structural case. . In this chapter deverbal nominals, under a process or a result reading, were the focus of the discussion. I briefly make here some comments on object/concrete nouns. Object nouns designate entities in the world as opposed to process nouns. According to Grimshaw, nouns that do not denote an event do not have an argument structure to satisfy. Thus, concrete nouns do not entail an event and do not have an argument structure. Therefore, they do not have specific theta-roles to discharge. They optionally take semantic participants, with which according to Grimshaw, they are in rather loose rela-
Functional structure in nominals
tions. That is concrete nouns are much like result nominals in their behavior. For the type of approach put forth here, object nouns are also seen as unspecified roots inserted under a set of nominal functional projections. Thus they receive a similar treatment to the one proposed earlier for result nominals. More needs to be said on the semantic classes of the roots involved in such a formation. . A number of contexts argue for a separation of these two features in T, which in fact can be checked by different elements. Collins (1997) discusses two cases: expletive constructions and locative inversion. Consider (i): (i) there are people in the garden In (i) the expletive checks the EPP-feature of T. If the expletive checked the nominative Case feature of T, then T would not have any Case feature to check against people. Hence, there does not have a Case feature, and the nominative feature of T is checked (covertly) by features of the NP. . Here I assume that EPP is not necessarily satisfied by Merge/Move XP. Merge/Move X° can also check EPP, as in VSO pro-drop languages (cf. Alexiadou and Anagnostopoulou 1998 for detailed argumentation). . However, de Wit (1997) provides a formal parallel to the EPP requirement in verbal clauses. According to de Wit, the conceptual equivalent to the EPP is the requirement to fill D. This is why possessive pronouns and possessive elements in general are located in D, and this is perhaps why some languages have expletive determiners. De Wit assumes, following Alexiadou and Anagnostopoulou (1998), that there are two ways in which EPP can be satisfied, either by merging/moving an X° or by merging/moving an XP. According to de Wit, this specific view on EPP-checking enables the formulation of a parallelism to the expletive associate facts extensively discussed in the generative literature within the nominal domain (for a specific implementation see Chomsky 1995 and Frampton 1995). However, as discussed in Section 1.3, the status of D varies across languages. Thus Spec,DP is unlike Spec,TP a thematic position in English, but it is like Spec,CP, that is an A¹-position, in Greek. If Spec,DP is a thematic position, then it is not clear whether an argument based on the EPP could go through, as the EPP is relevant for non-thematic positions. If it is an A¹-position, then clearly the argument does not go through under the standard way of looking at the EPP. A possibility that comes to mind is to say that the need for D to be filled is for the whole extended nominal projection to count as an argument (see the discussion in Section 1.1), which is certainly not the case for T. . Similarly, if Raising involves restructuring of the lower and higher INFL, as has been proposed for Raising constructions in Romance in Torrego (1989), and Sola (1992) among others again it is expected that Raising will not take place in nominals. . Greek has nominalized clauses investigated in Roussou (1991). These include a CP which enters in the complement domain of D (see also Chapter 4):
The functional architecture of nominalizations
(i) to oti irthe me ksiafniase the that came-3sg me surprised ‘The fact that he came surprised me’ If D and C are equivalent it is not immediately clear why the pattern in (i) arises. The reader is referred to Roussou’s work for extensive discussion. For the purposes of the present discussion note that (i) contains tense features, i.e. it involves a tensed clause that initially appears in the complement domain of C, as the correlation mentioned in the text actually predicts. . As pointed out to me by Henry Davis, an issue arises concerning the status of predicate nominals. If nominal formation involves the embedding of an unspecified root under D, then the issue arises how can one distinguish between player in (ia) and (ib): (i)
a. b.
John is a player The player arrived
My suggestion would be that nominals of the type (ia) are roots embedded under Number only and do not include D.
Chapter 3
Intransitivity in Nominalization
In the previous chapter I have argued that process nominals are roots appearing beneath the ‘verbal’ functional heads Aspect and v. In this chapter I demonstrate that the variant of v within event nominals of the destruction type is the one that does not project an agent., To substantiate this point, I examine data from several language families (Section 1). Specifically, I discuss process nominalizations in Greek, English, German, French, Italian, Spanish, Catalan, Hungarian, Slavic (Polish and Russian) and Semitic (Hebrew and Arabic). As I further point out in Chapter 4, this restriction holds for certain gerundial and infinitival constructions across languages as well. The evidence is based on the observation that the single genitive argument within such nominals is necessarily interpreted as a theme and agents, if syntactically realized, must be realized as PPs. Moreover, Spec,vP does not seem to be active in any way within nominals. As will be shown, there is no process of passivization within nominals (Section 2). Rather, such nominals are generally intransitive, in fact very similar to ergative constructions, since, when agents are present, they bear the role of affector/instrument. That is they are of a different nature than the agents found with verbal passives or transitive agents. Note that Grimshaw (1990) has also suggested that nouns are generally intransitive. However, for Grimshaw intransitivity is shared by process nominals and by result nominals which are not argument taking. But the notion of intransitivity developed in Grimshaw refers to the fact that nouns are defective theta-markers, i.e. they need some help in order to theta-mark their complements, for instance the presence of the special preposition of in English. From this a number of facts follow within Grimshaw’s system. For instance, the fact that process nominals do not take sentential complements, the fact that NP movement is not possible within argument-taking nominals, and the fact that ‘passive’ complex event nominals do not exist. I deal with these two last issues in Section 2 of this chapter.
Functional structure in nominals
.
Event nominals are ergative constructions
It has been noted in the literature that across languages event nominals are, when derived from transitive predicates, ‘passive’ and not transitive and that they are derived from unaccusative predicates, but not from unergative ones (see Picallo 1991, Bottari 1992, Alexiadou and Stavrou 1998a, Alexiadou 1999a,b among others). In ‘passive’ nominals the agent appears as an adjunct introduced by a preposition or an adverbial phrase. In what follows I present in detail the type of data that led researchers to this conclusion.
. Greek event nominals Consider the Greek examples in (1). In (1a), the agent is introduced by a prepositional phrase. This PP has the same form as the one we find with verbal passives (cf. 1b). The example in (1c) shows that the theme argument of nominals derived from unaccusative predicates appears in the genitive, as already discussed in the previous chapter; (1d) shows that in nominals derived from unergative verbs the single argument cannot surface as a PP (recall here that such nominals have only a result interpretation). (1) a.
i katastrofi tis polis apo tus varvarus mesa se tris meres the destruction the city-gen by the barbarians within three days ‘The destruction of the city by the barbarians within three days’ b. i poli katastrafike apo tus varvarus the city-nom destroyed-pass by the barbarians ‘The city was destroyed by the barbarians’ c. i sinehis ptosi ton timon anisihi tus pandes the constant fall the prices-gen worries everybody-acc ‘The constant fall of the prices worries everybody’ d. to kolimpi tu Jani/*apo to Jani the swimming the John-gen/by the John
Observe also the contrast between (2a) and (2b). In (2a) the single genitive is interpreted as the theme argument, i.e. it is the police that has been captured. In (2b) on the other hand, where the nominal appears in the plural to force a result interpretation, the genitive can be interpreted as the possessor, i.e. the series of captures belong to the police.
Intransitivity in nominalization
(2) a.
i silipsi tis astinomias the capture the police-gen b. i silipsis tis astinomias the captures the police-gen
As mentioned in Chapter 2, with nouns such as metafrasi ‘translation’, paratirisi ’observation’, the genitive can be interpreted both as an owner and as a theme. However, if the nominal is put into a context that discriminates between a process and a result reading, then the owner interpretation survives in the result context only, as in (3b) below. In (3a), on the other hand, the genitive is necessarily understood as the object of observation: (3) a.
i paratirisi tis Marias the observation the Mary-gen b. i paratirisis tis Marias ine panda akrivis the observations the Mary are always exact-pl
While omitting the PP-agent is possible, omission of the object while retaining the subject is not compatible with maintaining a process interpretation: (4) *i sihni katastrofi tu Nerona the usual destruction the Nero-gen
What is excluded is an agent bearing genitive case and co-occurring with a genitive theme, no matter in which order the two genitives appear (see Horrocks and Stavrou 1987): (5) a. *i silipsi tu Jani tis astinomias the capture the John-gen the police-gen b. *tis astinomias i silipsi tu Jani
Two genitives are possible but only with nouns of the type description, as mentioned in the previous chapter, although judged unnatural by several speakers (see Horrocks and Stavrou 1987, Mouma 1993): (6) i perigrafi tu topiu tis Marias the description the landscape-gen the Mary-gen ‘Mary’s description of the landscape’ (7) i metafrasi tis Odisias tu Kakridi the translation the Odyssey-gen Kakridis-gen ‘Kakridi’s translation of the Odyssey’
Functional structure in nominals
As will be shown in Section 1.3, something similar is found in Romance. . Event nominals in English and other Germanic languages With nominals of type destruction the presence of two of-phrases is ungrammatical in English: (8) *the destruction of the city of the barbarians
There is, however, a basic difference between Greek and English nominalizations, referred to several times here: English nominalizations derived from transitive verbs can be both transitive and intransitive (see Chomsky 1970, Kayne 1984 among others): (9) a. the/John’s destruction of the city b. the/the judge’s reversal of the decision
Safir (1987) points out that in such examples the pre-nominal genitive is not interpreted as the owner of the head noun, but rather it refers to the causer of an event of destroying, etc. However, even in English such transitive nominalizations are limited. For instance, they are acceptable with nominals derived from verbs like destroy, reverse, assassinate, invade, capture, but not with nominals derived from verbs like grow (see Pesetsky 1995, Marantz 1997 for recent discussion). (10) a. the growth of tomatoes b. *John’s growth of tomatoes
Nominals such as destruction are related to verbs that resist the causative/inchoative alternation. Pesetsky (1995) has pointed out that English nominalizations cannot be derived from causative verbs. However, given that nominalizations such as the ones in (9) above are not permitted in Romance and Greek, the relevant property that distinguishes them from their English counterparts has to be investigated. I come back to this issue in Chapter 4. German and Icelandic also permit prenominal genitives, although there is some speaker variation with respect to the acceptability of such examples: (11) Attilas Zerstörung der Stadt Attila-gen destruction the city-gen
In both languages such constructions are restricted and only possible with proper names.
Intransitivity in nominalization
In German, as pointed out to me by Gisbert Fanselow (personal communication) and as discussed in Audring (1999), certain nominals can take both NPs in the genitive and PPs introduced by von as their theme argument (cf. 12a). Moreover, the PP can be interpreted sometimes as a theme and sometimes as an owner, depending on whether the result or the process reading is triggered. Thus in (12c), where the plural form of the noun is used bringing about the result reading, the PP is understood as the owner of the observation. In (12b), however, the NP ‘the birds’ is understood as the object of the observation: (12) a.
die Entwicklung von großen Maschinen/großer Maschinen the development of big machines/big machines-gen b. die Beobachtung von Vögeln the observation of birds c. die Beobachtungen von Vögeln the observations of birds
The example in (13a) below is reminiscent of Greek ‘passive’ nominals. However, it seems that on a par with (13a), the pattern in (13b) exists as well. In (13b) the theme argument is introduced by a PP, while the agent bears genitive case. This pattern is reminiscent of one type of Hebrew nominalizations to be discussed in Section 1.5: (13) a.
der Angriff der serbischen Ziele durch die NATO the attack the Serbian targets by the NATO b. der Angriff der NATO von serbischen Zielen the attack the NATO-gen of Serbian targets
. Event nominals in Romance .. Catalan Picallo (1991) notes that event nominals in Catalan appear to be incapable of realizing the agent as an argument in the genitive, as the ungrammaticality of (14) suggests. (14) *l’afusellament de l’escamot d’en Ferrer Guardia the execution of the squad of Ferrer Guardia ‘the squad’s execution of Ferrer Guardia’
Genitive arguments are always interpreted as themes in such examples. Thus, the following strings are grammatical, but only in the interpretation where the
Functional structure in nominals
squad and the police are executed or captured respectively. The agent, when realized, must be realized as a PP (cf. 15c, see also Drijkoningen 1993): (15) a.
l’afusellament de l’escamot the execution of the squad b. la captura de la policia the capture of the police c. l’afusellament d’en Ferrer Guardia per part de l’escamot the execution of Ferrer Guardia on part of the squad
If the agent is realized in the genitive, then the result interpretation is obtained, since the nominal can appear in an environment requiring such an interpretation (cf. 16b). As discussed in the previous chapter, events or processes cannot be inconsistent: (16) a.
la demonstracio d’en Joan del teorema de Pitagores the proof of Joan of the theorem of Pythagoras b. la demonstracio d’en Joan del teorema de Pitagores the proof of Joan of the theorem of Pythagoras es inconsistent is inconsistent
Catalan unaccusative nominals are also ambiguous between an event and a result interpretation. The former is brought up by the inclusion of an aspectual modifier, as in (17): (17) a.
la tornada dels turistes durant l’estiu the returning of the tourists during the summer b. la crescuda del riu a la primavera the swelling of the river in spring c. la desapacicio d’en Joan a l’hora de sopar the disappearance of Joan at dinner time
Unergative nominals, on the other hand, are never ambiguous. They never have the event reading and always denote results. In these, agents can never appear as PPs: (18) a.
el salt de l’atleta the jump of the athlete b. *el salt per part de l’atleta the jump on the part of the athlete
Intransitivity in nominalization
The majority of these nominals is morphologically similar to non-deverbal nouns: they are unable to appear with a nominalization affix, such as -acio, -ada, -iment, -eig, while event nominals appear with such affixes (see example (17) above): (19) a. *la saltada, el saltament de l’atleta the jumpation, the jumpment (of the athlete) b. *el xisclement del nen the screaming of the child
For Picallo, the difference in nominalization affixes correlates with the fact that unergative nominals denote results and not processes. Furthermore, Catalan psych nominals lack passive counterparts, so experiencer arguments never appear within PPs (see the discussion on the nature of the by phrase in Chapter 2 and in Section 4 of this chapter). (20) a.
el temor d’en Joan als terratremols the fear of John to earthquakes b. *el temor als terratremols pare part d’en Joan the fear to earthquakes on the part of Joan
According to Picallo, the lexical, denotative, and syntactic properties of psych DPs suggest that these should be analyzed on a par with non-deverbal, result nouns and unergative nominals. .. French, Spanish and Italian In French, Spanish and Italian (see Cinque 1980, Milner 1982, Zubizarreta 1987, Bottari 1992 among others) two ‘de-phrases’ cannot appear within process nominals (cf. 21). (21) a. *la destruction de la ville des soldats the destruction of the city of the soldiers b. la destruction de la ville par les soldats the destruction of the city by the solders c. *la captura de los fugitivos de los soldados the capture of the fugitives of the soldiers d. la captura de los fugitivos por los soldados the capture of the fugitives by the soldiers e. *la cattura del soldato del enemico the capture of the soldier of the enemy
French
Spanish
Italian
Functional structure in nominals
f.
la cattura del soldato da parte del enemico the capture of the soldier by the enemy
In (21b, d, f) the agent is introduced by a PP that has a similar form to the PP introducing the agent in verbal passive sentences. In these languages, (21a, c and e) become grammatical again, when an adjectival possessive precedes the nominal (cf. (22a) from Milner 1982: 130). Moreover, as (22b) shows, similarly to Greek the presence of two genitive phrases is legitimate with ‘translation’ type nominals, although judged unnatural (cf. the Catalan examples in (16)). (22) a.
sa description d’un passant her description of a passant b. la traduction de Pierre de l’œuvre de Zola the translation of Peter of the work of Zola
In (22a), the adjectival possessor can only be interpreted as the owner. However, the prenominal possessor in Romance can never be interpreted as the theme, the agent being introduced by a de-phrase: (23) *sa démonstration d’un mathématicien grec her demonstration of a mathematician Greek
As can be clearly observed, in Romance and the Greek there is a crucial difference between nouns like destruction and nouns like translation: two genitives cannot appear within the former, but can within the latter. Incidentally, this group of nouns is the one that produces nominalizations that are ambiguous between the process and the result reading in all these languages. English, on the other hand, differs from these languages in permitting transitive nominalizations even with nouns of the destruction type. Extensive discussion of these patterns takes place in Chapter 4. Bottari (1992) argues extensively that Italian process nominals do not actually have an external argument slot. One of his arguments is repeated below. As (24) shows, while a nominal derived from an unaccusative predicate is compatible with the negation strategy (24a), a nominal derived from an unergative predicate is not (24b): (24) a.
il mancato arrivato del treno the non-arrival of the train b. *il mancato pianto di Luigi the non-mourning of Luigi
Intransitivity in nominalization
The possibility of a head noun to be modified by negation is sensitive to the interpretation of the nominal. According to Bottari, the negated nominal strategy is bound to [+PAS] nominals or the [+PAS] version of nominals, at least: (25) a. *la mancata descrizione the non-description b. la mancate descrizione della vita the non description of the city
Bottari explicitly argues that Romance passive nominals of the type (25b), i.e. nominals that include internal arguments only, are process nominals. According to Bottari’s analysis, the presence of negation is sensitive to the presence of an internal argument. For this reason, only passive and unaccusative predicates can be subject to it. On the other hand, unergative predicates are not acceptable as is manifested by the ungrammaticality of (24b). This fact suggests that only passive and unaccusative predicates give true process nominalizations in Italian.
. Slavic event nominals .. Russian In Russian process nominals (cf. Comrie 1976, Schoorlemmer 1995) the theme argument surfaces bearing genitive case, while the agent bears a different case, namely the instrumental. In (26a), the process reading of the nominal is highlighted by the presence of the aspectual modifier. As in Greek, only one argument bearing genitive can be present. This is illustrated in (26b), an ungrammatical example where both arguments bear genitive case: (26) a.
uni˜toþeni butylki piva za pjat’minut extermination bottle-gen beer-gen in five minutes dvumja studentami-ve㘠vpolne oby˜naja two-students-instr is thing entirely usual b. *Razruãenie Saraeva blokadnikov destruction Sarajevo-gen siege-holders-gen
Process nominals are also derived from unaccusative predicates, and as Schoorlemmer points out, also from unergative predicates. In both cases the single argument appears in the genitive. An example including an unergative nominal is given in (27):
Functional structure in nominals
(27) Ee polzanie po polu v te˜enie celogo ve˜era her crawling over the floor in course of the whole evening zabespokoilo menja worried me
The presence of event unergative nominalizations seems at first sight problematic for the view to be proposed here that process nominals are generally unaccusative. However, it could be argued that Russian does not distinguish between unergatives and unaccusatives, thus permitting intransitive structures that generally include an internal argument. .. Polish Rozwadowska (1995) discusses the distinction between argument taking nominals and non-argument taking nominals in Polish. She shows that Polish nominals denoting complex events (in Grimshaw’s terms) can be split into two groups: verbal nominals and derived process nominals. The former preserve overt morphological contrast between the imperfective and the perfective aspect, as already demonstrated in Chapter 2. The relevant examples are repeated below (from Schoorlemmer 1995: 321): (28) a.
ocenienie studentów przez nauczyciełi nastapiło evaluation-perf the students-gen by teachers occurred szybko quickly b. ocenianie studentów przez nauczyciełi ciagneło si˛e evaluation-imp students-gen by teachers lasted refl przez cały tydzien´ through the whole week
As discussed in Chapter 2, nouns terminating in -nie/-cie inherit their aspectual properties from their related verbs. Crucially they are marked morphologically for aspect. The inheritance of aspectual properties is also reflected in their choice of aspectual modifiers: (29) ich zreperowanie w ciagu dwóch tygodni their repairing-perf in course of two weeks ‘repairing them in course of two weeks’
Polish verbal nominals take obligatorily direct object arguments, but their inheritance of other arguments is not possible.
Intransitivity in nominalization
On the other hand, process derived nominals terminate in semi-productive and non-productive nominalizing affixes. These too can be modified by agentive adjuncts, and aspectual phrases (from Centarowska 1997): (30) a.
ochrona ciebie przed promieniowaniem przez najbliizszy protection your-gen against radiation during coming tydzien´ week b. celowa sprzedaz was do gorszego klubu pilkarskiego deliberate sale you-gen to worse club football
These nominals differ from the verbal nominals in that they do not permit modification by manner adverbs and no accusative marked adjuncts. In both types of nominals, however, as in Russian and Greek, only one argument can be marked with genitive case. What is important for the discussion here is the fact that Schoorlemmer acknowledges that both in Russian and in Polish there is only one structural case associated with the single argument of the verb (unaccusative or unergative). If a nominal is derived from a transitive verb then this must be passivized, i.e. the external role is expressed by a PP. . Semitic event nominals .. Hebrew Hebrew process nominals come in two types. The first one includes a by-phrase similar to the one used to form the passive (see Borer 1993). In this type, the theme argument appears either in the šel-phrase or is bare. The second construction is known in the literature as the construct state, while the former is known as the free state. Doron (1989) has claimed that with concrete nouns the complement is connected to the head of the construct state with some kind of inalienable possession relation. However, Siloni (1997) argues against this distinction. (31) a.
ha-Sikum šel ha-’ir ‘al yedey ha-Siltonot the rehabilitation of the city by the authorities b. Sikum ha’-ir ’al yedey ha-Siltonot the rehabilitation the city by the authorities
Borer (1993) extensively argues that (31) involves a passive within the nominal. In fact both Hazout (1995) and Borer (1993) propose that Hebrew process nominals contain a passivized verbal projection.
Functional structure in nominals
As in Greek, while omission of the subject is possible, omission of the object while retaining the subject is not compatible with maintaining the process reading for the nominal: (32) a.
hoxaxat ha-te’ana al yedey ha-matematika’it proof the claim by the mathematician kedey lizkot ba-pras in order to win the prize b. *ha-hoxaxa šel ha-matematika’it the proof of the mathematician
The second type of Hebrew process nominals is illustrated in (33). In (33) the agent is introduced either by the šel-phrase (33b) or bears bare genitive (33a). Most importantly the theme argument bears accusative case. This is manifested by the fact that it is introduced by the accusative marker ’et: (33) a.
ha-harisa ha-cava ’et ha-’ir the destruction the army acc the city b. ha-harisa šel- ha-cava ’et ha-’ir the destruction of the army acc the city ‘The army’s destruction of the city’
As (34) shows, this type of nominal can also be modified by adverbials: (34) harisat ha-cava ’et ha’ir bi-mehirut the destruction the army acc the city in quickness ‘The army’s destruction of the city quickly’
Thus the second type of Hebrew nominals poses a potential problem for the view that event nominals are crosslinguistically intransitive (cf. Siloni 1994: 67ff, Borer 1993). I come back to this in Chapter 4. .. Arabic Arabic masdars, much like Hebrew process nominals, also come in various types. In nominals of type (35a), the theme is introduced by a preposition, while the agent bears genitive. On the other hand, in (35b) the theme bears genitive (cf. Fassi-Fehri 1993: 234). aqlaqa-nii ntiqaad r-rajul-i li-l-masruuc-i annoyed-me criticizing-nom the man-gen to the project-gen b. hadm-u l-madiinat-i the destroying the city-gen
(35) a.
Intransitivity in nominalization
In nominals of type (36), the theme argument appears in the accusative and the agent in the genitive. In fact this construction has been argued to bear some resemblance to English gerunds (see Aoun 1981, and Fassi-Fehri 1993 for discussion and references). (36) ʔaqlaqa-nii ntiqaad r-rajul-i l-masruuc-a annoyed-me criticizing-nom the man-gen the project-acc
. Hungarian event nominals In Hungarian derived nominals surface with two suffixes: -as/-es, glossed DEV in the examples below (taken from Szabolcsi 1994). (37) hivatkoz-ás reference-dev
The internal argument of derived nominals appears in the possessor’s position bearing nominative case: (38) a háború egyév-ev belül való befejez-és-e the war-nom year-loc within being end-dev-poss.3sg ‘The ending of the war within one year’
Agent subjects appear as possessors or in an àltal by-phrase and can mostly remain unexpressed: (39) a.
Péter Mari ellen való felszolal-ás-a Peter-nom Mari against being speak-dev-poss-3sg ‘Peter’s speaking against Mari’ b. Peter Mari által való megszégyenít-és-e Peter-nom Mary by being humiliate-dev-poss-3sg
Theme subjects of intransitive nominals can also appear in the possessor position (40a). Experiencer subjects appear only in possessor position, and can never introduced by a by-phrase (cf. (40b) vs. (40c)): (40) a.
a problema tegnap delutan való the problem-nom yesterday afternoon being fölmerül-és-e emerge-dev-poss-3sg ‘The emergence of the problem yesterday afternoon’
Functional structure in nominals
b. Péter János által való megver-és-e Peter-nom Janos by being beat-dev-poss-3sg beating up of Peter by Janos c. *a János által való megver-és-e
Szabolcsi observes that when a noun is derived from a verb that has both a transitive and an intransitive construal, the process interpretation is only available on the intransitive reading: (41) a.
Mari mos Mari-nom wash-3sg b. a mos-ás the wash-dev c. Mari mos-ás-a Mari-nom wash-dev-poss-3sg *the fact that Mari washes clothes
In Hungarian if there is one argument in need of case, it can always get nominative in the possessor’s position. If there are two arguments, one of them, if an agent, resorts to being expressed as a by phrase. Inherent case on postpositions remains unaffected within nominalizations. In other words, it is never the case that the two arguments are expressed with the same case. One of them must be included in the form of a prepositional phrase. To conclude, in this section I examined process nominals in a number of languages. I have shown that crosslinguistically process nominals are either ‘passive’ or unaccusative. Before proceeding to a structural account of this fact, in the next section I turn to a detailed investigation of the properties of passive nominals.
. On the properties of passive nominals . Background The term passive nominal refers to two related but not quite identical types of nominal constructions. The standard example of a passive nominal is given in (43). Chomsky (1970, 1972), Kayne (1984), Pesetsky (1990) and others argue that such nominals have the same argument structure as active nominals of the type exemplified in (44). As a consequence, they also have the same D-structure
Intransitivity in nominalization
and are then derived by movement, akin to movement of arguments within passive verbal clauses for reasons of case-assignment. The theme argument would move to subject position so that it can be assigned case by Infl/D(=’s). Exactly as with verbal passives, the underlying object of the nominal becomes the surface subject (cf. also Anderson 1983, Giorgi and Longobardi 1991, Zubizarreta 1987, Cinque 1980). (42) D-structure (43) S-structure (44) John’s distribution of the drugs
[distribution the drugs] the drugs’ [distribution t]
Grimshaw (1990) notes that nominals such as the ones in (43) are similar not only to verbal passives, but also to cases of adjectival modification by a referential or ethnic adjective. In the examples below, (45) is a passive nominal whose interpretation is quite close to that of (46). (47) is also similar to (45): here the adjective appears to be related to the argument corresponding to the argument of the verb. (45) Reagan’s defeat (46) Reagan was defeated (47) The French defeat
In Greek and Romance, the term passive nominal refers to nominals such as destruction, capture and so on which are inherently passive in meaning and co-occur with by-phrases across languages, and also in English (see Cinque 1980): (48) i silipsi tu Jani apo tin astinomia the capture the John by the police
The Romance/Greek and the English constructions are similar in that they both have a D-structure similar to verbal passive structures, i.e. no agent is merged at the specifier of vP. Picallo (1991) argues that the Romance construction involves a movement akin to that of the English one in (43). In the Romance counterpart of (48), the theme argument moves to a functional projection within DP, Spec,GenP, for reasons of Case-assignment. In this respect, (48) differs from (43), in that only the latter involves fronting to Spec,DP. In Romance such instances of preposing involve only pronominal objects (see 2.2.2), while in Greek preposing has the properties of A¹-movement (see Chapter 2). But this is not the only difference between the two constructions.
Functional structure in nominals
Another difference relates to how they fare with respect to Affectedness (see the discussion in the next section). Before turning to these issues, note that it is by no means a settled issue whether nominals of type (43) involve movement. For instance, other researchers have claimed that passive nominals never have a D-structure representation as in (42) (see Williams 1982, Higginbotham 1983, Grimshaw 1990 and to some extent Safir 1987). Given that such nominals do not have an argument structure, no passive movement takes place NP-internally. On the contrary, these authors claim, (49) corresponds to the representation such nominals should receive: (49) a. a-structure: b. D and S-structure
distribution the drugs’ distribution
Yet another line of research, represented by e.g. Doron and Rappaport-Hovav (1991), argues that passive nominals involve lexical externalization of the internal argument, as in (50) below. This analysis crucially relies on the fact that Spec,DP is a thematic position in English, given that it is not licensed by any other principles such as the EPP, as discussed in Chapter 2. (50) a. a-structure: b. D- and S-structure
distribution (y) the drugs’ distribution
In the next sub-sections I first deal with the properties of English passive nominals. I discuss the notion of Affectedness in 2.2.1 and certain crosslinguistic asymmetries in the way nominals show sensitivity to this property in 2.2.2. In Section 2.2.3 I argue against the result analysis of passive nominals. Building on Tenny (1994) and Snyder (1998), I show hat English passive nominals are generally event nominals, i.e. have a terminative interpretation. Given that Grimshaw’s tests diagnose process readings, i.e. are sensitive to a durative interpretation, it is expected that nominals solely possessing a telic interpretation are not well behaved with respect to these tests. In Section 2.2.4 I discuss the derivation of this type of English nominals. In Section 2.3 I turn to the role of adjectival modification. I argue that nominals of the type (43) are event nominals, while adjectival modification is consistent only with result nominals. Finally, in 2.4 I show that nominals of the type in (48) do not involve a process of passivization.
Intransitivity in nominalization
. Affectedness and aspectual properties .. On the notion of affectedness It is well known that in English a certain class of nouns does not admit the passive construction (cf. Chomsky 1970 and references therein for a first attempt towards explaining this asymmetry). Those who do admit it have been argued to be subject to the so-called Affectedness constraint. In the literature there have been several attempts to define the Affectedness property. Anderson (1977) defines objects that cannot be realized in the prenominal position as ‘unaffected’ in the traditional semantic sense, i.e. as not being affected by the eventuality denoted by the predicate. She formulates the following generalization: if the head noun does not express an action that affects, i.e. modifies, the state of the object, the latter cannot occur in the prenominal position. The data in (51) seem to be consistent with this. (51) a. b. c. d.
*the fact’s knowledge knowledge of that fact *the book’s discussion the discussion of the book
It was later observed by Jaeggli (1986) (cf. also Roberts 1987, Tenny 1987, Zubizarreta 1987 among others) that similar restrictions hold in the domain of middles: (52) a. This wood slits easily b. *This cat chases easily
In particular, Jaeggli (1986) formulated the ‘Affectedness Constraint’ as follows. (53) If a complement of X is unaffected, it is impossible to eliminate the external role of X.
On this view, Affectedness is a property that should be derived from a general theory of theta-assignment. Other semantic characterizations of Affectedness have appeared in the literature. Roberts (1987), for example, identifies verbs with affected arguments as the class of accomplishment verbs. As will be shown, there seems to be evidence that there is a connection between achievement predicates and Affectedness. On the other hand, Tenny characterizes Affectedness as a semantic property, having to do with predicate meaning and the manner in which a predicate’s arguments participate in the event the predicate describes. For Tenny, affected objects are objects capable of delimiting the event denoted by the verb.
Functional structure in nominals
Doron and Rappaport-Hovav (1991) define Affectedness over event structure representations and claim that what characterizes it is the separation of the arguments of the verb into different sub-eventualities, with the external argument missing from one of the sub-eventualities. Destroy is such a predicate type, i.e. one can distinguish between a Cause and a Become part, know is not such a predicate type, as one cannot distinguish between sub-eventualities. (54) y is an affected argument of V (x, y) iff the event-structure of V contains a sub-eventuality e such that y, but not x, is an argument in e.
.. Affectedness across languages While in English the possibility of preposing within NPs is a context for identifying affected objects, the behavior of related nouns in other languages is not similar. For instance, Cinque (1980) observes that in Italian the class of nouns in question is bipartite: some of these nouns can be passivized, whereas others still cannot. Notice that there is a difference between Italian and English regarding NP-preposing within the DP in that there is no NP-fronting in Italian (55c). Hence the reference to NP preposing in Italian is understood in terms of pronominal preposing : (55) a.
la conoscenza dell’algebra the knowledge of algebra b. la sua conoscenza the its knowledge c. *la di Giorgio descrizione Giorgio’s description
In Italian it is perfectly possible to express the internal argument by means of a possessive with the same class of nominals that cannot undergo preposing in English. Other nouns displaying the same contrast pointed out in Cinque (1980) are discussione ‘discussion’, pereguimento ’pursuit’. Greek lacks possessive adjectives, and moreover, as pointed out in Chapter 2, DP-preposing is an instance of A¹-movement. Hence Greek does not exhibit any type of restriction in terms of NP-internal fronting: (56) a.
tis polis i katastrofi the city-gen the destruction ‘the city’s destruction’ b. tis algevras i gnosi the algebra-gen the knowledge ‘algebra’s knowledge’
Intransitivity in nominalization
Markantonatou (1992) points out that Greek passive nominals do not seem to be subject to an Affectedness restriction, as the examples in (56b) and in (57) below are grammatical in Greek, while the equivalent English passive constructions involving fronting are not: (57) i sizitisi tu vivliu apo tus idikus the discussion the book-gen by the specialists ‘the book’s discussion by the specialists’
Kolliakou (1995), however, attempts to establish a test for diagnosing Affectedness in Greek, namely the admissibility of a pronominal object possessive. Building on Tenny, she points out that nominals that lack a telic event interpretation cannot co-occur with such an object (see also Chapter 2): (58) a.
to mazema tis elias the harvesting the olive-gen b. *to mazema tis the harvesting cl-gen (59) a. i dilosi tis paretisis the statement the resignation b. *i dilosi tis the declaration cl-gen
(59b) is ungrammatical if the clitic is interpreted as the theme argument of the nominal. It is grammatical if the clitic refers to the individual that made the statement. Note, however, that both the nominals in (58) and (59) are passive. In other words, the above facts show that the passive character of the nominal is not directly related to Affectedness. Rather Affectedness seems to be a more general property characterizing the aspectual properties of the noun. Anagnostopoulou (to appear) points out that certain cases of Greek passive nominalizations are also subject to the Affectedness constraint, in a manner similar to English ‘passive’ nominalizations. Consider the examples in (60): (60) a.
*?to
kitagma tis thalassas jia ores me kurase the look the sea-gen for hours me tired made ‘Looking at the sea for hours made me tired’ b. to kitagma tu vivliou jia ores me kurase the look the book-gen for hours me tired made ‘Looking at the book for hours made me tired’
Anagnostopoulou observes that it is plausible to account for the contrast
Functional structure in nominals
between (60a) and (60b) in terms of the Affectedness constraint. The book can be construed as an affected theme more easily than the sea. For example, a book can change position while I am looking at it, I can turn its pages, I can write notes on it etc. On the other hand, there is no way I can affect the sea by simply looking at it. Note further that only (60b) can be accompanied by a PP agent. (61) to kitagma tu vivliu ja ores apo to Jani the look the book-gen for hours by John
Consider now Spanish. In Spanish, affected arguments are marked by the dative preposition a (Torrego 1998:18 and references therein). (62) a.
el guardia vió (a) un prisionero the guard sees to a prisoner b. el guarida golpeó *(a) un prisionero the guard sees a prisoner
The object of the verb ‘see’ in (62a) does not have to be marked by the dative preposition. The object of the verb ‘beat’ in (62b) must appear with the dative preposition. This marking shows up in nominalizations as well (Torrego 1998: 38f.): (63) a.
su acusacion al teniente her accusation to the lieutenant b. *su acusacion del teniente her accusation of the lieutenant
The genitive preposition is rejected in favor of the dative preposition a ‘to’. According to Torrego, such nominals are process nominals. When the agent is not present, the nominal can be followed by the a genitive phrase: (64) la acusacion del teniente the accusation of the lieutenant
In general the case marker a tends to be impossible when the object is not affected by the derived nominal: (65) su vision del/*al soldato his view of the/to the soldier
Torrego further observes that there is a difference between derived nominals
Intransitivity in nominalization
and their related verbs. While the affected object of a verb has to be animate, the affected object of the derived nominal is not subject to such a restriction: (66) a.
sus golpes a la mesa their blows at the table b. golpeó (*a) la mesa she hit the table
The absence of an animacy constraint within derived nominals is explained in terms of the structural conditions that yield this constraint. In the domain of verbs, object raising is forced to the outer Spec,vP. In the domains of nominals, however, such a movement is not possible. Again note that the Spanish facts discussed in Torrego, unlike their Greek counterparts are transitive, i.e. they involve the presence of a prenominal agent. According to the interpretation of the facts given in Torrego, Spanish seems to present evidence for Affectedness in a different context from the one in which we encounter such restrictions in English. .. Passive nominals are not result nominals Let us now take a closer look at the aspectual properties of passive nominals. According to Grimshaw (1990), passive nominals as in (43) and (47) never have argument structure, i.e. they do not qualify as complex event nominals in her terms. The evidence Grimshaw presents falls in six categories. First, she points out that nouns that appear to be unambiguously result nominals allow preposing of possessives. (67) a. this problem’s solution b. Jack’s murder
Second, unambiguously theta-marking nouns, such as gerunds, do not allow preposing of possessives: (68) a. *the tree’s falling b. *the city’s destroying
Third, disambiguation of nouns toward the result interpretation leaves the grammaticality untouched. Thus, the plural form of the passive nominal can be used without altering the well formedness of the example. Moreover, passive nominals fail to occur with aspectual adjuncts: (69) the French defeats (70) *the book’s translation for three months
Functional structure in nominals
Fourth, passive nominals permit modifiers like yesterday’s: (71) yesterday’s European defeat
Fifth, passive nominals do not show the same control behavior as active nominals. (72) *the book’s translation in order to make it available to a wider readership
Finally, the passive reading in which the referential adjective seems to be related to an argument disappears when a possessive is included: (73) *Reagan’s liberal defeat
The ungrammaticality of (73) is accounted for if defeat is a result nominal, since if it were an argument taking nominal adding the possessive should have no effect. According to Grimshaw, given that ‘passive’ nominals lack argument structure, Affectedness cannot be a property reflecting a property of argument structure. Neither can it be conceived as being a property of possessive modifiers, as not all possessive modifiers are subject to Affectedness: (74) John’s dog
It would be hard to argue that the possessor in (74) is affected by the possessee. Thus it is only the complement use of the possessive that is subject to an Affectedness requirement. For Grimshaw complements are present in the lexical conceptual structure, and they do not necessarily project as arguments in argument structure. On the other hand, Zubizarreta (1987) argues that affected objects are arguments incorporated into the noun and this is what distinguishes them from other arguments. In other words, the city’s destruction is interpreted as a compound. As such it receives only the result interpretation, and as expected the following example is ungrammatical: (75) *Reagan’s defeat during a whole year was accurately reported
However, there are reasons to cast doubts on treatments that necessarily associate passive nominals with the result interpretation. First, Doron and Rappaport-Hovav (1991) point out that the prenominal NP in result nominals is typically ‘free-thematic’ in its interpretation, but the prenominal NP in passive nominals can only be interpreted as corresponding to the internal argument of the verb. The drugs in (76a) must be internal argument of distribution, whereas under the result nominal analysis in (76b), John can be either the theme or the agent:
Intransitivity in nominalization
(76) a. the drugs’ distribution b. John’s analysis
Precisely in the case where the prenominal NP must be interpreted as internal, the nominals do not display any semantic drift, as shown in (77a), and are noncount as shown in (77b). Nouns lacking an argument structure are typically count nouns, and may undergo semantic drift as in (77c): (77) a. the city’s development b. *the city’s developments c. the city’s developments
(no semantic drift) (out as plural of (a)) (semantic drift to housing complexes, the city interpreted as possessor)
According to Doron and Rappaport-Hovav, these consideration suggest that the prenominal NP in nominals such as (77a) is indeed an argument of the verb, and that passive nominals differ from result nominals in having an a-structure. Recall that in Chapter 2, when discussing the aspectual properties of event nominals, I presented data from Tenny (1994) and Snyder (1998) showing that passive nominals can only be modified by punctual modifiers. Hence, the reason for the ungrammaticality of examples like (75) is related to the type of aspectual modification used. According to these authors, passive nominals denote only the culmination of the event, i.e. they are terminative, they denote events and not processes. In other words, such nominals assert that the goal has been achieved, while there is no such assertion in English active nominals. On this view, it does not come as a surprise that passive nominals are incompatible with durative modifiers. Passive nominals are permitted in a context containing a modifier bringing about the telic interpretation, as shown in (78): (78) a. The book’s publication in a month b. *The book’s publication for a month
(Tenny 1994: 161)
Crucially, examples that force a non-delimited reading, in Tenny’s terms, of the passive nominal are not grammatical, while examples that force a delimited reading are well-formed. Note that the Greek examples discussed in Anagnostopoulou (to appear) differ from the English passive nominals in the following way: unlike their English counterparts they can be modified by durative adverbials (compare (60b) to (78b) above). Thus, Greek ‘passive’ nominals seem to pattern like
Functional structure in nominals
English so-called active nominalizations, which do not exhibit preposing, with respect to their aspectual properties, i.e. they denote processes. In other words, Greek passive nominals do not denote the culmination of the event, but rather they can also refer to duration. Greek nominals that denote the culmination of an event and thus pattern like English passive nominals are nominals derived from achievement predicates. The relevant examples are repeated in (79). As shown in (79c), the nominal ‘assassination’ can be linked with a possessive clitic: (79) a.
i dolofonia tu Athanasiadi se 5 lepta the assassination the Athanasiadis in 5 minutes b. *i dolofonia tu Athanasiadi ja mia ora the assassination the Athanasiadis for one hour c. i dolofonia tu the assassination his
Greek presents us with an interesting puzzle. Although both (79a) and (60b) involve passive nominals, culmination of an event is denoted only in (79), i.e. only in the case of an achievement nominal. This state of affairs seems to support Kolliakou’s (1995) argument that nominals of the type presented in (60) involve process nominals, i.e. nominals denoting non-delimited events. This clearly shows that the ‘passive’ character of the nominal is only indirectly related to Affectedness, if at all. Moreover, this shows that, at least in Greek, affected arguments are not necessarily measuring-out the event in Tenny’s terms leading to a telic interpretation, but can also be included within nominals bearing durative readings. In turn this suggests that affectedness is a property derived from a combination of root, ‘constant’, meaning and functional structure that brings about event properties. Clearly, all the above facts cannot be accounted for if we assume Grimshaw’s analysis. .. On the derivation of passive nominals Having shown that English passive nominals are a special case of event nominals, I now turn to their derivation. Roberts (1987) and Doron and Rappaport-Hovav (1991) among others analyze passive nominals as involving externalization of internal arguments. For Doron and Rappaport-Hovav, this externalization is lexical and can only be done with verbs that include sub-eventualities of the total eventuality described by the verb and its arguments. Their proposal is illustrated below:
Intransitivity in nominalization
(80) destroy (x, y) Cause (DO (x), Become (destroyed (y)))
The passive nominal destruction has an event structure that corresponds only to the sub-constituent become, as shown in (81). (81) a-structure: e-structure:
destruction (y) Become (destroyed (y)))
On this view, the argument of the passive nominal would be projected in Spec,DP directly. The structures in (81) capture the fact that the original external argument is no longer part of the lexical representation of the passive nominals, a fact that, according to the authors, explains the impossibility of an adjunct modifying an external argument and the absence of control into purpose clauses. Merging of the argument with DP is possible in English, since Spec,DP is a theta-position, unlike its counterpart in the other languages under discussion. For Roberts (1987) passivization within DPs actually involves the process E(TH), i.e. externalize theme. Thus, it is a rule that applies to items with particular thematic properties. Whenever the argument does not bear the theme thematic role, it cannot be externalized (see the discussion in the previous section on Affectedness). Hence the ungrammatical examples in (82) are explained on this basis: the class of verbs whose derived nominals cannot form ‘passive in DP’ corresponds to the class of verbs which cannot form middles, i.e. to verbs that do not have themes as internal arguments: (82) a. *the penalty’s fear b. *the library’s gift of the money c. *John’s laughing at
psychological verbs double object verbs verb + preposition
On movement approaches, the genitive moves from its original position to Spec,DP, as outlined in Section 2.1. Within an approach to event nominals that recognizes that their internal structure includes a number of functional layers, as outlined in Chapter 2, one could argue that movement to Spec,DP is preceded by movement to Spec,AspP, as suggested in van Hout and Roeper (1998), identifying each step of the derivation as being linked to semantic/event interpretation. According to van Hout and Roeper, theme arguments in nominalizations move to Spec,AspP for telicity checking. In this case, Aspect must be specified as being [+perfective]. The DP in ‘passive’ nominals could be argued to undergo further movement to Spec,DP, the locus of specificity/definiteness.
Functional structure in nominals
However, there seems to be no conclusive evidence that movement takes place within passive nominals (see also the discussion in Williams 1982, Rothstein 1983). Take for instance the resultative construction. As known, resultative phrases generally appear with transitive verbs. Such phrases may be predicated only of the object of a transitive verb, never of the subject. This restriction predicts that if a verb has no object, then it cannot appear with a resultative phrase. And indeed, as is known, unergative predicates do not take resultative phrases. On the other hand, passive and unaccusative verbs may appear with resultative phrases predicated of their surface subjects, without challenging this restriction: given a movement analysis of passives and unaccusatives, their surface subject is an underlying object. If in passive nominalizations the thematic genitive were an underlying object, we would expect that it could be predicated of a resultative phrase, contrary to fact: (83) a. b. c. d.
the collapse of the building apart *the building’s collapse apart *John’s arrival nude (Rothstein 1983) the arrival of John late (Pustejovsky 1995)
The examples in (83) suggest that passive nominals cannot be derived via a movement analysis, as they are not sensitive to tests diagnosing underlying object-hood. This strongly supports analyses such as Robert’s and Doron and Rappaport-Hovav’s, and suggests that the genitive phrase is directly merged in Spec,DP. Since Spec,DP is a thematic position in English, this merger is licit. Potentially the status of Spec,DP (A vs. A¹) across languages explains the crosslinguistic asymmetries discussed earlier on in the sense that only in English can Spec,DP license arguments. On this view, passive nominals differ from non-passive ones, in that the theme DP is directly merged at Spec,DP and not in the complement domain of the root. In other words, passive nominals solely have external theme-arguments. Given that Spec,DP is not restricted to a single thematic role (see the discussion in Grimshaw 1990 and references therein: also the discussion in Chapter 5), such a direct merging is possible, without the DP necessarily receiving an agent interpretation. To conclude, note that it has been pointed out that there are certain similarities between passive nominals and middles, but also adjectival passives (see Roberts 1987, den Dikken and Sybesma 1998, and Longobardi 1999 among others). I will not discuss the similarities to adjectival passives, as these are taken to be generally stative (though see the appendix to Chapter 5), and
Intransitivity in nominalization
stativity is not a characteristic property of passive nominals (Roberts 1987). I briefly mention some arguments that ‘passive’ nominals have more in common with middle formation than with a verbal passive. Longobardi (1999) points out that a difference between verbal passives and nominal passives is that unaffected objects can passivize in the former, but not in the latter (see also the discussion in Roberts 1987). Moreover, in some languages other than English the preposition introducing the agent in nominals with a possessivized object is not the same as the one expressing the agent in verbal passives (cf. Italian da vs. da parte di, German von vs. durch). The difference might perhaps be related to the different intrinsic semantics of the prepositions: by displays some independent instrumental meaning not shared by da and von. (84) a.
Die Stadt ist von den Barbaren zerstört worden the city is by the barbarians destroyed b. die Zerstörung der Stadt durch die Barbaren the destruction the city-gen through the Barbarians
Another observation that Longobardi makes is that languages have quite distinct morphological forms for passives but not always for passive nouns (but see the Turkish data discussed in the previous chapter); again this is reminiscent of the formation of middles for verbs. . Referential adjectives In the previous section I established that English passive nominals are actually event nominals. Greek passive nominals, which can also be process nominals, differ from the English type passive nominals in that they permit a number of aspectual interpretations. Let me now turn to the role of adjectival modification by group/ethnic/referential adjectives and show how this relates to the result vs. event distinction and to the discussion concerning passive nominals. It has been suggested that adjectives as the ones in (85) appear to behave almost like NPs (cf. Zubizarreta 1985 and Kayne 1984). According to Kayne, these can only encode an unlinked theta-role. Adjectival modification is precluded if the role encoded in the adjectival form is linked, i.e. linked to that of an internal argument (85c): (85) a. the American invasion of Vietnam b. the French defeat of the English c. *the Austro-Hungarian disappearance of the political scene
Functional structure in nominals
The set includes adjectives expressing nationality and other adjectives such as local, national, liberal, and so forth. Grimshaw refers to these adjectives as group adjectives, since they seem to pick out groups with a defining characteristic. Giorgi and Longobardi (1991) label them referential adjectives. According to Grimshaw, group adjectives like possessives are ambiguous. Sometimes they must be analyzed as subject-like and occur with process nominals. This is illustrated in (86), where the inclusion of a group adjective makes the presence of a complement to the noun obligatory. In (86) the use of the adjective constant forces the process reading of assignment: (86) the constant American assignment *(of untrained officials)
Grimshaw’s position with respect to cases like (86) is that group adjectives are like argument-adjuncts and not real arguments of the noun, much like possessives. On the other hand, Kayne (1984) suggests that the external theta-role of nominals can be assigned also to this group of adjectives. When these adjectives are used in the passive, as illustrated in (87), they can be related to a linked theta-role, i.e. that of an internal argument. However, the head in (87) is unambiguously a result nominal, given that it can pluralize, as in (87c) (hence the ill-formedness of *an attack of a foreign country), yet the group adjective is permitted. Thus, in their passive use, group adjectives are modifiers of result nominals. (87) a. the French defeat b. the defeat of France c. the French attacks on
The prediction this makes is that nominals modified by the passive use of group adjective, should not be able to co-occur with a by-phrase licensed by argument structure. The data in (88) illustrate precisely this point: (88) a. *the liberal defeat by Reagan b. *a Central American invasion by the U.S. Army
On Grimshaw’s view then, group adjectives modify result nominals in their passive use, but are a-adjuncts to process nominals in their agentive interpretation. Let us consider the distribution of referential adjectives in Greek in some detail. Such adjectives can modify intransitive object nominals, but it appears that they can also modify ‘transitive’ nominals, as in (89b):
Intransitivity in nominalization
(89) a.
i germaniki apopsi the German point of view b. i amerikaniki kritiki stus hirismus tu Milosevits the American criticism to the acts the Milosevic-gen
However, in Greek the counterpart of (86) is ungrammatical: (90) a. *i sinehis amerikani anathesi ton aksiomatuhon the constant American assignment the officials-gen b. *i germaniki katohi tis Ellados the German occupation the Greece-gen
(89b) differs from the examples in (90) in one important respect: the complement of the noun does not bear genitive case, but rather is introduced by a preposition. If the complement of the noun appeared in genitive, the example would be ungrammatical. Moreover, consider the contrast in (91a–b): (91) a.
i katohi tis Ellados apo tus Germanus ja 3 hronia the occupation the Greece-gen by the Germans for three years b. *i germaniki katohi tis Ellados ja 3 hronia the German occupation the Greece-gen for three years
While the passive nominal in (91a) can be accompanied by an aspectual PP, this is not possible in (91b), where the nominal is modified additionally by a group adjective. This suggests that only (91a) qualifies as a process nominal, and not (91b). It can be shown that referential adjectives do not behave like syntactic subjects across languages. For instance, no implicit argument control is possible in the presence of a group adjective, while this is case when the NP subject is present: (92) a. Cleopatra’s seduction of Caesar in order to become pregnant b. *the Egyptian humiliation of Caesar in order to become the most powerful nation (Markantonatou and Oersnes 1999)
The contrast in (92) further suggests that nominals including referential/group adjectives pattern like result nouns. Moreover, while nominal subjects are able to bind a reflexive in object position, this is not possible with the referential adjective, as is shown by the examples in (93), taken from Markantonatou and Oersnes (1999):
Functional structure in nominals
(93) a. John’s destruction of himself b. ??the American destruction of themselves
Bosque and Picallo (1996) point out that the internal argument of an event nominal in Catalan cannot surface as a thematic adjective. Similar facts hold for Greek: (94) a. *la producción sedera por parte de la Chila the production silky on the part of China b. *i amerikaniki kritiki apo tin Ellada the American criticism by Greece meaning: the criticism of Americans by Greece
On the other hand, (95) is grammatical: (95)
i germaniki ipohorisi the German retreat
However, (95) is not a process nominal. Evidence for this comes from the ungrammatical (96): (96) a. *i germaniki ipohorisi se 1 ora the German retreat in an hour b. *i sinehis germaniki ipohorisi the constant German retreat
Whenever the process reading of the nominal is forced, either by the presence of an aspectual adjective or an aspectual modifier, the example becomes ungrammatical. For Bosque and Picallo, the ungrammaticality of the examples in (94) follows from a violation of the Projection Principle. The nominals production and criticism must have an argument syntactically mapped into a complement position. If thematic adjectives in general are always mapped into a Specifier position, then the ungrammaticality of (94) can be accounted for because the nominal head lacks a complement. Bosque and Picallo (1996) and Picallo (1991) suggest that thematic adjectives can appear only within result nominals, and the discussion here supports their claim. Note that both in Romance, and Greek, result nouns can appear in transitive form, see the discussion in Chapter 4: (97) a.
la proposta soviètica (de desarmament) the Soviet proposal (of disarmament)
Intransitivity in nominalization
b. i italiki protasi ja ton afoplismo the Italian proposal for disarmament
Finally, note that referential adjectives in Greek do not pattern like other adjectives or even other DPs such as the possessor that qualify as a-adjuncts in Grimshaw’s terms. First of all, they cannot be found in predicative position, they cannot appear together with degree modifiers, and they cannot be licensed in the environment of determiner spreading (see e.g. Alexiadou and Wilder 1998, Androutsopoulou 1995, Kolliakou 1995, Stavrou 1995, Alexiadou and Stavrou 1999): (98) a. *i apofasi ine amerikaniki the decision is American b. *i poli amerikaniki apofasi the very American decision c. *i apofasi i amerikani the decision the American
Thus, the sequences referential adjective-noun seem to pattern like syntactic compounds with respect to Ralli and Stavrou’s (1997) criteria. This might be taken to suggest that the structure of the NP that includes a referential adjective modifying a result nominal is actually of the type illustrated in (99):
(99)
N A
N
To sum up, in this section I discussed the environments in which modification by referential adjectives is licit. The Romance and Greek facts point to the conclusion that modification by referential adjectives is licensed by result nominals and not by process nominals. A similar point can be made for the English cases. . No process of passivization There is a further point in the discussion on the formation of passive nominals and nominalization in general which needs to be addressed in this context. Grimshaw (1990) and Borer (1993) argue that nominals such as the destruction of the city by the enemy are derived through a process akin to passivization, i.e. a process that involves demotion of an external argument. Borer specifically argues that a passivized VP is selected by a superordinate N° affix to yield the
Functional structure in nominals
derived nominal. A similar view is defended in Picallo (1991). Moreover, in some analyses of passive, it is argued that the external theta-role is absorbed by the passive morphology (cf. Jaeggli 1986, Baker, Jaeggli and Roberts 1989). van Hout and Rooper (1998) suggest that nominalization affixes might play a similar role, i.e. that of absorbing a theta-role. While this proposal is quite attractive, and could account for the fact that process nominalizations are generally intransitive, as described in Section 1, it will not be pursued here. First, such a proposal fails to capture the fact that unaccusative predicates also give grammatical process nominals. If nominalization generally involved demotion of an external argument, then the formation of unaccusative event nominals should be impossible for the same reason that impersonal passivization does not occur with unaccusative verbs. Second, as Roberts (1987) points out, nominalization forms include various affixes, a fact that makes a treatment similar to the one offered for verbal passives (where the passive morpheme absorbs the subject thematic role) difficult. Moreover, note that it is not clear whether one can assume a syntactic analysis even for verbal passives. For instance, syntactic treatments of verbal passives in general run into a series of problems discussed at length in Embick (1997). To just mention one, it is not clear what the role of passive-like morphology is in deponent verbs, as the one shown in (100) below. It definitely does not absorb an external role, thus it does not block the assignment of accusative case to the direct object: (100) a.
metahirizome to leksiko use-non-active the dictionary ‘I use the dictionary’ b. katastrafike i poli destroyed-non-active the city ‘The city was destroyed’
Thus even if one were to make the strict parallelism between nominal and verbal passive, the above discussion suggests that, in general, presence of special morphology cannot be used as evidence for arguing that demotion takes place. There is also a further asymmetry between nominal and verbal ‘passives’ that seem to suggest that the former belong to a different type of process. Marantz (1997) points out that nominals including in their meaning reference to an agent are derived generally from predicates such as destroy which do not undergo the causative-inchoative alternation. Greek nominals derived from
Intransitivity in nominalization
verbs participating in the causative-inchoative alternation, as briefly discussed in the previous chapter, give either intransitive or ‘passive’ nominal forms. Sometimes the nominalization form has distinct forms for the ‘passive’ and the intransitive: (101) a.
o Janis alakse ta ruha John changed the clothes b. ta ruha /i katastasi alaks-an/e the clothes/the situation changed-3pl/3sg c. ta ruha alahtikan (apo to Jani) the clothes were changed from the John d. to alag-m-a ton ruhon (apo to Jani) the change the clothes-gen (from the John) e. i alag-i tis katastasis (*apo to Jani) the change the situation-gen (from the John)
The interesting observation that can be made here is that the nominals which can license a by phrase also include the infix -m-, while this is not possible for the ones which do not include the infix. But, sometimes a ‘passive’ nominal is possible, although the corresponding alternating verb does not passivize: (102) a. b. c. d. (103) a. b. c.
o Janis halarose tis vides John loosened the screws i vides halarosan the screws loosened *i vides halarothikan the screws were loosened to halaroma tis vidas (apo to Jani) the loosing the screw-gen (from the John) to tzami espase the glass broke *to tzami spastike apo ton aera the glass broke-pass-3sg by the wind ?to spasimo tu tzamiu apo ton aera the break the glass by the wind
In (103c) above the NP introduced by the preposition carries what Fox and Grodzinsky (1998) call the affector role or is causative in meaning. The verb
Functional structure in nominals
‘break’ or the verb ‘loose’ cannot form a verbal passive. However, the noun can appear in the putative passive construction. Formation of passive forms restricted to nominals is not particular to Greek. In Hebrew, as Hazout (1990) signals, there are deverbal nouns in the putative passive construction, although their source verb cannot passivize. For instance, the verb calax ‘crossed’ cannot undergo passivization, but its corresponding noun clixa (crossing) does appear in the nominal passive construction. (104) a.
ha-cava calax ’et ha-te’ala the-army crossed acc the-canal b. *ha-te’ala nicleca ’al yedey ha-cava the-canal was crossed by the-army c. clixat ha-te’ala al’yedey ha-cava crossing the-canal by the-army
Borer (1993) argues that these cases involve impersonal passivization, a process that is generally found in Hebrew. A similar argument could not be made for Greek, as Greek lacks impersonal passivization: (105) *edo horeftike ti nihta here was danced the night
A related pattern is found in Hixkaryana, a Carib language. This language lacks agentive passives, but in process passive nominals the agent is introduced by the relator wya ‘to, by’, which is the marking following the indirect object (and also the causee NP in causative constructions derived from transitive verbs, cf. Koptjevskaja-Tamm 1993: 156f.). (106) Waraka wya honyko wonir xe wehxaha Waraky by peccary 3.poss:shoot:an:pres::poss desire I:am ‘I want Waraka to shoot peccary’
Finally, another type of argument that has been brought up in the literature in order to suggest that nominalization is akin to passivization relates to the presence of implicit arguments in both constructions. This is visible by the existence of pairs like the following (cf. Roeper 1987): (107) a. They opened the door in order to let air in b. The door was opened in order to let air in c. the opening of the door in order to let air in
Intransitivity in nominalization
Event nominals like passives allow control into an infinitival purpose clause (see Chapter 1). Roeper (1987) and van Hout and Roeper (1998) (see also Harley 1995) take this to indicate that an implicit argument is present within process nominals. However, Lasnik (1988) and Williams (1985) argue against this. For instance, Lasnik pointed out that the controller need not be the agent, as is illustrated in (108a). Williams has observed that the purpose clause is legitimate in instances where no agent is conceivably present grammatically, as in (108b): (108) a. The boat was sunk by the torpedo [PRO to prove a point] b. The thermostat is on low [PRO to save money]
The above remarks can be taken to show that no agent needs to be syntactically active in nominals. Engelhardt (1999), in her discussion of Hebrew nominalization, points out that if derived nominals hosted an external argument in the form of PRO, the predicate should have retained its ability to license accusative case, as is the case in infinitival constructions. However, the unavailability of accusative case marking points to the absence of null subjects within nominal constructions. I take such arguments to convincingly show that no implicit argument is present within nominalizations. In other words, nominals in general lack external arguments, implicit or non-implicit, altogether: there is no suppressed external argument in process nominals. I come back to this issue in Section 4. To conclude, the discussion so far established the following: (a) process nominals of the destruction type are intransitive, (b) English ‘passive’ nominals are event nominals and thus have argument structure, and (c) intransitive nominals are not derived by a process of passivization. This is consistent with the assumption that transitivity is a derived property and with the general framework presented in the introduction. In the next section I propose attempt my account for the patterns discussed above.
. The structure of process nominals revisited Given the results of the previous sections the structure proposed in Chapter 2 for process nominals and repeated here in (109) should be re-considered:
Functional structure in nominals
(109)
DP FP (NumbP/AgrP)
D° the AP
FP F° AspectP AspectP¹
Aspect°
vP v
LP
L° Comp (=theme) √DESTROY the city Specifically, the accurate structural representation of process nominals should make reference to the fact that no thematic agent can be projected. As mentioned in the introduction, in the recent literature it has been argued that agents are introduced by a functional head situated immediately above the projection hosting the internal argument of the verb. This head has been labeled v in Chomsky (1995), or Voice in Kratzer (1994a,b). Moreover, the recent literature distinguishes two types of light vs: a transitive light v, and an intransitive one. The former combines with the external argument, the latter does not (110): (110) (i) transitive v [+external argument] v1 = Cause (ii) intransitive v [−external argument] v2 = Become/Happen
Given my discussion here, the situation that emerges with respect to nominalizations is the following. On the one hand, a functional head of the type v needs to be present within certain nominals, otherwise we would have no account for the process/event reading a group of nominals is associated with. On the other hand, the type of v included within such nominals has to be one that does not license an external argument. In other words, it has to be an intransitive light v. This is precisely the proposal I want to make. To account for the fact that process nominals are generally intransitive, I propose that the type of v found in process nominalizations is the one found in unaccusative structures, i.e. the one that does not project an external argument/agent. In other words, only v of type (110ii) is licit within process nominals of the destruction type.
Intransitivity in nominalization
On this proposal the burden of nominal derivation lies on the combinatory properties of functional heads and roots, i.e. it relies on the selectional/ compatibility requirements between v and specific roots. A causative v introduces the causer argument. A non-causative v does not. Both can combine with various semantic classes of roots. For instance, both vs can combine with roots like √DESTROY denoting a change of state, not internally caused and implying an external cause or an agent and with roots like √GROW denoting a change of state that is internally caused. In both cases embedding the roots under transitive v yields a transitive structure. When non-causative v combines with √DESTROY the result is a verbal passive or a nominalization, depending on whether the structure will appear under T or D. In both cases the agent appears in the form of a PP. When non-causative v combines with √GROW the result is an unaccusative verb, in the domain of T, or a nominal in the domain of D. Crucially, the presence of an agent is dictated by both the semantics of the root and the properties of v. Naturally, not all roots will be able to appear in all syntacticosemantic environments with the same result. Within DM illicit combinations, though well formed syntactically, cannot be interpreted Encyclopedically (I will not tackle this issue here, see Embick 2000c and Marantz 1999 for discussion). This view on the derivation of nominals accounts for the fact that specific nominalization types exist with only certain root types in combination with specific functional layers. For instance, unergative predicates never give ‘passive’ nominalizations. It has been argued that such predicates have the internal structure depicted in (111), i.e. they include a type (110i) v, as agentivity constitutes an integral part of their meaning. Unaccusative structures differ from unergative ones in that their v is of type (110ii), i.e. it does not bear agentivity features:
(111)
vP Subject
v¹ v
LP
The root involved in this formation must be of the type that simply denotes an activity, the agentivity component being a property of the functional layer. Within the nominal system, (111) could only be included in the formation of agentive nominalizations of the dancer type (see the discussion in Chapter 4). Note that my analysis crucially assumes that the manner component associated with v is active, i.e. it can license manner adverbs, even though v is
Functional structure in nominals
deficient (contra Hale and Keyser 1993, and Levin and Rappaport 1995). I take v to be deficient only with respect to its role of introducing an external argument. It is not deficient with respect to its possibility of licensing manner modification. Recall that in Chapter 1, I pointed out that there is a distinction between the transitivity content of v, and its semantic content. On the basis of this distinction, I propose that the property responsible for the licensing of manner modification is to be located in the semantic content of little v, and not its transitivity property. That is, eventive predicates only license manner (see Ernst 1998, Katz 1999b). The proposal that light v in (109) is an unaccusative light v leads to an answer to the question why process nominals cannot assign accusative case to their DP arguments. As has been already mentioned, light v performs the two requirements of Burzio’s generalization: it introduces the external argument and licenses accusative case. Given that in (109) no external argument is projected, no accusative case can be assigned within nominals. Assuming that only v = Become/Happen is present within nominalizations explains the restrictions on nominalizations with the Genitive construction in Greek. If one adopts the proposal that the light v introducing the Goal argument is actually identical with the v introducing the agent (Marantz 1993), then the ungrammaticality of nominalizations with Goal arguments is accounted for. What can appear in the genitive in nominalizations are theme arguments only and v is structurally absent. Marantz, building on Larson (1988), argues that double object constructions involve ‘complex’ predicate structures in which the verb takes as its complement a ‘predicate phrase’, in this case a VP. In such a ‘stacked’ VP structure, the object affected by the predicate is projected as the specifier of the lower VP, while the goal argument is projected as the specifier of a higher, ‘applicative’ or light v head (cf. Anagnostopoulou to appear, McGinnis 1998):
(112)
vP goal
v¹ v
VP
theme V¹ V
X
If what I have said so far is along the right lines, then the fact that goal
Intransitivity in nominalization
nominalizations are ungrammatical is accounted for under the analysis suggested here, as nominalization involves a defective v, i.e. one that does not project a specifier (see the discussion in Chapter 4). A final remark on word order within the DP is in order to conclude the discussion in this section. In the previous chapter, while discussing noun morphology, I suggested that √ must reach F° so that it is in the complement domain of D and can be Spelled-out as a noun. This would permit the DPtheme to move to AspectP for reasons that have to do with Case-checking. In principle there is no problem for such an analysis (and see also van Hout and Roeper 1998). The only problem that this analysis might face is the location of adverbial modifiers that follow the DPtheme. Assuming that manner adverbs are located in Spec,vP, and aspectual adverbs are located in Spec,Asp, we would be forced to propose that the DPtheme targets the external specifier of Aspect, adverbs not blocking the movement of a DP argument. According to this view, adverbs can occupy specifier positions, but the system as such does not preclude an extra specifier to locate the arguments (see Laenzlinger 1998).
. By-phrases and more on the lack of external arguments An issue that has not yet been touched upon concerns the role of the by-phrase within derived nominals. In this chapter I argued that external arguments of any type are absent from process nominals. Recall from that discussion that in a number of languages, agents within nominals take the form of PPs or adverbial phrases. I repeat some of the relevant data below: (113) the destruction of the city by the barbarians (114) i katastrofi tu dasus apo tus polites the destruction the forest-gen by the citizens
Pesetsky (1995: 312) observes that in general, the by-phrase is possible whenever a verb with an external argument undergoes morphological change that prevents normal case assignment to the object. This includes passivization, but also nominalization and adjectivization of all relevant sorts (-able). However, apart from the discrepancies between verbal passives and ‘passive’ nominals discussed in this chapter, it has been noted in the literature that there are some differences between the by-phrase encountered with verbal passives and the by-phrase found with derived nominals. A first difference concerns the status of the by-phrase within nominal
Functional structure in nominals
and verbal passives (originally observed in Hornstein 1977). Building on this observation, Rappaport (1983) states the by-phrase in verbal-passives can bear any thematic-role. By-phrases found in nominals differ from by-phrases found in verbal passives in that they are thematically limited. This point is extensively discussed in Roberts (1987), Grimshaw (1990) and recently in Fox and Grodzinsky (1998) among others. The data below illustrate the acclaimed difference (from Fox and Grodzinsky, henceforth F&G 1998: 325). F&G point out that by-phrases in derived nominals can only be interpreted as agents, instruments or creators. In other words, the object of by must be an affector. This is not the case with all by-phrases found with verbal passives, which can bear any semantic role. Thus, although (115a) is grammatical, (116b) is not. In German, the PP introduced by durch is used instead: (115) a. the imprisonment of refugees by the government b. the destruction of the city by lightning c. a book by Chomsky (116) a. the package was received by John b. *the receipt of the package by John c. die Entgegennahme des Pakets durch Hans/*von Hans the receipt the package through Hans/by Hans
Following Jaeggli (1986), Roberts (1987), Baker, Jaeggli and Roberts (1989), F&G argue that in verbal passives the thematic role of the subject is transmitted to the by-phrase. This operation involves the transmission of a theta-role that is never completely unsuppressed, and hence is unlimited. On the other hand, theta-transmission cannot take place in nominals. If this were the case, the above contrast would remain unaccounted for. For F&G, by has certain semantic properties that allow it to assign an affector thematic-role to its complement. In this case the by-phrase can be interpreted independently of the availability of theta-transmission. If this is correct, it supports the view put forth here that there is no external argument present in nominalizations. It further signals that in general all nominals, whether they appear to be passive or not have an intransitive base. The presence of a by-phrase is licit, when this is compatible with the semantics of the root (see Marantz 1997). This is so for destruction and capture type nouns crosslinguistically. A similar conclusion is also drawn in Bottari (1992) for derived nominals in Italian and den Dikken and Sybesma (1998) for English. However, note that in English by-phrases do not only refer to agents, but
Intransitivity in nominalization
can refer to the author or creator within result nominals. There are languages, e.g. Spanish, Hebrew, Russian where the counterpart of the by-phrase is only found with verbal passives and process nominals and not with other nominals. In other words, these languages formally distinguish between authorship by and passive by. This is illustrated below for Spanish: (117) a.
La captura de los fugitivos por les soldados the capture of the fugitives by the soldiers b. *un libro por Chomsky a book by Chomsky c. un libro de Chomsky
In Spanish, both the derived nominal and the verbal passive contain the form por. However, authorship in (117c) is denoted by means of the preposition de, while English uses by in such contexts as well. Similarly, in Russian both in passives and in derived nominals the agent theta-role is expressed via an instrumental phrase. In Hebrew, the preposition ’al-yedey introduces the external argument only with ‘passive’ nominals and verbal passives: (118) a.
ha-ir Sukma ’al yedey ha-Siltonot the city was rehabilitated by the authorities b. *tmuma al yedey David picture by David c. tmuma me’et David picture by David
Italian is also similar: it distinguishes between markers for agents da parte di and for authorship di. Greek also uses the same agent marker for verbal passives and ‘passive’ nominals, while it uses the genitive case to denote the author or the creator: (119) a.
i katastrofi tu dasus apo tus polites the destruction the forest-gen by the citizens b. to vivlio tu Chomsky the book the Chomsky-gen
It is precisely this set of facts that led Grimshaw to conclude that suppression of external arguments takes place in passives, and also in process nominals. In order to account for the difference in the distribution of by-phrase in nominals and verbal passives, Grimshaw suggests that the reason why the by
Functional structure in nominals
phrase in nominals is uniquely linked with the role of agent, and not with that of any external argument is because nominals do not have suppressed external arguments, although they have suppressed agents. F&G, on the other hand, distinguish between languages that only license by-phrases via theta-transmission (Spanish) from those that only do so in the passive (English). Thus, in F&G’s terms apart from English, in all other languages discussed here thetatransmission would be involved in nominals as well as in verbal passives. However, I presented arguments against the view that such a process is possible in nominalizations. Furthermore, it would be mysterious why only English were not to allow theta-transmission, if it is such a general process. Let us then assume that in all the languages discussed in this section no theta-transmission takes place. Prepositional phrases introduced by ‘by’ or da parte di or por could be considered as being similar to ergative or instrumental markers, a fact which Williams attempts to capture in terms of (120) (see Williams 1987: 365): (120) The agent is assigned to a by phrase if there is an internal theme
Nominals are a clear illustration of (120): theme arguments bear genitive and agents must be marked differently, as it is not possible to have two de- or genitive phrases within a process nominal. This, in correlation with the fact that nominalizations lack external arguments, can be taken to suggest that in nominalizations both arguments are internal. This view echoes Williams’ (1981) claim that nominalization involves internalization of an external argument. It is perhaps for this reason that the two arguments bear distinct markers: since they are both internal they are distinguished by means of special marking. In fact one could suggest that ergative marking is the result of the following principle (see Marantz 1991): (121) No ergative-case on a non-thematic subject
In other words, the by-phrase is internal to the verbal root and is thematically licensed by the preposition by or apo or por etc. Note that (120) or (121) seem to follow from a more general restriction. Given the results of the discussion here, nominalizations fall under the generalization established in Alexiadou and Anagnostopoulou (to appear). The authors, after examining a wide range of data, point out that whenever a sentence contains both a subject and a direct object, one of the arguments must vacate the VP by Spell-Out. They propose that argument externalization is related to Case. It is due to the fact that movement of both arguments to a single head T°
Intransitivity in nominalization
which contains two active Case features in the covert component is banned. This restriction is not obeyed when the second argument is a PP. In fact they note in passing that nominalizations provide a further environment where this generalization is attested. The bulk of facts examined here clearly point to the validity of the claim: given that in nominalizations both arguments are internal, and there is only one functional head checking structural case, one of them must surface as a PP. On this view, the by-phrase behaves like an independent phrase that creates a specific semantic relation depending on the presence of an affected internal theme, i.e. it is marked semantically (and morphologically) for its function. This seems to be close to the notion of passive agent in Croft (1991). In this respect, the by-phrase is strictly parallel to ergative case, if this is analyzed as a prepositional/lexical case (see the discussion in Chapter 5). In fact according to Nash (1995) only nominative-accusative languages have external arguments, ergative ones have only VP internal agents. Structurally, the by-phrase can be represented either as a specifier or as complement of LP:
(122)
LP DP
(123) L¹
LP PP
L¹
L PP L DP Note that (122) above is very close to the proposal in Marantz (1984) that ergative languages differ from nominative-accusative languages in that in the former the projection of the arguments at D-structure is the reverse. Also it is similar to Pesetsky’s (1995) representation of the ‘low’ cause argument, and in agreement with the intuition that agents in nominalizations derived from predicates that do not passivize behave like causers/instruments (see below). In any case, both structures bring ‘passive’ nominals close to inalienable possession/partitive/theme-instrumental structures (see Marantz 1993) and express the intuition that the affected theme and the affector are part of the same sub-event. These structures reflect a binary relation between the two arguments, as opposed to structures including agents introduced by light v. In this sense, the affecting of an instrumental or an affector agent is taken to occur within the same event that affects the theme. This restriction on the type of agent phrases in nominalizations enables us to understand why we find process nominals accompanied by by-phrases, although their corresponding verbal passives do not exist, as discussed in Chapter 2 and here. This is so, because the
Functional structure in nominals
agent in the nominalization structure has a different status than the agent in the verbal passive. If this view is correct note that one can account for the fact that by-phrases do not co-occur with ‘transitive’ agents as in *John’s destruction of the city by the barbarians by assuming, following Pesetsky’s analysis of experiencer predicates, that the low cause moves to the higher cause position, i.e. the two positions (complement of L and Spec,DP) are transformationally related (see Anagnostopoulou to appear). I come back to this point in the next chapter. Finally, let me come back to the genitive of authorship. The view adopted in this work concerning the status of this genitive is the one put forth in Grimshaw. For Grimshaw, nouns such as book do not have an argument structure, as they refer to objects and not to processes. In this case, the genitivephrase would be interpreted as a possessor, i.e. it would be a non-thematic subject. The ergative generalization would not be activated and these phrases would not receive an ergative marking. Now in English it seems to be the case that by can have both an ergative and a non-ergative use, while most of the other languages distinguish clearly between the two contexts.
Notes . Picallo (1991) has claimed that process nominals are passive in Catalan; see also Bottari (1992) for Italian and Borer (1993) for Hebrew. . Note that under Grimshaw’s (1990) and Borer’s (1993) analysis, nominalization always affects external arguments. Thus, it is unexpected that unaccusative predicates, which lack external arguments, also feed process nominalizations. . In earlier stages of Greek it was possible to express the agent via a genitive, so that two genitives were included within a derived nominal. . Note that the following German examples are grammatical (Gisbert Fanselow personal communication): (i)
a.
b.
dieser Vorschlag einer Haushaltskürzung der Grünen this proposal a budget-cut-gen the Greens-gen ‘The Green Party’s proposal of a budget cut’ dieser Versuch eines Rückritts des Umweltministers this attempt a resignation-gen the environment minister-gen ‘The environment ministers’ attempt of a resignation’
However, the nominals in (i) qualify as result nominals on the basis of Grimshaw’s criteria. Both are preceded by demonstratives which can only precede result nominals:
Intransitivity in nominalization
(ii) they observed the/*that assignment of the problem (Grimshaw 1990: 54) As will be discussed in Chapter 4 the possibility of having two genitives inside a nominal is also licit with result nominals in Greek. . In Spanish, as in Greek (see Chapter 2), certain nominalizations suffixes are sensitive to the type of verb (see also the discussion in the introduction). Bever and Sanz (1997) point out that in Spanish the affixes -íon, -ada specify that the argument of the nominalization is the patient of the verb, while affixes like -ador specify the agent of an action. When applying these suffixes to monadic verbs, agent suffixes attach to unergative verbs, while patient ones to unaccusatives: (i)
a. b.
(ii) a. b.
Mario habla mucho Mario talks a lot Mario lloro mucho Mario cried a lot
→ es muy hablador he is a great talker → *la llorada de Mario fue larga the crying of Mario was long
Maro llega siempre muy temprano → *es buen llegador Mario always arrives very early he is a good arriver Mario llego a tiempo → la llegada de Mario fue muy esperada Mario arrived on time Mario’s arrival was very much awaited
. In fact the argument concerning negation might turn out to be more complicated than Bottari presents it. Kaiser (1968) has pointed out that negated nominals are not acceptable in all contexts. For example, the sentence in (ia) is seems to be ill-formed: (i)
a. b.
*the non-arrival of the train lasted an hour the arrival of the train lasted an hour
Baeuerle (1987) has suggested that the contrast in (ia-b) is evidence that events are not a subset of propositions. The reason one should be driven to this conclusion is the following. Suppose that the semantic function of the negation operator is that of applying a set of possible worlds to yield the complements of these sets. In this case, the NP the non-arrival of the train will denote a proposition (a set of worlds in which the train didn’t arrive). Suppose that the predicates lasted an hour, as Vendler (1967) proposed for independent reasons, semantically select for events but not for propositional entities. In this case, the contrast in (i) is expected: these predicates should not be able to combine with the NP the non-arrival of the train, since the presence of negation forces the propositional reading of the NP, but they should be able to combine with the NP the arrival of the train, since this NP may denote an event. Now, if eventualities were a subset of propositions, then the arrival of the train, like the negated NP the non-arrival of the train will denote a proposition. But then, the contrast in the examples above cannot be accounted for. A similar view is defended in Zucchi (1993). However, it is by far not settled whether propositions belong to the set of eventualities or not. For instance, for Pustejovksy (1995) what contributes to propositional readings is the presence of an operator that quantifies over events. I leave the issue here open. . This has led Schoorlemmer (1995) to concluded that Aspect phrase is present within a verbal nominal but not within a process nominal. If this is taken to be an extra criterion
Functional structure in nominals
for the presence of an Aspect phrase, note that modification by such nominals is possible within a Greek process nominal. See also the discussion in the next chapter. (i)
i katastrofi tis polis ta mesanihta the destruction the city-gen the midnight-acc
. Recall that result nominals, on the other hand, have been argued at least by Picallo (1991) to be transitive. In Catalan, predicates that select results happily co-occur with two genitives. This is never possible with passive nominals. The presence of such patterns could be taken to argue that the distinction between event vs. result nominals maps into the distinction passive vs. active nominals, in the sense that result nominals include transitive agents, i.e. possessors while event ones do not (cf. Picallo 1991). I will come back to this issue in Chapter 4. . It has been argued by Shin (1998), that formation of -ung nominals in German is sensitive to the Affectedness property. In other words, the theme arguments of such nominals must be changed quantitatively or qualitatively: (i)
a. b.
*die Verkorkung der Flasche the corking up of the bottle die Verkorkung der Pflanzengewebe the corking up of the plant tissues
In (ia) the theme argument, the bottle, does not undergo the change in state of affairs denoted by the predicate, it has only acquired a new property, namely to be closed with a cork. On the other hand, in (ib) the theme argument has into changed into a cork after the termination of the process denoted by the head noun. Ehrich (1991) also points out that -ung nominalizations in German are compatible with achievement nominals only. See the discussion in the previous chapter . However, the situation was different in Ancient Greek, where constructions such as i tis poleos katastrofi were acceptable. This suggests that there was an Affectedness constraint active in earlier periods of the language. . Gisbert Fanselow (personal communication) observes that one reason for the oddity of (60a) might be that looking at the sea does not denote an action. Hence, the object of the predicate cannot be affected given that no action is involved which could result in changing something in the object. . The Spanish facts presented here will be seen under a different light in Chapter 4. It will be argued that these nominals are actually result nominals, i.e. the marker à will be analyzed as an oblique marker. Thus, the fact that the genitive is not possible within the types of nominals discussed in Torrego will be accounted for by showing that there is no environment for a structural case to be assigned within such nominals. Perhaps this is the reason why these nominals do not conform with the animacy requirement. . However, Roberts (1987: 249) gives a set of examples where such strings are grammatical: (i) the city’s destruction in order to prove a point
Intransitivity in nominalization
According to Graham Katz (personal communication), the text’s example does not create a strong ungrammaticality effect. Moreover, he points out that in general there seems to be a preference for NPs associated with relational nouns to appear in the pre-nominal position. Tenny (1987) points out that certain passive nominals seem to be able to co-occur with adjunct modifiers such as completely: (ii) the door’s opening completely . Interestingly predicates that are characterized by the separation property tend to be achievement predicates. . Many thanks to Alec Marantz for bringing these data to my attention. . Elena Anagnostopoulou (personal communication) suggests that the reason why (91b) is ungrammatical in Greek is related to the fact that the counterpart of John’s destruction of the city is also ungrammatical, i.e. to the fact that there is a thematic position in the English DP which is not active in Greek. Note, however, that there are a number of differences between the adjectival and the genitive construction even in English. Moreover, the position in which the adjective is merged is not identical with the position in which the genitive is merged, the latter being projected into Spec,DP, as will be discussed later on in the text. . A similar pattern exists in Italian. Bottari (1992: 72) notes that in Italian not all nouns yield felicitous results when modified by an adjective. Moreover, there is a contrast between nominals ending in -zione and those belonging to the -ura class. The latter give even worse results: (i)
a. b.
??la
restitiuzione giapponese dei prigioneri americani the restitution Japanese of the American prisoners *la lettura tedesca della nuova ordinanza ebbe luogo alle cinque the reading German of the new decree took place at five.
Bottari interprets these facts as suggesting that the external argument cannot be realized in Italian nominals in the form of a group adjective, i.e. these are passive in character. Note that -ura nominals in Italian also resist adjectival possessives, in fact they hardly allow the realization of an external argument. . Alexiadou and Stavrou (1998b, 1999) actually claim that there are no genuinely thematic adjectives as such, in the same sense as there are thematic DPs. The explanation given there is as follows: so-called ethnic/group nouns give rise to derived adjectives through affixation of the suffix -ik-. This suffix turns a noun (for example, Italos ‘Italian’) into the corresponding adjective Ital-ik. It absorbs the thematic role and at the same time allows the derived category to agree with the noun in all the relevant phi-features through the adjectival endings that can now be added to the -ik- infix. Hence, such adjectives bear only a lexical (etymological) relation to their original nominal parent, and this is what gives the impression of their being assigned a thematic role. In other words, modification by a referential adjective of this type provides us with a further test that distinguishes
Functional structure in nominals
between result and event nominals: only the former can be modified by these adjectives. . Here the Turkish facts presented in the previous chapter could be relevant. One could argue that in Turkish the presence of passive morphology is sufficient to absorb and transmit the theta-role to the adverbial subject. . According to Embick (1997), special morphology is used when v is not associated with an external argument. . The Carib pattern is reminiscent of Romance causative formation. Bok-Bennema (1991: 18) notes that in many accusative languages infinitival complements of causative and perception verbs are characterized by an ergative case system. The French examples in (i) illustrate this: (i)
a. b.
Marie le fera chanter/tomber Mary him will make sing/fall Marie le fera acheter à Paul Mary it will make buy to Paul ‘Mary will make Paul buy it’
In (i-b) we see that the object of acheter has the same case as the subject of the intransitive verbs chanter/tomber in (i-a), while the subject of the transitive bears dative case. For the analysis to be presented in the next chapter, this means that both the French causative construction and the Carib pattern involve deep agents. Interestingly, the à phrase in French is also used to mark the indirect object of a di-transitive verb and to express possession as in la voiture à Jean ‘the car to John’ (see Kayne 1994 for a recent discussion). . Schoorlemmer (1998) makes a similar point to the one made here on the basis of Russian facts. . Thanks to Elena Anagnostopoulou for raising this issue. . Picallo (1991) also proposes that syntactic nominals are another instance of Burzio’s generalization: they cannot assign structural case to their internal argument, thus no thematic role can be assigned to the subject position. The internal argument does not receive inherent case as the head nominal itself is a product of a syntactic derivation. The external argument is absorbed by the nominalization affix (see also Roberts 1987 for verbal passives). . A similar view with respect to the status of the of phrase has been advanced in Williams (1987). . Fukui and Speas (1986) suggest that the by-phrase is simply the VP-internal subject position. . Interestingly, this structural analysis of the agent within nominalizations and ergative languages is in agreement with proposals about the development of ergative marking, such as the one put forth in Garrett (1990). Garrett (1990: 264f) states that ergative and instrumental case marking are identical in many languages. In fact several typologists have speculated that ergatives have actually emerged from old instrumentals. According
Intransitivity in nominalization
to Garrett, instrumental NPs were reanalyzed as subjects. Such reanalysis is permitted because of the functional overlap between instrument and agents in clauses with transitive predicates, e.g. John opened the door with the key and the key opened the door. Similar reanalysis does not occur with intransitive predicates, as thematic instruments are rare or absent altogether in the subject position of intransitive clauses, e.g. John walks with a cane, but *a cane walks. Note here that the fact that German uses the preposition durch or that Russian denotes agents with instrumental case in passive nominalizations is supporting an analysis of this preposition as being an instrumental marker. . Thanks to Elena Anagnostopoulou for pointing this out to me. Safir (1987) argues that the fact that *John’s destruction of the city by the barbarians is ungrammatical suggests that the possessive and the by-phrase are arguments of the head noun. This specific claim is refuted in Grimshaw (1990: Chapter 4). . In order to determine which of the structures corresponds exactly to the underlying structure of nominalizations, note that there is a binding asymmetry between the two examples in (i): (i)
a. b.
the destruction of himself by John *the destruction of John by himself
To account for this asymmetry, several linguists have suggested that (ia) is acceptable because the anaphor himself is c-commanded by John, but does not bind John. (ib) is ungrammatical, because the anaphor himself is not bound by its potential antecedent John, but it c-commands it. Rouveret (1994) suggests that the contrast in (i) suggests that the agent within nominals can be projected in two ways: either in Spec, NP, or when the specifier of NP is not projected, as a sister to N¹. This is depicted in (ii): (ii) a. b.
[NP [Spec PPagent] [N’ PPtheme]] [NP [N¹ N PPtheme ] PPagent]]
In (iib) both the arguments of the noun are internal to N’, the theme argument is a sister to N° and the agent a sister to N¹. This line of argumentation seems somehow to support the view that the agent is located under the same intermediate projection as the theme.
Chapter 4
Variation in Functional Structure
In the previous chapters, I argued that (Greek) process nominals of the type destruction include the functional heads Aspect and v in addition to nominal functional heads. As is well known, however, not all nominalizations share the same properties. For instance, as discussed in Chapter 1, gerunds in English have distinct characteristics from derived nominals. Moreover, nominalizations are not uniform across languages: while adverbs can appear within process nominals in Greek, this is not possible with process nominals in English. In this chapter, I show how this lack of uniformity has its origin in the variation in the number and the type of (verbal and nominal) functional projections included within DPs. The reasoning is as follows. The properties of constructions that generally exhibit a mixed verbal/nominal character are the result of an association of an unspecified root with nominal functional categories in addition to the verbal ones (see also Borsley and Kornfilt 2000). But in principle the number of these categories could vary. It could be minimal or the whole set of projections encountered in other clause types. Moreover, the type of these projections, i.e. A vs. A¹, transitive vs. intransitive, is also subject to variation. Thus first, one should not expect nominalizations to be uniform across languages or nominal constructions in general to behave alike. Second, different nominalization affixes are expected to attach to different set of projections. The chapter is organized as follows. In Section 1 I discuss variation linked with the number and type of verbal functional projections. In Section 2 I turn to variation related to the number and type of nominal projections. Finally, in the appendix I demonstrate how the system proposed here accounts for various types of nominal infinitives.
.
Variation in the set of verbal projections
. Presence vs. absence of C In Chapter 2 I put forward the proposal that deverbal formation involves the
Functional structure in nominals
combination of AspectP with a nominal head. However, in principle D can also combine with higher verbal projections, giving rise to nominalized clauses across languages. In fact it seems to be the case that AspectP combines with both D and T, but only when a C level is present. The result of this combination is nominalized clauses; these have distinct properties from derived nominals; notably they include predicates that inflect for tense and agreement. (1) and (2) illustrate this type of nominalization for Greek and Turkish (Greek example from Roussou 1991, Turkish from Borsley and Kornfilt 2000): (1) to oti irthe ‘the that he came’ (2) Ben [siz tatil-e çik-acag-iniz-i I you-gen vacation-dat go-out-fact.fut-2pl-acc duy-du-m hear-past-1sg ‘I heard that you will leave for vacation’
In the next section I turn to some of the nominalization types discussed in Chapters 1, 2 and 3, namely -er nominals and gerunds, and show how the differences between these and nominals of the destruction type can be accounted for in terms of the presence vs. absence of an AspectP. . Presence vs. absence of Aspect .. -er nominals vs. non -er nominals (in Greek) As has been noted in the literature, there are certain parallels between eventive -er and process nominals of the destruction type (see Rappaport and Levin (R&L) 1992 for extensive discussion). For instance, eventive -er nominals, much like process nominals, can take complements (cf. (3) and (4)): (3) the defenders *(of human rights) (4) a. a devourer of fresh fruit/*a (good) devourer b. she devours fresh fruit/*she devours
For R&L (1992) inheritance of complement structure is the crucial factor for the division of -er nominals in two major subclasses — eventive and noneventive. The distinction between agentive and instrumental -er nominals is secondary and relies on this first major distinction. Moreover, eventive -er nominals can be modified by adjectives such as
Variation in functional structure
frequent. This type of modification is possible only in the presence of the internal argument (much like the examples we saw in Chapter 1 with nominals of the destruction type): (5) a. constant defenders *(of the government’s policies) b. frequent consumer *(of tobacco)
When the nominal is non-eventive, i.e. when no complement is required, then the aspectual modifier is not allowed: (6) a. this machine continues to be our only (*frequent) transmitter b. this machine continues to be our only transmitter (*to headquarters)
On the basis of these facts, R&L conclude that both eventive deverbals and -er nominals include an event position (along the lines of Larson 1983, Higginbotham 1985, Rothstein 1983, Larson 1998). At first sight, these similarities could lead one to suggest that -er nominals share the same structure with destruction type nominals. However, there are significant differences between -er nominals and the type of nominals discussed so far, reviewed recently in Alexiadou, Anagnostopoulou and Stavrou (2000) (see also R&L 1992). First, unlike other process nominals, -er nominals do not permit manner adverbial modification. This is illustrated in (7) with an example from Greek, a language where adverbial modification is licit with destruction type nominals, as has been extensively discussed in Chapter 2: (7) *o katharistis tu ktiriu prosekitka the cleaner of the building carefully
Second, as the examples in (8) show, -er nominals cannot bear aspectual modification. This means that -er nominals, contrary to destruction type nominals, are not interpreted aspectually, as they cannot be modified by any type of aspectual adjunct, perfective or imperfective: (8) a. *i damastes ton fotonion mesa se/gia enan eona . . . the tamers the-gen photons-gen within/for a century b. *o katharistis tu ktiriu epi ena mina telika apolithike the cleaner the-gen building for a month finally got fired
These contrasts suggest that -er nominals have a diminished verbal character, even in their agentive reading.
Functional structure in nominals
This property of -er nominals will be captured in term of variation in their functional architecture. What type of projections can be argued to be included within such nominals? Given that these are agentive, they must include v (see Marantz 1999, van Hout and Roeper 1998 among others). But the type of v they contain is different from the one found with other types of process nominals: while v in destruction lacks agentivity features (see Chapter 3), v in dancer contains agentivity features. Moreover, given that -er nominals do not permit aspectual modification and do not show the aspectual distinctions found with other process nominals, they must lack AspectP. As a result of their functional structure, -er nominals never denote specified events, but the agent of an event. If correct, this suggests that both Aspect and v contribute to the licensing of arguments. As a result, both -er nominals and process nominals of the destruction type can license internal arguments, in spite of the differences between them with respect to their functional architecture. The question that arises next is the following: if v is present, why isn’t the licensing of manner adverbs possible, since, as I argued in Chapter 2, this is linked to v? Note that -er nominals do have a manner component. This is clearly present on one reading of (9), which denotes that Olga dances beautifully (Larson 1998, 1999 for discussion). (9) Olga is a beautiful dancer
Alexiadou, Anagnostopoulou and Stavrou (2000) give a partial answer to this question. These authors propose that there is a link between the morphology of adverbial (manner) modification and the presence of AspectP. Only when AspectP is present is adverbial manner modification licit. Hence in both types of nominals a root √BEAUTIFUL can be included in the structure. In the environment of functional layers that include Aspect, such as dancing, it will be spelled out as an adverb. In the case of dancer, where AspectP is not included, it will be spelled-out as an adjective. Consider now how this proposal relates to the semantics of -er nominals. -Er nominals denote individuals (or classes of individuals) and more specifically the agent of an event, while process nominals denote events and the agent, when present in the form of a PP, is a participant in the event. Evidently, AspectP is relevant for events and not for individuals (see also Davis 2000). In the first case, the event is a property of an individual, while in the second case the individual is a participant in the event. The next question is what blocks the presence of adverbial aspectual modification, but legitimates the presence of adjectival aspectual modification.
Variation in functional structure
Clearly, the absence of aspectual adverbials is related to the absence of AspectP. The presence of aspectual adjectives, on the other hand, is potentially related to NumberP. If one considers Number and Aspect to be categories of the same type (see de Wit 1997, Alexiadou and Stavrou 1998a,b, Alexiadou 1999b and references therein), aspectual adjectives are licensed by Number. .. Polish vs. Russian nominals Support for the suggested link between the presence of adverbial manner modification and Aspect comes from the following asymmetry between Polish verbal nouns and Russian derived nominals. As discussed in Chapter 2, Polish verbal nouns exhibit true aspectual oppositions in the sense of perfective vs. imperfective. The relevant data are repeated below. (10) a.
ocenienie studentów przez nauczyciełi nast˛apiło szybko valuation-pf the students-gen by teachers occurred quickly b. ocenianie studentów przez nauczyciełi ciagn˛eło si˛e evaluation-imp students-gen by teachers lasted refl przez cały tydzien´ through the whole week
This is not the case with derived nominals in Russian, as is illustrated in (11). Although the Russian verb destroy has both a perfective and an imperfective form, the nominal destruction exhibits no such opposition. In fact Schoorlemmer (1995) provides a detailed argumentation for the lack of Aspect from the structure of Russian nominals: (11) razrušit-razsušat destroy-perf/destroy-imp
razrušenie destruction
As reported by Vinogradov and Švedova (1964), diachronically -nie nominals in Russian are by-products of passive verbal formation. Moreover, they have long been considered part of the verbal paradigm. Nie was added to the passive participle in an unrestricted way. When the passive form was non-existent a dummy passive formative was added to the verbal stem in order to keep the nominalizing pattern consistent. In fact in nineteenth-century Russian -nie nominalizations were formed from both members of the aspectual verbal pair, perfective and imperfective verbal bases, as illustrated in (12) below: (12) a.
vypisanie writing-out perf
Functional structure in nominals
b. vypisyvanie writing-out imperf
Vinogradov and Švedova mention a gradual diminishing of verbal characteristics of -nie nominals that started with the loss of the aspectual distinction. Interestingly, Russian nominals do not permit manner modification, while Polish ones do so (data from Schoorlemmer 1995): (13) a.
uzucie noza spryntie use knife-gen cleverly b. *upotreblenie noža xitro use knife-gen cleverly
Polish Russian
Again in (13) a correlation between the presence of AspectP and the presence of manner adverbial modification is observed. Finally, recall that Polish has two types of nominalizations: verbal nominals and derived nominals (Chapter 3, Section 1.4.2). The latter show no aspectual opposition and do not license manner adverbial modification. This pattern provides further evidence for the link between AspectP and manner adverbial modification. .. Gerunds vs. derived nominals in English A similar contrast is found if one compares English gerunds to English derived nominals. As pointed out in Chapter 2, (14a) differs from (14c) in that -ing nominalizations entail an imperfective event. However, this is not the only difference between the two types of nominals. As Lees (1960) observed, gerunds permit auxiliaries and adverbs, but disallow articles and prenominal adjectives. On the other hand, derived nominals (and mixed nominalizations) do not license auxiliaries or adverbs, but allow for prenominal adjectives and articles. The examples in (14) and (15) illustrate these differences (Zucchi 1993: 21): (14) a. b. c. d.
the performance of the song the beautiful performance of the song *the having performance of the song *the performance of the song beautifully
(15) a. b. c. d.
*the performing the song *his beautiful performing the song his having performed the song his having performed the song beautifully
Variation in functional structure
This contrast may also be attributed to the presence vs. absence of AspectP. If Aspect is considered to be absent from English derived nominals of the destruction type, one could attribute their ‘aspectual’ variable behavior to this structural property. They are aspectually unmarked, and hence can take both perfective and imperfective readings. .. Nominals from di-transitive predicates in Greek Consider finally nominalizations from di-transitive predicates in Greek. As Anagnostopoulou (to appear) points out these nominals appear to have deficient verbal structure in that they often seem to resist aspectual adverbial modification and that they do not easily co-occur with manner adverbials. Starting with aspectual adverbial modification, the following examples seem to be deviant or ungrammatical: (16) a.
To harisma enos vivliu s-tin Maria ??se ena lepto/ the gift a book-gen to-the Mary in one minute/ *epi deka lepta for ten minutes b. To dosimo ton ruhon s-tin Maria ?*se ena lepto/ the giving the clothes-gen to-the Mary in one minute/ ?*epi deka lepta for ten minutes
Manner adverbs which are generally licensed with event nominalizations in Greek, are not easily tolerated with the nominals under discussion: (17) a.
??To
harisma enos vivliu s-tin Maria me enthusiasmo the gift a book-gen to-the Mary with enthusiasm b. ??To dosimo ton ruhon s-tin Maria me hara the giving the clothes-gen to-the Mary with pleasure
On the other hand, adjectival modification denoting repetition with singular nouns seems to be acceptable, as is generally the case with process nominalizations: (18) a.
To tahtiko harisma vivlion s-tin Maria the regular gift books-gen to-the Mary b. To sihno dosimo ruhon s-tus ftohus the frequent giving the clothes-gen to-the poor
Functional structure in nominals
Nominalizations based on double accusative roots are much better behaved. They happily co-occur with different kinds of aspectual modification (19) as well as manner adverbs (20): (19) a.
i didaskalia ton mathiton ??se mia ora/epi mia ora the teaching the students-gen in an hour/for an hour b. i pliromi ton ipalilon se deka lepta the payment the employees-gen in ten minutes
(20) a.
i didaskalia ton mathiton me enthusiasmo the teaching the students-gen with enthusiasm b. i pliromi ton ipalilon me aprothimia the payment the employees-gen without pleasure
It thus appears that nominalizations based on give-verbs are more deficient than their counterparts based on pay-verbs. Nominalizations based on pay-verbs seem to show the properties of Russian derived nominals and -er nominals, i.e. to show deficient verbal structure. Specifically, given that they lack aspectual and manner adverbial modification, they must also lack AspectP. Recall that nominals derived from stative psych-verbs cannot be modified by manner adverbs; this is shown again in (21a). Moreover, nominals derived from non-causative forms of psych predicates are not completely acceptable in such environments either. Both, however, permit aspectual adverbial modifiers: (21) a.
i agapi tu Jani ja ti Maria *trela the love the John-gen for the Mary-acc madly b. o fovos tu Jani ja ti thalasa ??mehri trelas the fear the John-gen for the sea-acc to madness
On the assumption that stative environments generally lack vP, the above pattern seems to point to the following possibilities: either stative nominals lack AspectP as well, and the availability of durative adverbial aspectual modificaTable 1. Type of Nominal Nominalized Clause Derived Nominals -er nominals/certain derived ones? Stative-nominals?
Language Greek Greek/Polish English/Greek/Russian
Structure D embeds CP D embeds AspectP D embeds vP
Greek
D embeds LP
Variation in functional structure
tion is linked to the semantics of the roots, as these has been defined in Chapter 2, or they contain an AspectP which bears the specification [stative] (see Embick 2000b for such a proposal concerning stative participles). Table 1 summarizes the patterns of variation discussed so far. . Variation depending on the type of v/Voice (transitive vs. intransitive) More nominal types emerge depending on whether the v included in the structure is transitive or not. As already discussed in the previous section, -er nominals differ from derived nominals in that they include an agentive v. In this section I examine more types of nominalizations whose properties can be attributed to the status of v. .. Gerunds vs. derived nominals in English As was mentioned in the introduction and in Section 1.2.3, English gerunds contrast with derived nominals of type destruction. The differences between these two are summarized in (22)., (22) Adjectives Article DP-distribution Overt Subject Accusative Object
NPs √ √ √ * *
Gerunds * * √ √ (gen) √
In fact English has several types of gerunds illustrated in (23a–b) with similar meaning and distribution: (23) a. John’s performing the song b. I imagined John performing the song
The gerunds in (23a) and (23b) have frequently been considered not to form a unified class, the former is referred to as POSS-ing, while the latter as ACC-ing. It has been argued that the gerund in (23a), with its subject in the genitive is syntactically different from the gerund in (23b) with an objective case subject (Rosenbaum 1967, Roeper and Wasow 1972, Horn 1975, Abney 1987 among others). Here I concentrate mainly on the former type. Chomsky (1970) has suggested that gerunds are NPs dominating Ss. Zucchi (1993), however, points out that since there are no sentences with a genitive subject, the presence of a genitive NP indicates that the preverbal NP is not a
Functional structure in nominals
sentential subject either. That is the presence of a genitive subject is consistent with a DP structure and not with a verbal sentential one. Moreover, gerunds, unlike sentences, do not license sentential adverbial modifiers. This suggests that they do not involve ‘high’ functional projections that can host such adverbs of the type found with verbal clauses, as discussed in Chapter 2: (24) a. *John’s probably being a spy b. *John’s fortunately knowing the answer
Recently, Harley and Noyer (1998b) attempted to account for these differences within the framework of Distributed Morphology. They argued that in the gerundive context a vP is necessarily present, as gerunds take the verbal from of the root, rather than the nominal one. As gerunds cannot take a determiner (*the destroying the city), Harley and Noyer proposed that gerunds are special instances of small clauses with vP predicates, as illustrated in (25) below. (25) [SC Belushi [vP mixingj [FP drugs and alcoholi [LP tj ti]]]]
According to Harley and Noyer, the subject of the gerund is not merged in Spec,vP. In their system it is the presence of FP that makes the licensing of accusative case on the internal argument possible. With Harley and Noyer, I assume that vP is present within gerunds. In fact as I argued in Chapter 2 and above, AspectP is also included in their functional architecture. However, there are problems with Harley and Noyer’s account having to do first with the position of the subject and second with the availability of accusative case. According to the system assumed in the introduction, the presence of an FP licensing accusative is not necessary; v can be argued to project a specifier, so that it can also license accusative case on the internal argument. However, matters are not that simple. Portner (1992:94) discusses one argument in favor of assuming that the subject of gerunds is directly projected in Spec,DP (he makes the argument for both the POSS-ing and the ACC-ing type). This involves scope. Consider (26): (26) Everyone(’s) not smiling bothered me
Here, everyone obligatorily has wide scope over not. This fact is in contrast to the state of affairs with a clause: (27) Everyone did not smile
If (26) contains a structure similar to (27) as a subpart, in the sense that subjects are generated in Spec,vP, it is quite difficult to see how to prevent all the
Variation in functional structure
scope relations possible in (27) from being available in (26). However, if everyone is generated in Spec,DP wide scope for everyone seems inevitable. There are two ways one might seek to derive wide scope for not in (26). Either not is able to raise at LF adjoining to IP, or everyone reconstructs at LF to a position below negation, i.e. to Spec,vP. The former possibility is not available with (26). The latter possibility cannot be the correct structure for (26), given that wide scope for the negation is not possible. Thus in this case the subject is generated external to the vP, it has not been moved there from anywhere and therefore no scope reconstruction can occur. In fact Portner assumes that PRO is generated in Spec,vP. Here I adopt Portner’s proposal and I take v in gerunds to be transitive. Given that gerunds appear in nominal contexts in English and take possessor subjects, they are embedded under a D° head which is the locus of the possessive affix. Spec,DP is also the position where the subject is generated. There is a remaining issue concerning the absence of adjectives from gerunds. This fact can be taken to correlate with the absence of any other nominal functional projections from gerunds. In other words, English gerunds are AspectPs embedded directly under D, i.e. they lack NumberP. Finally, English mixed nominalizations, e.g. John’s destroying of the city, pattern with English derived nominals with respect to the status of v and presumably the presence of AspectP (see Harley and Noyer 1998b for extensive discussion). Table (28) summarizes the variation discussed thus far. (28)
Transitive v
Intransitive v
Agentive v Gerunds in English -er nominals in English/Greek Non agentive v — destruction nominals across languages, English mixed nominalizations
A note of clarification is in order here. I have been assuming that there are two sources of variation related to v: agentivity and transitivity (see Embick 2000c). This means that in principle v could be agentive, and not transitive, for instance as in verbal passives, or -er nominals. But it could also be agenitive and transitive as in transitive verbal constructions and gerunds. .. Nominalizations in Hebrew In the previous section the possibility of licensing an accusative object within a nominal construction was attributed to the presence of a transitive v. In this
Functional structure in nominals
section I concentrate on Hebrew nominalizations, one variant of which also licenses accusative case on the internal argument. The relevant data are repeated below: (29) harisat ha-cava ’et ha’ir bi-mehirut the destruction the army the city in quickness
In (29) the internal argument, ‘the city’, is preceded by the accusative marker ’et and the agent bears genitive case. As far as I can see, there are two possible accounts for this pattern. The first possibility is to suggest that Hebrew, like English, has two types of nominals. One type which includes a transitive v, and one type which contains an intransitive v. Given that the two types do not show any other significant differences between them, the difference in transitivity must be related to the feature specification of v. The second possibility is to suggest that the accusative case found in these Hebrew nominals is not a structural accusative and that in (29) the genitive is the only structural case present. In fact this has been independently argued for in Siloni (1997). Here I summarize two of her arguments. Siloni points outs that there are significant distinctions between the accusative case of transitive verbs and that found in nominal contexts which point to the conclusion that the accusative within event nominals is not a structural case, but rather an inherent one. The most salient distinction concerns the particle ’et. This particle appears with a definite object only. That is, when a verb takes a definite complement, ‘et must precede the complement. When the verb takes an indefinite accusative complement, ’et cannot appear. Nouns can assign accusative case only in the presence of ’et: (30) a.
ha-cava hara *(et) ha’-ir the-army destroyed acc the city b. ha-cava haras (*et) ’ir ’axat the-army destroyed acc city one c. harisat ha-cava ’et ha-’ir destruction the-army acc the-city
When the objects of nouns receive genitive case they can be indefinite. According to Siloni, as there is no reason to assume that there is a definiteness requirement on the objects of event nominals, this distinction in the use of ’et in nominal contexts is unexpected, if the ‘et found in nominal contexts were the same as the one found in verbal contexts.
Variation in functional structure
A second difference relates to the fact that event nominals are unable to realize their object as an accusative pronoun: (31) a.
ha-cava haras ‘oto the-army destroyed him b. *harisat ha-cava ‘oto destruction the-army him
Assuming that the accusative case borne by pronouns is structural, the fact that this cannot occur within a nominal context follows from the hypothesis that the accusative case in nominals is inherent and not structural. Other instances of inherent accusative case have been discussed in the generative literature for instance, in German (van Riemsdijk 1983), in Italian (Belletti and Rizzi 1988), in Icelandic (Zaenen et al. 1985), and in Spanish (Torrego 1998). As mentioned in Chapter 3, Torrego (1998) argues that affected objects in Spanish bear inherent accusative. Such affected objects are introduced by the dative marker a: (32) a.
su acusacion al teniente her accusation to the lieutenant b. *su acusacion del teniente
However, there exists a further complication both in Hebrew and in Spanish, as far as the case of the theme argument is concerned. Recall that in Spanish when the possessive adjective is not present, as in (33), the de-phrase becomes grammatical again. Contrast (33) to the examples under (32). Moreover, in the type of Hebrew nominals referred to as passive, the theme argument surfaces with genitive case and not with accusative. This is illustrated in the (34a–b). (34c), where the theme argument bears accusative case, is ungrammatical: (33)
la acusacion del teniente the accusation of the lieutenant
(34) a.
ha-harisa šel ha-’ir the-destruction of the-city b. harisat ha-’ir destruction the-city the city’s destruction’ c. *ha-harisa ‘et ha-’ir the-destruction acc the-city
Functional structure in nominals
In the above examples when a genitive agent is present the theme argument bears inherent case. When the agent is absent, the theme surfaces with genitive case. This pattern is expected if (a) genitive is a structural case, and (b) only one structural case can be licensed within the DP. What is not immediately clear is why there should be such a distinction between structural genitive objects and inherent accusative ones related to the absence vs. presence of an agent. One would not expect structural case to be present in the absence of an agent, and inherent case to be possible in the presence of a genitive agent, under the standard interpretation of Burzio’s generalization, unless the genitive arguments (theme and agent respectively) compete for the same head assigning case (but see Siloni 1997 and Torrego 1998 for alternative views). If nominalizations constitute an environment that falls under the generalization established in Alexiadou and Anagnostopoulou (to appear), as suggested in the previous chapter, this state of affairs is expected: as both arguments are internal to the same domain, and there is only one head checking structural case, one of them must surface bearing a prepositional/inherent marking. The situation in Hebrew is a bit more complex. Consider the Hebrew nominalization patterns once again. (35) a.
ha-harisa šel ha-cava ’et ha’ir the-destruction of the-army acc the city b. harisat ha-cava ’et ha’ir destruction the-army acc the city c. bxinat ha-mismaxim ’al-yedey ha-rašut ha-musmexet examination the-document by the-authority the-competent d. ha-bxina šel ha-mismaxim ’al yedey ha-rašut the-examination of the-documents by the-authority ha-musmexet the competent
(35a–b) illustrate the two ways in which a genitival relation between a head noun and a noun phrase can be indicated in Hebrew. While in (35a) genitive case is realized via šel ‘of ’, in (35b) šel ‘of ’ does not appear and abstract genitive seems to be assigned by the head noun itself. The second construction is known in the literature as the construct state, while the former is known as the free state. In (35c–d) the agent is introduced by a by-phrase. The complement of the head noun appears either in the construct (35c) or in the free (35d) state. Now a noun that takes an inherent accusative theme cannot realize an al-yedey ‘by’ phrase.
Variation in functional structure
According to Siloni, inherent accusative is available within NPs only if the external argument has been discharged. This is the case in (35a–b), but not in (35c–d), where the by-phrases do not bear the external thematic role in the same way genitive arguments do. In further support of this claim, Siloni points out that ‘et phrases are possible within nominals that contain referential adjectives: (36) a.
ha-hafcaca ha-yisre’elit ’et levanon the-bombing the-Israeli acc Lebanon b. ha-sixzur ha-sini ’et ha-xoma the-reconstruction the-Chinese acc the-wall
Genitive case is illicit in the above examples. Siloni’s conclusion seems puzzling. Even more so since inherent accusatives are included within passive and unaccusative constructions in colloquial Hebrew (cf. Siloni 1997: 84 and references therein): (37) a.
noda li ’et ze ‘etmol was known to me acc this yesterday b. kara li kvar ’et ha-te’una ha-zot happened to me already acc the-accident the-this
While it could be argued that (37a) involves an implicit agent, no such possibility arises for (37b), an unaccusative predicate. In fact, as McCloskey (1996) points out in his discussion of ‘salient’ unaccusative constructions in Irish, such patterns are anticipated by the general framework developed in Burzio (1986). Consider the Irish unaccusative verbs in some detail. Irish has two types of unaccusative predicates: ‘salient’ unaccusatives, i.e. verbs which take a single argument marked with what traditionally is taken to be a preposition and which entirely lack a structural subject (38a). McCloskey establishes by various tests that this single oblique argument remains internal at all levels of representation. Verbs belonging to the second type, referred to as ‘putative’ unaccusatives, although semantically belong to the class of unaccusative verbs, bear no formal mark of being different syntactically from any other intransitive (38b). Many of these verbs exist in systematic alternation with the salient unaccusative type. As McCloskey convincingly shows, the single argument of these verbs is a surface subject, and is not introduced by a preposition:
Functional structure in nominals
(38) a.
neartaigh ar a ghlór strengthened on his voice ‘His voice strengthened’ b. neartaigh a ghlór strengthened his voice ‘His voice strengthened’
The general framework developed in Burzio leads one to expect the existence of unaccusatives that subcategorize for a PP, i.e. the salient unaccusative in Irish. Burzio suggested that such verbs do not exist. However, the Hebrew and Irish facts show that Burzio was right in his proposals: there exist unaccusative predicates single arguments of which bear inherent/prepositional case. A similar point can be made on the basis of unaccusative predicates in German, where the single argument bears either inherent accusative or dative as shown in the examples in (39) (see Grewendorf 1989 among others for discussion): (39) a.
mich friert me is cold b. mir schwindelt to me is dizzy
On this interpretation, the Hebrew examples can be used as evidence in favor of the lack of an external argument. That is the fact that the theme-argument bears inherent accusative case is not linked to the presence of an external argument. Rather it suggests the lack thereof. In other words, inherent accusative is licit in the absence of an external argument, as expected under Burzio’s generalization. Inherent accusative is present whenever there is a structural genitive in the clause, and only one functional head where case could be checked (see also the discussion in Chapter 5). One final remark concerning the data in (36) is in order here. Recall that in Chapter 3 I argued that the licensing of referential adjectives is only possible within result nominals. Based on this, one could suggest that the nominals in (36) are result nominals and not event nominals. If this is correct, (36) should be analyzed on a par with the Romance data discussed later on in this chapter. The data in (35a–b), however, still pose a problem, since these clearly involve process nominals. Taking Siloni’s analysis of the ‘et phrase to be on the right track, I believe that the Hebrew constructions involve a type of antipassive in the nominal structure. A similar proposal should be made for the German cases briefly discussed in Chapter 3, i.e. the alternation der Angriff der
Variation in functional structure
Zielen durch die NATO vs. der Angriff der NATO von der Zielen, both meaning ‘Nato’s attack of the targets’. The anti-passive, as this is found in a number of ergative languages, has the following properties: the underlying agent becomes subject of the antipassive and surfaces in the absolutive, i.e. the case standardly associated with internal arguments in ergative languages. The underlying object goes into a peripheral function being marked by an inherent/oblique case. This is shown below with an example from Squamish (Jacobs 1994:131). (40a) illustrates the ergative/transitive pattern, while (40b) illustrates the antipassive clause, in which the patient is marked in the oblique case: (40) a.
na ch’em’-t’-Ø-as ta Tam ta Pita RL bite-trans-3abs-3erg det Tom det Peter ‘Tom bit Peter’ b. na ip’a7-ím-Ø alhi Qál’qalilh t-ta súkw’am RL hold:dt:3abs det Q.(name) obl-det bark ‘Q. had some cedar bark with her’
The Hebrew examples in (35a–b) are exactly parallel to this. Viewed this way, (35a–b) involve a theme bearing oblique case and an (internalized) agent bearing genitive case (assuming that there is no semantic distinction between the free and the construct state). On the other hand, (35c–d) involve the passive construction, i.e. an agent being prepositional, and a theme bearing genitive case. To sum up, in this section I examined the nature of accusative case within Hebrew nominals. Following Siloni, I suggested a possible analysis of the Hebrew examples according to which (a) accusative case within nominals is not structural, (b) languages differ (even internally) with respect to whether the nominals which include non-structural accusative are process or result. For instance, the Spanish cases involved result nominals, while some of the Hebrew nominals were clearly process nominals; (c) Hebrew process nominals with accusative marked objects are actually anti-passive constructions, i.e. they involve internalized agents and oblique marked objects.
. Variation in the set of nominal projections . Presence vs. absence of number: licensing of adjectival modification In the previous section I examined how the differences in the type and the
Functional structure in nominals
number of verbal functional projections can account for the various nominal constructions one finds across languages and within a language. In this section I turn to variation related to the number and the nature of nominal functional projections. A first source of variation is the presence vs. absence of the functional category Number. As discussed in Chapter 2, Ritter (1991) argued that Number is included within the functional architecture of nominals on the basis of the fact that plural endings have the same status as tense and agreement affixes do in the verbal domain. If one takes this proposal seriously, the obvious conclusion is in the absence of a singular vs. plural opposition, NumberP should be absent as well. In the first section of this chapter, it was signaled that -er nominals can pluralize and permit adjectival modification. In fact the presence of adjectival modification was argued to bear a close relationship to Number. The relevant data are repeated below. (41) the faithful defenders of the government’s policies
On the other hand, gerunds lack plural marking, and also do not license adjectival modification. This is accounted for, if we assume that Number is licensed within -er nominals, but its presence is not permitted within gerunds. The question that arises next concerns the presence of Number within process nominals of the destruction type, which, according to Grimshaw, do not pluralize, but do license adjectival modification. Interestingly, it can be shown that Grimshaw’s claim is not correct. As mentioned in Chapter 2, Bierwisch (1989) observed that pluralization of nominals in German does not always bring about the result interpretation, as in examples such as die Besteigungen der beiden Gipfel dauerten 6 Wochen ‘the climbings the two tops lasted 6 weeks’, where the noun clearly has an event interpretation. Moreover, Greek nominals related to unaccusative predicates do receive an event interpretation when in the plural e.g. i afiksis turiston oli ti nihta ’the arrivals tourists-gen all the night’. This set of data clearly shows that Grimshaw’s criterion distinguishing process from result nominals in terms of the availability of plural marking does not hold. Hence process nominals include the functional category Number. . Variation depending on the type of D In all the nominal types discussed so far, D is present in the structural representation. The presence of this functional head turns the construction into an ar-
Variation in functional structure
gument. What I claim in this section is that the A vs. A¹ status of Spec,DP is also a source of parametric variation. As will be shown, languages in which Spec,DP is an A-position permit transitive nominalizations e.g. John’s destruction of the city, while languages in which Spec,DP is an A¹-position do not allow them. As mentioned in Chapter 3, Greek process nominalizations cannot be transitive, in the sense that it is never the case that both the theme and the agent argument of the nominal can surface bearing genitive case. The relevant data are repeated below: (42) a. *i katastrofi tu vivliu tis Marias se pende lepta the destruction the book-gen the Mary-gen in five minutes b. *tis Marias i katastrofi tu vivliu se pende lepta c. *i perigrafi tu Jani tis Marias se pende lepta the description the John-gen the Mary-gen in five minutes
The examples in (42) are all ungrammatical. When both arguments of the nominal are present, agents necessarily surface as PPs: (43) i katastrofi tis polis apo tus ehtrus se pende lepta the destruction the city-gen from the enemies in five minutes ‘The destruction of the city by the enemies in five minutes’
However, as it has been pointed out in several places throughout this study, (44) below is possible. In (44) the first genitive, the one immediately following the head noun, is understood as the object of description, while the second one is understood as the individual to whom the description belongs, i.e. the author of the description. (44) differs from (42c) in that the genitive DP understood as the author of the description refers to a known poet and thus makes the interpretation easier to get. (44) ?i perigrafi tu topiu tu Seferi (*se pende lepta) the description the landscape the Seferis-gen in five minutes
Other nouns that behave like this are metafrasi ‘translation’, anaskopisi ‘review’, and also certain picture nouns agalma ‘statue’, portreto ‘portrait’. At first sight, (44) seems to challenge the view that process nominalizations in Greek are intransitive and raises some questions the claim that two arguments bearing genitive case are illicit within nominalizations. As mentioned in Chapter 3, there exists a difference between Greek, Romance, and Germanic with respect to the availability of ‘transitive’ nominalizations. Consider again some data from English and German:
Functional structure in nominals
(47) a. the Vandals’ destruction of Rome b. John’s translation of the book (48) a.
Kolumbus Entdeckung Amerikas Columbus’ discovery of America b. Attilas Zerstörung der Stadt Attila’s destruction the city-gen c. Marias Übersetzung des Buches Mary’s translation of the book
The examples in (47) and (48) show that English and German generally permit transitive nominalizations, where the agent appears in prenominal position even with destruction type nouns. However, as already mentioned in the introduction, English destruction type nominalizations are transitive, but English nominalizations related to verbs entering the causative-inchoative alternation such as grow are never transitive. (49) a. John’s destruction of the city b. *John’s growth of tomatoes
If ‘transitivity’ in nominalization is generally possible, the fact that this does not happen with all types of predicates is mysterious. This asymmetry had been taken in Chomsky (1970) as an argument for the lexicalist treatment of derived nominals as opposed to gerunds. Recall now the situation in French (50) and Italian (51). (50) a. *sa destruction de la ville her destruction of the city b. sa description d’un passant her description of a passant c. la traduction de Pierre de l’oevre de Zola the translation of Peter of the work of Zola (51)
la tua descrizione di Giorgio the your description of G.
As pointed out in Chapters 2 and 3, French and Greek disallow transitive nominalizations with destruction type nouns, while they allow them with translation type nominals. What is interesting here is that the agent can either appear in the form of an adjectival possessive, or a de-phrase. Italian permits
Variation in functional structure
similar constructions as well: in (51) the agent appears in the form of an adjectival possessive and the theme is introduced by di. In what follows I show that the nominals in the languages under discussion split into two groups. In the one group, which includes Greek and Romance, the nominals involved are result nominals that can license a possessor argument. In the second group, which includes English, the nominals involved are not transitive in the sense that they involve agents introduced by a light v. Rather agents are introduced by D, a projection which has thematic properties in a language like English. Greek does not license such a position, as Spec,DP behaves like an A¹-position in this language. In the next subsection I turn first to the distribution of the two genitives in Greek and Romance. .. Two genitives are allowed only within result nouns In the data discussed above we saw that in languages such as Greek and Romance, transitive nominalizations are restricted only to translation type nouns and are impossible with destruction type nouns. Zubizarreta (1987) notes that this relates to a semantic difference between the two groups: nouns like description, translation and so on can denote a process or an event but they can also denote the result of a process. Nouns like destruction, on the other hand, can only refer to a process and never to the result of a process. With respect to Greek, Markantonatou (1992) and Kolliakou (1995) both present arguments that the second genitive in the examples in (44) bears a somehow loser relationship to the possessed noun. Markantonatou claims that the noun and the thematic genitive form some kind of compound, and the second genitive functions as a possessor. On the other hand, Kolliakou argues that the two genitives instantiate different types of possessives, the innermost being a pseudo-possessive. Kolliakou actually makes the point that pseudopossessive genitives are analogous to non-intersective adjectives. On this view, the innermost genitive in (44) determines the kind of description. Let us look at the syntactic behavior of the two genitives in Greek in some detail. While the two genitives behave as normal constituents when they occur in the DP alone, when they occur together the first genitive seems to be attached to the deverbal noun in a compound form. As the examples below show, there is a contrast in possessor-extraction contexts. While extraction of the creator (52b) is grammatical, extraction of the theme argument is impossible when the creator genitive is present (52a). In (52c) the wh-phrase can refer both to the work, Odyssey, and to the translator, Kakridis:
Functional structure in nominals
(52) a. *tinos i metafrasi tu Kakridi ine sto rafi deksia? whose the translation the Kakridi-gen is on the shelf on the right b. tinos i metafrasi tis Odisias ine sto rafi deksia? whose the translation the Odyssey-gen is on the shelf on the right c. tinos i metafrasi ine sto rafi deksia whose the translation is on the shelf on the right
Similarly in the case of possessor fronting, movement of the theme argument in the presence of the creator is ungrammatical: (53) a. *tis Odisias i metafrasi tu Kakridi ine sto rafi the Odyssey-gen the translation the Kakridi is on the shelf deksia on the right b. tu Kakridi i metafrasi tis Odisias the Kakridi-gen the translation the Odyssey-gen c. tis Odisias i metafrasi the Odyssey the translation
In (52) and (53) the theme argument cannot move over the Creator genitive to Spec,DP. Markantonatou (1992) attributes this contrast to the fact that the head noun together with the theme genitive forms one constituent that can only move as a unit. This conclusion is supported by data such as the ones given in (54). As the contrast in these strings shows, modification of the theme genitive is not permitted, while modification of the creator genitive is well formed: (54) a. *i metafrasi tis makroskelus Odisias tu Kakridi . . . the translation the lengthy Odyssey the Kakridi b. i metafrasi tis Odisias tu eksohu Kakridi the translation the Odyssey the great Kakridi c. i metafrasi tis makroskelus Odisias the translation the lengthy Odyssey
The two genitive phrases also contrast with respect to modification by relative clauses: (55) a. *tu Kakridi i metafrasi tis Odisias, the Kakridi-gen the translation the Odyssey-gen, i opia ine ena oreo piima the which is a beautiful poem
Variation in functional structure
b. i metafrasi tis Odisias tu Kakridi, the translation the Odyssey-gen the Kakridi-gen o opios ipirkse, daskalos mu who was teacher my c. i metafrasi tis Odisias, i opia ine ena oreo piima the translation the Odyssey the which is a nice poem
In (55) only the genitive bearing the creator role can be modified by a relative clause. Of course, relative clause modification of the deverbal noun is also possible. Given these differences, it seems correct to follow Markantonatou and claim that in examples like the ones above metafrasi and Odisias are interpreted as being one single unit. This phenomenon is well attested in Greek where constructions of the sort Noun +NP [gen] that behave as units are in really common use e.g. faki epafis ‘lenses contact-gen’. Further support for this comes from the use of possessive-clitic test established in Kolliakou (1995). Kolliakou points out that possessive clitics, when present within the DP, are always referential, i.e. they denote an extensional, never an intensional or ‘kind’ modifier of the noun. Hence the genitive DP in (56a) cannot be replaced by a clitic: (56) a.
to vivlio tis istorias the book the history-gen b. *to vivlio tis the book hers
Consider now (57): (57) a.
i perigrafi tu topiu tu Seferi the description the landscape-gen the Seferis-gen ‘Seferi’s description of the landscape’ b. i perigrafi tu tu topiu the description his the landscape-gen ‘his description of the landscape’ c. *i perigrafi tu tu Seferi the description its [the landscape] tu Seferi-gen d. i perigrafi tu apo ton Seferi the description its [the landscape] by Seferis
In (57a) the creator follows the head noun and the genitive complement. In (57b) the clitic is clearly interpreted as the creator of the description. On the
Functional structure in nominals
other hand, in (57c) where the creator is in the genitive and the possessive clitic refers to the theme is out. The only option to express the theme relation for the possessive is (57d), with the PP expressing the agent. (57c) is parallel to (56b) above, i.e. the theme genitive together with the head noun denote a type of description. To sum up, the above examples constitute evidence for considering the head nominal together with the innermost genitive as a type of compound. The outer genitive bears the role of the possessor; it is presumably situated in the specifier of a PossP (Delsing 1993). In other words, its licensing follows from the general pattern of the syntax of possession (see for instance the proposals in den Dikken 1997 and Alexiadou and Wilder 1998). Note here that, if possessors bear structural case, the fact that noun head forms a compound together with the theme argument enables the construction to escape Alexiadou and Anagnostopoulou’s (to appear) generalization, by leaving only one argument inside the DP bearing structural case, namely the possessor. In other words, the two genitives do not compete for the same Case checking head, as one of them forms a compound with the head noun, and hence does not need to check its case (see Baker 1988). Zubizarreta (1987) made a similar suggestion concerning the Romance examples discussed in the previous subsection. According to her, result nouns can include a possessor variable, while process nominals cannot carry such a variable. As a result, the possessive pronoun that appears is interpreted as a possessor, i.e. as carrying a modifier role. The difference between English and Romance follows from the fact that in English the prenominal position is not position that is restricted to [+poss] elements. Hence the item that appears there bears a number of roles. On this view, the nouns that permit the double genitive constructions are not process nominals. As expected both Greek and Romance translation type nouns behave alike with respect to the tests distinguishing process from result nominals. As the examples below show, Greek nominals that permit the double genitive construction do not permit aspectual modification of the type licensed inside process nominals: (58) a.
i sihni metafrasi tis Odisias the frequent translation the Odyssey-gen b. *i sihni metafrasi tis Odisias tu Kazandzaki the frequent translation the Odyssey-gen the Kazandzakis ine sto rafi is on the shelf
Variation in functional structure
Furthermore, as shown in (59), modification by aspectual adjectives is possible for the noun accompanied by the two genitives, but only when the plural form of such nominals is used. This is expected, if plural nominals tend to denote results, keeping in mind the refinements made in Chapter 2 and earlier on here: (59) i sihnes perigrafes tu topiu tu Seferi the frequent descriptions the landscape-gen the Seferis-gen ‘Seferis’ frequent descriptions of the landscape’
When the second genitive is replaced by the apo-phrase, the string is interpreted as a passive nominal, and thus can appear in an environment which licenses event/process readings, such as that involving the predicate begin (60a). On the other hand, the nominal with the two genitives is ungrammatical (60b), as expected if it only has the result interpretation: (60) a.
i metafrasi tis Odisias apo ton Kakridi arhise the translation the Odyssey-gen from the Kakridis started-3sg prin dio mines before two months ‘The translation of the Odyssey by Kakridis started two months ago’ b. *i metafrasi tis Odisias tu Kakridi arhise prin the translation the Odyssey-gen the Kakridis started-3sg before dio mines two months
Similar contrasts can be observed in Romance (French data from Siloni 1997:98; Catalan data from Picallo 1991: 293f.): (61) a. b. c. (62) a. b.
la présentation de livres de ce journaliste est toujours raffinée the presentation of books of this journalist is always refined *la présentation frequente de livres de ce journaliste . . . the presentation frequent of books of this journalist la présentation frequente de livres par ce journaliste la demonstracio de’en Joan del teorema de Pitàgores és inconsistent ‘the proof of Joan of Pythagoras theorem is inconsistent’ *la demonstracio del teorema de Pitàgores per part de’en Joan és inconsistent ‘the proof of the theorem of Pythagoras on the part of Joan is inconsistent’
Functional structure in nominals
As (61b) shows, aspectual modification, which brings about the process reading of the nominal, is not licit when two de-phrases are present. (61c), on the other hand, in which the agent is introduced by a PP, is grammatical under the process reading. The same holds for the Catalan examples (62a–b). Predicates that generally select for results, such as being inconsistent, happily allow the nominal construction with the two de-phrases, while they disallow the passive one, where the agent is introduced by a PP. A similar point has been made by Engelhardt and Trugman (1999) for double genitives in Russian nominals. According to these authors, double genitives are only allowed with some non-process derived nominals and some nominals derived from verbs that assign inherent genitive. On the other hand, the presence of double genitives is disallowed within process nominals, as discussed in Chapter 3: (63) a.
konspekt lekcii brata summary lecture-gen brother-gen ‘brother’s summary of the lecture’ b. konspektirovanie lekcii *brata/bratom summarizing lecture-gen brother-gen/instr ‘brother’s summarizing the lecture’
Note in the examples above that the two forms differ. Only the second one, bears the affix -nie. In fact only -nie monimalizations permit modification by frequent and constant: (64) postojannoe vyvešivanie/*vyvyska bel’ja constant hanging-out sign laundry-gen
To conclude, in this section I discussed the presence of transitive patterns of nominalizations across languages. I presented evidence that these constructions do not actually involve process nominals, but rather result ones. Moreover, I showed that the complement genitive appears to form some sort of compound with the head noun. In this sense, such nominalizations do not run into the problem of having two DPs with the same case competing for the same checking head. In the next section I turn to English transitive nominalizations. .. Agents in Spec,DP Recall that in English the string John’s destruction of the city is grammatical. Unlike the constructions discussed in the previous subsection, such nominals are clearly process nominals. In order to maintain the proposal that in destruc-
Variation in functional structure
tion type nominalizations v is always deficient, I have to suggest that the agent in English transitive nominalizations cannot be merged in Spec,vP, but rather it is located in the specifier of a different functional head, one that can license such arguments. The functional head that comes to mind is D. Recall that in Chapter 2, I presented arguments that Spec,DP is a thematic position in English. According to Abney (1987), evidence for this fact comes from the observation that the thematic argument in passive nominalizations can occupy this position, as exemplified in (65): (65) the city’s destruction
On this view, the DP expressing the ‘actor’ is never generated in Spec, vP, but in Spec,DP. Crucially, the type of agent involved in such constructions is not the type associated with a light v, as in the case of transitive verbal clauses. In principle there are two options for the agent to be situated in Spec,DP. The first possibility is that the DP is directly merged there. Spec,DP is a thematic position, one that can host thematic DPs, hence such a direct Merging is possible (see the discussion on passive nominals in Chapter 3). The second alternative, suggested to me by Elena Anagnostopoulou, is that the DP can be moved there from its base position within the LP (see Chapter 3 and (66a)). This analysis is in fact the one Pesetsky (1995) proposes for deriving causative experiencer predicates. According to Pesetsky, constructions of the type this annoys John involve movement from a lower causer position to a high causer position, as depicted in (66b). A similar derivation could be proposed for the transitive nominalizations discussed in this subsection.
(66) a.
LP Theme L¹ L
b.
PP
VP Causeri V¹ V
VP
Exper V¹ V Causeri
Functional structure in nominals
As noted in the previous chapter, such a derivation would account for the fact that sequences of the type *John’s destruction of the city by the barbarians are ungrammatical. However, as already mentioned, Spec,DP in English is not a position strictly related to an agent theta-role. The data in (67a–b) illustrate that possessives are not thematically restricted, in the way by-phrases within nominalizations are. In fact Grimshaw (1990) took this lack of restrictiveness as an argument for the view that in nominalization the possessive is licensed by any suppressed argument, i.e. in principle it can be associated with any theta role: goal, theme, agent, experiencer, location, source etc. (67) a. John’s destruction of the city b. Many people’s fear of flying c. yesterday’s destruction of the city
As (67c) shows, a temporal possessive can also occur in prenominal position. Naturally such DPs cannot be associated with any thematic role. In other words, the prenominal DP in (67c) could not be argued to have moved there, in the sense that yesterday could never have the function of a deepcauser/instrument of the type *the destruction of the city by yesterday. In fact one could argue, on the basis of the above analysis, that only those DPs that are initially merged as by-phrases keep their agentive thematic-role, as only these undergo movement from a thematic position to a thematic position. According to Pesetsky, the thematic roles assigned to the two positions must be of the same type for this process of movement to be licit in the first place. DPs which are merged in Spec,DP directly are never linked with an agentive thematic role, if this view is correct. Assuming that agents in transitive nominalizations are located in Spec,DP and are never merged in Spec,vP, explains why in German only proper names can appear within these constructions (see Chapter 3). These are arguably situated in D (see Longobardi 1994). Two questions still remain, however. First, why do English nominals not behave uniformly with respect to the possibility of including an agent e.g. John’s destructions of the books vs. *John’s growth of tomatoes? Second, why is there a crosslinguistic asymmetry with respect to the availability of transitive nominalizations? Turning to the first issue, there are two possibilities. According to the first one, whenever a root denotes an externally caused event such as √DESTROY, the specifier of D can be interpreted as an external causer (agent). On the other
Variation in functional structure
hand, roots like √GROW which denote an activity that occurs spontaneously, thus being internally caused, do not permit the external causation interpretation of the specifier of D (see also the discussion in Chapter 3). This accounts for the unavailability of transitive nominalizations with verbs like grow entering the causative alternation. Crucially, √GROW can receive an agentive interpretation only if it is inserted under a syntactic causative head. This is not the case with √DESTROY which inherently possesses such an agentive interpretation. Alternatively, both roots can give transitive nominalizations, but formations with √GROW will not be felicitous, as these will receive no Encyclopedic interpretation. With respect to the crosslinguistic distribution of the availability of agents introduced by D, it is necessary to recall from Chapter 2 that the status of Spec,DP varies across languages While it has an A status in English, it has an A¹-status in e.g. Greek. As mentioned in Chapter 2, in English DP corresponds to IP, while in Greek DP corresponds to CP. Spec, DP is not an argument position in Greek and thus wh-movement naturally takes place within this domain. The relevant data from Horrocks and Stavrou (1987) are repeated below: (68) a.
b. c. d. e.
mu ipes oti diavases [to vivlio tinos] me told-2sg that read-2sg the book whose ‘You told me you read whose book?’ mu ipes oti diavases [tinosi [to vivlio ti] tinosi mu ipes oti diavases to vivlio ti [tinosj [[to vivlio tj]]i mu ipes oti diavases ti [to vivlio tinos]i mu ipes oti diavases ti
Example (68a) is taken to represent the structure prior to movements. In (68b), the genitive phrase has moved to Spec, DP. From this position subsequent movement to Spec, CP as in (68c) is possible. (68d) and (68e) involve fronting of the whole DP to Spec,CP. In the former case wh-movement has not applied DP internally, in the latter it has. In (69) I repeat Szabolcsi’s (1994) data that offer further confirmation for aligning Spec,DP with Spec,CP. The Hungarian pre-nominal possessor may occupy two positions. It may follow the determiner, in which case it has nominative case (69a), or it may precede the determiner, in which case it has dative case, as shown in (69b). (69) a.
a Mari kalap-ja the Mari-nom hat-3sg
Functional structure in nominals
b. Mari-nak a kalap-ja Mari-dat the hat-3sg
As discussed in Chapter 2, it seems reasonable to propose that Marinak in (69b) occupies the specifier position of an A¹-position. Interestingly in Hungarian, agents appear either as nominative possessors or within an àltal ‘by’ phrase: (70) a.
Péter Mari ellen való felszolal-ás-a Peter-nom Mari against being speak-dev-poss-3sg. ‘Peter’s speaking against Mari’ b. Peter Mari által való megszégyenít-és-e Peter-nom Mary by being humiliate-dev-poss-3sg
On the basis of the above, I conclude that transitive nominalizations are only possible, when a thematic specifier, other than Spec,vP is licensed in the structure. Hence in Greek such nominalizations are never possible, while they are in English. Finally, let me turn to the discussion of the Romance facts that include an adjectival possessor. Recall that in Romance it is not possible to have an adjectival possessor with destruction type nouns, while this is the case with translation type nouns. The relevant data are repeated below. (71) a. *sa destruction de la ville her destruction of the city b. sa description d’un passant her description of a passant
In the previous section I pointed out that according to Zubizarreta (1987) nominals with two de-phrases were actually result nominals. She has also argued that sequences of the type (71b) involve result nominals. Translation type nouns allow for the adjectival to be interpreted as a possessor and not as an agent. Thus we would expect that whenever such nouns are put in a context that clearly disambiguates between the result and the event reading, it is always the case that nominals preceded by an adjectival possessor receive the result interpretation. Evidence for this comes from the facts discussed in Bottari (1992), who shows that nominals introduced by possessive pronouns actually lack the event interpretation. Bottari notes that when selected by certain predicates (factives) derived nominals can be modified by negated element. In this case the derived nominals receive a propositional interpretation. Thus, accord-
Variation in functional structure
ing to Bottari, the negation strategy is bound on the argument supporting property of these nominals: only argument taking nominals, i.e. process nominals are compatible with the negated nominal strategy, non-argument taking nominals i.e. result nominals are not. In this light, consider the following examples: (72) a.
la non-descrizione della vita campestre mi aveva sorpreso the non-description of country life surprised me b. *la non-descrizione mi aveva sorpreso c. *la sua non-descrizione della citta li aveva irritati his non description of the city irritated them
The ungrammaticality of (72b) is attributed to the fact that the nominal in this environment has the result reading that is incompatible with negation. Interestingly, when a prenonimal possessive adjective is present then the nominal cannot be negated, patterning like a non-process noun (72c). Given that only such constructions can have a possessive reading, it is only in these that the possessor adjective can be present. To conclude, both Romance and Greek differ from English in that the English transitive nominalization pattern crucially involves process nominals with an extra thematic position. In other words, the transitive nominalizations are not of the same type in the languages under discussion: they involve result nominals in Romance/Greek, and such nominals are analyzed in terms of possession structures, but involve process nominals in English, where Spec,DP can be licensed as a thematic position irrespectively of the deficient character of v, which is common for all process nominals of the destruction type across languages.
. Summary In this chapter I have made three points. First, I demonstrated that the presence of Aspect contributes to event specification, and moreover, it is linked with the availability of adverbial modification Second, I proposed that while the presence of v contributes to eventive interpretation, the transitivity of v regulates the agentive vs. non-agentive character of the nominal and the availability of accusative case. Third, I argued that the A vs. A¹ status of D regulates the availability of transitive nominalizations.
Functional structure in nominals
Appendix: Nominal infinitives In languages such as Italian, Dutch and German infinitival constructions can appear in the complement domain of a determiner. This is illustrated in (73) with an Italian example: (73) Giovanni udì il mormorare John heard the whisper-inf
Much like other nominals, nominalized infinitives can appear in argument positions. This is illustrated below with an Italian example, but similar facts hold for the Dutch and German constructions: (74) il suo mormorare parole dolci spiega la reazione di Maria the his whisper sweet words explains the reaction of Mary
Grimshaw and Selkirk (1976) and Salvi (1983) have observed that Italian infinitivals of kind (73) contain at least a VP node, given that they can be modified by adverbs (75). Similar observations hold for Dutch (see Zwart 1987) and German (see Ehrich 1991 among others): (75) il suo scrivere quella lettera improvissamente the his write that letter suddenly (76) het annhoudend appels eten begint mij te vervelen the continuously apples eat-inf begins me to bore (77) das oft kleine Katzen streicheln the often small cats caress
In terms of the model developed here, the above facts suggest that nominalized infinitives include AspectP and vP. Nominalized infinitives show evidence for the presence of a more elaborated nominal domain, since they allow adjectival modifiers: (78) a.
il suo continuo eseguire la canzone the his continual perform-inf the song b. het aanhoudende appels eten begint mij te vervelen the continual apples eat-inf begins me to bore c. das ständige Belästigen der Frau the continuous annoying the woman-gen
However, if one looks closer, several sub-types of infinitivals present them-
Variation in functional structure
selves. For instance, in Italian, the nominal infinitival seems to have three subtypes, listed in (79), from Zucchi (1993): (79) a.
l’avere egli scritto quella lettera the have-inf he written that letter b. il suo mormorare parole dolci the his whisper sweet words c. il mormorare sommesso del mare the whisper-inf soft of the sea
Grimshaw and Selkirk (1976), Salvi (1983), Zucchi (1993) give evidence that the NPs in (79b) and (79c) differ in their internal structure. Whenever the argument of the infinitive is introduced by the preposition di ‘of ’, adjectives but not adverbs modify the infinitive: (80) a. *il cessare improvvisamente delle ostilitá the cease-inf suddenly of the hostilities b. il cessare improvisso delle ostilitá the cease-inf sudden of the hostilities
Moreover, the infinitive cannot occur with a direct object when the subject of infinitive is introduced by di. Both adverbs and direct objects occur when no of-subject is present: (81) a. *il rievocare la guerra degli anziani the re-evoke-inf the war of the elders b. il rievocare degli anziani the re-evoke of the elders c. il suo mormorare sommessamente the his whisper-inf softly d. il suo mormorare parole dolci the his whisper soft words
The syntactic behavior of the infinitive in the (a) and (c) examples is similar to the behavior of deverbal nominals in Italian: (82) *la rievocazione la guerra degli anziani the re-evocation the war of the elders
The pattern in (79a) is yet again different, in the sense that it includes a nominative subject, while in (79b) the subject is introduced in the form of
Functional structure in nominals
a possessive pronoun. Moreover, no adjectives are allowed in the infinitival type (79a): (83) a. *il continuo essere egli a corto di denaro the continuous be-inf he short of money b. il suo continuo esserse a corto di denaro the his continuous be-inf short of money
Zucchi (1993) claims that there is a semantic distinction between the three classes. Type (79a) appears in contexts where that-clauses occur. Type (79c) is compatible with event predicates. Type (79b) denotes propositional entities. This difference in interpretation seems to correlate with the different syntactic properties these various classes of infinitivals have. From the point of view of the analysis presented in the previous sections, (79a) has the status of a fullfledged clause. Note that it includes weak pronominal subjects, such as egli, which according to Cardinaletti (1997) are located in Spec,AgrSP. On the other hand, construction (79b) patterns much like English gerunds, as accusative objects and adverbs are licensed within the construction. Finally, (79c) seems to lack AspP, as it does not tolerate adverbial modification and to involve a deficient vP projection. German nominal infinitivals also seem to split in two sub-types, one that patterns very much like process nominalizations, and one that seems similar to English gerunds, since it permits accusative objects and adverbial modification (see Ehrich 1991 for discussion). The first type is illustrated in (84). The theme argument bears genitive case, the agent is introduced either by a PP or is in initial position: (84) a.
das Belästigen der Frau durch Peter the annoying the woman-gen by Peter b. Peters Belästigen der Frau Peters annoying the woman-gen
Thus, whatever analysis one assumes for process nominals, carries over to these constructions (see Reuland and Kosmeijer 1993 for a detailed description and an analysis). The second type is illustrated in (85): (85) das oft kleine Katzen streicheln the often small cats caress
Note that adverbs are not licit within the first sub-type of infinitivals. This
Variation in functional structure
suggests that the internal structure of type (b) includes AspectP, while this projection lacks from the first type. Moreover, v in type (b) licenses accusative case, while it is deficient in type (a). Ehrich (1991) suggests that there is a semantic difference between the two types. Type (a) refers to specific events, while type (b) refers to generic events. For this reason, infinitivals of type (b) are not compatible with temporal expressions like yesterday: (86) *die Mülltonne gestern leeren the carbage-bin yesterday empty
In this sense the distinction between the two types is parallel to the distinction discussed in the case of Italian infinitives and English gerunds. Turning to infinitival clauses in Welsh, it should be noted that these are referred to as verbal nouns (cf. Rouveret 1994). Verbal nouns are used to form periphrastic forms of tenses together with the auxiliary be and also to form complement clauses. However, as shown by Rouveret, they show mixed verbal and nominal properties. What is interesting for our discussion here is the fact that the internal argument of the verb noun appears to have the same dependence on the head element, as genitive arguments do in nominal contexts: (87) a.
geiriau cân words of song b. dysgu cân learning a song
(Rouveret 1994: 252)
In Irish, which has a rich inflectional morphology, the direct object visibly bears genitive case: (88) Nuair a bhí siad ag ceannach an tí when c were they Prog buy the house-gen ‘when they were in the process of buying the house’
Similarly to a nominal in the genitival construction, the verbal noun does not cause mutation of the initial consonant of the dependent term. Interestingly, sequences that include two arguments bearing genitive case are ungrammatical: (89) *ddarllen y plant lyfr arall read the children book other
Example (89) becomes grammatical again, once the subject DP is introduced
Functional structure in nominals
by the marker í. According to Rouveret, this shows that there is only one case available in the lower DP. (90) i r’plant ddarllen llyfr arall to the children read book other
This pattern is strikingly similar to the one of process nominals observed Greek, and discussed in the previous Chapters.
Notes . I will not be concerned with this type here. One could derive the difference between eventive and non-eventive -er nominals in a system that makes use of roots in multiple environments, by assuming that the latter include a v that is only agentive, but lacks event features. This is possible if one assumes that v is split into two layers, one containing event information, and one containing only agentivity features. Both of these could be included in the structure, or only one of them could combine with the root giving us the different types of nominals. . Notice that in Greek -er nominals, literally, -t- nominals, being animate, have a variable gender (di Domenico 1995, Harris 1991, Alexiadou and Stavrou 1999). Another distinctive property of the -er nominal in Greek, differentiating it from both instrumental nominals and derived event nominals (katastrofi) is the fact that agentive -er nouns are modified by adjectival ethnic nouns and not by ethnic adjectives: (i) a. b.
o Elinas tragudistis vs *o Elinikos tragudistis the Greek singer i eliniki katastrofi the Greek destruction
This distinction between the two ‘adjectives’ is sensitive to the animacy of the modified noun involved. . A different type of explanation is given in Larson (1998). Larson points out that the adjective in (9) in the text is ambiguous between an intersective and a non-intersective reading; these readings are paraphrased in (ia,b): (i) a. b.
Olga is dancer and Olga is beautiful Olga dances beautifully
On the non-intersective reading beautiful applies to Olga as a dancer, i.e. her dancing is beautiful even if she herself is unattractive. Larson claims that the reading in (ib) only involves the presence of an event argument. . Hamm (1999:33ff.) discusses the nominalization patterns found in Akatek Maya. In this language there are three types of nominalizations. Type 1 has a more sentential structure in the sense that aspect morphology is contained and adverbs can modify the head noun, but adjectives are not allowed. There are two further types, both of them lacking aspect
Variation in functional structure
inflection. They both disallow licensing of adverbs. It seems then that there exists a strong correlation between the availability of such projections and licensing of adverbs crosslinguistically. . This of course raises the question why although v is included in the structure in both cases, we end up denoting an individual in the case of -er nominals, but an event in the case of process nominals. Richard Larson (personal communication) suggests that the semantics of these nominals can be derived compositionally in the following way. Take -er and -ing as giving scope to the relevant arguments with respect to lambda abstraction. Suppose dance’ first combines with an event argument and an agent: dance (x,e). More generally, suppose that unergative nouns N first combine with an event argument and then an agent. Finally, suppose the translation of N into IL is N¹. Then the effect of adding -er can be given as follows: (i) N-er ==> lx G e[C(e) f N¹(x,e)] dancer ==> lx G e[C(e) f dance’(x,e)] ‘‘the set of individuals x, such that generally for events e, if e is a context of the right kind C, x dances in e’’. This account seems compatible with Marantz’s (1999) system that makes use of ‘nominalizing’ heads, but it is not immediately compatible with the assumptions made here which do not rely on the presence of such heads. Perhaps one way to implement syntactically the semantics of (i) is to assume that v breaks into two layers each hosting a different feature: a lower v encoding eventivity and a higher one encoding agentivity (Alec Marantz personal communication). . Engelhardt and Trugman (1999) point out that licensing of PP modification is licit: (i) konspektirovanie lekcii vpopyxax summarizing lecture-gen in haste . Note here that since Greek lacks nominal gerunds altogether, it is not surprising that in most instances Greek nominalizations share properties with English gerunds. . In Alexiadou (1999a) nominal gerunds were contrasted with verbal gerunds of the type found in Hebrew (see Siloni 1997) and in Greek (see Rivero 1994). These constructions fall outside the scope of the investigation here. . Arabic masdars have been argued by Fassi-Fehri (1993) to be very similar to English gerunds. As (i) shows, the masdar can appear in the position of a direct object. As is the case with English gerunds, the agent bears genitive case and the object bears accusative case: (i) ?aqlaqa-nii ntiqaad r-rajul-i l-masruuc-a annoyed-me criticizing-nom the man-gen the project-acc Constructions that show a similar pattern, i.e. genitive subjects and accusative objects are found across languages, that is they are not limited in gerundial or nominal contexts. In fact such patterns generally reflect a nominal system, even though they are encountered in verbal-like domains (see Borsley and Kornfilt 2000 for discussion of a number of patterns across languages). For instance, there are languages where genitive subjects appear
Functional structure in nominals
inside verbal clauses. Ouhalla (1997) signals that in VSO orders in Berber, the postverbal subject bears genitive case, while the object bears absolutive case. This is illustrated in (ii-a). And in fact Ouhalla takes this pattern to suggest that the syntax of the verbal clauses is ‘nominal’. In Turkish nominalized clauses the subject also bears genitive case and agrees with the verb (cf. (ii-b) from Kornfilt (1998)), while the object bears accusative case: (ii) a.
Zra-n ifruxn Hamish see-3pl boys-gen Hamish-abs ‘The children have seen Hamish’ b. bizi-im oku dug-umuz kitap-lar we-gen read-nonfut-p 1pl book-pl
. Abney (1987) has proposed that POSS-ing gerunds involve nominalizations of VPs, while ACC-ing gerunds involve nominalizations of sentences, i.e. at least TPs. Evidence that the two gerund forms have a different internal structure comes from the fact that there exists a number of environments where they two do not behave alike (cf. Horn 1975, Portner 1992 and references therein). For instance, only poss-ing gerunds permit reciprocal subjects, and can be topicalized. On the other hand, only acc-ing gerunds can be clefted, and it is only possible to extract out of acc-ing gerunds and not out of poss-ing: (i) a. each other’s giving up the ship b. *each other giving up the ship c. *Fred singing the national anthem everyone imagined d. Fred’s singing the national anthem everyone imagined e. It’s John kissing Mary that would upset me f. Which city do you remember him describing? g. *Which city do you remember his describing? Portner (1992) argues extensively that these differences between the two reflect first the difference in structure and follow from an analysis according to POSS-ing gerunds are definite while ACC-ing gerunds are indefinite. According to Portner, the POSS element plays exactly the same role as the definite determiner. . Note that this makes the structure of English gerunds look similar to the structure of participles. This is a welcome result in view of the fact that -ing gerunds grew out of participles historically (see Demske 1999 for discussion and references). German -ung nominalizations would receive a treatment analogous to English gerunds. . Engelhardt (1999) contains arguments against Siloni’s view. If Engelhardt’s points are along the right track, then the difference between the two types of nominals in Hebrew is derived by the proposal in the previous paragraph in the main text. . A case can be made that at least in Spanish the nominals containing inherent accusatives are not event nominals. In the next section I argue that Romance nominals including an adjectival possessor are actually result nominals. If this is correct, then the Spanish facts in (32a) should be analyzed as instances of result nominals. Thus, the theme argument bears inherent case, as there is no functional projection licensing a structural case.
Variation in functional structure
In fact, the Spanish paradigm gives the most convincing argument for the non-structural nature of case (genitive or accusative) within result nominals. Specifically, Spanish presents us with a clear distinction between the domains where structural genitive is licensed, i.e. in process nominals which are intransitive (33), and those in which nonstructural case is permitted, i.e. in result nominals which are transitive (32a), in the sense that they include a functional head which licenses the possessor argument. . Thanks to Gisbert Fanselow for pointing this out to me. . The Chinese de-construction presents a pattern somehow related to the discussion on anti-passive within nominal structures. As Niina Zhang pointed out to me in this construction, all arguments of a de-verbalized phrase must be raised to the left of de; on the other hand, if both the agent and the theme DPs are present with a verb, it is always the agent DP raises and remains as a DP, while the theme DP merges with a preposition to form a PP. Although de cannot initiate a phrase, the pre-de phrase does not need to be nominal. Thus the raising of an argument is not required by de itself. Instead, it is required by the post-de verb. Thus it seems that de makes its dominated verb unable to keep the arguments in their in situ positions. When both an agent and a theme argument raise, the theme argument is reformed as a PP. This is similar to the antipassive operation found in ergative languages, where an ergative subject is changed into an absolute subject and an absolute object is changed into an oblique nominal (in dative case, for example), as discussed earlier in connection with Hebrew nominals. . Squamish has two types of antipassive clauses. The first one is the one shown in (40b). The second type involves incorporation of the patient into the theme as a suffix, marked bold in (i) below: (i) ses men sát-qwuy’ach-í7m-Ø kwetsi manílh 1sg:nom next give-finger-st:3abs that same ‘Then that same one extended his finger’ Most of incorporated objects are body parts. . Arabic seems to exhibit an anti-passive construction as well. In (i) the preposition li, which has the same form as the dative preposition, may be used to case mark the direct object (from Fassi-Fehri 1993: 247): (i) aqlaqa-nii ntiqaad r-rajul-i li-l-masruuc-i annoyed-me criticizing-nom the man-gen the project-gen . Although the fact that the nominal cannot be modified by an aspectual phrase in (44) would be enough to establish that this has only the result reading, and hence does not constitute a problem for the analysis proposed here, in what follows I go systematically over Greek and Romance data to establish this point. I show that (44) does not question the internal structure proposed for process nominals. In fact my argumentation will turn out to support the view put forth in Picallo (1991) that result nominals are transitive and also support the proposal that the structural case features inside process nominals are located in Aspect and not in a nominal category. . As is well known, adjectives generally split into two types: pure intersective adjectives
Functional structure in nominals
of the type illustrated in (ia) and non-intersective ones of the type illustrated in (ic). The contrast between the two types is illustrated in the paraphrases in (ib) and (id) respectively (from Larson and Segal 1995: 499). Non-intersective adjectives lack predicative counterparts: (i) a. b. c. d.
that is a yellow kitchen that is yellow, and that is a kitchen Marya is a former dancer Marya is a former and Marya is a dancer
. A similar view is defended in Borer (1993) who argues that in agents in transitive nominalizations could be analyzed as receiving the possessor theta role from ‘s, located under D° (see also Marantz 1997, Harley and Noyer 1998a,b). Although, Borer and Marantz and Harley and Noyer all suggest that the agent is projected in SpecDP, Borer’s analysis differs from the analysis in Harley and Noyer in that she argues that the input to the examples such the Vandal’s destruction of the city is again a passivized DP: the Vandals’ is situated in Spec,DP and is co-referential with the suppressed subject. My analysis crucially differs from Borer’s in that I assume that no suppressed argument is present within process nominals, and that the structure involved is an unaccusative one. . In a sense this relates to Grimshaw’s view that the possessive in these examples functions like an argument-adjunct. The possessor shares some properties with arguments and some with adjuncts. On this view, Spec,DP has such a status that permits it to host adjuncts as well as thematic DPs. Note that the view expressed in the text is linked to the discussion in Chapter 3 concerning the derivation of passive nominals, which involve ‘external’ themes. . In recent work, Pesetsky and Torrego (2000) argue that D is the counterpart of C across languages. In their view, possessors occupy the specifier position of a functional projection below D. If this is correct, then it means that the specifier of this projection is not licensed in Greek, much like Spec,IP is not an A-position in the language (see Alexiadou and Anagnostopoulou 1998). . Incidentally note that the Spanish facts discussed in Chapter 3 in connection with affectedness can be seen in a new light here. These nominals involve a possessive adjective and an object introduced by the marker a. If these nominals are analyzed as result nominals, as suggested, then the question arises what the status of the object is in this case, i.e. does it bear inherent or structural case.
Chapter 5
Nominalization and Ergativity
The previous chapters studied in detail the internal structure of nominals. In this chapter I show how the view on nominalization put forward here enables us to capture certain similarities between nominals in nominative-accusative languages and certain patterns found in ergative languages. In order to do that I examine two different aspects of the link between nominal structure and ergative patterns. In the first part of this chapter I deal with certain similarities related to Case (Sections 1 to 3). In the second part I turn to two views on ergative marking, the first one of which was briefly discussed in Chapter 3, namely that ergative case is either analogous to a PP or to a possessor. These two ergative patterns are seen here as a reflex of the two ways of denoting agenthood within nominalizations (Chapters 3 and 4): as either merged in Spec,DP or included in the form of a by-phrase within the lexical phrase containing the root.
.
The Case patterns in nominalization and ergative languages
According to Dixon (1994), the term ergative is used to describe a grammatical pattern in which the subject of an intransitive clause is treated in the same way as the object of a transitive clause. The term was first used to refer to the case marking on constituents of a noun phrase: ‘ergative’ is the case of the transitive subject, while ‘absolutive’ is the case of the intransitive subject and the transitive object. Ergativity is thus complementary to the familiar grammatical pattern of accusativity, in which one case (nominative) marks both intransitive and transitive subject, with another case (accusative) being used for the transitive object. The examples in (1) illustrate this superficial characteristic for English, a nominative-accusative language and for Tonga, an ergative language (Tonga examples from Bok-Bennema 1991: 2): (1) a. She kicked her b. She worked/she arrived c. *Her worked
Functional structure in nominals
(2) a.
Na’e tamate’i ’a-Kollaiate ’e Tevita aux-past kill abs-Goliath erg-David ‘David killed Goliath’ b. Na’e alu ’a-Tevia ki Fisi aux-past go abs-David to Fiji ‘David went to Fiji’
Specifically, in certain ergative languages the agent argument, referred to as A-argument, is marked with ergative case, while the patient argument, referred to as P-argument, and the sole argument of intransitives (referred to as the S-argument) are marked with absolutive case. This state of affairs is similar to the case patterns in nominalizations, as shown in (3). The theme argument ‘city’ in (3a) and the intransitive subject ‘John’ in (3b) both have genitive case. The agent in (3a) is introduced by a preposition. (3) a.
i katastrofi tis polis (apo tus Italus) the destruction the city-gen by the Italians b. i afiksi tu Jani the arrival the John-gen
The similarities between these two patterns are summarized in table (1) below (see also Alexiadou 1999a,b, Bok-Bennema 1991, Williams 1987 and de Wit 1997 among others)., Table 1. A-argument S-argument P-argument
N/A system nom nom acc
E/A system erg abs abs
Nominalizations pp gen gen
It is an old intuition in the literature that ergative languages differ from accusative languages in that ergative constructions are actually passive constructions (see e.g. Fillmore 1968, Hale 1970). Subsequent syntactic research on ergative languages has led to various competing proposals, which are briefly discussed in note 5 and in the next subsection. Here I concentrate on a specific view on the syntax of ergative languages, namely one that offers a natural way of capturing the analogy between nominalized clauses in nominative-accusative languages and ergative languages. According to Bok-Bennema (1991), ergativity is characterized by the inabil-
Nominalization and ergativity
ity of transitive verbs to assign structural case, that is it is sub-sumed under the more general pattern of unaccusativity (see the discussion in Grewendorf 1989). Bok-Bennema in fact claims that ergative patterns arise as a solution to the Case problem posed by unaccusativity. In an ergative language, one way to solve the Case problem is to have an exceptional case for the subject, so that the nominative case becomes free to be assigned to the object. This is very much reminiscent of the analysis of nominalizations presented in Chapter 3. In the next section I present in detail the analysis of ergative languages that brings the two patterns together. According to this analysis, ergative case is not to be a structural case, but rather a lexical/prepositional case, much like the prepositional phrase introducing agents within nominalizations and ergative languages, like process nominals, have a deficient v. I discuss these two points in turn.
. Unaccusativity in ergative languages . Ergative as a lexical case The literature on ergativity contains several attempts to reconcile the well known Case and Agreement patterns of ergative languages with a constrained theory of Universal Grammar (see note 4). Some of the analyses within the GB theory equate absolutive case to the nominative of nominative-accusative languages (see Bok-Bennema 1991, Nash 1995, 1996, Woolford 1997). The ergative case is analyzed as a by phrase (Hale 1970), a PP or a KP (Bittner 1994), a genitive NP (Bok-Bennema (1991) or an abstract equivalent of the accusative (Murasugi 1992). In particular Nash (1996) proposes that ergative languages differ from accusative languages in that the former lack vP, which is the structural position assumed in Chomsky (1995) to host the transitive subject. According to this proposal, agents are not thematically projected in ergative languages, in the sense that they are not thematically projected as specifiers of a ‘light’ v. They can be introduced in the structure as adjuncts. I come back to this in the next section. Nash claims that ergative is not a structural case parallel to nominative (and see also Mahajan 1993, Woolford 1997 and references therein). Rather it is a lexical case, parallel to dative. On the other hand, Bobaljik (1992) and Laka (1993) argue that absolutive case is the counterpart of accusative case, while ergative case is the counterpart of nominative case.
Functional structure in nominals
Let us consider the view that ergative is a lexical case in some detail. First, ergative case is morphologically marked, much like dative case, a property which might bring into question its structural nature, under the view that morphologically marked cases are standardly not considered to be structural (see Nash 1995). Second, as Woolford (1997) points out, there is a gap in the current inventory of lexical cases available in UG, because there is no lexical case associated with agents. Ergative case fills this gap. However, although it is well known that ergative case is associated with the agent theta role, this association is not perfect. For instance, ergative marking is used on locative arguments in Avar, on instrumentals in Udi, and it is homophonous with the genitive case marking in Inuit (Nash 1996: 198). In this respect, the apo phrase that introduces agents within derived nominals in Greek is quite similar: apo is not exclusively used to denote the agent, it can be used to denote the causer or the source, as is illustrated in the examples (4a) and (4b) respectively. Nevertheless, when apo combines with an animate phrase, it denotes the agent (5): (4) a.
o Janis arostise apo ti stenahoria tu John-nom got sick from the troubles his ‘John got sick because of his troubles’ b. o Janis irthe apo tin Ameriki John-nom came from the America-acc ‘John came from the U.S’
(5)
o Janis skotothike apo ton Petro John-nom killed-pass-3sg from Peter-acc
Although lexical cases such as dative or ergative usually mark predictable theta-roles, there is enough idiosyncratic behavior involved to conclude that a verb’s ability to assign lexical case to one of its arguments has to be specified in that verb’s lexical entry. Woolford notices that although some scholars accept the idea that there is a link between ergative case and the thematic role of agent in the active type of ergative languages where even intransitive agent subjects are marked with ergative case (e.g. Basque), they reject this idea for the type of ergative languages where only transitive subjects get ergative case. If ergative case were really a lexical case associated with agents, then it would not be limited only to transitive clauses. However, note that dative subject languages also divide in a similar manner: there are those that permit dative subjects in intransitive clauses e.g. Icelandic, and there are those that allow dative
Nominalization and ergativity
subjects only in transitive clauses e.g. Japanese. Second, Laka (1993) has argued that intransitive ergative agents in Basque are actually not intransitive, but rather transitive. Absolutive subjects are never associated with the agent role, except in the anti-passive construction. Evidence supporting the above claims comes from the following facts. As shown in (6) and (7), ergative subjects, like dative subjects, cannot appear in constructions with structural accusative objects, but rather they appear with absolutive and nominative objects respectively. Thus in Hindi, a split ergative language, in nominative accusative constructions only the nominative subject triggers agreement while the object cannot: (6) Raam rotii khaataa thaa Ram-masc.nom bread-fem-acc eat imp.masc be-past.masc ‘Ram habitually ate bread’
When the subject is dative or ergative, the object becomes absolutive and triggers agreement: (7) Raam-ne rotii khaayii thii Ram-erg bread-nom-fem eat perf-fem be past-fem
Similar facts can be observed in Icelandic, a language that has constructions with a dative subject and a nominative object. When the subject is dative, the object cannot be accusative (from Woolford 1997: 193–194): (8) Konan þýddi bókina woman-nom translated book-acc ‘The woman translated the book’ (9) Barninu batnaði veikin child-dat recovered from disease-nom (*acc) ‘The child recovered from a disease’
If the ungrammaticality of a structural accusative object in a dative subject construction is due to the fact that dative is a lexical case (as argued for in Yip, Maling and Jackendoff 1987), then this parallelism between dative and ergative subject constructions supports the view that ergative is a lexical case. . v is deficient in ergative languages In the previous section I presented arguments that ergative case is a lexical/ prepositional case and not a structural one. Here I turn to the hypothesis that
Functional structure in nominals
ergative languages do not contain a transitive v by concentrating on Nash’s (1995) proposal. According to Nash, given that ergative case is a lexical case, ergative and accusative languages do not share the same D-structure, or rather they do not look alike at a point prior to syntactic movement, after Merging of arguments has taken place. In accusative languages the subject is projected external to the VP, as the specifier of a functional projection that selects a VP. In ergative languages, the subject is projected VP-internally, as the highest adjunct of the lexical VP projection. (10a and b) below represent the two structures for nominative-accusative and ergative-absolutive languages respectively from Nash (1995, 1996):
(10) a.
vP Su
b. v¹
v
VP agent
VP V
V¹ V
OB
OB
The Internal Ergative Subject Hypothesis argues for the validity of Marantz’s (1984) proposal, according to which ergative and accusative languages have different D-structures (see note 5). However, unlike Marantz, there is no implication in Nash’s approach that objects are differently projected in ergative and accusative languages. In both language types, they are merged as sisters of the lexical verb. I re-interpret Nash’s analysis here as suggesting that ergative languages actually involve a deficient v. Hence the proposal in Nash and the one advanced here for nominalizations share an important common property: they both analyze certain nominalizations and ergative patterns as involving only unaccusative structures with a single theme argument. In other words, the two patterns are reflections of the same structure: one that involves a single theme argument that appears as sister of the lexical root, and an adjunct type of
(11)
AspP Asp° vP v
LP L Theme
Nominalization and ergativity
phrase that introduces the agent. Crucially, both nominalization and ergative clauses contain the partial tree in (11), representing mono-valent constructions, i.e. constructions that lack agents. (11) may then be embedded under D, giving rise to a nominal structure or under T giving rise to an unaccusative/passive/ ergative structure. When agents are included, these appear either as PPs within LP or, in the case of transitive nominalizations as possessors in Spec,DP (for ergative languages see the discussion in Section 4). In what follows I deal with case assignment/checking within nominals, in the sense of locating the functional category responsible for the licensing of the case associated with the single theme argument. I first argue that genitive case in process nominals is not a lexical case, but a structural one much like nominative or absolutive case. I then proceed to identify the projection responsible for its checking.
. Remarks on Case within DPs . Genitive is a structural case Many instances of morphologically case-marked DPs that can occur with [+N] categories, such as dative, genitive and partitive case have traditionally been analyzed as inherent cases. In contrast to V and P, A and N are taken to be the prototypical examples of inherent case assigners. Chomsky (1986) assumes that genitive case is assigned under theta-government to the right of N at D-structure and that subsequently case realization allows for the genitive to be spelled-out in prenominal position through the use of ‘s. However, Corver (1990) points out a number of problems with Chomsky’s treatment. First of all, it is not clear why ‘s is only realized in prenominal position. Second, it is not clear why the of-phrase cannot be realized on both sides of the head noun. Furthermore, as mentioned in Chapter 4, the prenominal in genitive is not necessarily theta-related to the noun. Consider (12) below: (12)
yesterday’s weather
Yesterday’s cannot be argued to be in a thematic relation with the noun. Hence it is not clear how the inherent case assignment mechanism actually works in this case. The assumption that the post-nominal genitive is an inherent case has also
Functional structure in nominals
been challenged by a number of researchers (see de Wit 1997 and references therein). Recent studies suggest that postnominal genitive is the nominal equivalent to accusative case. In this sense genitive is a structural case or at least it shares some characteristics with the structural cases (see the contributions to Alexiadou and Wilder 1998). Here I concentrate on the arguments that specifically show that the genitive case within derived nominals is not an inherent case (see also Alexiadou and Stavrou 1998c for a brief discussion on the basis of Greek data). The most typical instance of an inherent case is one in which a particular case is associated with a particular theta-role or related set of theta-roles. The feature [+genitive] is then linked to a specific interpretation. On the other hand, the structural cases, nominative and accusative, have the characteristic property that there are no thematic restrictions on the elements that bear these cases. For instance, any grammatical subject will get nominative case, regardless of the type of thematic role that it has been assigned and what its base position is. In the preceding sections we have seen that the genitive argument in nominalizations bears the thematic role of theme. Under a configurational theta-theory, this means that any DP being merged as the sister of an unspecified root will have this role. The thematic role in itself says nothing about the morphological case that this DP will surface with at PF. In the view adopted here, nominalizations, as well as verbal clauses, are formed derivationally. Thus an argument for genitive being an inherent case is not easily formulated, since only at PF we have a noun or a verb. Given that the same argument surfaces once with genitive and once with accusative case, depending on the domain it is included in, we cannot maintain the view that themes are always associated with genitives. In other words, we know that when the lexical roots found in nominalizations appear as verbs they do not assign inherent case. Since inherent case is attached, or must be seen as attached to certain roots, and since there is no category distinction between those categories assigning structural and those assigning inherent case, genitive case is not inherent. Thus it seems that the actual morphological affix associated with the theme argument is a more general property of the overall structure. Moreover, if genitive were an inherent case then it would not be immediately clear why the presence of a second genitive, bearing another thematic role, is banned, as the ungrammaticality of (13) shows (and see Chapter 3 for a detailed discussion). (13) *the destruction of the city of the barbarians
Nominalization and ergativity
This fact suggests that only one DP can bear genitive, which in turn means that there is only one head within DPs that bears the feature [assign genitive]. Furthermore, even under a view that takes nominalizations to involve some degree of argument structure change (Williams 1981), if the theme-genitive were inherent, then we would expect this to be constant across clause-types. Inherent case or PPs are never affected by argument structure changing processes. We have seen in Chapter 3 that this is the case in Hungarian. Here I further exemplify this with Russian (from Schoorlemmer 1995: 324): (14) a.
zloupotrebljat’ vlastj’u abuse-inf power-instr b. Zloupotrebljenie vlast’ju abuse power-instr
As discussed in Chapter 1, the impossibility to surface with accusative case correlates with the status of light ‘v’. If ‘v’ is deficient, as is the case in unaccusatives and passives, then ‘v’ lacks an accusative case feature. This is the case also in nominalizations. ‘v’ is deficient; hence no DP can surface with accusative case. Moreover, since the whole structure is embedded under D and not under T, the single DP will not surface with nominative case either. The question that arises next is where the feature [assign genitive] is located, a case feature not related to T and v. Consider the structural configuration proposed for nominalizations. The functional layers contain v, AspP and potentially there is a higher functional projection below D, NumberP or AgrP. With v deficient, there are two options, assuming that D, a C type of projection, can bear no case feature: genitive case is related to Number/Agr, or genitive case is related to Aspect. de Wit (1997) assumes, without further argumentation, a version of the first option, i.e. that the feature [assign genitive] resides within an Agr type of phrase, labeled Possessor Phrase. That something like this might be on the right track is further supported by the Hungarian facts described in Chapter 3, where the single argument bearing structural case surfaces with nominative, i.e. the case of possessors. I come back to this point. Before turning to case assignment in the DP, let me briefly outline my assumptions concerning Case Theory. . Case as a morphological property The theory of abstract Case together with the extended projection principle
Functional structure in nominals
have played a central role in the analysis of the syntax of arguments within the generative literature. However, recently several researchers (see Marantz 1991, Harley 1995, Haeberli 1999 among others) have pointed out that the theory of abstract Case can be eliminated from the Grammar. In particular, in systems such as the one put forth in Marantz (1991), and Harley (1995), case realization is seen as a morphological property of the clause as a whole. Which case is assigned to which DP arguments depends on how many DPs check structural case in the clause. In principle checking of case can proceed in any functional projection, as case is part of the syntactic configuration and is not linked with a specific head in the functional domain. Specifically, Marantz (1991) suggests that morphological realization of Case follows the hierarchy in (15): (15) Case realization disjunctive hierarchy lexically governed case (dative, genitive, e.t.c) dependent case (accusative and ergative) unmarked case (environment sensitive: nominative, absolutive, genitive in DPs) default case
The mechanism regulating which case appears if only one argument is syntactically realized has been given various formulations. For instance, Yip et al. (1987) propose a characterization of clause-bound case assignment depending on the number of NPs that are available for case assignment. In the same vein, Harley (1995) proposes the Mechanical Case Parameter: (16) a.
if one case feature is checked structurally in a clause it is realized as Nominative/Absolutive (mandatory case) b. if two case features are checked structurally in a clause the second is realized as Accusative/Ergative c. the mandatory case in a multiple-case clause is assigned in the top/bottom functional projection
Under this reasoning, if only one case feature is checked structurally within nominals, then morphologically this case would be realized as genitive, irrespectively of where the case features as such are situated. In other words, morphological case is dissociated from a specific structural position (and see Harley 1995, Alexiadou 1999a,b for further discussion). Based on the hypothesis that within destruction type nominals only the theme argument must check structural case, let us examine the possible checkers for it.
Nominalization and ergativity
. Locus of genitive As already mentioned, in principle there are three functional projections that could be related to genitive case: DP, NumberP and AspectP. For Abney (1987) D in (17) dominates the abstract nominal AGR that in languages such as English assigns genitive case to the DPs in Spec, DP.
(17)
DP DP
D¹
the teacher’s
D NP AGR book
But with the exception of phrases in nominalizations, in no other instance is the genitive phrase situated in D. Moreover, D has a mixed status, i.e. in some languages it is an A-position while it is an A¹-position in others. Because of these reasons, I do not consider the possibility of locating the genitive case feature in D, although arguably the morphological realization of this feature is determined by the presence of D, which determines the nominal character of the clause. Number could be argued to be the locus of the genitive case features (see e.g. Valois 1991, Ritter 1991). Specifically, Rouveret (1994) argues that Number has the same function within DP as Tense does within IP in the sense that it determines the reference of the nominal phrase and it licenses the case of the genitive complement. However, it seems rather to be the case that Aspect is the counterpart of Number and not Tense. As argued for in Alexiadou (1999b, 2000), a link between Aspect and Number in DPs can be established on the basis of the observation that the mass vs. count properties of the noun phrase have a direct influence on the aspectual character of a sentence (see e.g. Krifka 1992 among others). For instance, a count noun phrase can refer to something that is specially delimited, while a mass noun phrase describes something of undefined extent or quantity. These thoughts, in combination with the remarks made in Chapter 2 with respect to the role of nominal tense as merely establishing reference, lead to the conclusion that NumberP has a role similar to AspectP, while TP is similar to DP: Tense and D establish reference, while Number and Aspect relate to properties of (de)limitation. There is a further difference between Number and Tense. Although Number is present in all nominal clauses, it only bears a Case feature when the
Functional structure in nominals
structure contains a genitive complement. Tense is present in all propositions, but bears a Case feature only in finite ones. In other words, the case vs. [−case] properties of Tense are determined by its own feature specification. Those of number, on the other hand, are determined by the presence vs. absence of an internal complement. This asymmetry is puzzling. But now note that one could suggest that Number bears genitive case features by virtue of being the nominal counterpart of Aspect. In particular, Aspect has been argued to be responsible for accusative case checking (see Laka 1993), but also for nominative case checking in e.g. intransitive dativenominative constructions in Icelandic (see Taraldsen 1995) e.g. Mer lika bækurnar ekki ‘I-dat like the books-nom not’ (see Alexiadou 1999b, 2000 for discussion and further references). In these constructions the dative argument checks the EPP feature of T, while the nominative arguments checks its Case features in Aspect. Thus Rouveret’s argument for Number could go through, if one equates Aspect with Number. In this sense the symmetry between the verbal and nominal clauses would be complete: in both Case features in clauses where only one argument checks structural case are associated with ‘low’ functional projections. Moreover, the fact that there are nominalizations that lack AspectP seems to suggest that Number might play a role in Case checking. As we have seen in the previous chapter, there is an asymmetry between nominals of type discoverer and of the type destruction in that the former contain Number and lack Aspect, while the latter seem to contain both categories. Both, however, appear together with arguments bearing genitive. Thus it could very well be the case that the genitive within -er nominals at least is located in Number. What happens though when both categories are present, as is the case with most process nominalizations? Is it Number or Aspect that is responsible for Case licensing? It is not clear how one can decide on that. On the one hand, van Hout and Roeper (1998) have argued structural case on the argument DP of the process nominal is checked on Aspect in destruction type nominals. On the other hand, if Aspect and Number are indeed semantically related, it might be impossible to answer the question with respect to the location of Case features in an insightful way. Let me consider the following Hungarian data in an attempt to shed some light into this puzzle. Recall that in Hungarian argumental DPs bear nominative case. The relevant data are repeated below: (18) Péter János által való megver-és-e Peter-nom Janos by being beat-dev-poss-3sg beating up of Peter by Janos
Nominalization and ergativity
Szabolcsi (1994) suggests that the locus of the structural case within DPs is ArgP, given that in Hungarian possessors inflect much like verbal arguments. The existence of such paradigms would make it possible to argue that actually nominative–genitive arguments are associated with Number or AgrP. Consider now the paradigm in detail. In Hungarian possessive inflection is almost identical to verbal inflection, with the following twist: with singular possessors it corresponds to the definite object conjugation, and with plural possessors to the indefinite object conjugation. Interestingly, though, number agreement while total in finite clauses, is only partial in the nominal domain. Full lexical possessors show anti-agreement in number with the possessive affix. More specifically, when the possessor is third person plural, the plural marking is only on the possessor and not on the noun. To account for this pattern den Dikken (1998) proposes the following analysis. Possessive constructions involve a small clause whose head, a dative preposition, takes the possessor as its complement, as in (19). The Possessor can undergo movement to an A-specifier just outside the small clause, as in (20): (19) [SC Possessum [PP Pdat Possessor]] (20) [DP D [AgrP Agr [FP [PP DP] [F¹]]]]
Building on Rouveret (1994), den Dikken argues that the possessor is situated below Agr, in Spec,FP. On this account, the anti-agreement phenomenon observed within the Hungarian DP is due to the fact that the possessor and Agr are not found in a Spec, head configuration. Note that den Dikken’s analysis is intended for possessors, however, it could hold for thematic arguments as well, as there is no difference in the case and agreement patterns in the two cases. I take this analysis to indicate that nominative case within the Hungarian DP, and hence genitive within DPs in general, can also be associated with a lower functional projection, such as Aspect, and not necessarily with NumberP. Clearly the issue awaits further research. Thus far I claimed that the case patterns found in nominalization mimic those found in ergative languages. In what follows I turn to the two patterns in which agents appear in nominalizations and their reflex in ergative, and not only, languages.
. Agents, ergativity and the perfect While discussing nominalizations across languages in Chapters 3 and 4, it was pointed out that agents within such constructions are expressed either by a prepositional phrase within LP or as possessors in Spec,DP, as in (21).
Functional structure in nominals
(21) a. John’s destruction of the city b. the destruction of the city by John
In Section 2 of this chapter I presented Nash’s view that agents in ergative languages are similar to PP agents within nominals of the type (21b). However, this is only one view on the nature of ergative subjects. According to another one (see Mahajan 1997), ergative agents are much like possessors. In fact transitive sentences in Inuit have been argued to be very similar to (21a). Specifically, Johns (1992) proposes that the derivation of the ‘transitive’ Eskimo clause undergoes three stages. At stage I, which is prior to insertion in the syntax, a nominal is lexically derived from a verb. This is the passive nominal, illustrated in (22a) below (cf. Johns 1992: 62f.). At stage II, at D-structure, the derived nominal is the complement of a functional category that takes a subject, and the combination results in a nominal phrase (22b). This is formally a possessive phrase. Finally, at stage III, the nominal phrase is the complement of another functional phrase, which, according to Johns has its own subject in the absolutive, resulting in a matrix transitive clause: (22) a.
kapi-jaq stab-pass.part ‘the stabbed one’ b. anguti-up kapi-ja-a man-erg stab-pass.part.3s ‘the one that the man stabbed’ or ‘the man’s stabbed one’ c. anguti-up nanuq kapi-ja-a-Ø man-erg bear-abs stab.pass.part.3s/3S ‘the man stabbed the bear’ or ‘the bear is the man’s stabbed one’
In Eskimo linguistics, a significant similarity in case and agreement between transitive clauses and possessive phrases has been observed (see Johns 1992: 63f.). First, the case assigned to the possessor is the ergative (relative) case, which is the same case that is assigned to the actor of the transitive construction, as shown in (23) below: (23) a.
Jaani-up taku-ja-a-nga John-erg see-pass.part-3s/1S ‘John saw me’ b. Jaani-up nasa-a John-erg hat-3s ‘John’s hat’
Nominalization and ergativity
Second, the agreement found on the possessed noun in the possessive construction is very similar to the agreement that refers to the actor argument in the transitive construction: (24) a.
taku-ja-ra verb-pass.part-agr ‘I see it’ b. nasa-ra noun-agr ‘my hat’
According to Johns, the role of the patient is linked to the referential index of the nominal and will be assigned to the referential index of the passive nominal. The fact that the role of patient is already used up explains why, unlike the specifier of an English example such as (25) below, the Inuktitut specifier can only be the actor: (25) John’s destruction = John was destroyed (26) Jaani-up taku-ja-a John-erg see-pass.part-3s ‘the one John sees’ *John’s being seen’
The specifier of an English event nominal must be interpreted as the external argument when the internal argument has been assigned, as in (21a) above. According to this view then, transitive sentences in certain ergative languages patterns like transitive nominalizations. In fact, this proposal relates to Allen’s (1964) view of the emergence of transitive constructions through the possessive structure. According to Allen, possessors or possessive case are generally used to express the subject of a transitive sentence, as is the case in the Eskimo examples discussed here. In sum (27) illustrates the configurations that have been argued in the literature on ergative languages to be responsible for the introduction of ergative-subjects. (27) a. Possessor Predicate Theme b. Predicate Theme PP
In the next section I investigate the distribution of the patterns in (27) in some detail. Since in most of the cases my data come from perfect formation across languages (see Abraham 1993, 1998, 2000, Benveniste 1966, Kayne 1993,
Functional structure in nominals
and many others), I first provide a brief overview of the interpretation and the form of the perfect, so that I can refer back to its properties in the course of the discussion. . Background The perfect is rather different from other aspects e.g. perfective vs. imperfective, as it tells nothing directly about the situation itself, but rather relates some state to a preceding situation. According to Comrie (1976b), the two English sentences I have lost my pen and I lost my pen differ in meaning. With the perfect, there is an implication that the pen is still lost, while with the non-perfect there is no such implication. More generally, the perfect indicates the continuing present relevance of a past situation. This difference between the perfect and the non-perfect has led many linguists to doubt whether the perfect should be considered an aspect at all. One way in which the perfect differs from other aspects is in that it partakes of both the present and the past. A sentence such as ‘Mary ate’ says that a certain event took place in the past. The perfect construction ‘Mary has eaten’ says that Mary is now in a certain state, a state of having eaten in the past. ‘have eaten’ is thus true of a resultant state, a state that holds at a given time if and only if the agent in question is the agent of an eating event that culminated earlier than that time (see Comrie 1976b, Parsons 1990 among many others and references therein). As is well known, there are two forms in which the meaning of the perfect is expressed crosslinguistically: analytic and synthetic. For instance, the perfect active in Latin is a synthetic form with an aspectual affix and a specific set of endings varying for person and number (cf. 28a). On the other hand, the present perfect in English is an analytic from consisting of a special auxiliary and a participle (28b): (28) a.
amavi ‘I have loved’ b. I have built this house
Languages that express the perfect analytically show a further split. There are languages, e.g. English, Greek, which use auxiliary HAVE in the formation of the perfect with all types of predicates, transitive and intransitives, unergatives and unaccusatives: (29) a. I have bought the book b. I have sung
Nominalization and ergativity
c.
I have arrived
There are also languages such as Italian or German which use auxiliary HAVE in the formation of the perfect with transitive and unergative verbs, while they use auxiliary BE with unaccusative verbs. These languages are referred to as auxiliary selection languages. In languages such as Italian, the unaccusative participle agrees with the derived subject, while transitive and unergative participles show no agreement. That is participle agreement and the presence of HAVE are in complementary distribution (see Kayne 1993, however, who discusses some exceptions to this). Adjectival inflection with unaccusative predicates was possible in Old English and in Old German (Renate Musan personal communication), as shown in (30d,e), where the participle bears plural inflection. As Burzio (1986) states, one finds both BE and participial agreement in passives, cases of reflexive si and unaccusative verbs. (30) a.
Gianni ha capito la questiona Gianni has understood the question b. Gianni ha telefonato Gianni has called c. Gianni e arrivato Gianni is arrived d. we wæron gecumene we were come e. Er ist gekommen > er ist ein gekommener He is come he is one who came
Finally, there are languages which use auxiliary BE with all types of predicates, transitives and intransitives alike, e.g. Bulgarian, Irish, Hindi and so on. In Bulgarian the perfect participle always agrees with the subject, derived or non-derived. The latter case is illustrated in (31a) from Iatridou &al. 2000. In Hindi ‘transitive’ perfects, the participle agrees with the theme argument (31b) from Mahajan (1997). (31) a.
Maria e pisal-a knigata Maria is write-fem-sg the book ‘Maria has written the book’ b. Raam-ne bhindiiyaa pakaayii he Ram-erg okra.fem.pl cook.perf.fem.pl be-pres.fem.pl ‘Ram has cooked okra’
Functional structure in nominals
In general the above patterns show that whenever the auxiliary is BE the participle is a verb that agrees in the adjectival pattern, while whenever the auxiliary is HAVE the participle is more verbal. I come back to this issue. Before concluding, a few remarks on the morphology of the perfect are necessary. In languages with analytic perfects, usually the special auxiliary carries Tense information and the participle is specified for aspect in the sense of perfective vs. imperfective. For instance, in Greek, the perfect participle is based on the perfective aspect stem, even though Greek verbs show a perfective/ imperfective distinction outside the perfect. (32) a.
o Janis ehi diavasi to vivlio John-nom has read-perf the book-acc ‘John has read the book’ b. o Janis diavase/diavaze to vivlio John-nom read-perf-3sg/imp-3sg the book
In Bulgarian, the perfect participle is based both on the perfective and on the imperfective stem. There is also a participle based on a neutral aspect (see Iatridou et al. 2000 for discussion and references therein): (33) a.
Marija (*vinagi) e obiknala Ivan Maria (*always) is love-perf.part Ivan (*ot 1980 nasam). (*from 1980 towards-now) ‘Maria has fallen in love with Ivan.’ b. Marija vinagi e obicala Ivan. Maria always is love-imperf.part Ivan ‘Maria has always loved Ivan’
Iatridou et al. (2000) point out that since the participle contains (im)perfective specification (as we saw in Bulgarian and Greek) and since Tense is clearly above the perfect morphology, and is expressed by the auxiliary, then the order of the affixes in a tree representation must be as in (34): (34)
Tense [Perfect [Asp [VP]]
(34) seems to be an accurate representation for a number of languages as is clear from Cinque’s (1999) detailed discussion (see also von Stechow 1999). Specifically, Cinque observes that in inflectional languages suffixes typically close off the word disallowing further affixation. As a consequence of that, when more suffixes occur in a sentence more verbs are needed to bear them,
Nominalization and ergativity
typically one for each suffix. In languages such as English and Spanish, where no successive leftward movements of lower parts of the clause take place, the sequence of the various V + suffix combinations may thus be taken to provide direct evidence for the relative order of the corresponding functional heads. If so, a sentence such as (35a) in English and its equivalent (35b) in Spanish, give evidence for the order of functional heads in (36): (35) a. These books have been being read all year b. Esos libros han estado siendo leidos todo el año (36) Tense > AspectPerfect > AspectProgressive > Voice > V
That is perfect participles are at least inflected for Voice and Aspect (much like process nominalizations; see the discussion in the appendix). After this brief introduction, let us take a closer look at the distribution of the patterns in (27) crosslinguistically. . The crosslinguistic distribution of the possessive agent It has been observed that across languages there is a strong preference to use a different marking to express the agent/subject in the perfect, namely a marking usually associated with possessors. For instance, the perfect tense in OldIndo-European languages manifested a pattern, which was described as ergative. Specifically, in Old Persian (cf. Allen 1964, Bok-Bennema 1991 among others) the subject in the perfect was nominative with intransitive verbs, but genitive with transitive verbs, whereas the direct object surfaced with nominative case (data from Bok-Bennema 1991: 19f): (37) a.
Bardiya avajata bardiya-nom killed ‘Bardiya was killed b. ima tya mana kartam Pasrthavaiy this-nom what-nom I-gen done Parthia-loc ‘This is what I have done in Parthia’
In Old Persian genitive is also the case of the agent in the passive construction. Something similar is found in Ancient Greek, in examples such as the following: (38) ta toutoi pepragmena the+these-dat things done ‘lit. Their acts’
Functional structure in nominals
In (38), the agent is expressed by the dative and not as in a normal passive by the preposition hupo. In Latin, constructions of the type illustrated in (39) are documented (from Jacob 1998: 108): (39) quid tibi nos tactio est what you us touched is
These constructions were formed on the basis of an action nominal or an infinitival and had also a deontic meaning associated with them. Again the agent here is introduced with a dative, i.e. a case standardly associated with the possessor in Latin and in Greek. An instance of a possessor subject is manifested in Northern Russian, and particularly in Northern Russian passives (NRPs). Timberlake (1976) describes the following pattern. NRPs are formed from intransitive as well as transitive verbs. The underlying subject is expressed by a prepositional phrase u +genitive, which is actually the form for the possessor in NR. The underlying object is expressed either in the accusative or genitive, or in the nominative. When it appears in the nominative, it can both appear agreeing with the finite verb and without any agreement: (40) a.
U menja bylo telenka zarezno me aux calf-acc slaughtered-part ‘by me there’s been slaughtered a calf ’ b. Pereexano bylo doroga tut crossed-part-n-sg aux road-nom.f-sg ‘there’s been crossing over the road’ c. U rybaka byl sxavacem by the fisherman aux caught.part.m.sg medvezij jazyk bear’s tongue.nom.m.sg. ‘by the fisherman was caught a bear’s tongue’
Timberlake shows that in NRPs the agent phrase has strong subject properties. The underlying object can optionally acquire the subject properties for case and agreement. According to Timberlake, the NRP facts argue for a partial demotion of the subject, and an incomplete promotion of the object. The prepositional subject of this construction behaves actually as a possessive phrase. Related patterns are found in ergative languages that exhibit tense-splits. That is in the perfect subjects appear as possessors. Consider Georgian. In the
Nominalization and ergativity
present Tense the subject of a transitive or an intransitive carries nominative case. The object of the transitive carries dative case. In the aorist, the subject of the transitive carries ergative case. In the perfect, the subject carries dative case (cf. King 1994: 94f.). (41) a.
deda cerils cers mother-nom letter-dat write-3s.3o ‘Mother is writing a letter’ b. dedam cerili daCera mother-erg letter-nom write.3s.3o ‘Mother wrote a letter’ c. dedas cerili dauCeria mother-dat letter-nom write-3s.3io ‘Mother apparently wrote a letter’
Present
Aorist
Perfect
The subject in the perfect is marked with a case standardly associated with experiencers/beneficiaries. According to Mahajan (1997) a similar pattern can be observed in Hindi. In the perfective, Hindi subjects of transitive verbs carry ergative case (42), while in the imperfective Hindi has a nominative-accusative syntax: (42) Raam-ne bhindiiyaa pakaayii he Ram-erg okra.fem.pl cook.perf.fem.pl be-pres.fem.pl ‘Ram has cooked ocra’
To form a regular possessive construction, Hindi uses the genitive or the dative. But according to Mahajan (1997), these cases are related. He claims that ergative, genitive and dative constitute specific choices of morphological form of a lexical preposition and not different items. The presence of ergative in the perfective is tied to the lexical properties of the perfective morphology. Now for both the Hindi and Georgian perfect constructions, it has been argued that the participle cannot assign Case (see Mahajan 1997, King 1994 among others). King (1994) in fact explicitly argues that there is no external argument present in the perfect tense in Georgian. King proposes that these constructions actually involve a dative argument being introduced by a light semi-functional head. This structure is very close to the structure of possessors and that of double object constructions, as analyzed in Marantz (1993), McGinnis (1998), and Anagnostopoulou (to appear). Apart from containing possessor subjects, what all these examples have in common is that they all involve a passive participle (but see Abraham 1998) or
Functional structure in nominals
an action nominal. In the system developed in this monograph the structure of participles and action nominals is closely related (see the appendix). In fact most of the patterns discussed here appear to be similar to transitive nominalizations presented in Chapter 4, where I argued that agents are merged in D. According to Kayne (1993), the structure of perfect participles contains such a projection universally. In fact this head plays a crucial role in restoring transitivity and in auxiliary selection. When D incorporates into the universal auxiliary BE, the result is perfect formation with auxiliary HAVE. The following questions arise. Why do the subjects in the perfect surface as possessors? What are the differences between nominalizations and perfect participles? I turn to the former issue in the next section. As for the latter, see the discussion in the appendix and several remarks made in the following subSections. . Possessor subjects in the perfect As already mentioned, a perfect sentence such ‘I have eaten’ is true of a resultant state, a state that holds at a given time if and only if the agent in question is the agent of an eating event that culminated earlier than that time. According to several authors (see the references above), this is taken to suggest that the perfect is stative in meaning. But in fact the perfect bears the interpretation of a state reached as the result of a prior eventuality. What is of interest here is that it has been pointed out that subjects in the perfect have properties distinct from subjects in non-perfect constructions: namely they appear to pattern on a par with subjects of stative verbs. Before turning to the behavior of perfect subjects, let me note here some of the characteristics of stative subjects. As Kratzer (1994b) recently discusses, such subjects are not true causers. Consider (43): (43)
Mittie owns the dog
In (43) the verb is stative and the external argument denotes, according to Kratzer, the person who is the holder of the state consisting in owning the dog. Kratzer calls this other type of external argument ‘holder’. Under the assumption that external arguments are introduced by ‘light’ heads, to compute the meaning of (43), we need a head adding the external argument, which is different from that introducing the argument of an eventive predicate. In Kratzer’s system one cannot combine the ‘holder’ function with the denotation of an action predicate or the agent function with the denotation of a stative predi-
Nominalization and ergativity
cate. This explains why there is a connection between the Aktionsart of a verb and the thematic role of its external argument. Thus, subjects of stative verbs can be understood as possessors, experiencers and not as true causers. Further evidence for the difference between possessors and causers and their interaction with the various aspectual classes of predicates comes from other language families as well. Kibrik (1993) describes processes of transitivity increase in Athabaskan languages. Apart from causativization, i.e. the introduction of a new agent to the original propositional structure resulting in a formation of a single, though derivationally non-elementary propositional structure, across Athabaskan languages a typological variety of causatives is spread, the socalled possessives. These are derived exclusively from states indicating motionless location or existence and designate possessing the object in a certain state. Consider a Hupa example of a possessive formed from an existential/locative state verb (from Kibrik 1993: 55): Ø-sa- Ø- ʔa:n 3/a-aff-TI-lie ‘It is lying’ b. Ø-s-eh-l-ʔa: 3/g-aff-!sg/A-TI-lie ‘I have it lying there’
(44) a.
In Hupa, the causatives as such and the possessives are in complementary distribution depending on the lexical class of the initial verb: the possessives are derived from the verbs of existential/locative state and the causatives form other types of verbs allowing Actor supplementation. Similar patterns occur in other Athabaskan languages, e.g. Slave and Navajo. There is a great deal of parallelism between these two processes. In the recent literature it has been pointed out that the perfect shares certain characteristics with stative predicates. Specifically, Iatridou et al. (2000) observe the following patterns. As is well known, statives cannot appear in the imperative mood while there is no such restriction with eventives: (45) a. write a book b. *know French
Like statives, the perfect cannot appear in the imperative: (46) a. Write a book b. *have written a book
Functional structure in nominals
Pseudocleft formation is another environment where the perfect patterns on a par with stative predicates. (47) a. what I did was kiss Mary b. *what I did was love Mary c. *what I did was have kissed Mary
Note now that the Georgian perfect shows properties similar to stative predicates in the language. According to Nash (1994), in order to augment their valency, stative verbs in Georgian combine with applicative morphemes. In (48b), the verb is combined with u- that is glossed as BEN/POSS, which signals the presence of a benefactive-possessive DP: (48) a.
Nino tax tze cev-s Nino-nom sofa on lie-pres:3sg ‘Nino lies on the sofa’ b. Am Kac-s svili u-Cev-s saavadmq’oposi this man-dat child-nom ben/poss-lie-pres:3sg hospital-in ‘lit. To this man the child lies in the hospital’
What this shows is that subjects of stative verbs and perfect subjects have a common property: they both surface with a case related to possessors. On the other hand, another set of thematic suffixes is included in the nominative pattern, which includes non-stative verbs and external arguments that behave like causers. Nash (1996) identifies these suffixes as the instantiation of causative v in the language, as they are only present in the nominative pattern (see 48b). However, ergative subjects, i.e. aorist subjects, do not relate to any of these patterns: (49) a.
Gogo-m xe da=xaT-a girl-erg tree-nom prev-draw-aor-sg ‘The girl drew a tree’ b. Gogo xe-s da=xaT-av-s girl-nom tree-obj prev-draw-pres-sg ‘The girl will draw a tree’
The above patterns show the following: first, the semantic difference of stative subjects has a morphological reflex in some languages. Second, the perfect shares common properties with stative predicates. Third, subjects in the perfect in certain languages surface with morphology related to possessors. In
Nominalization and ergativity
other words, in certain languages the morphology of the subject is consistent with its semantic function. This is in agreement with the long-standing intuition in the generative literature that the perfect is in principle a possessive construction. Building on Kratzer and Iatridou et al., I take the perfect to correspond roughly to a passive construction which comes into one’s possession. The head introducing the subject in the perfect could be taken to be a nominal functional head, in a way similar to the type of head introducing the possessor in the case of transitive nominalizations in English. This is actually what King’s proposal about the perfect in Georgian amounts to as well. On this view then, both perfect structures and transitive nominalizations include the internal structure of an action nominal of the type discussed in Chapter 4. Thus it seems that Georgian has a clear reflex of the semantic relations between the various predicates and their respective agent/subject arguments across Tenses: subjects in the perfect surface bearing the possessor case, i.e. they are introduced by a nominal head (pattern 27a). Subjects in the aorist, on the other hand, are deep causers, and thus surface with ergative case (pattern 27b). Finally, external causers introduced by v surface with nominative. That is the three tense series include different internal structures. In turn this suggests that the selectional relationships between different types of Tense/Aspect/v heads and their complements trigger these asymmetries. Recall that in the Hindi perfect, the agent argument surfaces with ergative case. In view of the above remarks, this could mean that its thematic relation to the predicate is similar to the one found in the aorist in Georgian, i.e. ergative subjects are deep agents, as will be suggested in the next section, or else it is an accident of morphology. Note here that the Hindi perfective construction has evolved from constructions in the earlier language that showed inflected agentive forms in initial position, and past participles in concord with non-initial grammatical subjects, in other words, i.e. ‘he wrote the book’ was expressed as ‘by him the book was written’. Certain usages in the Dehli dialect and the Panjab dialect still indicate the historical origin of the construction (source: McGregor, 1995:78). If Hindi perfect agents are deep causers, then one would expect that the semantics of the Hindi perfect relate actually to those of the Georgian aorist/perfective. This might not be so unexpected, as we know independently that there are languages where the perfect has the semantics of the perfective, e.g. Italian (see Giorgi and Pianesi 1997 for extensive discussion). With this in mind, let me turn to a brief discussion of the phenomenon of auxiliary selection.
Functional structure in nominals
. Auxiliary selection . HAVE-BE and the perfect As discussed in the previous section, subjects in the perfect are actually possessors. This fact, if true, brings us to two related claims that have received much attention in the literature: (a) verb HAVE is basically BE + P e.g. ‘I have a book is derived from’ to me there is a book and (b) the use of HAVE in the perfect tense is derived from a BE possessive construction (see Benveniste 1966, Kayne 1993). In the Principles and Parameters literature this link between possession and the emergence of transitive structures is understood in the following sense: auxiliary HAVE restores the transitivity of the predicate (see Hoekstra 1984, Roberts 1987 among others). Others have argued that the selection of a specific auxiliary is correlated with the Aktionsart properties of the predicate, at least in some languages e.g. German (see Grewendorf 1989, Abraham 2000 and references therein). On this view, predicates that denote an uncompleted action take auxiliary HAVE, while predicates that denote a completed action take auxiliary BE. More recently, it has been argued that transitivity is restored in languages that use auxiliary HAVE in the formation of the perfect, and of the possessive construction by means of incorporation of the P that introduces the possessor to the universal copula BE (see e.g. Kayne 1993, Mahajan 1997). According to this view, transitivity is not restored in ergative languages which lack the HAVE possessive construction and only use the BE one, and form the perfect with auxiliary BE only. Let me first consider in some detail Mahajan’s account, as it states a very important generalization. Mahajan has attempted to put the fact that a lexically distinct form of the verb HAVE is generally missing in verb peripheral languages together with the fact that ergative case patterns are found only in peripheral languages, SOV, and VSO, while verb medial languages are never ergative. He offers an explanation of this correlation on the basis of the structures in (50).
(50) a.
AuxP
VP Aux
Aux VP XP
V¹ V
AuxP
b.
YP
XP YP
V¹ V
Nominalization and ergativity
In SOV languages, the preposition, which introduces the subject cannot incorporate into Aux to yield HAVE. According to Mahajan, this accounts for the aspect-split factor, i.e. the fact that in most ergative-split languages it is in perfect where such a split arises and HAVE is also the auxiliary used to form the perfect tense. However, if one were to assume, as I did in the previous section, that ergative subjects in Hindi are located in the complement domain of the lexical head, then incorporation cannot take place. That is the subject is too deeply embedded to incorporate into the auxiliary BE. A further case in point is the perfective passive in Irish. As mentioned above, in Irish the perfect is formed by means of auxiliary BE. Consider the example in (51) from McCloskey (1996): (51) t’a teach ceannaithe agam be house bought by me ‘I have bought a house’
Here the argument corresponding to the subject argument in a transitive clause appears as the object of a preposition, while the argument corresponding to the object in a transitive clause appears as a surface subject. As McCloskey points out, this is actually an unaccusative pattern. Both the progressive and the perfective passive contain auxiliary BE. Mahajan’s account cannot explain the fact that Irish (or Bulgarian for that matter) has auxiliary BE although it is not an ergative, head-final, language. However, such facts follow straightforwardly from the assumption that there are two structures involved in introducing agents in these constructions: one which locates them in the specifier of a nominal head (pattern 27a) and one that locates them inside the lexical projection (pattern 27b). If Irish participles contained ‘deep’ prepositional agents, as is suggested by (51), then we would have an explanation for the fact that Irish uses only auxiliary BE in the formation of periphrastic tenses, since again the subject is too deeply embedded, and cannot incorporate into the auxiliary. Note that Irish uses for the expression of statives and psychological states exactly the same structures used to express possession. The experiencer appears in a prepositional phrase in object position (see Noonan 1993): (52) a.
tá gaeilge ag Fliodhais be Irish at Fliodhais ‘Fliodhais knows Irish’
Functional structure in nominals
b. tá eagla roimh an bpúca ag Ailill be fear before the Puca at Ailill ‘Ailill fears the Puca’
Crucially, the language uses nominal predicate constructions to express psychological states. Hence the underlying structure of both (51) and (52) involves (27b) and not (27a). Let me mow turn to some harder cases: these include auxiliary selection languages, and finally only HAVE languages, which seem to be the hardest case for any account that treats HAVE as a derivative form of BE. . Auxiliary selection languages and only HAVE languages In the previous section I briefly mentioned the derivation of HAVE in accounts that follow Kayne’s analysis of auxiliary selection. Another version of such an account is given in (53) from Iatridou et al. (2000). This is very close to the structure of transitive nominalizations discussed in Chapter 4.
(53)
BEP BE
XP (+N) X PerfP Perf° AspP en VP
In (53) there are two derivations that take place. X is a nominal head. In one possible derivation it may incorporate into BE resulting in HAVE. In such a case, the participle stays where it is, i.e. in Perf° and it does not show agreement. In the second derivation, the participle raises to X and therefore it shows nominal inflection (number, gender). In such a case, the auxiliary remains BE. In languages like English, only the first derivation is possible. In languages like Bulgarian, only the second derivation is possible. Languages in which both derivations are possible are auxiliary-selection languages. For Iatridou et al. X is actually a stativity head that can be seen as introducing the possessor subject, comparable to D in transitive nominalizations, as I mentioned above. However, in this account it is not entirely clear what is the property that forces one or the other derivation in HAVE only or BE only languages.
Nominalization and ergativity
Let me first point out that serious problems for the incorporation analysis arise with only HAVE languages. If a Kaynian type of analysis were on the right track, there would be no explanation for the reason why some languages show up with HAVE also in unaccusative structures. In these, there is no oblique possessor/agent that could incorporate into a higher auxiliary to yield HAVE. At least for these languages then another source from the one discussed so far should be considered. A possibility that suggests itself is that in only HAVE languages, the path followed involved first the formation of the possessive/main verb HAVE, perhaps via P incorporation, as suggested. This item then was used to form the perfect combined with a subordinate (perhaps infinitival or other small) clause. A derivation like this is suggested in Hoekstra (1984), Harris and Campell (1995), and Iatridou (1996) among others. According to Harris and Campbell (1995: 182ff.), Old French and Old English were languages which formed the perfect by means of constructions that had more or less the following meaning: ‘one possesses that which has been done’. Examples of this construction are given below for Old French and Old English respectively: (54) a.
et chis empereres avoit letres seur lui escrites qui and this emperor has letters on him written which b. gyt ge habbaþ eowre heortan geblende yet you have your hearts blinded
The characteristic properties of such constructions are stated as follows. They consist of a possessor subject, the verb HAVE, the possessed object and a clause that included the object and a verb realized as a passive participle. The subordinate clause could be construed with a subject, not necessarily co-referential with that of the main clause. The existence of two subjects each verb is accounted for by positing two clauses, i.e. a biclausal structure. This subordinate/small clause involves actually a Control structure in the transitive variant. That is the meaning of the Old French and Old English examples is (55a) and not (55b) below. Interestingly, Parsons (1990) observes that sequences as in (55a) above are quite similar to the structure of perception verbs and could be analyzed the same way, i.e. as a verb taking a small clause complement (55c). According to Parsons, the source of the modern English perfect lies in the meaning of HAVE that makes me in (55c) responsible for his present state of bindedness in the most direct way.
Functional structure in nominals
(55) a. Sara has the finished report b. To Sara is the finished report c. I saw him bound
Harris and Campbell (1995) also report that in English the perfect HAVE developed out of the possessive HAVE through a process of reanalysis, which in their terms involves a biclausal structure which becomes a monoclausal one. A similar process took place also in German (Renate Musan personal communication.). In other words, the main verb HAVE becomes an auxiliary (see Roberts 1993). In view of the fact that there are languages, e.g. Greek, where very few relics of an auxiliary selection are detected, then perhaps in these languages the perfect developed from structures of this type. In other words, perhaps HAVE is derived from BE, but in the formation of the perfect Tense this is not visible anymore. In fact the perfect construction in Greek seems to have developed from the combination of a main verb HAVE together with an infinitive irrespective of transitivity of the predicate. According to Browning (1969), such formations were common round 1300. Some examples are given below: (56) a.
ehi elthi etote o Mega Kiris have-3sg come-inf then the great Lord-nom (Moreas Chronicle 4365) b. sas exo ipi you have-1sg told-inf (Moreas Chronicle 6773)
However, Horrocks (1997) points out that Browning’s examples are not supported by the manuscript tradition, or by the sense required in the relevant contexts. Horrocks (1997: 228f.) states that the formation of a perfect with HAVE took place in the modern period and that this formation originated from constructions involving the past tense of the verb HAVE in Greek and an aorist infinitival, crucially a counterfactual conditional. This form was later used as the pluperfect and subsequently the perfect form involving the present tense of the verb HAVE was formed. An example illustrating a conditional interpretation is given in (57): (57) ekini an se ihan vri these men if you would have found
This periphrasis came to be interpreted as a hypothetical pluperfect rather than
Nominalization and ergativity
a true modal, through the assumption that it represented a hypothetical pastof-past. From there this periphrasis began to be used in real times past-of-past contexts as a true pluperfect. According to Horrocks, perfect formation was still rare in the vernacular Greek of the 19th century, and was at the time beginning to being used by writers. The only true perfect forms in Medieval Greek were periphrases using the perfect passive participle in combination with first the verb BE and later the verb HAVE, though the latter seems to be a formation that took place after Romance influence. That is the perfect at some stage had a form similar to the English example in (55a). Now, if the origin of the perfect construction, at least in the cases discussed above, was actually a Control infinitival structure, then the lower v was/is not deficient, and hence its accusative case assigning property is not impaired. The only thing that happened was that a lexical verb became a functional one. The development of the perfect in certain Romance languages could also be argued to have followed a similar path. For instance, in languages such as Spanish, it seems to be the case that it developed from a clause embedded under main verb HAVE. This would explain the fact why in Spanish there is no auxiliary selection irrespectively of the transitivity of the predicate. Moreover, in this case main verb HAVE developed from a verb meaning ‘carry’ or ‘hold’. So far I have discussed cases in which the perfect developed from a construction that involved the main verb HAVE. I have also mentioned cases in which the perfect has developed from a passive construction (Hindi). The different derivations correlate with the type of auxiliary used in the respective languages. The question that arises now is why it is the case that we find auxiliary selection languages, i.e. languages where the selection of an auxiliary correlates with the transitivity of the predicate. For instance, French and Spanish are both Romance languages but only French is an auxiliary selection language. I restrict myself in just saying that the HAVE construction in principle could be derived either via incorporation of P to BE or follow a derivation similar to the path followed in English and Greek. As for the use of BE with unaccusatives, this follows straightforwardly from a Kaynian type analysis. I refer the reader to the articles cited in the text for an extensive discussion.
. Conclusion In this chapter I examined two main similarities between nominalization and patterns of ergativity: case and perfect formation. Specifically, I investigated
Functional structure in nominals
the possible sources for the formation of the perfect tense to the extent that these are related to the patterns of nominalization. Second, I discussed briefly the issue of auxiliary selection. The first observation made is that the ‘transitive’ perfect has a semantic interpretation according to which the subject is no longer a causer. This led to the idea that languages are expected to express the perfect subject similar to possessors. As discussed in Chapters 3 and 4 subjects in nominals come in two ways: they are either merged in Spec,DP as possessors or appear within the domain of the root. If one transfers these patterns to perfect formation, then HAVE languages, which result from incorporation, involve possessor subjects, while BE languages involve deep-agents introduced as PPs. The second observation is that in certain cases the perfect is formed by a main verb HAVE and a complement clause. The main verb HAVE is later reanalyzed as an auxiliary. Although these languages also express the perfect as a possession structure, they never make a difference between transitive and intransitive verbs. The third point made is that auxiliary selection languages perhaps make a distinction between an adjectival/resultant state construction and a ‘transitive’resultant state one. The HAVE pattern could either have developed out of a biclausal source or could be the result of P incorporation into BE, as suggested in Kayne (1993) and work inspired by Kayne.
Appendix: Participle formation The literature recognizes that in languages such as English there are three types of participles, illustrated in (58), namely the perfect participle, the passive participle and the so-called adjectival participle: (58) a. I have written three poems b. Three poems were written by me c. Three poems in this book are well-written
These participles show the same morphology as the table in (59) illustrates: (59) Perfect sung written fought
Passive sung written fought
Adjectival sung written fought
Nominalization and ergativity
Lieber (1980) based on this fact proposes that English adjectival passives are formed from verbal participles (passive and perfect participle) by zero affixation. The table in (60) summarizes the properties they have in common, and where they differ: (60)
Perfect Morphology: -en + Stative + (on some accounts) Eventive – (on some accounts) Verbal + Adjectival – Passive meaning – Active meaning +
Passive + – + + – + –
Adjectival + + – – + + –
In what follows, I concentrate mainly on the differences between adjectival and passive participles (see also Embick 2000b). For perfect participles see Iatridou, Anagnostopoulou and Izvorski (2000), and some comments made in the main text and later on here. The standard view on adjectival passives has been that adjectival passives are first of all passive and second of all lexical. For instance, Bresnan (1982) argued that the passive participle only (and not the perfect participle, contra Lieber 1980) constitutes the input to the adjectival passive formation rule. Therefore, the eaten dog means the dog that *was* eaten and not the dog that *has* eaten. Levin and Rappaport (1986) follow Lieber and Bresnan in assuming that there is a category-changing rule of type in (61): V[part] →[V[part]]A
(61)
The authors follow Bresnan in taking the adjectival passive participle to be passive (that is as involving some form of suppression of the external argument). Since Wasow 1977, it has been assumed that adjectival passive participles are lexical. This is expressed in Jackendoff (1977) and Abney (1987) in structural terms as follows. The adjectival passive affix is base generated as a sister of V and makes V into an adjective: Lexical participle=adjectival passive participle
(62)
Z V
Z participle a³x
Functional structure in nominals
On the other hand, Jackendoff and Abney assume that the participle affix of verbal participles adjoins to the whole VP (i.e. verbal participles are phrasal): Verbal Participle = passive participle
(63)
Z VP
Z participle a³x
Recently, Kratzer (1994b) argued in detail that adjectival passives are neither passive nor always lexical. Kratzer argues for a different way of looking at adjectival passives by making two points (embedded within the view that the external argument is not an argument of the verb, Marantz 1984). First of all, as Kratzer points out, adjectival participles are not passive. The view that adjectival passives have passive meaning does not account for the fact that while passives necessarily have an implicit agent, adjectival passives do not (i.e. they are compatible with reflexive actions). This is clearly shown in German where the two types of passive are distinguished through the presence of a different auxiliary, werden vs. sein: (64) a.
das Kind war gekämmt Adjectival Passive the child was combed Stative: Compatible with reflexive action (no agent) b. das Kind wurde gekämmt Verbal Passive the child was combed Eventive: Incompatible with reflexive action (necessarily agent)
Second, adjectival passives are not always lexical, but can also be phrasal. Evidence for the existence of phrasal adjectival passives comes from the availability of manner adverbial distribution. As shown in (65), adverbs may not modify adjectives (98b), but they may modify certain adjectival passives (65a): (65) a.
das Haar war ziemlich schlampig gekämmt the hair was rather sloppily combed b. *das Haar war ziemlich schlampig fettig the hair was rather sloppily greasy
Kratzer proposes that the adverbs modify the verb in a lower VP that is embedded within the adjectival passive morpheme. She further points out that negated adjectival passives are incompatible with manner modification:
Nominalization and ergativity
(66) *das Haar war hässlich ungekämmt the hair was ugly uncombed
The fact that modification is impossible in (66) is taken to mean that the affix cannot attach to a verbal maximal projection, i.e. the participles of the type in (66) are not phrasal. Greek provides morphological evidence in favor of Kratzer’s partition. Negated participles systematically exhibit different morphology than participles without negation, and this correlates with a difference in the availability of manner modification, as shown in (67). (67) a.
to pedhi ine htenismeno askima the child is combed ugly b. to pedhi ine ahtenisto/ *ahtenisto askima the child is uncombed/*uncombed ugly c. gra-meno: written a-graf-to: unwritten ftiag-meno: made a-ftiah-to: unmade diavas-meno: read a-diavas-to: unread
This leads to the view that there are three types of participles (abstracting away from perfect participles): (i) passive, (ii) lexical adjectival and (iii) phrasal adjectival, whose properties are summarized below: (68) Same morphology Eventive Verbal Agent Manner adverbs
Passive + + + + +
Lexical Adjectival + (not in Greek) – – – –
Phrasal Adjectival + – – – +
For Kratzer, phrasal adjectival participles are built by attachment of the adjectival participle affix to the phrasal category. Lexical adjectival participles are built by attachment of the participle affix to the lexical category. Let us see how this distinction translates in the system proposed here, where different types of ‘‘categories’’ are derived from the combination of roots with different kinds of functional heads. Starting from adjectival participles, the
Functional structure in nominals
presence of manner adverbs in the phrasal ones means that light v is present licensing a manner component:
(69)
vP v
LP
L Comp(theme) write This v is an intransitive v (Embick 2000b), and therefore the participle is compatible with a reflexive action. Moreover, the fact that this manner component is realized with an adverb and not with an adjective suggests that an Asp head is also present (see Embick 2000b; see Chapter 4 for arguments that the presence of Asp is crucial for the realization of manner adverbials in the Morphology Component). In fact phrasal adjectival participles include an Asp° specified as [Perfect of Result], thus receiving the structural representation in (70): (70)
AspP Asp
vP v
LP
L write
Comp(theme)
Alexiadou, Anagnostopoulou and Stavrou (2000) suggest, following Kratzer (1994b), that this Asp head encodes the meaning expressed in the Perfect of Result (Comrie 1976b for discussion). They further propose that the perfect of result is not encoded n the meaning of the true perfect which in such a system would be attached higher than Asp and would be compatible with any aspectual specification (progressive, non-progressive, etc.) (see Iatridou, Anagnostopoulou and Izvorski, op.cit., for an analysis of the syntax of perfect). The formula in (104) below, from Kratzer (1994b: 36), gives the semantics associated with this participial form: (71) 1. build=8xe8es [building (e) and exist (x) (f target (e))] 2. PERFECT of RESULT 8P8s s∃es [P(e) and s = ftarget(e) ] 3. (build PERFECT of RESULT) = 8x8s∃e [building (e) and exist (x) (s) & s= f target (e)) and & s= f target (e) ] = 8x8s∃e [building(e) and exist(x)(s) & s= f target (e) ]
Nominalization and ergativity
Kratzer’s analysis translated into this framework takes the presence of v to be responsible for making √build into [building (e)] This in turn means that the stative meaning of adjectival passives does not follow from the category A(djective), but from the semantic information on Asp (cf. Embick 2000b, who argues that the meaning expressed in English by adjectival participles, in Creek it is expressed by verbs). On the other hand, lexical adjectival participles do not include v. Therefore, there is no manner modification that is licensed by v. Moreover, lexical Adjectival Participles do not include Asp. Since there is no v, the meaning of lexical adjectival passives cannot be that of the perfect of Result that requires an event. Negated participles do not have event implications and therefore they are always lexical. Lexical adjectival passives are taken to be bare roots, assuming that roots express always resultant states (Harley and Noyer 1998a).
(72)
LP L
Comp(theme)
This analysis makes a prediction: since lexical adjectival passives are bare roots and therefore stative, it is predicted that double object verbs which, on some accounts cf. for instance Marantz (1993), Collins (1997), McGinnis (1998), Anagnostopoulou (to appear), include a light v introducing the indirect object can form phrasal but no lexical adjectival passives.
(73)
vP Goal
v¹
vAPPL
VP V
Theme
This prediction is borne out in German. Kratzer (1994b) shows that while double object verbs may yield phrasal adjectival passives (74d) they may not yield lexical adjectival passives (74e): (74) a.
weil sie ihm because they him-dat b. *weil sie ihm because they him-dat c. *weil sie ihm because they him-dat
die Antwort the answer-acc die Antwort the answer-acc die Antwort the answer-acc
flüsterten whispered flüsterten whispered zu-flüsterten to-whispered
Functional structure in nominals
d. die Antwort war ihm zugeflüstert the answer was him-dat to-whispered e. *die Antwort war ihm unzugeflüstert
The obligatoriness of the prefix -zu in the example (74c) suggests that it instantiates vAPPL in (74). The ungrammaticality of the form with negation (74e) shows that lexical adjectival passives including vAPPL are impossible. Finally, verbal passives include v [+agent] and Asp. v is Agentive, while Asp can contain all possible specifications similarly to verbs, contra Embick (2000b) who argues that Asp carries only a perfective specification. There is a last point that must be addressed in this discussion and that relates to the agreement properties of perfect participles in auxiliary selection languages and how participles differ from adjectives in their formation. Although participles receive adjectival morphology, this does not mean that they also possess the semantics of the adjective. A soup that is cool does not have to be cooled, but a soup that is cooled requires a cooling event. From the point of view adopted here, it must be the case that adjectives and participles differ in that the former are bare roots, while the latter include certain layers of functional structure. Crucially, the formation of the participle may involve v, and Aspect, while the formation of the adjective does not. While agreement on adjectives is a result of copying/matching operations with a nominal head, agreement on perfect participles is a result of the nominal character of the construction. In cases where the auxiliary is HAVE there is no presence of agreement. Here one must assume that participles behave like non-finite verbal constituents. That is they do not combine with a nominal projection and they do not combine with Tense either. As a result, they are not included in the AGR-set and they surface with no agreement whatsoever, i.e. the default case. This seems to be in agreement with the historical development of the perfect, at least in Greek.
Notes . Although ergativity generally manifests itself in morphological case-systems, the term is also used for languages that have no morphological case, but in which the unique argument in intransitive sentences agrees with the verb in the same way as the theme in transitive sentences and in which the agent of the latter shows another form of agreement (see Marantz 1984, Levin 1985, Bok-Bennema 1991, Murasugi 1992, Dixon 1994, Nash 1995 among others for more details). . Massam (to appear) discusses the consequences of this hypothesis for nominalizations
Nominalization and ergativity
in ergative languages. According to her, one might expect that the nominal case system is identical to the verbal phrase, or one might find a sort of antipassive situation. The latter is what she argues to be the case in Niuean. . Williams (1987) in fact suggests that genitival of in English is an absolutive case marker. . Interestingly, in Archi, an ergative language, nominalizations retain the ergative pattern of finite clauses (cf. Koptjevskaja-Tamm 1993:93): (i) labu armili girman Ralq’-b-amul el we-gen army-erg Germans:cl1:2pl:nom defeat:ac:cl1:2pl:nom ko we:nom learned ‘We learned about the defeat of the Germans by our army’ N/A languages exhibit this pattern only in the domain of nominal syntax. . de Rijk (1966) and Marantz (1984) propose that the projection of arguments are reversed in ergative and accusative languages. Levin and Massam (1984) were the first to outline the proposal that ergativity patterns are related to a parameter in Case theory. They propose the following conditions on Case assignment: (i) a. b. c.
Cx must be assigned Cy (y≠x ) can be assigned under theta-government Case is assigned only under government
The parameter they propose is as follows: (ii) a. b.
x = I (Nominative/Accusative) x= V (Ergative/Absolutive)
The notion of obligatory case assignment is adopted in Bobaljik (1992) (see also Laka 1993, Harley 1995, Fanselow 1998 among others), within the framework of Chomsky (1995). . The details of her analysis can be summarized as follows. In her system INFL assigns exceptional genitive (i.e. ergative) to transitive subjects which move to Spec,IP. Objects raise and adjoin to I’ from where they receive nominative (or absolutive) case. An alternative solution is to express the agent or the theme role as an oblique NP either by passivizing or antipassivizing the clause. Bok-Bennema assumes that accusative case is in principle available in ergative languages. Whether or not this case can be assigned follows from the specific properties of transitive verbs. The proposal in Bittner and Hale (1996) also shares some basic common ground with Bok-Bennema’s analysis. Bittner and Hale (1996) propose a general system of CaseBinding, according to which case assignment in ergative languages works as follows: a transitive verb cannot case-bind the direct object as there are no other elements that compete with it for case. As a result the object moves to INFL to get nominative. In other words, in ergative languages the verbs do not see the nominative case, and hence cannot assign accusative.
Functional structure in nominals
. Nash (1995) further claims that such an analysis of ergativity enables us to capture two related phenomena: the perfectivity of ergative constructions and person-splits. I will not discuss person-splits here, but I will make a few observations with respect to the perfectivity of ergative constructions. In ergative configurations, as Nash observes, the transitive sentence expresses a completed action. This is also a property independently characterizing unaccusative predicates (see the discussion in the previous section). According to Nash, this directly follows from the fact that ergative languages differ from nominative accusative ones in having a deficient v. . Recall further from the discussion of the data in the previous chapters that certain passive nominals in Greek were also interpreted as perfective, i.e. as expression of an action that has been completed. Given the reasoning in Nash (see previous footnote), this can be explained in a similar manner. . A line of argumentation that I will not discuss here is the one taken by a number of researchers (starting with Chomsky 1986) and presented more recently in Harley and Noyer (1998b). According to this, of insertion in English takes place as a last resort, as there is no functional projection within process nominals that could assign case to the theme argument. I suppose that a similar argument could be made for genitive, i.e. genitive case could be analyzed as a last resort case, but I will not go into that here. . Irene Rapp pointed out to me that in German there are several nouns whose complements are preceded by a preposition, which is not selected by the corresponding verb: (i) a. b.
ich wünsche einen Apfel I wish an apple-acc der Wunsch nach einem Apfel the wish after an appl-dat
Rapp takes such cases as illustrating that Case is dependent on the semantics of the noun and it is not structurally assigned. I leave this issue open here. . Haeberli (1999) differs from the Marantz and Harley in that he argues that EPP can also be dispensed with. On the other hand, Marantz and Harley argue that arguments are licensed by (extended) projection and not by abstract Case. . In other systems, such as Wunderlich’s (1997), Case is determined on the basis of the properties of the thematic role borne by a noun phrase, and directly linked to entries in the thematic grid of a verb. . Thanks to Elena Anagnostopoulou and Gisbert Fanselow for making this suggestion. . Another case that seems similar in this context is the necessive construction in Finnish, discussed by Laitinen and Vilkuna (1993). Necessive predicates in Finnish are impersonal and consist of a verb like ‘must’ or the copula and the passive participle of the verb. They express modal concepts, such as necessity and obligation. The single argument bears genitive case, as in (i), but can also bear nominative. Neither has all the properties of the grammatical subject. When a ‘transitive’ structure is used the agent bears genitive case.
Nominalization and ergativity
(i) lehmien pitää tulla kotiin cow-pl:gen must-3sg come-infl home ‘The cows must come home (ii) sinun pitää tuoda lehmät kotiin you-gen must bring-infl cow-pl:nom home According to Laitinen and Vilkuna, when the nominative case is used in the intransitive context, then the NP refers mostly likely to an inanimate entity. The genitive refers to an animate. As far as I understand from the description of the data, such constructions also consist an ergative pattern in an otherwise non-ergative language and reflect an unaccusative system. What is particularly interesting, in view of the discussion in Section 3, is the fact that these environments involve a passive participle and an oblique agent. From this point of view they resemble possessive perfects in Northern Russian discussed in Timberlake (1976). Thanks to Gisbert Fanselow for bringing this pattern to my attention. . These proposals are in agreement with certain views on the development of transitive structures presents itself (cf. Horrocks 1998 for a recent overview). At an early stage, which is preserved in nominalizations in nominative-accusative languages, languages lack a formal expression of an external argument. In other words, the functional projection vP, which introduces the external argument and is responsible for the assignment of accusative case is inactive. One single argument, the single argument of intransitives and the one (theme) argument of transitives, is represented as the internal argument of the verb. The remaining arguments (if any) are represented as optional adjuncts marked semantically for function, e.g. dative for location/experiencer, ‘genitive’ for source/agent. . Many thanks to Elena Anagnostopoulou for bringing Kratzer’s argument to my attention. . Sometimes possessive constructions are translated into English with the help of the verb keep. . Here I follow a suggestion by Sabine Iatridou (personal communication). For Iatridou et al. this nominal head is a stativity head that is located higher than AspPPerfect in (36). . A problem that Mahajan’s account faces is that it cannot explain the fact that there are head-final languages that are auxiliary selection languages, e.g. German. Moreover, Mahajan’s account becomes more problematic, if ones assumes, following Zwart (1993), Kayne (1994) and others, that there are no head final languages. In this case, the reason for the parameter splitting have to be located somewhere else. . Mahajan points out that Basque is a language that has ergative subjects, is head final but still uses HAVE. The existence of such type of languages further strengthens the conclusion that the factors that determine auxiliary selection are rather complex. It might be that in Basque the construction had a derivation similar to the one in English (cf. below in the main text). . Note that this correlates also with the fact that Irish is a language that also lacks the double object construction. Assuming that the correct analysis of this construction is
Functional structure in nominals
one that involves the Goal argument being introduced by a light head, much like possessors, then this suggests that Irish only has the structure where agents are PPs, and possessors appear only as locatives, i.e. again as PPs. That is Irish lacks auxiliary HAVE because, unlike English, it lacks the structure responsible for the introduction of Goal arguments, as proposed in Harley (1995). . Another possibility that suggests itself, at least for some cases (progressive), is that the subject is introduced by a light head, which is actually the copula. Evidence for this comes from the following facts. In the Skye dialect of Scots Gaelic (see also Hendrick 1991, Carnie 1995), subjects of infinitival clauses are only licensed when a light verb be is present: (i)
be thoigh leam sibh/Màiri a bhith a’coiseachd don sgoil c want with 1s you/Maryprt be prt walk to school
(ii) *be thoigh leam sibh/Màiri a’coiseachd don sgoil c want with 1s you/Mary prt walk to school ‘I want you/Mary to walk to school’ The discussion of these constructions and of the perfective construction in the Celtic literature offered in Carnie (1995) concludes that these involve split VPs: a light v, headed by the auxiliary introducing the subject and a participle which involves certain aspectual projections and an object which carries genitive case. Suppose that the participle has a variant of the structure suggested in the appendix, i.e. that it is a lexical root together with an internal theme that is dominated by Aspect. This part of structure combines with a higher light head that introduces the agent. Given that this light head is actually the auxiliary, no incorporation could take place yielding HAVE in this case, since the subject is located in the specifier of the copula. . In the Old English example cited in (30), the participle carried inflection. There are also cases where one finds no inflectional markings on the participle. However, according to Harris and Campell these instances were soon replaced by auxiliary have. Moreover, in Hellenistic Greek there are some cases where the verb to be together with a perfect participle is used to denote the perfect, carrying person and number agreement. Transitive verbs on the other hand form the perfect by using the aorist participle and the auxiliary have. However, as Browning (1969) notes these uses never became general. . Such constructions also exist in modern Greek (cf. Iatridou 1996 for discussion): (i) eho grameno to grama have-1sg written the letter The construction in (i) is quite close to the one of the ordinary perfect construction in Modern Greek, but differs from it in that it involves a different type of participle, one that agrees with the object. According to Iatridou, these constructions can be analyzed as follows: HAVE takes a passive small clause as its complement, as in (ii). The object receives accusative from the verb HAVE, which is the main verb used in this occasion.
Nominalization and ergativity
(ii) eho [to grammai grammeno ti] have-1sg the letter written However, (ii) differs from the possessive construction with HAVE as it lacks a possessive reading. In (ii) there is no thematic relation between the subject of the small clause and the verb HAVE. As Iatridou points out, if there is a small clause below HAVE, the predicate must be stage-level, while the subject of the small clause can be definite or indefinite (iiia). When the verb takes only an NP complement, the latter cannot be indefinite (iiib). When the complement of HAVE is a definite NP (iiib), the sequence means something like ‘I have the car in my temporary custody’: (iii) a.
eho to pedi arosto/*kondo have-1sg the child sick/short ‘I have the child sick/*short’ b. eho to aftokinito have-1sg the car
Iatridou links this interpretation with presence of an unpronounced stage-level predicate, i.e. (iiib) has the interpretation of a small clause with a covert predicate. Iatridou points out then that the complement of HAVE does not show a behavior similar to that of BE in there is . . . environments, but rather it behaves more like when-clauses, in that a definite NP is possible as long as the predicate is a stage-level one. This possibility does not exist for the there is construction. This is used as an argument against the decomposition approach to HAVE. . Note here that this is very close to the two structures that have been proposed in the literature for possessors: possessors can either be locational PPs within a small clause that includes the possessed item, or they can be introduced by a light head (see den Dikken 1997, Alexiadou 1999b for references and discussion). PP/locational possessors are what I label here deep agents/causers. . This section is based on joint work together with Elena Anagnostopoulou and Melita Stavrou.
Chapter 6
Conclusions
This study was devoted to an examination of the internal functional architecture of nominals under the following viewpoint: what we think of as the syntactic categories N, V and A are actually morphological categories created by the syntax, i.e. post-syntactically realized. Certain functional heads play a central role in defining a domain for syntax and for phonological and semantic interpretation. It is syntax that creates the words, and it is syntax that makes sense of the constraints on word formation. My main results can be summarized as follows. Functional projections fully determine the categorial status of lexical projections (Marantz 1997), and not vice versa. The behavior and appearance of verbs/nominals/participles follows from general processes operating in specific syntactic structures, and both are linked to the presence or absence of functional layers and of their feature specification (T, D, Aspect, v). The categorical distinction on the basis of primitive features was replaced by a system where such distinctions are defined by functional heads. Moreover, I showed that nominals/participles split into several types depending on whether they include certain layers of functional structure, e.g. light v and Aspect, Number and D, or not and depending on the feature specification of these projections. Importantly, the study established a typology of derivational processes denoting events in terms of functional architecture. In fact not only do different type of nominals arise depending on the feature specification and presence vs. absence of functional heads, but also different clauses and different language types (Chapter 5). Concerning the distinction between process and result nominals, I argued that while result nominals are inserted directly under nominal projections, event nominals of the destruction type include a set of verbal functional projections, namely AspP and a vP of the type that does not license an external argument (cf. 1). The presence of AspP explains the aspectual properties that these nominals have been argued to possess (cf. Grimshaw 1990). The presence of vP accounts for the eventive reading of these nominals.
Functional structure in nominals
(1)
DP FP (NumbP/AgrP)
D° the AP
FP F° AspectP Aspect¹
Aspect°
vP v
LP
L° Comp (=theme) √DESTROY the city On my proposal, LP is included in the internal structure of result nominals as well, meaning that both result and process nominals can have complements. In other words, the crucial difference between the two groups of nouns lies on the presence of functional layers that are responsible for event interpretation within event/process nouns and the lack thereof from result nominals. A great deal of the discussion in Chapter 4 concentrated on showing how variation dependent on the number and the type of the functional layers given in (1) is responsible for the various nominalization types across languages and within a language. The second goal of my investigation was to pursue the hypothesis that aspects of the syntax of DPs of nominative-accusative languages are strikingly similar to aspects of the syntax of ergative languages. Specifically, after identifying the functional layers which should be assumed within process nominals (Chapter 2), and showing that process nominals are actually intransitive (Chapter 3), I argued in Chapter 5 that the Case patterns in process nominalizations in NA languages of the type the destruction of the city by John mirror those of E languages, in the sense that the nominalization string constitutes an ergative pattern. I proposed that this parallelism can be explained in terms of a common underlying structure. The unifying point was the proposal that both nominalizations in N/A languages and E languages are in fact unaccusative systems. This view was in agreement with one of the views on ergativity that have been proposed in the recent literature. In both contexts, the by-phrase and
Conclusion
the ergative phrase respectively were analyzed on a par as instantiations of deep-causers/instrumental phrases (Chapter 3). Chapter 5 also offered some remarks against the one to one link between projections and case features couched within the framework of Marantz (1991). Furthermore, in Chapter 5 the Inuit transitive construction was related to the transitive nominalization patterns found in English, e.g. John’s destruction of the city. In fact both constructions involving possessor Agents and those involving PP Agents in ergative languages were examined to the extent that they are related to nominalization patterns. Given that such patterns are found in the perfect tense in several languages, it was suggested that perfect formation across languages depends on the means the individual languages have to express the nominal pattern in general. An analysis of ergative splits was suggested, which capitalizes on the variation related to the feature specification of functional heads that build the functional architecture of the clause in each case. In sum, I hope to have shown how the differences and similarities among the various (nominalization) patterns across languages follow from differences in the functional architecture and how nominalized clauses in turn interact with the syntax of the perfect and that of ergative languages. I also hope to have made a convincing case for the claim that there are no differences between word formation and syntax. Evidently, several problems have not been addressed and certain issues and questions arise from my analysis. Pursuing them goes well beyond the scope of this study. Hopefully they will be addressed in further research.
References
Abney, S. 1987. The English Noun Phrase in its Sentential Aspect. Ph.D. dissertation, MIT. Abraham, W. 1993. ‘The aspect-case typology correlation: perfectivity triggering split ergativity’. Folia Linguistica 30/1–2: 5–34. Abraham, W. 1998. ‘Das angebliche Passivstatus des Perfektspartizips’. Groninger Arbeiten zur germanistischen Linguistik, 42. Abraham, W. 2000. ‘Überlegungen zum Passiv im Deutschen und anderen Sprachen ‘Argumenthypothese und Aspekthypothese’. ZASPIL 15: 1–35. Alexiadou, A. 1997. Adverb Placement: a Case Study in Antisymmetric Syntax. Amsterdam: John Benjamins. Alexiadou, A. 1999a. ‘Remarks on the Syntax of Process Nominals: an Ergative Pattern in Nominative Accusative Languages’. In Proceedings of NELS 29, 1–15. Alexiadou, A. 1999b. On the Syntax of Nominalization and Possession: Remarks on Patterns of Ergativity. Habilitationsschrift, Universität Potsdam. Alexiadou, A. 2000. ‘Checking of Nominative Case’. Paper presented at the Workshop on Case at the DGfS Conference in Marburg. Alexiadou, A. and E. Anagnostopoulou. 1997. ‘Notes on ECM, control and raising’. ZAS Papers in Linguistics 8: 17–27. Alexiadou, A. and E. Anagnostopoulou. 1998. ‘Parametrizing Agr: Word Order, Verbmovement and EPP-checking’. Natural Language and Linguistic Theory 16.3: 491–539 Alexiadou, A. and E. Anagnostopoulou. 1999. ‘Tests for Unaccusativity in a Language without Tests for Unaccusativity’. In the Proceedings of the 3rd international conference on Greek Linguistics. Alexiadou, A. and E. Anagnostopoulou. To appear. ‘The Subject in situ Generalization, and the Role of Case in driving Computations’. Linguistic Inquiry. Alexiadou, A., L. Haegeman and M. Stavrou. In progress. Nouns Phrases: a Generative Approach. London: Longam. Alexiadou, A. and M. Stavrou 1998a. ‘On Derived Nominals in Greek’. In B. Joseph, G. Horrocks, and I. Philippaki-Warburton (eds) Themes in Greek Linguistics II. John Benjamins, 101–129. Alexiadou, A. and M. Stavrou 1998b. ‘(A)symmetries in DPs and Clauses: Evidence from Derived Nominals’. The Linguistic Review 15.2–3: 257–276. Alexiadou, A. and M. Stavrou 1998c. ‘Partitives and Pseudopartitives in Greek’. To appear in Studies in Greek Linguistics 19. Alexiadou, A. and M. Stavrou 1999. ‘On Thematic Adjectives’. Ms. University of Potsdam and AUTH.
Functional structure in nominals
Alexiadou, A. and C. Wilder (eds). 1998. Possessors, Predicates and Movement in the DP. Amsterdam: John Benjamins. Alexiadou, A. and C. Wilder. 1998. ‘Adjectival Modification and Multiple Determiners’. In A. Alexiadou and C. Wilder (eds) Possessors, Predicates and Movement in the DP. Amsterdam: John Benjamins, pp. 303–332. Allen, W.S. 1964. ‘Transitivity and Possession’. Language 40: 337–343. Anagnostopoulou, E. 1999. ‘On Experiencers’. In A. Alexiadou, G. Horrocks and M. Stavrou (eds) Studies in Greek Syntax. Dordrecht: Kluwer, pp. 67–93. Anagnostopoulou, E. To appear. Double Object Alternations and Clitics. Berlin: Mouton de Gruyter. Anderson, M. 1977. ‘NP Pre-posing in Noun Phrases’. Proceedings of NELS 8: 12–21. Anderson, M. 1983. ‘Prenominal Genitive NPs’. The Linguistic Review 3: 1–24. Androutsopoulou, A. 1994. ‘The Distribution of the Definite Determiner and the Syntax of Greek DPs’. Proceedings of CLS 30. Aoun, Y. 1981. ‘Parts of Speech: a Case in Redistribution’. In A. Belletti, L. Brandi, and L. Rizzi (eds) Theory of Markedness in Generative Grammar. Scuola Normale Superiore di Pisa, Pisa, pp. 3–24. Arad, M. 1999. ‘On the Nature of v: Evidence from Object Experiencer verbs’. Paper presented at the 22 GLOW Colloquium, Berlin March 1999. Audring, J. 1999. ‘Nominalization and Argument Structure’. Term paper, FU Berlin. Baeuerle, R. 1987. Ereignisse und Repraesentationen. Habilitationschrift, Universität Konstanz. Baker, M. 1988. Incorporation: a Theory of Grammatical Function Changing. Chicago: The University of Chicago Press. Baker, M., O. Jaeggli and I. Roberts. 1989. ‘Passive Arguments raised’. Linguistic Inquiry 20: 219–251. Belletti, A. and L. Rizzi. 1988. ‘Psych-verbs and Theta Theory’. Natural Language and Linguistic Theory 3: 291–352. Bennis, H. and T. Hoekstra. 1989. Generatieve Grammatica. Dordrecht: Foris. Benveniste, E. 1966. Problèmes de linguistique générale. Gallimard. Berman, R. 1978. Modern Hebrew Structure. University Publishing Projects, Tel Aviv. Bernstein, J. 1993. Topics in the Syntax of Nominal Structure across Romance. Ph.D. Dissertation, CUNY. Borsley, R. and J. Kornfilt. 2000. ‘’Mixed Extended Projections’’. In R. Borsley (ed.) The Nature and Function of Syntactic Categories. Academic Press, 101–131. Bever, T. and M. Sanz. 1997. ‘Empty Categories access their Antecedents during Comprehension: Unaccusatives in Spanish’. Linguistic Inquiry 28: 69–91. Bierwisch, M. 1989. ‘Event Nominalizations: Proposals and Problems’. In Grammatische Studien 194. Berlin: Akademie der Wissenschaften der DDR 1–73. Bittner, M. 1994. Case, Scope and Binding. Dordrecht: Kluwer. Bittner, M and K. Hale. 1996. ‘The Structural Determination of Case and Agreement’. Linguistic Inquiry 22: 1–68. Blake, B. 1982. ‘The Absolutive: its Scope in English and Kalkatungu’. In P. Hopper and S. Thompson (eds) Studies in Transitivity, Syntax and Semantics 15. Academic Press, 71–94.
References
Bobaljik, J. 1993. ‘On Ergativity and Ergative Parameters’. MIT Working Papers in Linguistics 19: 45–88. Bobaljik, J. and D. Jonas. 1996. ‘Subject Positions and the Role of TP’. Linguistic Inquiry 27: 195–236. Borer, H. 1984. Parametric Syntax. Dordrecht: Foris. Borer, H. 1993. Parallel Morphology. Ms., University of Massachusetts at Amherst. Borer, H. 1994. ‘The Projection of Arguments’. Ms., University of Massachusetts at Amherst. Borer, H. 1999. The Form, the Forming and the Formation of Nominals. MS., USC. Bottari, P. 1992. ‘Romance ‘Passive Nominals’. Geneva Generative Papers 0.0: 66–80. Bok-Bennema. R. 1991. Case and Agreement in Inuit. Dordrecht: Foris. Bosque, I. and M.C. Picallo. 1996. ‘Postnominal Adjectives in Spanish Indefinites DPs’. Journal of Linguistics 32. Bresnan, J. 1982. ‘The passive in Lexical Theory’. In J. Bresnan (ed.) The Mental Representation of Grammatical Relations. Cambridge: Mass, MIT Press, 3–86. Browning, R. 1969. Medieval and Modern Greek. London: Hutchinson and Co. Ltd. Burton, S. 1997. ‘Past Tense on Nouns as Death, Destruction, and Loss’. Proceedings of NELS 27: 65–78. Burzio, L. 1981. Intransitive Verbs and Italian Auxiliaries. Ph.D. Diss. MIT. Burzio, L. 1986. Italian Syntax. Dordrecht: Reidel. Cardinaletti, A. 1997. ‘Subjects and Clause Structure’. In L. Haegeman (ed.) The New Comparative Syntax. London: Longman. Carnie, A. 1995. Non-verbal Predication and Head-Movement. Ph.D. Diss. MIT. Cetnarowska, B. 1997. ‘The Position of Object Pronominals in Polish Verbal Nominals’. Ms., University of Silesia. Chomsky, N. 1970. ‘Remarks on Nominalization’. In R. Jacobs and P. Rosenbaum (eds) Readings in English Transformational Grammar. Waltham, Mass.: Ginn and Company, pp. 184–221. Chomsky, N. 1972. Studies on Semantics in Generative Grammar. Mouton. Chomsky, N. 1986. Knowledge of Language. New York: Praeger. Chomsky, N. 1995. The Minimalist Program. Cambridge: MIT Press. Chomsky, N. 1998. ‘Minimalist Inquiries: the Framework’. Ms., MIT. Cinque, G. 1980. ‘On extraction from NP in Italian’. Journal of Italian Linguistics 5: 47–99. Cinque, G. 1993. ‘’On the Evidence for Partial N-Movement in the Romance DP’’. University of Venice Working Papers in Linguistics 3.2: 21–40. Cinque, G. 1999. Adverbs and Functional Heads. Oxford: Oxford University Press. Collins, C. 1997. Local Economy. Cambridge, Mass.: MIT Press. Collins, C. and H. Thráinsson. 1996. ‘VP internal Structure and Object Shift in Icelandic’. Linguistic Inquiry 27: 391–344. Comrie, B. 1976a. ‘The Syntax of Action Nominals: A cross-language Study’. Lingua 40: 171–201. Comrie, B. 1976b. Aspect. Cambridge: Cambridge University Press. Comrie, B. 1978. ‘Ergativity’. In W. Lehman (ed.) Syntactic Typology. The Harvester Press limited, pp. 329–394. Cornilescu, A. 1993. ‘Notes on the Structure of the Romanian DP and the assignment of Genitive Case’. Working Papers in Linguistics, Univ. of Venice 3.2: 1–21.
Functional structure in nominals
Corver, N. 1990. The Syntax of Left Branch Extraction. Ph.D. Diss. University of Tilburg. Crisma, P. 1993. ‘On Adjective Placement in Romance and Germanic Event Nominals’. Ms., University of Venice. Crisma, P. 1995. ‘On the Configurational Nature of Adjectival Modification’. Ms., University of Venice. Croft, W. 1993. Syntactic Categories and Grammatical Relations. Chicago: The University of Chicago Press. Demirdache, H. 1996. ‘The chief of the United States’ Sentences in St’át’imcets (Lilloet Salish): a Crosslinguistic Asymmetry in the Temporal Interpretation of Noun Phrases’. Papers for the 31st International Conference on Salish and Neighboring Languages, 79–100. University of British Columbia. Davis, H. 2000. ‘On Nouns and Nominalizations in Salish’. Ms., University of British Columbia. Davis, H. and L. Matthewson. 1997. ‘Determiners, Tense and the Entity/Event Parallel’. To appear in Proceedings of NELS 28. Davis, H. and L. Matthewson. 2000. ‘On the Functional Determination of Lexical Categories’. Ms., University of British Columbia. Demske, U. 1999. ‘Nominalisierungen im Deutschen und Englischen: Überlegungen zu einer Theorie sprachlichen Wandels’. In S. Kanngießer and P. Vogel (Hrsg.) Elemente des Sprachwandels. Westdeutscher Verlag, pp. 98–138. Delsing, L.O. 1988. ‘The Scandinavian Noun Phrase’. Working Papers in Scandinavian Syntax 42: 57–79. Delsing, L.O. 1993. The Internal Structure of the Noun Phrase in the Scandinavian Languages. Ph.D. Diss. University of Lund. den Dikken, M. 1997. ‘The Syntax of Possession and the verb Have’. Lingua 101: 129–150. den Dikken, M. 1998. ‘Anti-agreement in DP’. Ms., CUNY. den Dikken, M. and R. Sybesma 1998. ‘Take serials light up the Middle’ Ms., CUNY and Leiden. Dixon, R. 1994. Ergativity. Cambridge: Cambridge University Press. Doron, E. 1989. ‘Derived Nominals in Hebrew’. To appear in J. Payne (ed.) The Syntax of Noun Phrases. The Hague: Mouton. Doron, E. and M. Rappaport-Hovav. 1991. ‘Affectedness and Externalization’. Proceedings of NELS 21: 81–94. Dowty, D. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel. Dowty, D. 1989. ‘On the Semantic Content of the Notion Thematic Role’. In G. Chierchia, B. Partee, and R. Turner (eds) Properties, Types and Meanings. Dordrecht: Kluwer. Drijkoningen, F. 1993. ‘Movement Theory and the DP-hypothesis’. Linguistics 32: 813–863. Ehrich, V. 1991. ‘Nominalisierungen’. In A. von Stechow and D. Wunderlich (eds) Semantik; Ein Handbuch der internationalen Forschung. Berlin: de Gruyter, pp. 441–458. Embick, D. 1997. Voice morphology and its interfaces with syntax. Ph.D. Diss. University of Pennsylvania. Embick, D. 1998. ‘Syntax and Categories: Verbs and Participles in the Latin Perfect’. Ms., MIT. Embick, D. 2000a. ‘Features, Syntax and Categories in the Latin Perfect’. Linguistic Inquiry 31: 185–230. Embick, D. 2000b. ‘Participial Structures and Participial Asymmetries’. Ms. MIT.
References
Embick, D. 2000c. ‘Unaccusative Syntax and Verbal Alternations’. Ms. MIT Enç, M. 1987. ‘Anchoring Conditions for Tense’, Linguistic Inquiry 18: 633–657. Engelhardt, M. 1999. ‘Nominalizations and Control Theory’. Ms., MIT. Engelhardt, M. and H. Trugman. 1999. ‘Double Genitive Constructions in Russian’. Ms. Hebrew University, Jerusalem and CTEH, Holon. Ernst, T. 1998. ‘Manners and Events’. Ms., Rutgers University. Fanselow, G. 1998. ‘Case and the Structure of Lexical Entries’. Ms., University of Potsdam. Fassi-Fehri, A. 1993. Issues in the Structure of Arabic Clauses and Words. Dordrecht: Kluwer. Fox, D. and Y. Grodzinsky. 1998. ‘Children’s Passive: a View from the by-phrase’. Linguistic Inquiry 29: 311–332. Frampton, J. 1995. ‘Expletive Insertion’, paper presented at the Workshop on the Role of Economy Principles in Linguistic Theory, Berlin February 1995. Fraser, B. 1970. ‘Some Remarks on the Action Nominalization in English’. In R. Jacobs and P. Rosenbaum (eds.), Readings in English Transformational Grammar. Waltham, Mass.: Ginn and Company, pp. 83–98. Fu, J. 1994. On deriving Chinese derived Nominals. Ph. Diss. University of Massachusetts at Amherst. Fu, J., T. Roeper and H. Borer. 1998. ‘The VP within Nominalizations: evidence from Adverbs and the VP Anaphor do-so’. Ms., University of Massachusetts at Amherst. Fugita, K. 1994. ‘Middle, Ergative and Passive in English: a minimalist perspective’. MIT Working Papers in Linguistics 22: 71–90. Fukui, N. and M. Speas. 1986. ‘Specifiers and Projection’. MIT Working Papers 8: 128–172. Garrett, A. 1990 ‘The Origin of NP Split Ergativity’. Language 66: 261–296. Giorgi, A. and G. Longobardi. 1991. The Syntax of Noun Phrases. Cambridge: Cambridge University Press. Giorgi, A. and F. Pianesi. 1997. Tense and Aspect: from Semantics to Morphosyntax. Oxford: Oxford University Press. Giusti, G. 1996. ‘Is there a FocusP and a TopicP in the noun phrase structure?’. University of Venice Working Papers in Linguistics 6: 105–128. Grewendorf, G. 1989. Ergativity in German. Dordrecht: Foris. Grimshaw, J. 1990. Argument Structure. Cambridge, Mass.: MIT Press. Grimshaw, J. 1991. ‘Extended Projection’. Ms., Rutgers University. Grimshaw, J. and L. Selkirk. 1976. ‘Infinitival Noun Phrases in Italian’. Ms., University of Massachusetts, Amherst. de Hackbeil, H.W. 1984. ‘On two Types of derived Nominals’. Proceedings of CLS 20: 308–332. Haeberli, E. 1999. Features, Categories and the Syntax of A-positions: synchronic and diachronic Variation in the Germanic Languages. Thèse de doctorat. Université de Genève. Haegeman, L. and J. Gueron. 1999. English Grammar: A Generative Perspective. Oxford: Blackwell. Haider, H. 1988. ‘Die Struktur der deutschen Nominalphrase’. Zeitschrift für Sprachwissenschaft 7: 32–59. Haider, H. 1992. ‘Die Struktur der Nominalphrase-Lexikalische und funktionale Strukturen’. In L. Hoffmann (ed.) Deutsche Syntax: Ansichten und Aussichten. Berlin: Mouton de Gruyter, 304–333.
Functional structure in nominals
Hale, K. 1970. ‘The Passive and Ergative in Language Change: the Australian case’. In S. Wurm and D. Layscock (eds), Pacific Linguistic Studies in Honor of Arthur Capell. Pacific Linguistic Series, pp. 757–81. Hale, K. and S. Keyser. 1993. ‘Argument Structure’. In K. Hale and S.J. Keyser (eds) The View from Building 20. Cambridge, Mass.: MIT Press, pp. 53–109. Halle, M. and A. Marantz. 1993. ‘Distributed Morphology and the Pieces of Inflection’. In K. Hale and S.J. Keyser (eds) The View from Building 20. Cambridge, Mass.: MIT Press, pp. 111–176. Hamm, F. 1999. Modelltheoretische Untersuchungen zur Semantik von Nominalisierungen. Habilitationschrift, University of Tübingen. Harley, H. 1995. Subjects, Events and Licensing. Ph.D. Diss. MIT. Harley, H and R. Noyer 1998a. ‘Licensing in the non-lexicalist Lexicon: Nominalizations, Vocabulary Items and the Encyclopedia’. MIT Working Papers in Linguistics 32: 119–137. Harley, H. and R. Noyer. 1998b. ‘Mixed Nominalizations, Short Verb Movement and Object Shift in English’. In Proceedings of NELS 28. Harris, J. 1991. ‘The Exponence of Gender in Spanish’. Linguistic Inquiry 27: 27–62. Harris, A. and Cambell, L. 1995. Historical Syntax in cross-linguistic perspective. Cambridge University Press. Hazout, I. 1990. Verbal Nouns: Theta-theoretical Studies in Hebrew and Arabic. Ph.D. Diss. University of Massachusetts at Amherst. Hazout, I. 1992. ‘The Verbal Gerund in Hebrew’. Natural Language and Linguistic Theory 10: 523–553. Hazout, I. 1995. ‘Action Nominalizations and the Lexicalist Hypothesis’. Natural Language and Linguistic Theory 13: 355–404. Hendrick, R. 1991. ‘The Morphosyntax of Aspect’. Lingua 85: 171–210. Higginbotham, J. 1983. ‘Logical Form, Binding and Nominals’. Linguistic Inquiry 14: 395–420. Higginbotham, J. 1985. ‘On Semantics’. Linguistic Inquiry 16: 547–593. Hockett, C. 1958. A Course in Modern Linguistics. New York: MacMillan. Hoekstra, T. 1984. Transitivity: grammatical relations in Government and Binding Theory. Foris. Hoekstra, T. 1994. ‘HAVE as BE plus or minus’. In G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi and R. Zanuttini (eds) Paths Towards Universal Grammar: studies in Honor of Richard S. Kayne. Georgetown Studies in Romance Linguistics, pp. 199–215. Hoekstra, T. 1995. ‘To HAVE to BE Dative’. In H. Haider, S. Olsen and S. Vikner (eds) Studies in Comparative Germanic Syntax. Dordrecht: Kluwer, pp. 119–138. Hoekstra, T. 1999. ‘Parallels between Nominal and Verbal Projections’. In D. Adger, S. Pintzuk, B. Plunkett, and G. Tsoulas (eds) Specifiers: Minimalist Approaches. Oxford: Oxford University Press, 163–187. Holmberg, A. 1987. ‘The Structure of NP in Swedish’. Working Papers in Scandinavian Syntax 33. de Hoop, H. 1992. Case Configuration and Noun Phrase Interpretation. Ph. D. Diss. University of Groningen. Horn, L. 1975. ‘On the Non-sentential Nature of the POSS-ing Construction’. Linguistic Analysis 1: 333–387.
References
Hornstein, N. 1997. ‘S and X’ Convention’. Linguistic Analysis 3: 137–176. Hornstein, N, S. Rosen and J. Uriagereka. 1994. ‘Integrals’. University of Maryland Working Papers in Linguistics 2: 70–90. Horrocks, G. 1997. Greek: a History of the Language and its Speakers. London: Longman. Horrocks, G. 1998. ‘Greek Voices- Indo-European Echoes: toward a unified theory of the ‘‘middle’’’. Paper presented at the Syntax meeting of the University of Thessaloniki, May 1998. Horrocks, G and M. Stavrou. 1987. ‘Bounding Theory and Greek Syntax: Evidence for wh-movement in NP’. Journal of Linguistics 23: 79–108. van Hout, A. and T. Roeper. 1998. ‘Events and Aspectual Structure in Derivational Morphology’. MIT Working Papers in Linguistics 32: 175–220. Jacobs, P. 1994. ‘The Inverse in Squamish’. In T. Givón (ed.) Voice and Inversion. Amsterdam: John Benjamins, pp. 121–146. Jacob, D. 1998. ‘Transitivität, Diathese und Perfekt: zur Enstehung der romanischen habenPeriphrasen’. In H. Geisler and D. Jakob (eds) Transitivität und Diathese im romanischen Sprachen. Tübingen: Niemeyer, pp. 105–126. Jackendoff, R. 1977. X’-Syntax. Cambridge, Mass.: MIT Press. Jackendoff, R. 1990. Semantics and Cognition. Cambridge, Mass.: MIT Press. Jaeggli, O. 1982. Topics in Romance Syntax. Dordrecht: Foris. Jaeggli, O. 1986. ‘Passive’. Linguistic Inquiry 17: 587–663. Jelinek, E. 1993. ‘Ergative Splits and Argument Type’. MIT Working Papers in Linguistics 18: 15–42. Johns, A. 1992. ‘Deriving Ergativity’. Linguistic Inquiry 27: 57–87. Iatridou, S. 1996. ‘To Have and Have not: on the Deconstruction Approach’. Proceedings of WCCFL 14: 185–202. Iatridou, S., Anagnostopoulou, E., and R. Izvorski. 2000. ‘On the Morpho-Syntax of the Perfect and how it relates to its Meaning’. To appear in M. Kenstowicz (ed.) Ken Hale: a Life in Language. MIT Press. Kaiser, G. 1978. ‘Materialen zur Diskussion der Ereignisse’. Papiere der SFB 99, Nr. 31, Universität Konstanz. Karanassios, G. 1992. Syntaxe comparée du group nominale en grec et dans d’autres langues. Doctorat d’Etat Université de Paris VIII, Vincenne. Katz, G. 1999a. ‘Anti neo-Davidsonianism: against a Davidsonian semantics for state sentences’. Ms., University of Tübingen. Katz, G. 1999b. ‘Accounting for the Stative Adverb Gap’. Ms., University of Tübingen. Kayne, R. 1975. French Syntax: the Transformational Cycle. Cambridge, Mass.: MIT Press. Kayne, R. 1984. Connectedness and Binary Branching. Dordrecht: Foris. Kayne, R. 1993. ‘Toward a modular Theory of auxiliary Selection’. Studia Linguistica 47: 3–31. Kayne, R. 1994. The Antisymmetry of Syntax. MIT Press. Keyser, S. and T. Roeper. 1984. ‘On the Middle and Ergative Formation in English’. Linguistic Inquiry 15: 381–416. Kibrik, A. 1995. ‘Transitivity Increase in Athabaskan Languages’. In B. Comrie and M. Polinsky (eds) Causatives and Transitivity. Amsterdam: John Benjamins, pp. 47–67.
Functional structure in nominals
King, T. H. 1994. ‘Spec,AgrP and Case: evidence from Georgian’. MIT Working Papers in Linguistics 22: 91–110. Klein, W. 1994. Time in Language. London: Routledge. Kolliakou, D. 1995. Definites and Possessives in Modern Greek: an HPSG Syntax for Noun Phrases. Ph. Diss. University of Edinburgh. Koptjevskja-Tamm, M. 1993. Nominalizations. London: Routledge. Kornfilt, J. 1998. ‘On the Morphosyntax of Relative Clauses in Certain Turkic Languages: Past and Present’. Paper presented at ZAS, Berlin. Kratzer, A. 1994a. ‘On External Arguments’. University of Massachusetts Occasional Papers 17: 103–130. Kratzer, A. 1994b. The Event Argument and the Semantics of Voice. Ms., University of Massachusetts at Amherst. Krifka, M. 1992. ‘Thematic Relations s Links between Nominal Reference and Temporal Constitution’. In I. Sag and A. Szabolcsi (eds) Lexical Matters. Center of the Study of Language and Information, Stanford. Laenzlinger, C. 1998. Comparative Studies on Word Order Variation. Amsterdam: John Benjamins. Laitinen, L. and M. Vilkuna. 1993. ‘Case-Marking in Necessive Constructions and Split Intransitivity’. In A. Holmberg and U. Nikanne (eds) Case and Other Functional Categories in Finnish Syntax. Berlin: Mouton de Gruyter, pp.23–48. Laka, I. 1993. ‘Unergatives that assign Ergative’. MIT Working Papers in Linguistics 18: 149–172. Lapointe, S. 1979. A Theory of Grammatical Agreement. Ph.D. Diss. MIT. Lasnik, H. 1988. ‘Subjects and the theta-criterion’. Natural Language and Linguistic Theory 6: 1–17. Lasnik, H. 1993. ‘Case and Expletives revisited’. Ms., University of Connecticut. Larson, R. 1988. ‘On the Double Object Construction’. Linguistic Inquiry 19: 335–391. Larson, R. 1998. Lecture Notes. SUNY Larson, R. 1999. Event Modification in Nominals. Paper presented at SALT 8. Larson, R. and G. Segal. 1995. Knowledge of Meaning: an Introduction to Semantic Theory. Cambridge, Mass., MIT Press. Lebaux, D. 1986. ‘The Interpretation of derived Nominals’. Proceedings of CLS 22: 231– 247. Lecarme, J. 1996. ‘Tense in the Nominal System: the Somali DP’. In J. Lecarme, J. Lowenstamm and U. Shlonsky (eds) Studies in Afroasiatic Syntax, Holland Academic Graphics, pp. 159–178. Lecarme, J. 1998. ‘Nominal Tense and Tense Theory’. In F. Corblin, J.-M. Marandin and C. Sorin (eds) Empirical Issues in Formal Syntax and Semantics 2. Peter Lang. Lees, R. 1960. The Grammar of English Nominalizations. The Hague: Mouton. Lefebvre, C. and P. Muysken. 1988. Mixed Categories: Nominalizations n Quechua. Kluwer Academic Publishers. Levin, B. 1985. On the Nature of Ergativity. Ph.D. Diss. MIT. Levin, B. 1993. English Verb Classes and Alternations. Chicago: The University of Chicago Press.
References
Levin, B. 1999. ‘Objecthood: An Event Structure Perspective’, Proceedings of CLS 35, volume 1: The Main Session, Chicago Linguistic Society, University of Chicago, Chicago, IL, 223–247. Levin, B. and D. Massam. 1984. ‘Surface Ergativity: Case/Theta Roles re-examined’. Proceedings of NELS 15. Levin, B. and M. Rappaport. 1992. ‘The Formation of Adjectival Passives’. Linguistic Inquiry 17: 623–662. Levin, B. and M. Rappaport. 1995. Unaccusativity. Cambridge, Mass.: MIT Press. Lieber, R. 1980. On the Organization of the Lexicon. Doctoral Dissertation, MIT. Longobardi, G. 1994. ‘Reference and Proper Names: a Theory of N-Movement in Syntax and LF’. Linguistic Inquiry 25: 609–665. Longobardi, G. 1999. ‘The Structure of DPs: some Principles, Parameters and Problems’. To appear in M. Baltin and C. Collins (eds) Handbook of Syntactic Theory. Oxford: Blackwell. Massam, D. To appear. Niuean Nominalizations. Proceedings of AFLA VII. McCloskey, J. 1996. ‘Subjects and Subject Positions in Irish’. In B. Borsley and Ian Roberts (eds.), The Syntax of Celtic Languages, Cambridge University Press, Cambridge, 241–283. McCloskey, J. 1997. ‘Subjecthood and Subject Positions’. In L. Haegeman (ed.) Elements of Grammar. Dordrecht: Kluwer, pp. 197–235. Mahajan, A. 1990. The A/A’ Distinction and Movement Theory. Ph.D. Diss. MIT. Mahajan, A. 1993. ‘The Ergativity Parameter: have-be alternation’. Proceedings of NELS 24. Mahajan, A. 1997. ‘Universal Grammar and the Typology of ergative Languages’. In A. Alexiadou and T. Alan Hall (eds) Studies on Universal Grammar and Typological Variation. Amsterdam: John Benjamins, pp. 35–57. Marantz, A. 1984. On the Nature of Grammatical Relations. Cambridge, Mass.: MIT Press. Marantz, A. 1991. ‘Case and Licensing’. Proceedings of Escol, 234–253. Marantz, A. 1993. ‘Implications of Asymmetries in Double Object Constructions’. In S. A. Mchombo (ed.) Theoretical Aspects of Bantu Grammar, pp. 113–149. Marantz, A. 1997. ‘No escape from Syntax: Don’t try a morphological analysis in the privacy of you own lexicon’. Ms., MIT. Marantz, A. 1999. ‘Creating words above and below little v’. Ms., MIT. Markantonatou, S. 1992. The Syntax of Modern Greek Noun Phrases with a derived nominal head. Ph.D. Diss. University of Essex. Markantonatou, S. 1995. ‘Modern Greek deverbal nominals: an LTM approach’. Journal of Linguistics 31: 267–299. Markantonatou, S. and B. Oersnes. 1999. ‘Group adjectives’. Paper presented at the syntax meeting of the University of Thessaloniki, January 1999. McGinnis, M.J. 1998. Locality in A-movement. Ph.D.Diss MIT. McGregor, R. 1995. Outline of Hindi Grammar. Oxford: Oxford University Press. Milner, J.C. 1982. Ordres et raisons de langue. Éditions du Seuil. Mouma, E. 1993. ‘On some Properties of DPs in Modern Greek’. UCL Working Papers in Linguistics 5: 75–102. Müller, G. 1998. ‘Parallel Movement’. Ms., University of Stuttgart.
Functional structure in nominals
Murasugi, K. 1992. Crossing and Nested Paths: NP-movement in accusative and ergative languages. Ph.D. Diss, MIT. Musan, R. 1995. On the Temporal Interpretation of Noun Phrases. Ph.D. diss, MIT. Musan, R. 1998. ‘The core Semantics of the Present Perfect’. ZAS Papers in Linguistics 10: 113–145. Nash, L. 1994. ‘On BE and HAVE in Georgian’. MIT Working Papers in Linguistics 22: 153–171. Nash, L. 1995. Portée argumentale et marquage casuel dans les langues SOV et dans les langue ergatives: l’exemple du géorgian. Thèse de Doctorat, Université de Paris VIII. Nash, L. 1996. ‘The Internal Ergative Subject Hypothesis’. Proceedings of NELS 26: 195–210. Nichols, J. 1992. Linguistic Diversity in Space and Time. Chicago: The University of Chicago Press. Noonan, M. 1993. ‘Statives, Perfectives and Accusativity: the importance of being Have’. Proceedings of the West Coast Conference on Formal Linguistics 11: 354–370. Oehrle, R. 1976. The Grammatical Status of the English dative Alternation. Ph.D. Diss. MIT. Ouhalla, J. 1988. The Syntax of head Movement: a study of Berber. Ph.D. Diss. UCL. Ouhalla, J. 1997. ‘Genitive Subjects and the VSO Order’. In A. Alexiadou and T. A. Hall (eds) Studies ion Universal Grammar and Typological Variation. Amsterdam: John Benjamins. Parsons, T. 1990. Events in the Semantics of English. Cambridge, Mass.: MIT Press. Payne, T. 1997. Describing Morphosyntax: a Guide for Field Linguistics. Cambridge: Cambridge University Press. Pesetsky, D. 1995. Zero Syntax. MIT Press. Pesetsky, D. and E. Torrego. 2000. T-to-C Movement: Causes and Consequences’. To appear in M. Kenstowicz (ed.) Ken Hale: a Life in Language. MIT Press. Picallo, M.-C. 1991. ‘Nominals and Nominalizations in Catalan’. Probus 3: 279–316. Pollock, J-Y. 1989. Verb Movement, UG and the Structure of IP, Linguistic Inquiry 20: 365424. Portner, P. 1992. Situation Theory and the Semantics of Propositional Expressions. Ph.D. Diss. University of Massachusetts at Amherst. Postal, P. 1977. ‘Antipassive in French’. Lingvisticae Investigationes 1: 2: 333–374. Pustejovsky, J. 1995. The Generative Lexicon. Cambridge, Mass.: MIT Press. Ralli, A. 1994. ‘Feature representations and feature-passing operations: the case of Greek Nominal inflection’. In the Proceedings of the 8th International Symposium on English and Greek, School of English Dpt of Theoretical and Applied Linguistics. Ralli, A. 1997. ‘Inflectional Features and the Morphological Component Hypothesis’. GLOW Newsletter 38: 46–47. Ralli, A. 1998. ‘On the morphological Status of inflectional Features: Evidence from Modern Greek’. In B. Joseph et al. (eds) Themes in Greek Linguistics. John Benjamins, pp. 57–74. Ralli, A. and M. Stavrou. 1997 ‘‘Morphology-Syntax Interface: two cases of phrasal Compounds’’. To appear in Yearbook of Morphology. Rappaport, M. 1983. ‘On the Nature of Derived Nominals’. In B. Levin, M. Rappaport, and A. Zaenen (eds) Papers in Lexical-Functional Grammar. Indiana University Linguistics Club, pp. 113–142.
References
Reuland, E. and W. Kosmeijer. 1993. ‘Projecting inflecting verbs’. In G. Fanselow (ed.) The Parameterization of Universal Grammar. Amsterdam: John Benjamins, pp. 37–72. van Riemsdijk, H. 1983. ‘The Case of German Adjectives’. In F. Henny and B. Richards (eds), Linguistic Categories: Auxiliaries and Related Puzzles. Dordrecht: Reidel, pp. 223–252. de Rijk, R. 1966. ‘Redefining the Ergative’. Ms., MIT. Ritter, E. 1987. ‘NSO nouns in Modern Hebrew’. Proceedings of NELS 17: 521–537. Ritter, E. 1991. ‘Two Functional Categories in Noun Phrases: Evidence from Modern Hebrew’. In S. Rothstein (ed.) Syntax and Semantics 26, Academic Press, San Diego, pp. 37–62. Ritter, E. 1993. ‘Where’s Gender?’ Linguistic Inquiry 24: 795–803. Rivero, M-L. 1990. ‘The Location of Non-Active Voice in Albanian and Modern Greek’. Linguistic Inquiry 21: 135–146. Rivero, M-L. 1994. ‘Clause Structure and Verb Movement in the Languages of the Balkans’. Natural Language and Linguistic Theory 12: 63–120. Rizzi, L. 1997. ‘On the fine Structure of the left Periphery’. In L. Haegeman (ed.) The New Comparative Syntax. Longman. Roberts, I. 1987. The Representation of Implicit and Dethematized Subjects. Dordrecht: Foris. Roberts, I. 1993. Verbs and Diachronic Syntax. Dordrecht: Kluwer. Roeper, T. 1987. ‘The Syntax of Compound Reference’. Ms., University of Massachusetts at Amherst. Roeper, T. and T. Wasow. 1972. ‘On the Subject of Gerunds’. Foundations of Language 8: 44–61. Rosenbaum, P. 1967. The Grammar of English Predicate Complement Constructions. Cambridge, Mass.: MIT Press. Rothstein, S. 1983. The syntactic forms of predication. Ph.D. diss MIT. Roussou, A. 1991. ‘Nominalized Clauses in the Syntax of Modern Greek’. UCL Working Papers in Linguistics 3: 77–100. Rouveret, A. 1994. Syntaxe du galois. Paris: CNRS Éditions. Rozwadowska, B. 1995. ‘The Duality of Polish -nie/-cie nominals’. In E. Gussmann (ed.) Licensing in Syntax and Phonology. PASE studies and monographs vol. 1. Lublin: Wydawnictwo Folium. Safir, K. 1987. ‘The Syntactic Projection of Lexical Thematic Structure’. Natural Language and Linguistic Theory 5: 561–601. Salvi, G. 1983. ‘L’infinito con l’articolo e la struttura del SN’. Rivista di Grammatica Generativa, 197–225. Schoorlemmer, M. 1995. Participial Passive and Aspect in Russian. Ph.D. Diss. Utrecht University. Schoorlemmer, M. 1998. ‘There is no such a thing as a passive nominal’. Ms., Utrecht University. Schoorlemmer, M and P. de Wit. 1996. ‘Prenominal Arguments in Russian, German and Dutch’. ZAS Papers in Linguistics 5: 184–202. di Sciullo, A. M. and E. Williams. 1987. On the Definition of Word. Cambridge, Mass.: MIT Press.
Functional structure in nominals
Shin, S. 1998. ‘Some Derivational Constraints on Event Nominals: German Word-Formation on the Suffix -ung’. Ms., Seoul National University. Siegel, L. 1997. ‘Gerundive Nominals and the role of Aspect’. To appear in Proceedings of ESCOL. Siewierska, A. 1998. ‘Passive-to Ergative versus inverse to ergative’. In A. Siewierska and J. Song (eds) Case, Typology, and Grammar. John Benjamins, pp. 229–246. Siloni, T. 1997. Noun Phrases and Nominalizations: the Syntax of DPs. Dordrecht: Kluwer. Snyder, W. 1998. ‘On the Aspectual Properties of English Derived Nominals’. MIT Working Papers in Linguistics 25: 125–139. Solà, J. 1992. Agreement and Subjects. Ph.D. Diss. Universitat Autònoma de Barcelona. von Stechow, A. 1999. ‘Das deutsche Perfekt’. Zeitschrift für Literaturwissenschaft und Linguistik 113: 86–118. Stavrou, M. 1995. ‘Epexegesis vs. Apposition’. Scientific Yearbook of the Classics Dept. Aristotle University of Thessaloniki. Stavrou, M. 1999. ‘The Position and Serialization of APs in the DP: Evidence from Greek’. In A. Alexiadou, G. Horrocks and M. Stavrou (eds) Studies in Greek Syntax. Dordrecht: Kluwer, 201–225. Stavrou, M. and G. Horrocks. To appear. ‘Lexeme-based separationist morphology: evidence from Greek deverbal abstracts’. In the Proceedings of the Morphology Meeting in Malta. Stowell, T. 1981. Origins of Phrase Structure. Ph.D. Diss. MIT. Stowell, T. 1982. ‘The Tense of Infinitives’. Linguistic Inquiry 19: 561–570. Szabolcsi, A. 1983. ‘The Possessor that run away from home’. The Linguistic Review 3: 89–102. Szabolcsi, A. 1987. ‘Functional Categories in the Noun Phrase’. In I. Kenesei (ed.), Approaches to Hungarian 2, 167–190- JTE. Szeged. Szabolcsi, A. 1994. ‘The Noun Phrase’. In F. Kiefer and K. Kiss (eds) Syntax and Semantics 27 The Syntactic Structure of Hungarian. Academic Press, pp. 179–274. Taraldsen, K. 1990. ‘D-projections and N-projections in Norwegian’. In J. Mascarò and Marina Nespor (eds) Grammar in Progress. Dordrecht: Foris, pp. 419–431. Taraldsen, K. 1995. ‘On Agreement and Nominative Objects in Icelandic’. In H. Haider, S. Olsen and S. Vikner (eds) Studies in Comparative Germanic Syntax. Dordrecht: Kluwer, pp. 307–327. Tenny, C. 1987. Grammaticalizing Aspect and Affectedness. Ph.D. Diss. MIT. Tenny, C. 1994. Aspectual Roles and the Syntax-Semantics Interface. Dordrecht: Kluwer. Timberlake, A. 1976. ‘Subject Properties in North Russian Passive’. In C. Li (ed.) Subject and Topic. Academic Press, pp. 545–594. Torrego, E. 1989. ‘Experiencers and Raising Verbs in Spanish’. Ms., University of Massachusetts at Boston. Torrego, E. 1998. The Dependencies of Objects. Cambridge, Mass.: MIT Press. Travis, L. 1988. ‘The Syntax of Adverbs’. McGill Working Papers in Linguistics. Special Issue on Comparative Germanic Syntax 280–310. Travis, L. 1991. ‘Inner Aspect and the Structure of VP’. Ms., McGill University. Tsimpli, I.M. 1995. ‘Focussing in Modern Greek’. In K. Kiss (ed.) Discourse Configurational Languages. Oxford: Oxford University Press, pp. 176–206.
References
Valetopoulos, F. 2000. ‘Meleti tis emfanisis kai ton simasiosintaktikon idiotiton ton rimatikon usiastikon vasi somaton kimenon’ (A corpus based study of the properties of derived nominals). MA thesis, AUTH. Valois, D. 1991. ‘The Internal Structure of DP and Adjective Placement in French and English’. Proceedings of NELS 21: 367–382. Vendler, Z. 1967. Linguistics in Philosophy. Ithaca: Cornell University Press. Verkuyl, H. 1989. ‘Aspectual Classes and Aspectual Composition’. Linguistics and Philosophy 12: 39–94. Verkuyl, H. 1993. A theory of Aspectuality: the interaction between temporal and atemporal structure. Cambridge: Cambridge University Press. Wasow, T. 1977. ‘Transformations and the Lexicon’. In P. Culicover, T. Wasow and A. Akmajian (eds) Formal Syntax. New York: Academic Press. Williams, E; 1981. ‘Argument Structure and Morphology’. The Linguistic Review 1: 81–114. Williams, E. 1982. ‘The NP Cycle’. Linguistic Inquiry 13: 277–293. Williams, E. 1985. ‘PRO and subject of NP’. Natural Language and Linguistic Theory 3: 297–315. Williams, E. 1987. ‘The Thematic-structure of derived Nominals’. Proceedings of the Chicago Linguistic Society. Willim, E. 1995. ‘On the Grammar of Polish Nominals’. Ms, Jagiellonian University of Cracow. de Wit, P. 1997. Genitive Case and Genitive Constructions. Ph.D. Diss. University of Utrecht. Wunderlich, D. 1994. ‘Functional Categories’. Ms. University of Düsseldorf. Wunderlich, D. 1997. ‘Cause and the Structure of Verbs’. Linguistic Inquiry 28: 27–68. Yip, M., J. Maling, and R. Jackendoff. 1987. ‘Case in Tiers’. Language 63: 217–250. Zamparelli, R. 1994. ‘Pre-nominal Modifiers, Degree Phrases and the Structure of AP’. Unpublished manuscript, University of Rochester. Zubizarreta, M.L. 1985. ‘The Relation between Morphophonology and Morphosyntax: the Case of Romance Causatives’. Linguistic Inquiry 16: 247–289. Zubizarreta, M.L. 1987. Levels of Representation in the Lexicon and in the Syntax. Dordrecht: Foris. Zucchi, A. 1993. The Language of Propositions and Events. Dordrecht. Kluwer. Zwart, J. 1987. ‘The Nominal Infinitive in Italian’. Ms., University of Groningen. Zwart, J-W. 1993. Dutch Syntax: a minimalist approach. Ph.D. Diss. University of Groningen.
Index adjective 7, 68, 69, 123, 128, 130 agentive 41 aspectual 41, 44, 129 classifying 30 ethnic 91, 103 group 104 intersective 162, 166 thematic 30, 106 possessive 139, 156, 139, 147, 157 predicate 69 referential 91, 98, 103, 104, 105, 107 adverbs 14, 15, 16, 24, 47, 49, 56, 57, 61, 68, 69, 127, 136, 158, 159, 160, 162, 163 Affectedness 92, 93, 94, 95, 96, 97, 98, 100, 101, 122, 166 Agent 7, 11, 12, 17, 18, 21, 37, 38, 40, 54, 55, 77 ff., 88 ff., 102, 103, 108, 110 ff., 124, 125, 130, 138 ff., 150, 152 ff., 160, 163, 165, 166, 168, 170 ff., 182, 185 ff., 195, 200, 201, 204 ff. agreement 29, 68, 69, 169, 171,180, 181, 186, 204, 207, 208 assignment 68, 69 participial 183, 204 phrase 179 Aktionsart 18, 189, 192 Ancient Greek 43, 122, 185 Arabic 27, 77, 88, 163, 165 argument 14, 15, 19, 43, 87, 90, 98, 99, 145, 146, 148, 150, 168, 174, 176, 199 actor 181 agent 40, 54, 81, 113, 188, 124, 125, 154, 173 (see also agent) across Tenses 191 deep Agents 198 in passives 200 dative 178 event 12, 13 external 17, 18, 23, 24, 56, 84, 107, 108, 109, 111, 112, 113, 115, 115, 116, 117, 118, 119, 120, 140, 142, 187 and Aktionsart 189 genitive 38, 39, 42, 71, 141, 161 (see also genitive) goal 44, 45, 114 holder 188 internal 11, 20, 47, 86, 89, 100, 101 104, 105 implicit 110 movement 60, 61
in perfect passives 193 possessor 147 referential 13 structure 9, 10, 37 licensing of 16, 17, 59, 66, 67 change 175 theme 42, 44, 45, 54, 58, 78, 79, 85, 88, 122, 147, 168, 183 movement of 148 in -er nominals 129 Aspect 6, 27, 57, 58, 101, 157, 162, 175, 177, 178, 179, 182, 184, 185, 191, 204, 208, 211 imperfective 50,52, 86, 184 perfective 50,52, 86, 101, 184 phrase 16, 17, 19, 33, 47, 49, 50, 52, 57, 59, 61, 63, 66, 67, 71, 72, 115, 121, 122, 128, 130, 131, 133, 165, 177, 202, 204 Aspectual Hypothesis 42 aspectual 49, 115, 134 durative 134 frequency 48 interval denoting 48 manner 48, 49, 58, 87, 113, 115, 129, 130, 134, 200, 201, 202 modal 48 temporal 48, 71 speaker-oriented 48 Atomicity Thesis 3 auxiliary selection 183, 188, 191, 192, 194, 196, 197, 198, 204, 207 Berber 164 Bulgarian 183, 184, 193, 194 Burzio’s Generalization 18, 114, 124, 140, 142 Case: absolutive 167, 169 binding 205 ergative 169,192 as a lexical case 169, 170, 171 features 17 genitive: as a structural case 173, 174, 175 inherent 138, 140, 174 as a last resort 206 parameter 205 possessor 191 realization disjunction hierarchy 176 Catalan 13, 46, 71, 73, 77, 81, 82, 83, 84, 106, 120, 122, 151, 152
Index
Chinese 48, 165 constant 3, 67, 78, 104, 129, 152, 175 control 12, 98, 101, 105, 111, 195, 197 dative alternation 43 determiner: definite 65 in gerunds 136 in infinitives 158 phrase 1, 28, 29 A vs. A¹ status of 31, 32, 155, 156 spreading 107 double object verbs 43, 44, 71, 101, 114, 187, 203, 207 ergativity 1, 19, 21, 25, 167, 168, 169, 179, 197, 204, 205, 206, 212 features 3, 4, 6, 8, 17, 19, 20, 22, 23, 34, 55, 60, 67, 69, 71, 74, 75, 113, 119, 123, 130, 162, 165, 176, 177, 178, 211, 213 morphological 6 syntactic-semantic 6 Finnish 206 French 13, 14, 64, 77, 83, 91, 97, 103, 104, 124, 146, 151, 189, 195, 197
inflection class 70 instrument 77, 125, 154 Internal Ergative Subject Hypothesis 172 Inuit 51, 170, 180, 213 Irish 141, 142, 161, 183, 193, 207, 208 Italian 23, 30, 35, 64, 71, 77, 83, 84, 85, 94, 103, 116, 117, 120, 123, 139, 146, 158, 159, 161, 183, 191 Lexical Integrity Hypothesis 3 lexicon 3, 4, 5, 6, 8, 23 Maya 162 Mechanical Case Parameter 176 Mirror Principle 65 modifier 12 agent-oriented 11 aspectual 11, 39, 82, 85, 99, 106, 129, 150 punctual 99 extensional vs. intensional 73, 149 Morphological Structure 6, 67 processes in 67, 68, 69
Halkomelem 64, 65 Hebrew 15, 16, 24, 27, 29, 48, 52, 77, 81, 87, 88, 110, 111, 117, 120, 137 ff., 163, 164, 165 Hindi 171, 183, 187, 191, 193, 197 Hixkaryana 110 Hungarian 23, 29, 32, 33, 77, 89, 90, 103, 155, 156, 175, 178, 179 Hupa 189
nominalization 1, 3, 5, 7, 8, 15 ff., 42, 52, 54 ff., 66, 71, 73, 77, 83, 107 ff., 118, 120, 121, 124, 127, 128, 140, 162, 167, 173, 179, 197, 198, 212, 213 mixed 1, 3 transitive 80, 146, 154, 156, 157, 191, 194 nominals: accusative case in 15, 16, 138, 139 adverbs in 15, 16, 47, 48, 49 (see also adverbs) as ergative contructions 112, 167 ECM in 59, 60, 61 -er 24, 128, 129, 130 event 9, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89 complex 9 expletives in 30, 60 focus movement in 32, 33, 155 passive 52, 90, 91, 92, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 109, 110, 166, 181 plural 72 process 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 111, 166 VP node in 14, 15, 59, 73 raising in 30, 60 result 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 tense in 59, 60, 61, 62, 63, 64, 65, 66, 71 wh-movement in 32, 33, 155 number phrase 29, 131, 137, 177, 17
Icelandic 80, 139, 170, 171, 178 incorporation noun 165 preposition 5, 192, 193, 195, 197, 208
Old English 183, 195, 208 Old German 183 Old French 195 Old Persian 185
gender phrase 29 gender 6, 29, 33, 34, 36, 67, 68, 70, 162, 194 genitive 32, 34, 36, 38 ff., 45, 46, 54, 59, 71, 77 ff., 96, 101 ff., 114, 117, 118, 120, 122, 123, 135 ff., 155, 160 ff., 185 ff., 205 ff. Georgian 186, 187, 190, 191 German 23, 51, 64, 69, 71, 72, 77, 80, 81, 103, 105, 116, 120, 122, 125, 139, 142, 144, 145, 146, 154, 158, 160, 164, 183, 192, 196, 200, 203, 206, 207 gerund 1, 2, 22, 135, 136, 161, 164 Greek 15, 16, 21 ff., 27, 30 ff., 64, 70 ff., 80, 81, 84 ff., 91, 94 ff., 114, 117, 120 ff., 127, 128, 129, 133, 134, 137, 144 ff., 155, 156, 157, 162, 163, 165, 166, 170, 174, 182, 184, 185, 186, 196, 197, 201, 204, 206, 208
Index
passive 18, 22 ff., 39, 42, 50, 52, 71, 73, 83, 84, 85, 87, 105, 107, 108, 168, 171, 173, 185, 186, 187, 191, 195, 197, 198 (see also passive nominals) adjectival 4, 199, 199, 200, 201, 206, 207, 208 agent 119 participle 131 perfective 193, verbal 4, 107, 108, 115, 116, 117, 118, 199, 200, 201, 206, 207, 208 perfect 22, 50, 71, 170, 179, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 195, 196, 197, 198, 199, 201, 202, 203, 204, 208, 213 Polish 50, 77, 86, 87, 131, 132, 134 possessive constructions 179, 207 possessor subject 186, 194, 195 predicate 2, 12, 18, 23, 85, 93, 100, 102, 108, 113, 114, 119 ff., 128, 133, 134, 136, 144, 146, 152, 153, 156, 160, 182, 183, 189, 190 ff., 206 ditransitive 42, 43 psychological 45, 46, 47 unaccusative 40, 41, 42, 78 putative 141, 142 salient 141, 142 transitive 38, 39 unergative 40, 41, 42
Projection Principle 60, 73, 106, 175 root 7, 8, 23, 37, 55, 56, 58, 63, 67, 69, 75, 100, 102, 113, 116, 118, 127, 130, 136, 154, 162, 167, 172, 174, 198, 208 Russian 24, 77, 85, 86, 87, 117, 124, 125, 131, 132, 134, 152, 175, 186, 207 Small clause 179, 195, 208, 209 Somali 64, 65 Spanish 34, 64, 70, 77, 83, 96, 97, 117, 118, 121, 122, 139, 143, 164, 165, 166, 185, 197 Squamish 143, 165 Tense 6, 8, 19, 25, 29, 33, 59, 60, 62, 63, 64, 65, 68, 69, 71, 75, 128, 144, 177, 178, 184, 185, 186, 187, 191, 192, 193, 196, 198, 204, 213 chain 61 Theme 2, 6, 24, 39, 41, 42, 44, 45, 46, 54, 77 ff., 84 ff., 95, 96, 98, 101, 102, 114, 118 ff., 125, 139 ff., 145 ff., 150, 154, 160, 164, 165, 168, 172 ff., 181, 183, 204 ff., (see also argument) Tonga 167 Turkish 50, 51, 103, 124, 128, 164 Voice Phrase 27, 33, 49
In the series LINGUISTIK AKTUELL/LINGUISTICS TODAY (LA) the following titles have been published thus far, or are scheduled for publication: 1. KLAPPENBACH, Ruth (1911-1977): Studien zur Modernen Deutschen Lexikographie. Auswahl aus den Lexikographischen Arbeiten von Ruth Klappenbach, erweitert um drei Beiträge von Helene Malige-Klappenbach. 1980. 2. EHLICH, Konrad & Jochen REHBEIN: Augenkommunikation. Methodenreflexion und Beispielanalyse. 1982. 3. ABRAHAM, Werner (ed.): On the Formal Syntax of the Westgermania. Papers from the 3rd Groningen Grammar Talks (3e Groninger Grammatikgespräche), Groningen, January 1981. 1983. 4. ABRAHAM, Werner & Sjaak De MEIJ (eds): Topic, Focus and Configurationality. Papers from the 6th Groningen Grammar Talks, Groningen, 1984. 1986. 5. GREWENDORF, Günther and Wolfgang STERNEFELD (eds): Scrambling and Barriers. 1990. 6. BHATT, Christa, Elisabeth LÖBEL and Claudia SCHMIDT (eds): Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences. 1989. 7. ÅFARLI, Tor A.: The Syntax of Norwegian Passive Constructions. 1992. 8. FANSELOW, Gisbert (ed.): The Parametrization of Universal Grammar. 1993. 9. GELDEREN, Elly van: The Rise of Functional Categories. 1993. 10. CINQUE, Guglielmo and Guiliana GIUSTI (eds): Advances in Roumanian Linguistics. 1995. 11. LUTZ, Uli and Jürgen PAFEL (eds): On Extraction and Extraposition in German. 1995. 12. ABRAHAM, W., S. EPSTEIN, H. THRÁINSSON and C.J.W. ZWART (eds): Minimal Ideas. Linguistic studies in the minimalist framework. 1996. 13. ALEXIADOU Artemis and T. Alan HALL (eds): Studies on Universal Grammar and Typological Variation. 1997. 14. ANAGNOSTOPOULOU, Elena, Henk VAN RIEMSDIJK and Frans ZWARTS (eds): Materials on Left Dislocation. 1997. 15. ROHRBACHER, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and pro-drop. 1999. 16. LIU, FENG-HSI: Scope and Specificity. 1997. 17. BEERMAN, Dorothee, David LEBLANC and Henk van RIEMSDIJK (eds): Rightward Movement. 1997. 18. ALEXIADOU, Artemis: Adverb Placement. A case study in antisymmetric syntax. 1997. 19. JOSEFSSON, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998. 20. LAENZLINGER, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and clause structure in Romance and Germanic. 1998. 21. KLEIN, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. 22. ALEXIADOU, Artemis and Chris WILDER (eds): Possessors, Predicates and Movement in the Determiner Phrase. 1998. 23. GIANNAKIDOU, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. 24. REBUSCHI, Georges and Laurice TULLER (eds): The Grammar of Focus. 1999. 25. FELSER, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions. 1999. 26. ACKEMA, Peter: Issues in Morphosyntax. 1999.
° 27. RUZICKA, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. 28. HERMANS, Ben and Marc van OOSTENDORP (eds.): The Derivational Residue in Phonological Optimality Theory. 1999. 29. MIYAMOTO, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 1999. 30. BEUKEMA, Frits and Marcel den DIKKEN (eds.): Clitic Phenomena in European Languages. 2000. 31. SVENONIUS, Peter (ed.): The Derivation of VO and OV. 2000. 32. ALEXIADOU, Artemis, Paul LAW, André MEINUNGER and Chris WILDER (eds.): The Syntax of Relative Clauses. 2000. 33. PUSKÁS, Genoveva: Word Order in Hungarian. The syntax of È-positions. 2000. 34. REULAND, Eric (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. 35. HRÓARSDÓTTIR, Thorbjörg. Word Order Change in Icelandic. From OV to VO. 2000. 36. GERLACH, Birgit and Janet GRIJZENHOUT (eds.): Clitics in Phonology, Morphology and Syntax. 2000. 37. LUTZ, Uli, Gereon MÜLLER and Arnim von STECHOW (eds.): Wh-Scope Marking. 2000. 38. MEINUNGER, André: Syntactic Aspects of Topic and Comment. 2000. 39. GELDEREN, Elly van: A History of English Reflexive Pronouns. Person, ‘‘Self’’, and Interpretability. 2000. 40. HOEKSEMA, Jack, Hotze RULLMANN, Victor SANCHEZ-VALENCIA and Ton van der WOUDEN (eds.): Perspectives on Negation and Polarity Items. 2001. 41. ZELLER, Jochen : Particle Verbs and Local Domains. n.y.p. 42. ALEXIADOU, Artemis : Functional Structure in Nominals. Nominalization and ergativity. 2001. 43. FEATHERSTON, Sam: Empty Categories in Sentence Processing. 2001. 44. TAYLAN, Eser E. (ed.): The Verb in Turkish. n.y.p. 45. ABRAHAM, Werner and C. Jan-Wouter ZWART (eds.): Issues in Formal German(ic) Typology. n.y.p 46. PANAGIOTIDIS, Phoevos: Pronouns, clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. n.y.p.
