bowers- non-event nominals and argument structure
TRANSCRIPT
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
1/13
Non-event nominals and argument structure
John Bowers
Cornell University, Department of Linguistics, 203 Morrill Hall, Ithaca, NY 14853-4701, United States
1. Introduction
According to the standard theory of argument structure, argument DPs merge directly with lexical items of categories V,
N, etc. to form newsyntactic objects of thesame category. Bowers (2004, 2010a,b) proposes an alternative theory in which all
argument DPsare merged in the specifiers of light verb categories of different types. I argue in this paper that a theory of this
kind is given direct support by the existence of a range of non-event nominals in English whose morphosyntactic and
semantic properties correspond precisely to those of the basic argument categories assumed by the theory. The proposed
analysis, if correct, extends to derivational morphology the claim ofDistributed Morphology (DM) (Halle and Marantz, 1993)
that word formation is syntactic, not lexical (Marantz, 1997; Embick, 2004; Harley, 2009), thereby making syntax the onlygenerative component of the grammar.
2. Background
I first spell out explicitly the theory of argument structure assumed in this paper and illustrate how it works by deriving
active/passive pairs in English. I then set thestage for my analysis of nominals by discussing very brieflysome of the evidence
supporting the claim that so-called grammatical function changing (GFC) morphology is associated in the syntax with
particular argument heads.
Lingua 121 (2011) 11941206
A R T I C L E I N F O
Article history:
Received 1 April 2008
Received in revised form 2 August 2009
Accepted 20 January 2011
Available online 17 March 2011
Keywords:
Argument category
Derivational morphology
Distributed Morphology
Noun incorporation
Agentive nominal
Event nominal
A B S T R A C T
Bowers (2010a) proposes that all argument DPs are introduced in the specifiers of light
verb categories of different types. It is argued in this paper that such a theory is given
direct support by the existence of three types of non-event nominals in English whose
morphosyntactic and semantic properties correspond precisely to those of the basic
argument categories Ag(ent), Th(eme), and Aff(ectee). Assuming that these argument
categories are spelled outin derived nominal structures as the morphemes er/-or, -ment/-
ion/-ure, -ee, respectively, each type of non-event nominal can be derived by standard
syntactic processes such as head movement. The proposed analysis, if correct, extends to
derivational morphology the claim of Distributed Morphology (Halle and Marantz, 1993)
that word formation is syntactic, not lexical (Marantz, 1997; Embick, 2004; Harley, 2009),
thereby makingsyntax the only generative component of the grammar. It is further shown
that if noun incorporation is syntactic, then productive compounds such as deer hunter,
painting consignor, etc. can be derived syntactically as well and that the order of
incorporated nouns in such structures mirrors precisely the order in which argument
categories are merged.
2011 Elsevier B.V. All rights reserved.
A list of abbreviations is provided at the end of this article.
E-mail address: [email protected].
Contents lists available at ScienceDirect
Lingua
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / l i n g u a
0024-3841/$ see front matter 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.lingua.2011.01.007
http://dx.doi.org/10.1016/j.lingua.2011.01.007mailto:[email protected]://www.sciencedirect.com/science/journal/00243841http://www.elsevier.com/locate/linguahttp://dx.doi.org/10.1016/j.lingua.2011.01.007http://dx.doi.org/10.1016/j.lingua.2011.01.007http://www.elsevier.com/locate/linguahttp://www.sciencedirect.com/science/journal/00243841mailto:[email protected]://dx.doi.org/10.1016/j.lingua.2011.01.007 -
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
2/13
2.1. Theoretical assumptions
As stated above, all arguments are introduced above the root in Spec of light verb categories.1 There are three primary
argument categoriesAg(ent), Th(eme), andAff(ectee).2 Ag-arguments include subjects of transitive and unergative verbs; Th-
arguments include objects of transitive verbs and subjects of unaccusative verbs; and Aff-arguments include prototypically
animate dative arguments of various kinds. In addition, there are a number of secondary argument categories such as Goal,
Sou(rce), Ben(efactive), Instr(umental), and so forth, and I assume, following Cinque (1999), Larson (2004), and others, that
modifiers and adjuncts of various kinds are also introduced in the specifiers of functional categories. Argument categories, as
well as standard functional categories such as C, T, Pr,3 Voi,4 etc., merge with a root or with the output of previous merge
operations in an order determined by the following universal constraint5:
(1) Universal Order of Merge (UOM):
Ag< Ben< Goal< Th< Aff< Voi< Pr< T< C
The arguments required by a given root are determined by a(rgument)-selection features (e.g. [Ag], [Th], etc.) which are
checked and deleted when the root raises and adjoins to the head of the selected category. The argument categories
themselves, which are listed in the lexicon and selected for the Numeration, have c-selection features of the standard sort
which are satisfied by merging a phrase of the required category in Spec of the a-selected category. 6 An argument category
may also assign inherentCase to the DP in its Spec or require that it be contained in a PP headed by a particular preposition. In
English, for example, the argument category Ag may c-select a PP headed by by. Similarly, the category Aff may c-select a PP
headed by to orfor, depending on the particular verb.
7
The argument category Th, in contrast, does not select PP, but insteadhas the option of assigning inherent ACC case (whose phonetic realization is identical to that of structural ACC Case in English)
to a DP in its specifier. As an alternative to assigning inherent case or selecting a PP, a category head frequently has the option
of selecting a DP with an unvalued Case feature which must be valued through an Agree relation. In English all three primary
argument categories Ag, Th, and Aff have the option of receiving structural Case.
Subject and object relations arise solely from the operation of the Agree relation, defined in terms of the notions probe
and goal. A probe is a set of uninterpretablef-features that are valued and deleted by establishing an Agree relation witha goal containing matching interpretable f-features and an uninterpretable structural Case feature, which is also valuedand deleted by the Agree operation. Only two probes are available, one in T and one in Voi(ce), which assign structural
NOM and structural ACC Case, respectively. In English, T and Voi also contain an uninterpretable c-selection feature (the
so-called EPP feature), which can only be satisfied by internal merge or by merge of an expletive element with no
interpretation. Following Chomsky (2000, 2001), Move = Agree + EPP is a composite operation in the sense that its EPP
feature must be satisfied as soon as the Agree relation is established. However, I follow Collins (1997), Bowers (2002a),
and others, in assuming that an EPP feature associated with Agree does not necessarily have to be satisfied by moving toits specifier the same constituent with which the probe establishes the Agree relation. Hence, the EPP feature of a head H
can also be satisfied by moving the closest constituent of the required category in the domain of H or by merging an
expletive with H.
1 Broadly similar notions are explored in Borer (2005: Vol. 2, Chapter 2), and Ramchand (2008), though the details of their approaches are quite different
from that proposed here.2 For convenience, Iuse the terms Ag, Th, etc. as abbreviations for light verbs v with therespective features [Ag], [Th], etc. It should be emphasized that
these are syntactic categories, not -roles. See Bowers (2010a, Chapter 1), for discussion.3 ThecategoryPr isthe position to which verbs in English raise.In earlier work of mine (Bowers, 1993,2001), thiscategory,like Chomskys v, was alsothe
category in whose specifier subjects of transitive verbs originated. That is no longer thecase in thetheoryproposedhere. I suggest in Bowers(2010a) thata
better name for this category might be Mood.4 ThecategoryVoiis similar in many respects to thecategoryTr proposed in Bowers(2002a). In particular, itis the locus ofthe probe thatenters intoan
Agree relation with objects and assigns ACC Case to them. See also Koizumi (1993, 1995). It should not be confused with Kratzers (1996) category Voice,which is basically the same as v and Pr.5 The UOM thus states in strict bottom-to-top derivational terms the idea that there is a fixed universal hierarchy of functional categories, thereby
extending Cinques (1999) hypothesis that there is a universal ordering of adverbial modifiers contained in the specifiers of functional categoriesto the full
range of functionalcategories and, more radically, to a setof argument categories. That there is such a universalorderof Merge seems to be empirically true.
Whether it can be derived from more basic principles remains to be seen, though certain observations suggest that there may be a semantic basis for the
UOM. If,for example,the suggestion in Bowers(2010a) that thefunction of thecategoryPr is tobind theeventvariable with an existentialor modal operator
is correct, then Pr must necessarily be mergedafterall theargumentcategories.Similarly,Larson (2004) suggests, following Davidson (1967), that adverbial
causal-clauses are the innermost (i.e. merged earliest) adverbial modifiers of the verb because causal relations distinguish and individuate events, hence
have a privileged status for events.6 I thus depart from the usual minimalist assumption that selection takes place in the head-complement configuration. Selection of the standard sort is
here broken down into two distinct relations: a-selection, which specifies what argument categories a given root requires, and c-selection, which is a
relation between an argument category and an XP in its specifier. (But see footnote 7 below.).7 In casessuch as this where theroot itselfrequiresan argument of a specific category,or, even more restrictively, an argument headedby a specific LI of a
given category,both theroot andthe relevant argument category contain theappropriatec-selectionfeature. Thus, theroot throw contains the following a-
selection feature and associated c-selection feature: {[Aff], [__toP]}. At the same time, the category Aff contains the selection features [__toP]. When throw
adjoins to Aff, both features are checked anddeleted. In cases where the range of possible c-selection features is constrained only by the light verb itself, noc-selection features need be specified in lexical roots. Thus, Ag in English quite generally c-selects either D or byP, regardless of the particular root.
J. Bowers / Lingua 121 (2011) 11941206 1195
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
3/13
A DP with an unvalued Case-feature is active, while one whose Case-feature has been valued and deleted is inactive. An
inactive DP is frozen in place and cannot enter into another Agree relation of the same type ( Chomsky, 2000). Nothing,
however, prevents an inactive DP from entering into another type of agreement relation such as wh-Agree. Crucially, once a
DP has become inactive, it is no longer visible to another probe searching for a goal with matching f-features.Generalizing the analysis of transitivity proposed in Bowers (2002a), I assumea universal category Voi(ce) with one of two
values: active ([+act]) orpassive ([act]). In English, when Voi has the value [+act], it contains a probe that assigns structural
ACC Case. When Voi has the value [act], it has no probe, though it does retain an EPP feature (Bowers, 2002b).
Agree is constrained by the standard locality condition (2) (Chomsky, 2000:122):
(2) Locality Condition (LC):
Suppose P is a probe and G is goal. Then Agree holds between P and G just in case G is the closest set of features in
the domain D(P) of P that match those of P.The domain D(P)of P isthesister ofD, and G is closest to PifthereisnoG0
matching P such that G is in D(G 0).
I also assume the following generalized and relativized version of Chomskys 2001 Phase Impenetrability Condition (PIC):
(3) Relativized Phase Impenetrability Condition (RPIC):
Let P be a probe of a given type (f+EPP, pure EPP, or wh-Agree) that has been valued at some stage of thederivation and suppose that P0 is the next probe of the same type to be introduced into the derivation. Then P0 is
barred from searching for a matching goal within the domain of P.
The RPIC restricts the search space of the probes in such a way that once af-Agree, pure EPP, or wh-Agree relation has beenestablished, the domain of the head containing the probe is no longer accessible to a probe of the same type introduced later
in the derivation. The RPIC plays a crucial role in the derivation of active transitive sentences, as will be seen shortly, and also
in explaining the reconstruction properties of quantified DPs and PPs (see Bowers, 2010a,b, for details).
2.2. Derivation of actives and passives in English
I illustrate howthe theory works by deriving the activesentenceJohn threw the ball to Mary and its passive counterpart the
ball was thrown to Mary by John. Consider the derivation of the active sentence first:
(4)
[
TP
JohnNOM T PrP
Agree Pr VoiP
EPP
the ballACC Voi AffP
to Mary
Aff ThPAgree
Th AgP
Ag throw
[Ag],[Th],[Aff]
In derivation (4), the probe in the [+act] Voi head assigns the value ACC to the Case feature of the nearest active DP, in this
instance the Th-DP the ball, which then moves to [Spec, Voi], where it c-commands the Aff-PP to Mary. The probe in T is now
available to assign NOM Case to the Ag-DP John, the nearest constituent with matching f-features and an unvalued Casefeature. Notice that by the RPIC, the Ag-DP is unavailable to the probe in T since it is in the domain of the probe in Voi.
However, because pure EPP probes are orthogonal to Agree probes, nothing prevents the Ag-DP from first escaping to
[Spec, Pr] to satisfy the EPP feature of Pr, after which it is free to form an Agree relation with the probe in T. The Ag-DP thus
J. Bowers / Lingua 121 (2011) 119412061196
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
4/13
does not move to [Spec, T] in a single step. Rather, it must first move to [Spec, Pr] to satisfy the (possibly universal) EPP
feature of Pr. Only then is the probe in T able to form an Agree relation with the Ag-DP, assigning it NOM Case and moving it to
[Spec, T].8 Interspersed with these operations is head-to-head movement of the root to Ag, Th, Aff, Voi, and Pr (not shown in
(4)), accompanied by checking and deletion of the a-selection features of the root.
Consider next the derivation of the passive sentence the ball was thrown to Mary by John:
(5)
[
TP
the ballNOM T PrP
Agree Pr VoiPbe
Voi AffP-EN
[to Mary]
Aff ThP
Th AgP
[by John]
Ag throw[Ag],[Th],[Aff]
Here there is no probe in Voi, because it has the feature [act]. The probe in T must therefore assign NOM Case to the nearest
active DP, which in this case is the Th-argument the ball, leaving the AgP with no alternative but to select a PP headed by by. If
Ag selected a DP with unvalued Case feature, there would be no way for its Case feature to be valued and the derivation
would crash. Conversely, if Ag selected a by-phrase and the value of Voi were [+act], then the derivation would crash because
there would be no active DP available for the probe in T to form an Agree relation with, hence no way for its uninterpretable
f-features to be checked and eliminated. The end result is that in passive sentences, the AgtP must surface as a PP headed byby in a position lower than that of any other argument.9 As in the previous derivation, the Th-DP must move first to [Spec,
Voi] to satisfy its EPP feature, followed by movement to [Spec, Pr], and finally to [Spec, T].10
Suppose now that Aff selects a DP with unvalued Case feature and the Th-head assigns inherent ACC Case to the Th-DP.
Then the probe in Voi will assign structural ACC Case to the Aff-DP and move it to [Spec, Voi], resulting in the double object
construction: John threw Mary the ball. Alternatively, if Voi is [act], then Ag must select a by-phrase and the Aff-DP will be
assigned NOM by the probe in T, producing the passive sentence Mary was thrown the ball by John. (See Bowers, 2010a, Chapter
3, for further details.)
As is evident, there is no need for either Case absorption or -role transfer in this theory. The former reduces to
the fact that Voi in English lacks f-features when it has the value [act]. -role transfer is unnecessary, because thesubject of an active sentence and the by-phrase of a passive sentence derive from the same structural position, namely,
[Spec, Ag].
2.3. Morphological realization of category heads
In many languages there is direct morphological evidence for the existence of argument categories in the form of affixes
attached to the verbal root that mark the presence of an argument of a given type. In preparation for the discussion of
derivational morphology that follows, I discuss briefly a couple of well-known examples from the literature.
8 I argue in Bowers(2010a, Chapter 1), that it is theobligatory movement of theAgt-DP to [Spec,Pr] in transitive activesentences that creates theillusion
of there being an externalargument position. There is thus no contradiction between my claim that thesubject of transitive verbs startsout in thelowest
argument position (where it in fact remains in passive sentences) and work in the eighties such as Williams (1981), Marantz (1984), and many others,
arguing that subjects of transitives occupythe highest argument position. See also Bowers(2010a, Chapter 1) forargumentsagainstthe view that the-role
of the external argument is assigned compositionally by VP (Marantz, 1984, 1997).9 See Bowers(2010a,b), forextensive argumentation in support of theclaimthatthe by-phrase of thepassive(andlikewisethe understood AgtP in short
passives) is c-commanded by all other arguments.10 See Bowers (2010a,b), for direct evidence from existential sentences that the passivized Th-DP moves through [Spec, Voi].
J. Bowers / Lingua 121 (2011) 11941206 1197
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
5/13
In many languages, arguments of various types must be accompanied by an applicative morpheme attached to the verbal
root just in case the argument in question is a DP with an unvalued Case feature. The argument most commonly marked in
this way is AffP, as shown in the following example from Chichewa (Baker, 1988):
(6) a. Mbidzi zi-na-perek-a msampha kwa nkhandwe.
zebras SP-PAST-hand-ASP trap to fox
The zebras handed the trap to the fox.
b. Mbidzi zi-na-perek-er-a nkhandwe msampha.
zebras SP-PAST-hand-APPL-ASP fox trap
The zebras handed the fox the trap.
In (6) a. the AffP nkhandwe fox is marked with the preposition kwa to and the ThP msampha trap is the direct object.
Example (6) b., in contrast, looks exactly like an English double object construction, including the change in preferred
word order, except that the Aff-head in Chichewa must be spelled out as the suffix -ir attached to the verbal root perek-
to hand. Bowers (2010a,b) argues that (6) b. should be derived in virtually the same way as a double object
construction:
(7)
[
TP
DP
mbidzi T PrPNOM PAST
DPzi-na
Agree Pr VoiP
Case: -aDP
nkhandwe Voi AffP
ACC DP
Aff ThPAgree -ir
Case: DP
msampha Th AgPACC
DP
Ag perek
Case:
Here the AffP is assigned ACC Case by the probe in Voi and moved to [Spec, Voi], while the AgP is assigned NOM Case after first
moving to [Spec, Pr] to satisfy its EPP feature. The root in the meantime raises from head to head, picking up the applicative
morpheme ir- in Aff along the way.11 If the Aff-head selects PP, in contrast, its spellout is phonetically null and the Th-DP
must be assigned ACC Case by the probe in Voi and moved to [Spec, Voi], producing (6) a., derived in exactly the same way as
the English prepositional dative construction in (3).
Consider next a case where the Ag head is realized by a special morpheme. In languages such as Greenlandic Eskimo with
Absolutive-Ergative Case-marking systems, a special antipassive morpheme must be attached to the verbal root under
certain conditions, as shown in (8) b.c. below (Baker, 1988):
11
In strictDM terms there would be nothing butsyntactic features in such derivations which would eventually be replaced with vocabulary items by lateinsertion. However, for the sake of clarity I present derivations throughout this paper as if spell-out accompanied each operation of the syntax.
J. Bowers / Lingua 121 (2011) 119412061198
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
6/13
(8) a. Angut-ip arnaq unatar-paa.
man-ERG woman(ABS) beat-INDIC:3sS
The man beat the woman.
b. Angut arna-mik unata-a-voq.
man(ABS) woman-INSTR beat-APASS-INDIC:3sS
The man beat a woman.
c. Angut unata-a-voq.
man(ABS) beat-APASS-INDIC:3sS
The man beat someone.
Let us assume that in certain types of Absolutive-Ergative Case-marking systems (i) the probe in Voi assigning structural ACC
Case is missing altogether and (ii) Ag-DPs are optionally marked with inherent ERG Case.12 If the Ag-DP is marked with
inherent ERG, then NOM (=ABS) is assigned to the Th-DP, resulting in examples such as (8) a. The so-called antipassive
construction arises when Ag c-selects a DP with unvalued Case feature. In that case, the Ag-DP is assigned structural NOM/ABS
and the Th-DP must either be marked with an inherent oblique case, such as INSTR in Eskimo, or realized as PROarb. In such
constructions, an antipassive morpheme (-a-, in this instance) is often required in the Ag-head, where it is picked up by verb
as it raises to T. This is illustrated in the following derivation of (7) b. and c.:
(9) [ TP angut unata-a-voq [PrP Pr [[ ThP arna-mik/PROarbTh [AgP Ag ]]]]ABS APASS-INDIC:3sS INSTR -a-
Why should it be the case that in so many languages the presence of an argument of a particular type is marked by a
morpheme attached to the verbal root? From the perspective adopted here, this possibility simply follows from the fact that
functional heads (in this instance argument heads) are often realized as morphological affixes, together with the standard
assumption of head-to-head movement. I turn next to the main topic of this paper, showing that phenomena of exactly the
same kind exist in nominal structures as well.
3. Morphological realization of category heads in nominal structures
It so happens that there is no direct morphological realization of the primary argument categories Ag, Th and Aff in verbal
roots in English. I will argue, however, that each of them is morphologically realized in derived nominal structures. That this
is so has been somewhat obscured by the fact that discussion of nominal structures in the literature has largely focused on
event (or process) nominals such as the following13:
(10) a. The consignment to Sothebys of a large art collection by Nelson Rockefeller.
b. Nelson Rockefellers consignment to Sothebys of a large art collection.
c. The art collections consignment to Sothebys by Nelson Rockefeller.
d. *Sothebys consignment of a large art collection by Nelson Rockefeller.
The verb consign requires Ag, Th and Aff arguments, as does the event nominal consignment. The nominals in (10) a.c. thus
refer to an event of Nelson Rockefellers consigning a large art collection to Sothebys. In such forms all three arguments are
realized as DPs or PPs and the structure of the event nominal corresponds directly to that of the sentence Nelson Rockefeller
consigned a large art collection to Sothebys.14 However, closer examination reveals the existence of a wider range of
nominalizations in English than is usually considered, but before analyzing them, I first need to discuss briefly how nominals
in general are derived in the framework proposed here.
12 See Woolford (1997), for arguments that ERG is an inherent rather than a structural Case. See also Laka (2000) and Bowers (2010a), Chapter 4.2, for
further discussion.13 Throughout this discussion I ignore the fact that Th-arguments precede Aff-arguments in neutral or unmarked word order. See Bowers (2010a,
Chapter 5), for a possible explanation of this fact.14 There is by now a considerable body of work on this topic inspired jointly by Chomsky (1970) and Abney (1987). See, for example, Siloni (1997) and
Alexiadou (2001). Note that the discussion here is confined to what Chomsky terms derived nominals. English gerundive nominals (or gerunds), as has been
shown by Abney (1987), Bowers (1987), and others, are hybrid structures, consisting of a nominal superstructure and a verbal substructure. Specifically, I
assume that derived nominals contain nothing but lightn categoriesabove the root,whereas gerundivesconsist of light v categories up to thelevel of Pr,andlight n categories above that.
J. Bowers / Lingua 121 (2011) 11941206 1199
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
7/13
3.1. The structure of nominals
I assume that the structure of derived nominals and sentences is parallel in many fundamental respects, the main
difference being that a slightly different inventory of functional categories is required in theextended projection of nominals.
Minimally, nominals differ from sentences in the following ways15:
1. Nominals require the functional category D(eterminer) instead of T(ense), as well as an intermediate category I term Nom
that occupies a place in nominal structure corresponding to that of Pr in verbal structures.
16
In English, nouns obligatorilyraise as far as Nom but no further, while in many languages they may or must raise to D.17
2. Derived nominals do not project the category Voi (or the categories Perf and Prog) at all. In other words, there are simply
no items in the lexicon with the complex of features [n, Voi]. It follows immediately, since Voi features such as [+/active]
play no role in the derivation of derived nominals, that there is no way of assigning structural ACC Case, explaining why
there are no nominals of the form *the destruction the city, *my criticism him, and so forth. At the same time, this accounts
for the well-known fact that both raising-to-object (ECM) and double object constructions are impossible in nominals:
(11) a. *Johns belief (of) Bill to have left (cp. Johns desire for Bill to leave)
b. *the presentation (of) Mary of the trophy (cp. the presentation of the trophy to Mary)
3. The only structural Case in nominals is GEN, which I assume is assigned by an optional probe in D and is spelled out as s in
English. The probe in D also has an EPP feature, accounting for the position of genitive subjects in [Spec, D].
4. Nominals select the same primary argument categories Ag, Th, and Aff as verbs and the UOM is the same in nominals as itis in sentences. (The same is true for quasi-arguments such as Source, Goal, Ben, Instr and for modifiers such as Manner,
Place, Time, etc.) However, only Ag and Th have the option of selecting either PP or DP with an unvalued Case feature in
English derived nominals, whereas Aff may only select PP. Hence, Ag-DP may either be marked with the preposition by, as
in sentences, or assigned structural GEN and moved to [Spec, D]. Likewise, Th-DP may either be marked with the
preposition of or assigned structural GEN and moved to [Spec, D].
The reader can easily verify that these assumptions are sufficient to account for the familiar array of data in (10).
3.2. Non-event nominals
Turning now to non-event nominals in English, I show that they are of just three kinds: (1) agentive nominals; (2) result
or theme nominals; (3) affectee nominals. These are illustrated in (12)(14), respectively:
(12) a. The consignor (to Sothebys) (of this major painting collection) (*by N.R.) was later revealed to be N.R.
b. The collections consignor (to Sothebys) was later revealed to be N.R.
(13) a. The consignment (to Sothebys) (by N.R.) was revealed to be a Rembrandt.
b. N.R.s consignment (to Sothebys) was revealed to be a Rembrandt.
(14) a. The consignee (*to Christies) (of the painting collection) is rumored to be Sothebys.
b. The collections consignee is rumored to be Sothebys.
Consider first the agentive nominals in (12). The idea that the interpretation of such nominals is closely tied to the argument
structure of the corresponding verb is of course not new. Thus Fabb (1984), Keyser and Roeper (1984), Burzio (1986), and
Levin and Rappaport (1988) all argue that ernominals can only be formed from verbs that have an external argument and
that they necessarily refer to the external argument. In the framework proposed here, Ag is the head that introduces the
external argument. Let us suppose, then, that when the argument category Ag has nominal features, it may be realizedmorphologically as the agentive suffix er/-or.18 Suppose furthermore that the light noun Ag, in contrast to the light verb
Ag, does not c-select a DP specifier. This accounts for the fact that agentive nominals cannot co-occur with an explicit Ag-DP/
PP. Semantically, -er maps an event function onto an individual bearing the Ag relation to the event in question.19 An
agentive nominal of the form (12) a. can then be derived straightforwardly as follows:
15 See Bowers (2010a, Chapter 5), for a more detailed analysis of nominal structures along the lines sketched out here.16 Bowers (1991) argues that there is a category Nm in nominals that is the functional equivalent of Pr in sentences. The category I am here calling Nom is
virtually the same as Nm.17 See Bernstein (2001), for a useful survey.18 The agentive light n realized morphologically as er/or must of course be distinguished from the null light n that occurs in other nominal forms. In a
strict DM framework the content of the two heads would be distinguished in terms of syntactic features and replaced with the appropriate morphemes by
late insertion at the end of the derivation. However, in order to simplify the exposition, derivations are presented as if they contained lexical items of the
traditional sort (see also footnote 11).19
Asnoted by Levin and Rappaport(1988:10711072) agentive nouns canbe inanimateas well as animate, e.g. (meat) grinder, (corn) planter, (apple) corer,etc., reflecting the fact that Ag-arguments in English are quite generally either animate agents or inanimate causes.
J. Bowers / Lingua 121 (2011) 119412061200
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
8/13
(15)
[
DP
D NomP
Nom ApplP
PP
Appl ThP
PP
Th AgP
Agn consign
Agn
the to Sothebys of the collection -or
Reinterpreting Di Sciullo and Williams (1987) in current minimalist terms, I assume that the suffix or/eris a light noun of
the type Ag. Hence when the root consign adjoins to or/er, it forms the Ag noun consignor, which subsequently raises by
successive head movement to Th, Appl and finally Nom.Given this analysis, the observation ofBurzio (1986:161) and Keyser and Roeper (1984:389, 395), developed in detail by
Levin and Rappaport (1988) and Rappaport Hovav and Levin (1992), that unaccusatives resist ersuffixation (e.g. *arriver,
*dier, *appearer, *happener, etc.) follows automatically under the assumption that the er derivational suffix is uniquely
associated with the Ag category. Likewise, Levin and Rappaports (1988) observation that unergative verbs generally permit
er suffixation (e.g. walker, runner, climber, beeper, blinker, etc.) is also predicted.
Consider next a nominal such as (13) a. This is what Grimshaw (1990) refers to as a resultnominal, contrasting with the
nominals in (10), which she calls event nominals. I propose to account for the morphosyntactic form and semantic
interpretation of result nominals by assuming that the suffix ment is systematically ambiguous. In addition to forming
event nominals, -mentcan also be the realization of a light noun of category Th.20 Like the Ag suffix er/or, the Thsuffix -ment
cannot c-select a DP, hence may not co-occur with an overt Th-DP. At the same time, it saturates the Th-argument
semantically, mapping an event function onto an individual bearing the Th-relation to the event in question. The nominal
in (13) b. can thus be derived as shown in (16):
(16)
[
DP
D NomP
Nom AffP
PP
Aff ThP
Th AgPn
DP
Th Ag consignnment
to Sothebys N.R.
Given the assumption that derivational affixes can be associated with light noun argument categories, agentive and theme
nominals are exactly what we would expect to find.
20 Obviously, -ment is not the only derivational suffix that can be a realization of [Th, n]. The suffixes ion and ure (e.g. the donation was a Pollock, the
forfeiture was a villa in Maui, etc.), among others, as well as zero-suffixation (e.g. the transfer was a large sum, the display was a famous collection of etchings),
canall be used to formTh-nominals. Lexical idiosyncrasies areaccountedfor by theDM mechanismof late insertionof vocabulary items after specific rootshave combined in the syntax with [Th, n] (see also footnote 9).
J. Bowers / Lingua 121 (2011) 11941206 1201
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
9/13
What about the third primaryargument category Aff?Consider the suffix ee. Thoughthere area number of different uses of
this suffix,Barker(1998:704)showsthatthereisoneclearsubcaseinwhichthederivednounreferstotheindirectobjectofthe
stem verb (addressee, lessee);. . . This use is defined by Funk and Wagnalls Standard College Dictionary as follows: One who
undergoes, or benefits from, some action: used especially in legal terms, and opposed to er, -or, a s i ngrantor,grantee. Clearly
ee in this use is an Aff-suffix, as is confirmed by the fact that an overt Aff-PP cannot co-occur with it. Hence ee must be the
morphological realizationof a light nounof category Aff, which is pickedup by theroot before movingup to the Nomposition21:
(17)
[
DP
D NomP
the Nom AffP
Aff ThPn-ee PP
of the collection Th consign
Other uses ofee are less clear. There is, for example, one subclass characterized by Barker (1998:704, 705) as one in which
the ee noun refers to the direct object participant of the stem verb, e.g employee, adoptee, advisee, callee, consultee, detainee,
trainee, etc. There are a number of possible analyses that one might entertain here. One is that the direct objects of verbs such
as employ, adopt, advise, etc. are Aff-arguments rather than Th-arguments. This would have the advantage of accountingautomatically for the fact that the objects of such verbs are always animate, since Aff-arguments are canonically animate.
Another possibility is suggested by the analysis of the Affectedness Constraint proposed in Bowers (2010a), which relies
crucially on subdividing the class of Th-arguments in terms of the feature [+/affected]. The direct object use of the ee
suffix could then be associated uniformly with the category [Th, n, +affected]. One advantage of this approach is that it could
be extended to cover another subclass ofee nouns referring to the animate and affected subjects of unaccusative verbs, e.g.
escapee, standee, resignee, dinee, returnee, arrivee, sittee, etc. (Barker, 1998:705). Though I do not have the space here to
analyze in detail all the various uses ofee, it is evident that the framework proposed here provides a number of interesting
possibilities that can be explored in future research.
In conclusion, then, each of the syntactic categories Ag, Th, and Aff is morphologically realized in a distinct type of non-
event nominalization, providing direct evidence that the primary argument types must be introduced in projections of
argument categories rather than being merged directly with lexical items.
4. Non-event nominals with incorporated nouns
I next extend the proposed analysis of derived nominals to include Noun Incorporation (NI) structures. It will be shown
that not only do such structures provide independent evidence in support of a syntactic approach to derivational
morphology, but they also provide evidence in support of the particular ordering of arguments embodied in the UOM.
Bowers (2004, 2010a) argues that NI structures are not produced by movement of a bare noun from an argument position
into a complex verbal form, as argued by Baker (1988) and others, but rather result from the possibility of satisfying the c-
selection features of argument categories by adjoining a bare noun directly to the appropriate head. Consider, for instance,
the compound verb deer hunt in the following example:
(18) John deer hunts in winter.
I propose to derive the relevant part of (18) as follows:
(19)
[
ThP
Th AgP
N Th DPdeer John
Ag Th [__N] [Ag],[Th]
hunt Ag hunt Ag[Ag],[Th] [Ag],[Th] [__D]
The root [hunt first raises and adjoins to the Ag-head, checking and deleting the a-selection feature [Ag], followed by MergeoftheDPJohn in [Spec, Ag] to satisfy the selection feature of Ag. This is followed by raising and adjunction of the complex Ag-
head thus formed to the Th-head, accompanied by checking and deletion of the a-selection feature [Th]. At this point, instead
21 For reasons not presently clear to me, the ee suffix seems to exclude the possibility of Ag-phrase.
J. Bowers / Lingua 121 (2011) 119412061202
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
10/13
of satisfying the c-selection feature of Th by merging a full DP/PP in [Spec, Th], let us suppose that the Th-head has the option
of satisfying its c-selection requirements by adjoining the bare noun deerdirectly to the complex Th-head, producing the
compound verb deer hunt.
As is well known, there is a highly productive pattern of compounding in English that combines a bare noun bearing the
Th relation to the root with an agentive nominal, e.g. deer hunter, beer drinker, bus driver, painting consignor, etc. We can now
derive such compounds straightforwardly in the syntax by combining agentive noun formation with NI, resulting in
derivations of the following sort:
(20)
[
ThP
Th AgPn
painting Th nn n
consign AgAg Th n
-orconsign Ag
n-or
The agentive noun consignoris first formed by adjoining the root conjoin to the agentive morpheme or. It then raises and
adjoins to thenull Th head, whose c-selection feature is satisfied by adjoiningthe nounpaintingto it. If thec-selectionfeature
of the Th head had instead been satisfied by merging a PP such as of the paintingin [Spec, Th], the result would have been the
agentive nominal the consignor of the painting.
Similarly, the Aff argument of an agentive nominal can be incorporated in examples such as the following:
(21) N.R. was considered a good auction house consignor.
Here the noun auction house is adjoined to the Aff head after the agentive nominal consignorhas raised and adjoined to it. It
even seems to be possible to incorporate both a Th noun and an Aff noun in some cases22:
(22) N.R. was considered a good auction house painting consignor.
In this example, painting is understood as the Th argument ofconsign and auction house as the Aff argument. Crucially, the
opposite order is impossible:
(23) *N.R. was considered a good painting auction house consignor
The explanation is straightforward. Given the theory proposed here, the right to left order of incorporated nouns must reflect
the UOM. Hence the Th-argument paintingmust be immediately to the left of the raised agentive nominal consignorand the
Aff-argument auction house must be to the left of the raised nominal painting consignor.
Similarly, an Aff argument can be incorporated into a Th nominal:
(24) That Picasso was the best auction house consignment (by N.R.) that I know of.
However, any attempt to incorporate an argument first, followed by raising and adjunction to a Th or Aff suffix yields
awkward results, because the root is not in the right position to attach morphologically to the suffix:
(25) a. ??The best collector consignments (to auction houses) are paintings.23
b. ??The best painting consignee is Sothebys.
c. *The most popular painting collector consignee is Sothebys.
In contrast, incorporation of any number and combination of full or incorporated arguments is possible in ordinary event
nominals:
(26) a. The FBI is investigating consignments to auction houses of painting collection by museum curators.
b. The FBI is investigating auction house consignments of painting collections by museum curators.
c. The FBI is investigating painting collection consignments to auction houses by museum curators.
22 Such data seem to contradict Grimshaws(1990) observation (cited in Baker, 1997:106)that indirect objects are difficult to incorporate, e.g.gift-giving to
children vs. *child-giving of gifts. Baker (1997:94) also stars the following examples: *child-reading,*spy-telling. However,it isnot hard toimagine contextsin
which these might be perfectly acceptable and an agentive nominal such as child readerone who reads to children seems fine, as do examples parallel to
(22), e.g. child candy giving (but *candy child giving). Clearly, more extensive empirical investigation of the contextual and lexical factors that influence
acceptability judgments in these cases is necessary.23
Note that thenominal in (25) a. is perfectlyacceptable if construedas an event nominal, parallel to (26) d.,as in collector consignments (to auction houses)take place less often than they used to.
J. Bowers / Lingua 121 (2011) 11941206 1203
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
11/13
d. The FBI is investigating museum curator consignments to auction houses of painting collections.
e. The FBI is investigating auction house painting collection consignments by museum curators.
f. The FBI is investigating auction house museum curator consignments of painting collections.
g. The FBI is investigating painting collection museum curator consignments to auction houses.24
h. The FBI is investigating auction house painting collection museum curator consignments.
This follows directly from the fact that the event-forming suffix ment, unlike the other non-event forming suffixes, does not
saturate any of the predicates arguments other than the event argument. Hence the event nominal is free to take as manyovert arguments as it likes, each of which can be realized either as a full DP or as an incorporated bare noun.
Notice that so far we have not actually proposed an explicit derivation for event nominals of the sort illustrated in (10)
and (25). Consider first the fact that basic nouns such as picture, story, account, etc. can have the same argument structure as
derived nominals such as consignmentand that in such cases no morphological material at all is picked up in the course of the
derivation. Thus a nominal such as a picture of Aristotle by Rembrandt could be derived as follows:
(27)
[
DP
D NomP
a Nom ThP
PP
of Aristotle Th AgP
PP
Ag picture
by Rembrandt n
This might suggestthat the best way of deriving (10a) would be to convert the root [consign into the nominal form consignmentprior to merging with Ag, after which the derivation would proceed much as in (27). There are, however, twoproblems with this
approach. The first is that it is tantamount to a lexical theory of derivation. If initial merge of roots such as [consign with a lightnoun n, realized as the suffix ment, always precedes every other application of Merge, then the process of combining roots with
eventive suffixes might as well be carried out in the lexicon, prior to the application of any syntactic rules. The second is that the
suffixes that form event nominals would, under this analysis, differ from all the other suffixes we have considered so far in not
being associatedwith anygrammaticalfunctionhead. To avoid theseproblems, I propose insteadthatthe suffixes thatmark event
nominals are morphological realizations of the category Nom. Example (10a) would then be derived as follows:
(28)
[
DP
D NomP
the Nom AffPn
PP
consign Nomn to Sothebys Aff ThP
ment PP
of a large art Th AgPcollection
PP
by N.R. Ag
24 Note that reversing the order of the Th argument and Ag argument is not as bad as expected: The FBI is investigating museum curator painting collection
consignments to auction houses. This is because museum curator painting collection canbe interpreted as an incorporatedcompoundnoun, hence thenominal
in question would be equivalent to consignments to auction houses of museum curator painting collections. In general, judgments in these cases arecomplicated by the possibility of forming compound nouns of different types.
J. Bowers / Lingua 121 (2011) 119412061204
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
12/13
This immediately explains why event nominals are able to co-occur with a full set of overt arguments, whereas each of the
non-event nominals discussed earlier excludes overt realization of one or another of its arguments. In terms of basic
argument structure, the derivation of an event nominal such as the consignment of a large art collection to Sothebys by Nelson
Rockefelleris completely parallel to that of the corresponding sentence Nelson Rockefeller consigned a large art collection to
Sothebys. After the introduction of the argument categories Ag, Th, and Aff, however, the derivations diverge as a
consequence of the fact that derived nominals are closed off by the nominal categories Nom and D, whereas sentences are
closed off by the verbal categories Voi, Pr, and T. This shows that nominalization is not merely a matter of changing the
syntactic category of a lexical item but involves the entire syntactic derivation. In any case, the only difference between a
derived noun such as consignmentand a basic noun such as picture is that the phonetic realization of the category Nom is
-ment in the former but null in the latter.25
5. Conclusion
In conclusion, the existence of agent, theme, and affectee nominals, corresponding to the three primary argument types,
provides direct evidence in support of a theory in which arguments are introduced in projections of the syntactic categories
Ag, Th andAff, each of which can be realized by particular morphemes. Furthermore, these results argue strongly for theview
that derivational morphology, like inflectional morphology, is fundamentally syntactic in nature. Finally, the ordering of
incorporated nouns in NI structures provides additional evidence in support of the UOM.
Abbreviations
DM Distributed Morphology
GFC grammatical function changing
UOM Universal Order of Merge
EPP Extended Projection Principle
LC Locality Condition
PIC Phase Impenetrability Condition
RPIC Relativized Impenetrability Condition
a-selection argument selection
c-selection constituent selection
NI Noun Incorporation
Ag Agent
Th Theme
Aff Affectee
Sou Source
Ben Benefactive
Instr Instrumental
C Complementizer
T Tense
Pr Predication
Voi Voice
D determiner
NOM nominative case
ACC accusative case
ERG
ergative caseABS absolutive case
INSTR instrumental case
GEN genitive case
APASS antipassive morpheme
IND indicative mood
PAST past tense
SP subject pronoun
APPL applicative morpheme
ASP aspect
Nom nominal category corresponding to Pr
25
Or, in strict DM terms, there exist vocabulary items consignmentand picture that can be inserted in the structure [Nom [Hconsign. . .
] Nom], but none ofthe form *consign or *picturement.
J. Bowers / Lingua 121 (2011) 11941206 1205
-
7/28/2019 Bowers- Non-Event Nominals and Argument Structure
13/13
Perf perfect aspect
Prog progressive aspect
References
Abney, S., 1987. The English Noun Phrase in Its Sentential Aspect. Doctoral Dissertation. Massachusetts Institute of Technology, Cambridge, MA.Alexiadou, A., 2001. Functional Structure in Nominals: Nominalization and Ergativity. John Benjamins Publishing Company, Amsterdam/Philadelphia.
Baker, M.C., 1988. Incorporation: A Theory of Grammatical Function Changing. The University of Chicago Press.Baker, M.C., 1997. Thematic roles and syntactic structure. In: Haegeman, L.M.V. (Ed.), Elements of Grammar: Handbook of Generative Syntax. Kluwer
Academic Publishers, Dordrecht/Boston, pp. 73138.Barker, C., 1998. Episodic ee in English: a thematic role constraint on new word formation. Language 74 (4), 695727.Bernstein, Judy B., 2001. The DP hypothesis: Identifying clausal properties in the nominal domain. In: Baltin, M., Collins, C. (Eds.), The Handbook of
Contemporary Syntactic Theory. Blackwell, Malden, MA, pp. 536561.Borer, H., 2005. The Normal Course of Events. Oxford University Press.Bowers, J.,1987.ExtendedX-bartheory, theECP andthe Left BranchCondition. In:Crowhurst, M. (Ed.),Proceedings of theWest Coast Conference on Formal
Linguistics 6. The Stanford Linguistics Association, Stanford, CA, pp. 4762.Bowers, J., 1991. Thesyntax and semanticsof nominals.In: Moore,S., Wyner,A. (Eds.), Proceedings of theFirst Semanticsand Linguistic TheoryConference,
Cornell Working Papers in Linguistics 10. Department of Modern Languages and Linguistics, Cornell University, Ithaca, NY, pp. 130.Bowers, J., 1993. The syntax of predication. Linguistic Inquiry 24 (4), 591656.Bowers, J., 2001. Predication. In: Collins, C., Baltin, M. (Eds.), The Handbook of Contemporary Syntactic Theory. Blackwell Publishers Ltd., pp. 299333.Bowers, J., 2002a. Transitivity. Linguistic Inquiry 33 (2), 183224.Bowers, J., 2002b. Participial (non-)agreement in impersonal passives. Paper presented at the Department of Linguistics, Rutgers University, November 22,
2002.Bowers, J., 2004. Toward a unified theory of argument structure and grammatical function changing morphology. Paper presented at the Workshop on
Argument Structure, CASTL, University of Troms, November 4, 2004.Bowers, J., 2010a. Arguments as Relations. MIT Press, Cambridge, MA.Bowers, J., 2010b. Argument structure and quantifier scope. In: Duguine, M., Huidobro, S., Madariaga, N. (Eds.), Argument Structure from a Crosslinguistic
Perspective. John Benjamins, Amsterdam, pp. 151179.Burzio, L., 1986. Italian Syntax: A Government-Binding Approach. Reidel, Dordrecht.Chomsky, N., 1970. Remarks on nominalization. In: Jacobs, R.A., Rosenbaum, P.S. (Eds.), Readings in English Transformational Grammar. Ginn and Co.,
Waltham, MA, pp. 184221.Chomsky, N., 2000. Minimalist inquiries. In: Martin, R., Michaels, D., Uriagereka, J. (Eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard
Lasnik. MIT Press, Cambridge, MA, pp. 89155.Chomsky, N., 2001. Derivation by phase. In: Kenstowicz, M. (Ed.), Ken Hale: A Life in Language. MIT Press, Cambridge, MA, pp. 152.Cinque, G., 1999. Adverbs and Functional Heads. Oxford University Press, New York.Collins, C., 1997. Local Economy. MIT Press, Cambridge, MA.Davidson, D., 1967. Causal relations. Journal of Philosophy 64, 691703.Di Sciullo, A.M., Williams, E., 1987. On the Definition of Word. Linguistic Inquiry Monograph 14. MIT Press, Cambridge, MA.Embick, D., 2004. On the structure of resultative participles in English. Linguistic Inquiry 35 (3), 355392.Fabb, N., 1984. Syntactic Affixation. Doctoral Dissertation. Massachusetts Institute of Technology, Cambridge, MA.Grimshaw, J., 1990. Argument Structure. Linguistic Inquiry Monograph 18. MIT Press, Cambridge, MA.
Halle, M., Marantz, A., 1993. Distributed morphology and the pieces of inflection. In: Hale, K., Keyser, S.J. (Eds.), The View from Building 20. MIT Press,Cambridge, MA, pp. 111176.
Harley, H., 2009. The morphology of nominalizations and the syntax of vP. In: Giannakidou, A., Rathert, M. (Eds.), Quantification, Definiteness, andNominalization. Oxford University Press, Oxford/New York, pp. 321343.
Keyser, S.J., Roeper, T., 1984. On the middle and ergative constructions in English. Linguistic Inquiry 15, 381416.Koizumi, M., 1993. Object agreement phrases and the split VP hypothesis. In: Bobalijk, J.D., Phillips, C. (Eds.), MIT Working Papers in Linguistics 18: Papers
on Case and Agreement I. Department of Linguistics and Philosophy, MIT, Cambridge, MA, pp. 99148.Koizumi, M., 1995. Phrase Structure in Minimalist Syntax. Doctoral Dissertation. Massachusetts Institute of Technology, Cambridge, MA.Kratzer, A., 1996. Severing the external argument from its verb. In: Rooryck, J., Zaring, L. (Eds.), Phrase Structure and the Lexicon. Kluwer Academic
Publishers, Dordrecht, pp. 109137.Laka, I., 2000. Thetablind case: Burzios generalization and its image in the mirror. In: Reuland, E. (Ed.), Arguments and Case: Explaining Burzios
Generalization. John Benjamins, Amsterdam, pp. 103129.Larson, R., 2004. Sentence-final adverbs and scope. In: Wolf, M., Moulton, K. (Eds.), North East Linguistic Society (NELS) 34. Graduate Linguistic Student
Association, University of Massachusetts, Amherst.Levin, B., Rappaport, M., 1988. Nonevent er nominals: a probe into argument structure. Linguistics 26, 10671083.Marantz, A., 1984. On the Nature of Grammatical Relations. MIT Press, Cambridge, MA.Marantz, A.,1997. No escapefrom syntax: dont trymorphological analysis in the. privacy of your ownlexicon.Paper presentedat the21st Penn Linguistics
Colloquium, University of Pennsylvania Working Papers in Linguistics 4(2). University of Pennsylvania, Philadelphia.Ramchand, G., 2008. Verb Meaning and the Lexicon: A First Phase Syntax. Cambridge University Press, Cambridge/New York.RappaportHovav,M., Levin,B., 1992. ernominals:implications for a theoryof argument structure. In:Stowell,T.T.,Wehrli, E. (Eds.), Syntaxand Semantics
26: Syntax and the Lexicon. Academic Press, New York, pp. 127153.Siloni, T., 1997. Noun Phrases and Nominalizations: The Syntax of DPs. Kluwer Academic Publishers, Dordrecht.Williams, E., 1981. Argument structure and morphology. The Linguistic Review 1, 81114.Woolford, E., 1997. Four-way case systems: ergative, nominative, objective and accusative. Natural Language and Linguistic Theory 15, 181227.
J. Bowers / Lingua 121 (2011) 119412061206