You are on page 1of 22

Final version appeared in N. Dehé & A. Wanner (eds.

) (2001) Structural Aspects of


Semantically Complex Verbs. Frankfurt, Bern & New York: Peter Lang, 105-130.

The Syntax of Cognate and Other Unselected Objects*


Claudia Felser (University of Essex) & Anja Wanner (University
of Wisconsin-Madison)

1. Introduction
It is a characteristic property of intransitive verbs that they do not take any
overt complements. As far as their selectional or subcategorisation properties
are concerned, they are usually described as "single-argument" verbs, or one-
place predicates. While intransitive verbs of the unaccusative type have a
complement at D-Structure (cf. (1a)), which needs to raise to structural subject
position because it is not Case-licensed by the verb (Burzio 1986), intransitive
verbs of the unergative type are traditionally considered to be "Agent-only"
verbs - that is, they do not select an internal argument at all (cf. (1b)).
(1) a. arrive [ Theme ]
b. smile Agent [ ]
Syntactic tests that show different behaviour of unaccusative and unergative
verbs include passivization (Perlmutter 1978, possible with unergative verbs),
adjectival passive formation (Levin & Rappaport 1986, possible with
unaccusative verbs) and, arguably, auxiliary selection in languages that supply
two different auxiliaries for the formation of the perfect (Burzio 1986).
There are, however, a variety of constructions in which an intransitive
verb occurs with a non-subcategorized postverbal noun phrase. The following
examples illustrate the transitive use of unergative verbs:

*
We thank the audiences at the 22nd DGfS Meeting at Marburg, Germany, and at the
Department of Phonetics and Linguistics, University College London, for helpful comments
and discussion. The authors' names appear in alphabetical order.
1
(2) a. She smiled her happy smile. (cognate object)
b. She danced a piece from Swan Lake. (hyponym of cognate object)
c. She nodded her approval. (reaction object)
d. She laughed herself tired. (resultative construction)
e. She worked her way into the committee. (way-construction)
In the case of cognate objects - noun phrases that are morphologically related to
the verb itself - modification of the postverbal DP is virtually mandatory, while
resultative constructions typically involve a reflexive anaphor that is bound by
the matrix subject. In the case of the way-construction, which can be treated as
a quasi-idiomatic variant of the resultative construction, the head of the
postverbal noun phrase is fixed, but the determiner usually agrees with the
matrix subject.
A subset of unaccusatives sometimes called "ergative" verbs also occur
in transitive structures, including cognate object and resultative constructions:
(3) a. The ice-cream melted. / She melted the ice-cream.
b. He died. / He died a gruesome death.
c. The milk froze. / The milk froze solid. / He froze the milk solid.
The present paper examines the syntax of intransitive verbs of the unergative
type in English, focusing on constructions in which they are followed by a
noun phrase.1
If we take it that every unergative verb can, in principle, occur in one or
the other of these constructions, the question arises as to how such
complements are formally licensed. In each of these examples the original
meaning of the verb is preserved, yet its syntactic behaviour does not seem to
match its lexical specification. The noun phrases in question are not part of the
lexical representation of unergatives but nevertheless occur in an argument
position, which constitutes a prima facie violation of the Theta Criterion
(Chomsky 1981: 36). Basically, there are two possible ways out of this
dilemma: (i) One could argue that all unergative verbs have a transitive
counterpart, i.e. that there are two verbs smile, one intransitive and one
transitive, or (ii) the working hypothesis could be that the transitive uses listed
above indicate that unergatives are inherently transitive, i.e. that there is only
one verb smile, which is a transitive verb. We shall explore the second
possibility here.

1
We will not discuss the semantic restrictions between the verb and the postverbal DP
(for example, while you can smile a happy smile, you cannot smile a stupid grin, even
though the semantic content of the two nominal expressions is very similar). These
restrictions seem to be entirely semantic or pragmatic in nature, and are similar to those we
find in resultative constructions of transitive verbs (you can, for instance, eat your paper
plate soggy - at least according to a paper-plate dinner party guest recorded by Carrier &
Randall 1992 - but you cannot eat your mother happy).
2
Given the possibility of paraphrases such as smile = 'produce a smile',
Hale & Keyser (1993, 1997, 1999) have suggested that the lexical
representation of unergatives includes a nominal complement that is
incorporated into the verb at the (pre-)syntactic level of L-syntax. While we
share with Hale & Keyser the assumption that unergatives are intrinsically
transitive, we will argue against their claim that denominal verbs are derived
through syntactic incorporation. Instead, we suggest that the transitive
characteristic of unergatives is also reflected in syntax proper, in that it allows
them to combine with a Theme DP. Under this view, the verbs in (2) do not
have a more complex lexical-semantic representation than their intransitive
counterparts - they are simply specified as transitive.
The remainder of the paper is organized as follows. Section 2 discusses
the syntactic status of cognate and reaction objects of unergative verbs. In
section 3, we examine the consequences of Hale & Keyser's incorporation
hypothesis for the analysis of cognate object structures, before suggesting an
alternative treatment of these constructions in section 4. Finally, we discuss the
possibility of extending our proposal to resultative constructions in section 5.

2. The Syntactic Status of Post-Intransitive Noun Phrases


First, let us consider the syntactic status of the overt postverbal noun phrases in
the examples listed above. Do they occupy an argument position (as
complements of V, or possibly as subjects of complement small clauses), or
should they be analysed as adjuncts? Jones (1988) has argued that cognate
objects are adjunct predicates, on the basis of examples such as (4) below
which show that cognate objects cannot undergo argument movement when the
verb is passivized.
(4) a. *A gruesome death was died by John.
b. *A weary sigh was sighed by Bill.
According to Jones, analysing cognate objects as adjuncts is possible on the
assumption that the Case Filter (Chomsky 1981:49) only applies to noun
phrases that are arguments. Non-subcategorized adjunct DPs, on the other
hand, need not be assigned structural Case. Observe, however, that failure to
passivize can often be attributed to a lack of modification in the cognate object,
or to other pragmatic factors. Compare the ungrammaticality of (5a,b), whose
active correlates are unacceptable as well.
(5) a. *A smile was smiled.
b. *This life was lived.
However, not all cognate objects structures are incompatible with the passive
(cf. (6a) below). Reaction objects, which are not subject to quite such as strict

3
modification restriction, can move around much more freely than cognate
objects (compare (6b)). Under appropriate discourse conditions, the postverbal
noun phrase in an "intransitive" resultative construction such as (6c) can (and
has to) raise in a passive construction, too:
(6) a. A merry dance was danced by Sam. (Jones 1988: 91)
b. Warm thanks were smiled at the audience.
c. They had talked her into love and she was not easily to be talked out of it
again. (Jane Austen, Emma)
MacFarland (1994) provides further counter-examples to Jones' claim that
cognate objects are adjuncts. They are based on the observation that in English,
the subject is also the topic of the sentence. If the cognate object is modified
(that is, if it carries semantic content that is beyond, or more specific than, the
lexical meaning of the verb), it becomes a plausible topic, and thus is
acceptable as the subject of a clause.
(7) A weary sigh was sighed by the overworked field worker at the end of a long day.
Another argument against analysing the post-intransitive DP as an adjunct
phrase is the fact that cognate objects or reaction objects cannot be separated
from the verb by a manner adverb:
(8) a. *He sighed wearily a (heavy) sigh.
b. *She nodded gracefully her approval.
An further test for complementhood is a noun phrase's eligibility to become the
subject of a Middle Construction (compare e.g. Hale & Keyser 1993). Cognate
objects pass this test, too:
(9) Dreams of winter dream more easily when you are in Alaska.
Finally, the postverbal noun phrase can be modified by an attributive adjectival
passive. According to Levin & Rappaport (1986), this indicates that the DP in
question is a complement of the verb.
(10) a. The backlog of unflown flights was, of course, appalling. (MacFarland
1994)
b. a half-smiled goodbye
c. her fast-worked way into the committee
The idea that cognate objects are arguments rather than adjunct expressions is
supported by certain semantic properties of cognate object constructions. It is
well known (compare e.g. Verkuyl 1993, Tenny 1994) that argument noun
phrases play an important role in construing the aspectual structure of the
clause. Dowty (1991), for instance, discusses in some detail the "aspectual role"
(to borrow a term from Tenny 1994) of so-called Incremental Themes -
complement DPs that measure the progress of an event towards its culmination.
The external argument, on the other hand, makes no such contribution towards
4
telicity. In the following example it is the apple that changes as the event goes
on, not the girl.
(11) The girl ate the apple.
But noun phrases lose this capacity to "measure out" an event (to borrow
another term from Tenny 1994) when they are indefinite or - as Verkuyl puts it
- specified as [-Specified Quantity Aspect].
(12) a. The girl ate three apples in five minutes.
b. *The girl ate apples in five minutes.
In short, for an event to be interpreted as telic, there normally has to be a
nominal complement. This generalization does not hold the other way round,
though, since there are transitive structures that do not denote telic events:
(13) They pushed the cart (along the river).
It seems, then, that there is a correlation between a DP's syntactic position and
its potential to function as an aspectual delimitor of the event described. It has
been noted that cognate objects, too, can induce aspectual shifting towards
telicity (MacFarland 1994).2 Let us therefore conclude that cognate and
reaction objects - irrespective of their precise semantic status - function as
complements syntactically.
Under this view, sentences such as (14) are structurally analogous to
ordinary transitive structures.
(14) a. She smiled her happy smile.
b. They nodded their approval.
A cognate or reaction object, then, is a non-subcategorized DP that originates in
an argument position within the VP headed by an intransitive verb. If we follow
Chomsky (1995:315) in assuming that the specifier and complement positions
of V are reserved for the verb's internal arguments, we are left to conclude that
the postverbal noun phrases in (14) are actually θ-marked by the verb, and
grammatically function as direct objects. They will check Case just like other
direct objects do, that is, by raising (overtly or covertly) to the outer specifier of
a light verb v (Chomsky 1995), or to the specifier of some Agreement
projection.
(15) ... [vP [DP her happy smile ]k [v' she [v' smiledi [VP ti tk ]]]]

2
This is also captured by Quirk et al.'s (1985: 750) observation that cognate objects
are similar to the "resultant objects" of verbs of creation. Note that the resultative
construction (which is discussed in section 5 below) also involves a form of aspectual
shifting towards a telic event.
5
Note that cognate objects do not behave like canonical Patient or Theme
arguments in the sense of referring to some pre-existing participant in the event
described. Rather, they denote something like the product, or outcome, of the
action of V-ing - a property which is included in Dowty's (1991) characteristics
of Proto-Themes. As their presence often has a telicizing (i.e., aspectual) effect
on the predicate, we suggest that cognate and reaction objects are best
characterized as Aspectual Themes (see also section 4 below).3

3. Cognate Objects and the Incorporation Hypothesis


Among many others, Hale & Keyser (1993) assume that the external argument
role is not part of a verb's lexical specification, or thematic grid. If we assume,
following Hale & Keyser (1993) and Chomsky (1995), that the Agent or
Causer argument orginates in the specifier position of an (agentive or
causative) light verb v, and that it is θ-marked not by the verb directly, but
compositionally by the entire v-VP complex, then unergative structures present
us with the somewhat curious situation in which the core VP contains nothing
but a single, argumentless verb (cf. (16)).
(16) vP
tu
DPsu v'
tu
v VP
1 g
V v tV

According to Hale & Keyser, however, the situation is not quite like it seems:
They claim that all verbs have an internal argument as part of their lexical
representation, and that unergative verbs are in fact concealed transitives.
Based on the idea that "the proper representation of predicate argument
structure of verbs is itself a syntax", Hale & Keyser (1993:53) suggest that
unergatives are denominal verbs derived via incorporation of a nominal
complement into a verbal head (a process dubbed "conflation" by Hale &
Keyser 1999). The incorporated element leaves behind a trace in V-
complement position.

3
This idea bears a certain resemblance to Tenny's (1992: 2) suggestion that aspectual
roles are the ones that are relevant to syntax ("Only the aspectual part of thematic structure is
visible to syntax").
6
(17) V’ V’
2 ti
V N ==> V N
g g 2 g
∅ smile N V t
g g
smile ∅

We saw above that with regard to their syntactic behaviour, cognate and
reaction objects resemble direct objects of transitive verbs rather than
nonargument DPs. The presence of an Aspectual Theme argument in syntax,
however, is potentially problematic for a Hale/Keyser-style analysis of
unergatives, at least under the widely held assumption that a head cannot assign
the same theta-role - or, for that matter, the same aspectual role - twice. That is,
if the nominal complement of V is seen as denoting the "object", or product,
created by performing the action of V-ing (for example, a smile is created by
somebody’s smiling), then how come unergatives admit cognate objects in
addition to this? It is also implausible to analyse cognate objects as
incorporated objects, as full DPs are not normally available for head-
incorporation, and in addition must have their Case checked via raising into the
checking domain of an appropriate functional head. Also, on the assumption
that all internal arguments originate within the core VP, we may have trouble
accomodating an an additional Goal argument under the Hale & Keyser (1993)
approach:
(18) a. She smiled her most radiant smile at him.
b. At last, the committee nodded their approval at the candidate.
If Kayne (1994) is correct in that rightward movement is universally barred,
then heavy DP-shift data indicate that the the cognate object (the Aspectual
Theme) originates lower than the Goal argument, that is, in V-complement
position (cf. (19b)) (a fact that tends to be obscured by the application of overt
object raising with less heavy DPs).
(19) a. She smiled at him the most beautiful smile he had ever seen.
b. She smiledi [VP [PP at him ] [V' ti [ DP the most beautiful smile
he had ever seen ]]]

Hale & Keyser (1997) explicitly discuss cognate object structures, and by way
of accounting for the optional presence of an object DP, suggest a mechanism
7
of "index deletion" that renders the trace of the incorporated nominal invisible
to the syntax (and hence, to semantic interpretation).
(20) V’
tp
V [e]
2
N V
g g
smile ∅

Index deletion thus makes it possible for an "adverbial increment" (such as a


cognate object) to be generated in V-complement position, replacing the
nominal component of the verb. The authors admit, however, that "[t]he
acquisition of this increment is [...] totally mysterious to us at this point" (Hale
& Keyser 1997:44). We believe that the Hale/Keyser analysis of denominal
verbs should be treated with some caution, on the following grounds:
First, observe that some denominal verbs differ considerably in their
phonological form from their alleged incorporata (e.g. death > die, song >
sing), a kind of change which is not normally effected by syntactic head
movement.
Second, as Hale & Keyser note themselves (1993:95), they are forced to
allow for more or less complex VP constituents to undergo lexical insertion
(but see Jackendoff 1997:83-91, 161-164 for a similar proposal). However,
merging maximal projections that have been formed in accordance with
syntactic principles of structure-building conflicts with the commonly held
view that the lexicon and the syntax are two (functionally) distinct components,
with the lexicon listing all and only the idiosyncratic, non-predictable
properties of lexical items, and with the standard view on lexical insertion,
according to which a lexical element replaces a terminal symbol, or X0
category, in the phrase-structure representation.
Third, as has been pointed out by Kiparsky (1997), it appears that the
incorporation hypothesis incorrectly predicts the existence of certain types of
ungrammatical structures. If denominal verbs are derived via N-incorporation,
then what rules out, for instance, examples like the following that contain a
"stranded" modifier?
(21) a. *We saddled her horse Western.
(cf. We put a Western saddle on her horse.)
b. *They danced Western / rain.
(cf. They danced a Western / rain dance.)
Fourth, postulating a novel - and apparently optional - mechanism of index
deletion looks rather like an ad hoc solution. On the assumption that index
8
deletion involves trace deletion, Hale & Keyser's analysis also has trouble
correctly ruling out examples like (22) below (an observation attributed to Joan
Bresnan, see Hale & Keyser 1997:42n.10, 44):
(22) *She saddled a western saddle on the mare.
(Cf. She saddled the mare with a western saddle.)
Kiparsky (1997) cites further data that are potentially problematic for Hale &
Keyser's approach. Consider (23):
(23) a. to dump garbage by the roadside
b. to blanket an area with advertising
If in (23), the nominal component of the verb has been replaced by an adverbial
increment, then how come the two need not be semantically related? Or, to put
the question another way, how can we ever hope to provide a coherent semantic
characterisation of the verb if phrases such as by the roadside and with
advertising can take the place of the verb's L-syntactic arguments (at a) dump
and (with a) blanket?
In sum, there are several reasons for being skeptical about the syntactic
incorporation analysis. Instead, we prefer to adopt the traditional assumption
that denominal verbs are derived by a lexical conversion rule (which does not
face any problems with changes in the phonological form). Regarding their
subcategorisation properties, we suggest that they are like ordinary transitives
in that they license a Theme argument.

4. The Null Object Hypothesis


In this section, we propose that as far as their argument structure is concerned,
unergatives are like ordinary transitives. Specifically, we suggest that the
Theme argument of unergatives is necessarily associated with a syntactic
position, even though it will not normally be spelled out.
(24) V'
tp
V XP
g g
smile {her happy smile / ∅}
Observe that many transitive verbs can also occur without an overt Theme:
(25) a. Bill drank.
b. They were eating.
c. She gave to charity.

9
Mittwoch (1982) argues that sentences like the above always imply the internal
argument of the verb. That is, Bill drank entails Bill drank something. The same
appears to be true for unergative structures such as (26a), which entails (26b):
(26) a. Mary smiled.
b. Mary smiled some kind of smile.
The fact that in relation to (26a), it is legitimate to ask What kind of smile did
Mary smile? constitutes further evidence that the "missing" object is always
understood. According to Mittwoch, the understood Theme in (25) is best
characterized as an unquantified indefinite pronoun (a nonquantificational, or
partitive, counterpart of something). This idea is captured by the following
meaning postulate (adapted from Fodor & Fodor 1980, as cited in Mittwoch
1982):
(27) x SMILE ≡ (∃y) x SMILE y
As an alternative to a full DP, both unergatives and transitives can combine
with unspecified quantity pronominals such as a lot or too much.
(28) a. He drank a lot / too much / more than necessary.
b. She laughed a lot / too much / more than necessary.
Note that these pronominals have a partitive interpretation, and typically
modify bare indefinite nominals as in a lot of wine, too much chocolate. In (28)
above, they can be understood as modifying (or quantifying over) the
understood Theme argument (cf. more X than necessary).4
Next, observe that the implied Theme can also be referred to by an
anaphoric pronoun:5

4
Mittwoch (1982: 116) maintains that expressions such as a lot, when following an
intransitive verb, are ambiguous between duration and frequency, whereas they can be
interpreted as modifying the understood object of objectless transitive verbs. We disagree
with this, and instead claim that these expressions can also be interpreted as modifying the
understood unspecified object of unergatives. Compare
(i) ?She was capable of smiling a lot of different types of smile
(ii) ?He tended to dance too much foxtrot, but not enough waltz
5
Mittwoch (1998), who analyses cognate objects are overt realisations of Davidsonian
event arguments, claims that the anaphoric pronouns in sentences such as (28a) take the
event argument as their antecedent. While her proposal is not dissimilar in spirit to the one
defended here, it strikes us as potentially problematic: First, since cognate objects (in
Hebrew) also occur with transitive and stative verbs, Mittwoch is forced to assume, contra
Kratzer (1995), that an abstract event variable is part of the lexical specification of all verbs.
In that case, we may wonder why the distribution of cognate objects in languages like
English is limited to a comparatively small number of (non-stative) verbs. Secondly, given
the evidence for the relative hierarchical prominence of the event argument (see again
Kratzer 1995), it seems strange that it should be spelled out by an expression in V-
complement position.
10
(29) a. Bill sang too beautifully for us to ignore (it). (Mittwoch 1998: 311)
b. Mary laughed, but it wasn't a cheerful laugh.
Moreover, it is possible for adverbs to make reference to the understood object
(compare Mittwoch 1998).
(30) a. John wrote illegibly.
b. Mary laughed inaudibly.
In (30a,b), it is the writing produced by John that is illegible, and the laugh
uttered by Mary that is inaudible. Compare also Quirk et al.'s (1985) claim that
the sentence pairs like those in (31) are equivalent in meaning:
(31) a. She smiled beautifully. / She smiled a beautiful smile.
b. They fought desperately. / They fought a desperate fight.
In short, there is evidence that in objectless transitive and unergative structures,
the understood Theme argument is actually part of the sentence's semantic
representation. However, the Theme argument of unergatives is not an event
participant in the same way as the Theme selected by eat or drink is. Rather, it
usually serves to delimit the event described aspectually, and hence was
labelled an "Aspectual Theme" above.
Normally, the internal argument of an unergative verb will only be
spelled out if it is modified, that is, if it is informationally richer than the
corresponding (and tautological) bare noun. This generalisation holds for
transitive verbs too - compare the oddness of (32):
(32) a. ??She smiled a smile / smiles.
b. ??They donated a donation (to the museum).
Since cognate objects do not make any substantial contribution to the semantics
of the clause, they must usually be modified in order to be deemed acceptable.
As the examples in (33) below illustrate, the situation is different for reaction
objects. Roughly speaking, they are abstract nouns, expressing an attitude
which can be made visible by the action denoted by the verb. Minimally, the
semantics of the abstract noun must be compatible with the production of the
indefinite object of the verb (that is, selectional restrictions apply). For
instance, while the expression of one's approval does is not contingent on the
production of a smile, one can nevertheless easily imagine that it includes a
smile - hence the acceptability of smile one's approval. The expression of a
negative attitude such as disapproval, by contrast, would not normally be
associated with the act of smiling, which is why smile one's disapproval sounds
very odd.
(33) a. She smiled a thankful / an approving smile.
b. She smiled her thankyou / her approval.
Where the Aspectual Theme is unmodified (or fully tautological), its realization
is unnecessary, and hence will normally be precluded for reasons of economy.
11
We suggest that the understood Theme in apparently objectless transitive or
unergative structures is a null pronominal. For the sake of concreteness, we will
take the empty category in question to be "small" pro.6 The null object will then
check structural objective Case in exactly the same way as its overt
counterparts do.7
While the presence of an overt complement will often have a telicizing
effect, this effect is never triggered by a null Theme. This situation is parallel to
that of transitive verbs being used intransitively or with nonquantificational
complements (compare Mittwoch 1982).8
(34) a. John drank (beer) for/*in five minutes.
b. John drank three beers in/*for five minutes.

(35) a. Last night, Mary danced for/*in five minutes.


b. Last night, Mary danced the third piece from Swan Lake in/*for five minutes.
The telicizing effect of quantified cognate objects can also be illustrated by the
following framing construction:
(36) Being handicapped by her sprained ankle, it took her five minutes and ten seconds
to dance the third piece from Swan Lake.
In general, adding an overt definite or otherwise "quantificational" object turns
activities such as eat into accomplishments. Hence the following entailment
relations ("imperfective paradox") hold:
(37) a. John was eating. ⊃ John ate. 9
b. John was eating an apple. ⊃ John ate an apple. 8

6
Note that null objects are independently attested in languages like Italian (Rizzi
1986).
7
The idea that unergatives take a null Theme argument is consistent with the
observation that in colloquial German, it is possible to question the understood object of
unergative verbs such as lachen 'laugh', but not of unaccusatives such as ankommen 'arrive'.
(i) Was lachst du?
what laugh you
'Why are you laughing?' (Lit.: 'What are you laughing?')
(ii) *Was kommst du an?
what come you at (Lit.: 'What are you arriving?')
Even though was 'what' in (i) is pragmatically understood in the sense of 'why', its formal
status seems to be that of an object wh-pronoun questioning an (implicit) unspecified object
of lachen 'laugh'. The verb ankommen 'arrive', on the other hand, is a true unaccusative in
that it does not admit cognate or reaction objects.
8
Rapp (1997) points out that transitive verbs can occur intransitively only (at least in
German) when their direct object is an incremental Theme - which supports our assumption
that intransitive verbs are just like transitive verbs with an unspecified incremental Theme (it
should be noted, however, that Rapp's analysis captures this similarity by assuming a lexical
rule of argument deletion in the case of transitive verbs occuring intransitively - i.e. she
allows for transitive verbs to become intransitives, while we consider intransitives to be
underlying transitives).
12
Parallel entailment relations hold for unergative verbs:
(38) a. Mary was dancing. ⊃ Mary danced. 9
b. Mary was dancing a piece from Swan Lake.
⊃ Mary danced a piece from Swan Lake. 8
Parallels between the aspectual function of the object of unergative and
transitive verbs can also be illustrated by the halfway-test (see e.g. Tenny
1989). The ongoing change in the event is measured out by the ongoing change
in the object of the verb (compare the paraphrases in (39) below). The
measuring noun phrase roughly corresponds to Dowty's (1991) Incremental
Theme, which is found with verbs of creation, verbs of consumption and verbs
of definite change of state.
(39) a. He ate the apple halfway. / He ate half of the apple.
b. She sang the song halfway. / She sang half of the song.
c. They ran the race halfway. / They ran half of the race.
From a theoretical point of view, postulating that a null Theme is part of the
syntactic representation of the sentence has several advantages as well.
i. Single lexical entry. Assuming that unergatives are always transitive
makes it unnecessary to postulate multiple lexical entries for the same verb.
Rather, all verbs minimally require one internal argument (which, under certain
conditions, may be phonetically null), and different classes of verbs differ,
among other things, with respect to whether or not they additionally admit an
Agent (and/or Goal) argument. In fact, the presence versus lack of agentivity
seems to be the main difference between transitive/unergative predicates on the
one hand, and unaccusative verbs on the other (see Arad 1997 for a similar
view).
The basic requirement that a verb must minimally specify one (internal)
argument in its thematic grid apart, our proposal is compatible with the idea
that verb semantics and selectional restrictions permitting, argument structure
templates can be freely augmented up to a maximum thematic grid (compare
Rappaport Hovav & Levin 1998).
(40) a. Phil swept (the floor). (non-directed activty)
b. Phil swept the floor clean. (change of state)
c. Phil swept the crumbs off the table. (change of place)
It seems reasonable to assume that whenever a lexical head enters the syntax,
complement and specifier positions come for free, i.e. are potentially available.
According to Chomsky (1995), the operation Merge is computationally
costless, so that including a null Theme in the syntactic representation does not
violate Economy.
Regarding the argument status of the null Theme, we assume that it is
anchored in the verb's lexical semantic representation. Following Rappaport
13
Hovav & Levin (1998) we will assume that there are at least two different ways
for an argument to be licensed semantically: It may be either a variable or a
constant in the verb's semantic representation. While variables generally allow
for a wide choice of referential indices and functional specification (Mary hit
Bill, Tim, or the wall), the spelling out of argument constants is more restricted
because the constant carries (part of) the idiosyncratic meaning of a verb within
its semantic class. We will assume that the null Theme acquires its argument
status as part of the "Manner" component that is typical of unergative verbs
(compare e.g. Jackendoff 1990:88-90). That is, contrary to Hale & Keyser (and
with Kiparsky 1997), we assume that a verb's semantic or conceptual structure
does play an important role in determining the range of possible complements.
The sentence *She saddled a western saddle on the mare is ruled out, in our
system, because (i) the verb saddle is not derived by syntactic incorporation
(hence the possibility of index deletion, allowing subsequent insertion of an
adverbial increment, does not exist), and (ii) saddle s-selects a complement
denoting an entity that can be saddled (i.e., a substantive rather than a "light"
Aspectual Theme).
ii. Accounting for the presence of an external argument. Consider the
way arguments are mapped into the syntax: Many scholars have claimed that
the relative ordering of thematic roles is determined by a (supposedly
universal) thematic hierarchy. Compare e.g. the thematic hierarchy proposed by
Grimshaw (1990):
(41) (AGENT (EXPERIENCER (GOAL/SOURCE/LOCATION (THEME))))
If we assume a flexible approach to the mapping of argument structure to
syntax such as Larson's (1988:382) Principle of Argument Linking,
(42) If a verb determines θ-roles θ1 , θ2 , ... θn, then the lowest role on the Thematic
Hierarchy is assigned to the lowest argument in constituent structure, the next
lowest role to the next lowest argument, and so on.
then our claim that in VPs headed by an unergative verb, the V-complement
slot is blocked by a null Theme will account for the otherwise mysterious fact
that the Agent argument cannot possibly originate in V-complement position
(cf. There arrived/*laughed many new students).

14
(43) vP
tu
DP v'
AGENT tu
v VP
1 tu
V v tV DP
g
CO/∅
THEME
*AGENT
If we follow Chomsky (1995) in assuming that constituent structure is built up
in a bottom-up fashion, by successive application of a binary merger operation,
and that θ-role assignment is a property of merger, then it seems that the only
alternative ways of explaining the syntactic prominence of the Agent argument
in unergative structures would be (i) to claim that unergatives lack a thematic
grid altogether, and the Agent role is assigned independently by an abstract
light verb v (see Ritter & Rosen 1998), or (ii) to claim that the Agent role is
specially annotated in the verb's thematic grid (or on the Thematic Hierarchy),
which then "tells" the syntax that it should be inserted into a VP-external
argument slot (compare e.g. Grimshaw 1990). Neither of these two possibilities
seems particularly attractive though.
The null object hypothesis helps avoid the odd situation mentioned above
in which the core verb phrase consists of a single V only - a situation that
violates the antisymmetry requirement proposed by Kayne (1994). Also, to the
extent that it is a defining characteristic of predicates that they have argument
structure, the existence and theoretical status of argumentless verbs seems
questionable.
iii. Compositionality. Assuming the presence of an understood Theme in
syntax is consistent with the principle of compositionality, that is, with the view
that syntactic representations serve as the input to interpretation. Under this
view, an argument expression that contributes to interpretation (and which has
syntactic effects) must be part of the syntactic representation of a sentence.
Table (44) summarizes the different types of argument realisation
patterns that unergatives permit:
(44) THEME GOAL
She smiled pro --
She smiled at him pro PP
She smiled her radiant smile DP --
She smiled her radiant smile at him DP PP

15
Under this view, and as predicted by Burzio's Generalisation, there is a strong
correlation between the possibility of a Theme argument checking objective
Case and the presence of an Agent argument. In Chomsky's (1995) minimalist
framework, both are associated with properties of the functional category v.
Where no vP shell is projected (as is arguably the case for unaccusatives), the
Theme argument must raise to an external subject position in order to check
structural nominative. By definition, unaccusative verbs have only one
argument, which fills the lowest thematic and syntactic argument position.
There is no position available for an additional cognate (or other) object.

5. Resultative Constructions
Resultative constructions have frequently been analysed as involving a small
clause complement, with the postverbal DP functioning as the small clause
subject, and the adjectival or prepositional phrase denoting the "result state"
functioning as the small clause predicate (Hoekstra 1988, cf. (45a)).9 Within
predication theory (which does not recognize small clauses as constituents),
these constructions have been analysed as involving two independent but
coindexed arguments of the main verb (see e.g. Carrier & Randall 1992, cf.
(45b)). Larson's (1988) stacked-VP hypothesis presents us with a third option:
resultative complements can be analysed as syntactically derived complex
predicates, as indicated in (45c) below.
(45) a. She smiled [SC herself tired ]
b. She smiled [DP herself]i [AP tired ]i
c. Shek [vP tk [v' smiledi [VP herself ti [AP tired ]]]
Adopting a vP-shell structure as in (45c) for resultative constructions allows for
the principle of binary branching to be obeyed.
The fact that resultative small clauses do not alternate with infinitival or
finite complements indicates that unergatives do not admit fully propositional
complements (i.e., clauses of the category TP or higher).
(46) a. *They laughed themselves to be in a frenzy / to be silly.
(vs. They considered him to be foolish.)
b. *They laughed that they were in / got into a frenzy.
Intransitive resultative constructions such as (47b) below differ from transitive
ones like (47a) in that the postverbal DP cannot plausibly be analysed as a θ-
marked argument of the verb (compare e.g. Carrier & Randall 1992).

9
A more recent version of the small clause analysis has been proposed by Bowers
(1997), who assumes that the resultative small clause is headed by the functional category
Pr(edicate).
16
(47) a. She wiped the table clean. / She wiped the table.
b. She smiled herself tired. / *She smiled herself.
In (47a), She wiped the table clean entails that she wiped the table; the reading
that the table became clean by her wiping some other object is not available.
Herself in (47b), by contrast, is not a possible object of smile. Kaufmann &
Wunderlich (1998) label these "weak" and "strong" resultatives respectively,
with the latter, but not the former, involving a non-subcategorized DP
argument.
Observe further that the main verb imposes selectional restrictions on the
embedded predicate: the resultative phrase can only be of the category AP or
PP, but not DP or VP:
(48) They laughed themselves... a. [AP silly ]
b. [PP into a frenzy ]
c. *[DP complete fools ]
d. *[VP fall(ing) over ]
(49) He drank himself... a. [AP unconscious ]
b. [PP into a stupor ]
c. *[DP a wreck ]
d. *[VP die ]
Carrier & Randall (1992) argue convincingly that the XP denoting the result
state is an argument of the verb. From the point of view of our analysis, this
means that it is possible for the Aspectual Theme to be realized not only as a
noun phrase (i.e., as a cognate or reaction object), but also as an adjectival or
prepositional "result" phrase. AP and PP constituents are special, however, in
that they are not referential expressions, but are themselves interpreted as
predicates. That is, they may require a Theme argument themselves (cf. Carrier
& Randall 1992:177). Hale & Keyser (1997) have claimed that adjectival and
prepositional phrases, but not DPs or VPs, can function as predicates, and that
"VP may have a specifier only if its complement is a predicate", i.e., AP or
PP.10 The latter assumption, in turn, derives from the more fundamental
requirement that every predicate must have a subject (Rothstein 1995). The
observation that with unergative verbs, resultative constructions are possible
only if the resultative predicate is an AP or PP is consistent with Hale &
Keyser's (1997:39) claim. The assignment of the primary (Theme1) to the
complement of V and of the secondary Theme role (Theme2) to the noun phrase
in V-specifier position is illustrated in the diagram below.

10
Alternatively, the observation that unergatives do not take VP complements (cf. *She
laughed herself fall over) could be attributed to the idea that "bare" VPs are generally unable
to function as arguments (compare Felser 1999).
17
(50) vP
tp
DPsu v'
tp
v VP
1 tp
V v DP V'
THEME2 tu
tV AP
PP
*DP
*VP
THEME1
14243

She laughed herself silly


into a frenzy
*complete fools
* fall over
Semantically, the verb in (48) and (49) forms a complex predicate with its
adjectival or prepositional complement, so that the predicate laugh silly takes
the reflexive anaphor herself as its Theme.11 Evidence that "bare" adjectives or
prepositions may be syntactially incorporated into V is provided by heavy DP-
shift data such as (51) below, where subsequent to incorporation, the V-A
complex appears to have undergone verb movement (compare also Williams
1997):
(51) He [ ran threadbare ]i [ his brother's favourite pair of trainers ] ti
Compare also particle verbs as in He looked up the information, and lexicalized
resultative constructions with adjectives in German:
(52) a. sich kaputtlachen
REFL broken.laugh
'to kill oneself laughing'
b. sich sattessen
REFL full.eat
'to eat one's fill'

11
Resultatives are also analysed as complex predicates by Cormack & Smith (1999),
who suggest that they are semantically headed by an (asymmetric) two-place conjunction
operator.
18
The "fake reflexive" here is introduced because it is the only way to make the
resultative predicate take the referent of the subject argument as its Theme
(Levin & Rappaport Hovav (1995) call this the "direct object restriction",
stretching the term as far as to include the subject of a small clause
complement). Note that intransitive verbs of the unaccusative type do not occur
in reflexive resultative constructions because their subject argument originates
in the position of an internal argument and can therefore function itself as the
Theme of the resultative predicate:
(53) a. [The lake]i froze ti / *itselfi solid.
b. [The toasts]i burnt ti / *themselvesi black.
In sum, it appears that besides combining with cognate objects, unergatives
also take adjectival or prepositional result phrases as their complement (which
also describe, in some sense, the product or outcome of the action of V-ing).
Contrary to cognate objects, however, AP or PP complements are themselves
predicates, that is, they require a Theme argument as the subject of the
predication. The complex predicate formed by merging an AP or PP with an
unergative verb will therefore force the presence of an additional (secondary)
Theme argument in the specifier position of V.12

6. Conclusion
Summarizing, we have suggested (in the spirit of Hale & Keyser) that true
Agent-only verbs do not exist. That is to say, all verbs minimally have one
(internal) argument position specified in their thematic grid. Unlike Hale &
Keyser, however, we have proposed that the internal argument of intransitives -
a "light" or Aspectual Theme - is necessarily associated with a syntactic
argument position. Cognate and reaction objects are technically complements
of V, and associated with an aspectual θ-role. So-called unergative verbs, as
well as a number of transitives, permit their internal argument to be an empty
category (a null pronominal). Only if the Theme argument is spelled out can it
have a telizizing effect on the event described. A null Theme, by contrast, is
understood as an unquantified indefinite pronoun, and as such is incapable of
functioning as an aspectual delimitor. Regarding intransitive (or "strong")
resultative constructions, we have argued that they involve lexical or syntactic
complex predicate formation, and compositional assignment of a secondary
Theme role.

12
Though we do not share Ritter & Rosen's (1998: 142) view that "the single argument
of a monadic verb is freely inserted in object or subject position", we would not want to rule
out their analysis of resultatives, which involves embedding of the small clause in a
functional projection whose head encodes telicity (compare Borer 1993).
19
What are the consequences of our proposals for the syntactic
classification of verbs in general? If unergatives are transitive verbs after all,
does this mean that the traditional distinction between transitives on the one
hand, and intransitives on the other, is actually meaningless? This is not
necessarily the case. Rather, given our analysis, it seems that the focus of verb
classification should be shifted. There appear to be some verbs that can never
be used transitively, namely those unaccusative verbs that do not belong to the
subclass of "ergative" verbs.
(54) a. The guests arrived.
b. *The chauffeur arrived the guests.
c. *They arrived their usual noisy arrival.
d. *They arrived themselves breathless.
(55) a. Her earring disappeared.
b. *The children disappeared her earring.
c. *Her earring disappeared a mysterious disappearance.
d. *Her earring disappeared itself invisible.
While unergatives turn out to pattern semantically and syntactically like
transitives, true unaccusative verbs are genuinely different, and appear to
constitute a class of their own. They are genuine intransitives insofar as they
are the only "single-argument" verbs. Whenever a verb selects for a single
argument only, this must necessarily be mapped into a VP-internal argument
slot - the position usually associated with the Theme role. From our point of
view, then, genuine intransitive verbs are those which lack an external
argument, such as (such as arrive, disappear, or wilt).13 Syntactically,
unaccusatives fail to project a vP shell, and thus fail to provide the
configuration necessary for checking objective Case.14

7. References
Arad, M. 1997. Are unaccusatives aspectually characterized? (And other related questions).
In MIT Working Papers in Linguistics 32: Papers from the UPenn/MIT roundtable
on argument structure and aspect, ed. H. Harley, 1-20. MITWPL, Department of
Linguistics and Philosophy, MIT, Cambridge, MA.
Borer, H. 1993. The projection of arguments. In: University of Massachusetts Occasional
Papers 17: Functional projections, ed. E. Benedicto and J. Runner. GLSA,
University of Massachusetts, Amherst, MA.
13
Compare Arad's (1997: 15) claim that "unaccusatives are not characterized by having
a particular property, but by lacking something (agentivity of some kind)."
14
Under the assumption that transitive verbs that participate in the causative alternation
allow for a lexical rule that deletes or binds their external argument (Levin & Rappaport
Hovav 1995, Wanner 1999), one-argument verbs of the so-called "ergative" type (as in The
door opened) will be considered derived intransitives.
20
Bowers, J. 1997. A binary analysis of resultatives. In Texas Linguistics Forum 38: The
Syntax and Semantics of Predication, eds. R. Blight and M. Moosally, 43-58.
Department of Linguistics, The University of Texas at Austin.
Burzio, L. 1986. Italian syntax. A government-binding approach. Dordrecht: Reidel.
Carrier, J., and J. Randall. 1992. The argument structure and syntactic structure of
resultatives. Linguistic Inquiry 23:173-234.
Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris.
Chomsky, N. 1995. The minimalist program. Cambridge, MA: MIT Press.
Cormack, A., and N. Smith. 1999. Why are depictives different from resultatives? In
University College London Working Papers in Linguistics 11, 251-284.
Dowty, D. 1991. Thematic proto-roles and argument selection. Language 67: 547-619.
Felser, C. 1999. Verbal complement clauses. Amsterdam & Philadelphia: John Benjamins.
Fodor, J. A., and J. D. Fodor. 1980. Functional structure, quantifiers, and meaning
postulates. Linguistic Inquiry 11:759-770.
Grimshaw, J. 1990. Argument structure. Cambridge, MA: MIT Press.
Hale, K., and S. J. Keyser. 1993. On argument structure and lexical expression of syntactic
relations. In The view from building 20: Essays in Linguistics in honor of Sylvain
Bromberger, ed. K. Hale and J. Keyser, 53-110. Cambridge, MA: MIT Press.
Hale, K., and S. J. Keyser. 1997. On the complex nature of simple predicators. In Complex
Predicates, ed. A. Alsina, J. Bresnan, and P. Sells, 29-66. Stanford: CSLI
Publications.
Hale, K. and J. Keyser. 1999. A response to Fodor and Lepore, 'Impossible Words?'.
Linguistic Inquiry 30:543-466.
Hoekstra, T. 1988. Small clause results. Lingua 74:101-139.
Jackendoff, R. 1990. Semantic structures. Cambridge, MA: MIT Press.
Jackendoff, R. 1997. The architecture of the language faculty. Cambridge, MA: MIT Press.
Jones, M. 1988. Cognate objects and the case filter. Journal of Linguistics 24:89-110.
Kaufmann, I., and D. Wunderlich. 1998. Cross-linguistic patterns of resultatives, Arbeiten
des SFB 282 Theorie des Lexikons 109, Heinrich-Heine-Universität Düsseldorf.
Kayne, R. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.
Kiparsky, P. 1997. Remarks on denominal verbs. In Complex Predicates, ed. A. Alsina, J.
Bresnan, and P. Sells, 473-499. Stanford: CSLI Publications.
Kratzer, A. 1995. Stage-level and individual-level predicates. In The Generic Book, ed. G.
Carlson and F. J. Pelletier, 125-175. Chicago: University of Chicago Press.
Larson, R. 1988. On the double-object construction. Linguistic Inquiry 19:335-391.
Levin, B., and M. Rappaport. 1986. The formation of adjectival passives. Linguistic Inquiry
17:623-661.
Levin, B., and M. Rappaport Hovav. 1995. Unaccusativity at the syntax-lexical semantics
interface. Cambridge, MA: MIT Press.

21
MacFarland, T. 1994. Event structure and argument structure in cognate objects. In ConSole
I Proceedings, ed. P. Ackema and M. Schoorlemmer, 165-182. The Hague: Holland
Academic Graphics.
Mittwoch, A. 1982. On the difference between eating and eating something: Activities
versus accomplishments. Linguistic Inquiry 13:113-122.
Mittwoch, A. 1998. Cognate objects as reflections of Davidsonian event arguments. In
Events and Grammar, ed. S. Rothstein, 309-322. Dordrecht: Kluwer.
Perlmutter, D. 1978. Impersonal passives and the unaccusative hypothesis. In Proceedings of
the Berkeley Linguistic Society 4, 157-189. Berkeley Linguistics Society, University
of California.
Quirk, R., S. Greenbaum, G. Leech, and J. Svartvik. 1985. A comprehensive grammar of the
English language. London: Longman.
Rapp, I. 1997. Fakultativität von Verbargumenten als Reflex der semantischen Struktur.
Linguistische Berichte 172:490-529.
Rappaport Hovav, M., and B. Levin. 1998. Building verb meanings. In The projection of
arguments: lexical and compositional factors, ed. M. Butt and W. Geuder, 97-134.
Stanford: CSLI Publications.
Ritter, E., and Rosen, S. T. 1998. Delimiting events in syntax. In The projection of
arguments: lexical and compositional factors, ed. M. Butt and W. Geuder, 135-164.
Stanford: CSLI Publications.
Rizzi, L. 1986. Null objects in Italian and the theory of 'pro'. Linguistic Inquiry 17:501-557.
Rothstein, S. 1995. Pleonastics and the interpretation of pronouns. Linguistic Inquiry
26:499-529.
Tenny, C. 1989. The aspectual interface hypothesis. MIT Lexicon Project Working Papers
31. MITWPL, Department of Linguistics and Philosophy, MIT, Cambridge, MA.
Tenny, C. 1992. The aspectual interface hypothesis. In Lexical Matters, ed. I. Sag and A.
Szabolcsi , 1-27. Stanford: Stanford University Press.
Tenny, C. 1994. Aspectual roles and the syntax-semantics interface. Dordrecht: Kluwer.
Verkuyl, H. 1993. A theory of aspectuality. Cambridge: Cambridge University Press.
Wanner, A. 1999. Verbklassifizierung und aspektuelle Alternationen im Englischen.
Tübingen: Max Niemeyer Verlag.
Williams, E. 1997. Lexical and syntactic complex predicates. In Complex Predicates, ed. A.
Alsina, J. Bresnan, and P. Sells, 13-28. Stanford: CSLI Publications.

22

You might also like