Dualist Syntax

Stephen Wechsler
University of Texas at Austin

Proceedings of the 15th International Conference on

Head-Driven Phrase Structure Grammar

National Institute of Information and Communications Technology, Keihanna

Stefan Müller (Editor)
2008
CSLI Publications
pages 274–293
http://csli-publications.stanford.edu/HPSG/2008

Wechsler, Stephen. 2008. Dualist Syntax. In Müller, Stefan (Ed.), Proceedings

of the 15th International Conference on Head-Driven Phrase Structure Gram-
mar, National Institute of Information and Communications Technology, Keihanna,
274–293. Stanford, CA: CSLI Publications.
 Abstract
A dualist syntax has two components: (1) the lexicon, a structured set
of formatives (‘words’); and (2) rules for combining those formatives into
utterances. This paper defends syntactic dualism against three ‘monist’ chal-
lenges. First, evidence for lexical argument structure can be found in dever-
bal nominalization, which preserves that structure systematically. Second,
words represent the smallest units for idiom formation and contextual poly-
semy effects, which is expected on the dualist view but not if word meanings
are composed in the syntax. Third, the count/mass properties of nouns sug-
gest an interleaving of conceptual and grammatical information in semantic
composition.

1 The autumnal trees of monism

Like many theories, HPSG assumes that syntax is organized into two components:

(1) a. Lexicon: A structured set of formatives (‘words’).

b. Combinatory syntax and semantics: Rules for combining those forma-
tives into utterances.

I introduce the term dualist syntax for this grammatical architecture (cp. lexicalism,
on one interpretation of this term). Under this dualist conception, a lexical entry
contains, among other things, subcategorization information indicating the local
syntactic contexts in which the word can appear. Meanwhile, the combinatory syn-
tax and semantics specifies language-wide instructions on how to combine words,
e.g. the verb precedes its object in English, but follows it in Japanese. Here’s a
familiar HPSG style lexical entry, followed by a tree:
 D E 
(2) draw: SUBJ DPi
 D E 
 
COMPS DPj 
 
  
CONTENT draw′ i,j

(3) S

DP VP

Sue V DP

drew pictures

Recent years have seen the rise of certain monist challenges to dualism (Marantz
1997; Borer 2005a,b; Harley 2004, inter alia). Such approaches eschew sub-lexical
syntactic or semantic structure such as semantic decomposition and lexical cate-
gory specifications, positing instead that apparent sub-lexical structure is actually

275
built in syntax in the same process by which words are combined into utterances.
Marantz (1997) sums up this challenge aptly with the slogan ‘The lexicon is dead’.
On the most extreme monist assumptions, content words such as verbs are
featureless radicals. Some or all thematic roles are eliminated from the lexical
entries, instead assigned by silent ‘light verbs’ such as ‘little v’ that are assumed
to occupy functional head positions in elaborate phrase structures that typically
include only a few words among many phonologically empty terminal nodes. In
the following simpified structure, draw lacks thematic roles; the silent ‘light verbs’
v and r assign the agent and theme roles to their respective specifiers:
(4) vP

DP v′

Sue v VP

V rP

draw DP r’

pictures r

To borrow an evocative metaphor (from Anthony Woodbury, p.c.), these are au-
tumnal trees, with many bare branches, to which only a few words cling like dead
leaves, as the winter of transformational syntax ominously approaches.
Is there a substantive, empirically testable difference between the dualist and
monist approaches? My search for substantive arguments has turned up three puta-
tively pro-monist, anti-dualist arguments in the literature. In this paper I argue that
in all three cases, the facts actually favor, if anything, syntactic dualism.

2 Argument one: deverbal nominals

Certain English causative alternation verbs allow optional omission of the agent
argument (5), while the cognate nominal disallows expression of the agent (6):1

(5) a. that John grows tomatoes

b. that tomatoes grow

(6) a. *John’s growth of tomatoes

b. the tomatoes’ growth, the growth of the tomatoes

In contrast, nominals derived from obligatorily transitive verbs such as destroy

allow expression of the agent, as shown in (8a):

(7) a. that the army destroyed the city

1
This section is based on Wechsler 2007.

276
 b. *that the city destroyed

(8) a. the army’s destruction of the city

b. the city’s destruction

Following a suggestion by Chomsky (1970), Marantz (1997) argued that these data
show that the agent role is lacking from lexical entries. In verbal projections (5)
and (7) the agent role is assigned in the syntax by little v. Nominal projections like
(6) and (8) lack little v so they lack a structural source for the agent role. Prag-
matics takes over to determine which agents can be expressed by the possessive:
the possessive can express ‘the sort of agent implied by an event with an external
rather than an internal cause’ because only the former can be ‘easily reconstructed’
(from Marantz 1997; see also Harley and Noyer 2000): the destruction of a city
has a cause external to the city, while the growth of tomatoes is internally caused
by the tomatoes themselves (Haspelmath, 1993; Smith, 1970).
Marantz points out that this explanation is unavailable if the noun is derived
from a verb with an argument structure specifying its agent if there is one. The
problem for a dualist syntax is that nothing can be plausibly expected to block the
deverbal nominal from inheriting the agent of a causative alternation verb.
The empirical basis for this argument is the mismatch between the allowability
of agent arguments, across some verb-noun cognate pairs: e.g. grow allows the
agent but growth does not. But how general is the grow/growth pattern? If it is
the norm, as implied by Marantz and others, then this may indeed suggest that
the agent role is supplied by the syntactic configuration. But conversely, if exact
matches between noun and verb are the norm, and especially if the few mismatches
can be independently explained, then this becomes powerful evidence for exactly
the position Marantz seeks to attack. It would show that the verb does specify its
agent role (or lack thereof) in the lexicon, and the noun inherits the agent if and
only if the verb has one.
The facts strongly support the latter generalization: near-total parallelism be-
tween verb and noun, with ready explanations for the few counter-examples. First
consider non-alternating theme-only intransitives (‘unaccusatives’), as in (9) and
transitives as in (10). The pattern is clear: if the verb is agentless, then so is the
noun:

(9) arriv(al), disappear(ance), fall, etc.:

a. A letter arrived.
b. the arrival of the letter
c. *The mailman arrived a letter.
d. *the mailman’s arrival of the letter

(10) destroy/destruction, construct(ion), creat(ion), assign(ment), etc.:

a. The army is destroying the city.

277
 b. the army’s destruction of the city

Already this favors the dualist view. For the monist, the badness of (9c) and (9d)
would have to receive independent explanations: (9c) is disallowed because a fea-
ture of the root ARRIVE prevents it from appearing in the context of v (Harley and
Noyer 2000), while (9d) would be ruled out because the cause of an event of arrival
can’t be easily reconstructed from world knowledge. This implausible duplication
in two separate components of the linguistic system would be replicated across all
the intransitive and non-alternating transitive verbs.
What about causative alternation verbs? The claim that the grow(th) pattern is
typical of causative alternation verbs will be dubbed Chomsky’s Conjecture:

(11) Chomsky’s Conjecture: Noun cognates of causative alternation verbs lack

the agent argument.

Besides grow(th), Chomsky (1970, examples 7c and 8c) cited two other examples,
both experiencer predicates: John amused (interested) the children with his sto-
ries versus *John’s amusement (interest) of the children with his stories. But this
was later shown by Rappaport (1983) and Dowty (1989) to have an independent
aspectual explanation. Deverbal experiencer nouns like amusement and interest
typically denote a mental state, where the corresponding verb denotes an event in
which such a mental state comes about or is caused. These result nominals lack not
only the agent but all the eventive arguments of the verb, because they do not refer
to events. Exactly to the extent that such nouns can be construed as representing
events, expression of the agent becomes acceptable.
In a response to Chomsky (1970), Carlota Smith (1972) surveyed Webster’s
dictionary and concluded that Chomsky’s Conjecture is false: ‘There are many
counterexamples to this [Chomsky’s] claim: explode, divide, accelerate, expand,
repeat, neutralize, conclude, unify, and so on at length.’ (Smith 1972:137) Harley
and Noyer (2000) also noted many so-called ‘exceptions’: explode, accumulate,
separate, unify, disperse, transform, dissolve/dissolution, detach(ment), disengage-
(ment). The simple fact is that these are not exceptions because there is no gener-
alization to which they can be exceptions. These long lists of verbs represent the
norm, especially for suffix-derived nominals (in -tion, -ment, etc.).
As for zero-derived nominals, many of these also allow the agent, such as
change, release, and use: My constant change of mentors from 1992-1997. The
frequent release of the prisoners by the governor. The frequent use of sharp tools
by underage children. (examples from Borer 2003, fn. 13). Pesetsky (1995:79,
ex. 231) assigns a star to the thief’s return of the money, but it sounds fine to me,
the OED lists a transitive sense for the noun return (definition 11a), and corpus
examples like her return of the spoils are easily found.
Like the experiencer nouns mentioned above, many zero-derived nominals lack
event readings, and thus reject all the arguments of the corresponding eventive
verb: *the freeze of the water, *the break of the window, and so on. Others

278
marginally allow event readings, and to the extent that they do, agents are pos-
sible. In my judgment, his drop of the ball is slightly odd, but the drop of the ball
has exactly the same degree of oddness.
In short, the facts seem to point in exactly the opposite direction from what has
been assumed by the monists. Chomsky’s Conjecture is false.
Now, what is special about grow(th)? The answer is simple. When the noun
growth entered the English language, causative grow did not exist! There was only
intransitive grow. The OED provides these dates of the earliest attestations of grow
and growth:

(12) a. intransitive grow: c725 ‘be verdant’ ... ‘increase’

b. the noun growth: 1587 ‘increase’
c. transitive grow: 1774 ‘cultivate crops’

Thus growth entered the language at a time when transitive grow did not exist. The
argument structure and meaning were inherited by the noun from its source verb,
and then preserved into present-day English. This makes perfect sense from the
dualist perspective in which words have predicate argument structures. Nominal-
ization by -th suffixation is not productive in English, so growth is listed in the
lexicon. To explain why growth lacks the agent we need only assume that a lexical
entry’s predicate argument structure dictates whether it takes an agent argument or
not. So even this one word, cited repeatedly in the anti-lexicalist polemics, turns
out to provide evidence for dualism.

3 Argument two: sublexical scope

3.1 Two approaches to sublexical scope
Monist approaches eschew sub-lexical semantic structure such as semantic decom-
position, positing instead that apparent sub-lexical structure is actually built in syn-
tax. This move has reopened an old debate between ‘Generative Semantics’ and
lexical decomposition, and involves some of the same phenomena as the earlier
debate (Lakoff 1965, Dowty 1979). Verbs like get, give, and transitive want incor-
porate a possession component:
(13) a. John wants the car. ↔ John wants to have the car.
b. John got the car. ↔ John came to have the car.
c. Mary gave John the car. ↔ Mary caused John to have the car.
Durative adverbials can modify the implicit “have” state (McCawley 1974; Ross
1976; Dowty 1979, inter alia):
(14) a. John wanted the car (for two days). (want or have for two days)
b. John got the car (for two days). (have for two days)

This one can also be ambiguous. Consider the situation in

279 which John needs to compete everyday with someone else
to get the car for that day. Then, the sentence means "John
got it for two days", and in the third day, someone else got it.
 The same thing, considering cases like "John gave
me his seat in class for two days" under similar situation
as the previous comment.

c. John gave me the car (for two days). (have for two days)
This suggests these sentences have an underlying semantic ‘have’ formative. The
question is how this formative enters the picture.
On one view ‘have’ is in the lexical decomposition of the verb, as in (15b)
for want (a simplified version of the analysis in Dowty 1979). The verb want1
in (15a) takes a clausal (or controlled) complement, as in John wants very much
[for it to rain]. The verb want2 in (15b) is the transitive variant in (14a). Using
an underspecification semantics such as Minimal Recursion Semantics (Copestake
et al. 2005), we need to do little more than merely introduce the ‘have’ state as
an elementary predication, as in (15c). This alone makes it available for durative
adverbials to scope over.

(15) a. want1 := λP λx[want′ (x, P )]

b. want2 := λyλx[want′ (x, have′ (x, y))]
 D E 
c. want2 : SUBJ DPi
 D E 
 
COMPS
 DP j


     
 
CONTENT s1: want s, i, s2 , have s2, i, j

See Egg (1999) and Beavers et al. (to appear) for detailed formal accounts of sub-
lexical scope within underspecification semantics.
The other approach posits a silent syntactic formative (McCawley, 1974), as in
the analysis by Harley (2004):

(16) S

DP VP

John
V PP

wants DP P′

PRO Phave DP

the car

Durative adverbials can adjoin to this putative PP, thus explaining the scope facts.
Harley (2003) motivated the PP on the basis of controlled PP complements of want:

(17) John wants [PRO off the team].

Harley argued that since want allows this type of complement anyway, we need
only posit the silent preposition HAVE.

280
3.2 Evidence from idioms and contextual polysemy
Which approach is right? An argument that the ‘have’ formative is syntactic was
put forth by McCawley (1974), and more recently revived by Richards (2001) and
Harley (2004). They note the parallel verb+DP idioms across have, want, get, and
give, such as give/get the creeps and give/take/get flak:
(18) a. John gave everyone flak.
b. You get flak (when you take a stand)
They explain the parallelism by positing a single underlying idiom, “HAVE flak”,
which then combines with causal or inchoative semantic formatives:
(19) a. John CAUSE everyone [ HAVE flak ].
b. You BECOME [ HAVE flak ].
According to their account, the verb have is the spell-out of BE+HAVE, get is
BECOME+HAVE, and give is CAUSE+HAVE. So the idiom parallels follow from
the syntactic approach to sub-lexical scope. However, on the lexical decomposition
view, the ‘have’ formative is embedded in a lexical decomposition (see (15a,c))
and hence unavailable to form idioms, since it is not a syntactic formative. On that
view the idiomatic interpretations would have to be stipulated separately for each
collocation.
In a different theoretical setting, McCawley (1974) made essentially the same
argument regarding want+DP, an argument later revived by Harley (2004:258-9):

significantly, the various “readings” that any have DP expression can

have are all available with a want DP expression. When have’s com-
plement is a DP that denotes offspring, like daughter or child, as in
John has a daughter, have easily receives a ‘parenting’ interpretation,
and this is exactly the most felicitous interpretation for the covert have
in John wants a daughter. (Harley 2004:258-9)

The central empirical claim, then, is that the same idioms that can be formed from
have can also be formed with the ‘have’ component of verbs like get, give, and tran-
sitive want. As far as I know this quite interesting empirical claim has never been
explored fully and systematically, although I’ve taken some initial steps (Wechsler
2008). Put more broadly, the theoretical question is this: What are the minimal
units from which idioms are composed?
So far, the facts support the lexical decomposition view (Wechsler (2008)).
First consider the want (to have) DP cases discussed by McCawley and Harley.
When the DP is relational as in John has a sister, the main predicate comes from the that is, "have"
is a light verb.
noun, not the verb. Simplifying somewhat, analyses along the following lines have See below.
long been proposed (Partee 1999, citing a 1987 Landman and Partee unpublished
abstract; Tham 2006; Wechsler 2006; Beavers et al to appear):

(20) a. have = λP λx∃y[P (x, y)]

281
 b. a sister = sister ′
c. a headache = headache′
d. John has a sister = ∃y[sister ′ (John, y)]
e. John has a headache = ∃y[headache′ (John, y)]

Details vary but the key for now is that the same have appears with all relational
nouns, whether sister, headache, etc. This analysis can be extended to the other
verbs in (21):
(21) a. want = λP λx[want′ (x, ∃y[P (x, y)])]
b. get = λP λx[BECOM E(∃y[P (x, y)])]
c. give = λyλP λx[CAU SE(x, BECOM E(∃y[P (x, y))]
d. John wants a sister = want′ (John, ∃y[sister ′ (John, y)])
e. Eliza got a headache. = BECOM E(∃y[headache′ (Eliza, y)])
f. The music gave me a headache. =
CAU SE(music, BECOM E(∃y[headache′ (me, y)]))
Beavers et al (to appear) propose a unified analysis of relational and non-relational
DP complements of these verbs of possession, citing non-zeugmatic coordination
like John has a nice car and an even nicer sister who bought it for him. Both
variants are treated as the light verb have, roughly (20a). In John has a nice
car, the possession relation comes from car, extending Barker’s (1995) analysis
of genitives like John’s car, in which the noun car is type-shifted to select a pos-
sessor argument. Anyway, for the present purposes, the crucial point is that we
don’t need many have’s such as a ‘parenting have’, ‘kinship have’, ‘disease have’
(for headaches), and so on. There is just one have for all relational nouns, and if
Beavers et al (to appear) are right then the same one is used for true possession as
well.
For the same reason, the collocations exhibiting parallelism (get flak, give flak,
etc.) are not really idioms. They are compositional phrases involving figurative
senses of the DP plus the standard ‘light’ meaning of the verbs. For example, flak
refers to ‘a barrage of abuse or adverse criticism’ (OED), and frequently appears
without any of the support verbs get, take, or give ((22a-c) are cited in the OED;
(22d,e) are from the British National Corpus):

(22) a. 1968 N.Y. Times 20 May, 46. In spite of the current flak between Mayor
Lindsay and...the...administrator of Boston and New Haven..., the poten-
tial for the city is unlimited.
b. 1969 A. LURIE Real People, 163. Well, all right. So why all the flak?
c. 1976 T. STOPPARD Dirty Linen, 25. Isn’t that going to cause rather a
lot of flak in the... P.L.P.?
d. Just imagine the flak flying about if we have bad results.

282
 e. I expect the flak. If we get beat, it’s my fault

In short, collocations like get flak are no more idiomatic than get criticism. Similar
comments apply to the other putative idioms that distribute across support verbs.
On the other hand, English has many truly non-compositional idioms. Cru-
cially, they do not exhibit this parallelism across support verbs. For example, have
a baby on the ‘give birth to a baby’ meaning does not transfer to the other verbs,
as shown in (23) (from Wechsler 2008).
(23) a. Natalie doesn’t want to have a baby, so she’s going to adopt one.
b. #Natalie doesn’t want a baby, so she’s going to adopt one.
As shown by the contrast in (23), the phrase want a baby, in contrast to have a
baby, is general with respect to the ways of satisfying this desire. This phrase is
not ambiguous between ‘want to give birth to a baby’ and other possibilities such
as adoption.
Many more idioms can be added to this ((24a,b) are from McCawley 1974):

(24) a. I had a ball. (‘enjoyed myself’)

*I want a ball.
b. I had it out with Fred. (‘argued angrily’)
*I want it out with Fred.
c. C’mon, have a heart and give my kid an A. (‘be compassionate’)
*I don’t want a heart, and besides, he flunked the exam.
d. The okra is ready. Go ahead, have at it! (‘do something heartily’)
*But I don’t want at it! Yuck!
e. I’ve been had! (‘cheated’)
(*)I’ve been wanted! (‘someone wanted to cheat me’)
f. He had it away with his mistress. (had casual sex with’; Brit. dial.)
*He wanted it away with his mistress.
g. I’ll ‘ave you! (‘beat you, exact revenge on you’; Brit. dial.)
(*)I want you!
h. Don’t have a cow, man! (‘have an extreme reaction’; Bart Simpson)
(*)What if I want a cow?

As shown by these examples, true idioms do not extend from have to want.
More research is needed before we can generalize confidently from such data,
but there seems to be a discernable trend: Words represent the smallest level of
granularity for idiom-formation and contextual polysemy effects. The sublexical
formatives evidence by adverbial scope facts do not show contextual polysemy or
form idioms. Assuming for the sake of argument that this is a valid generaliza-
tion, then it has important implications for the dualist versus monist controversy.
Namely, this generalization is predicted on the dualist view, but not the monist
view.

283
3.3 Syntactic evidence
Different phrase structures are posited under the dualist analysis (25a) and Harley’s
(2004) monist analysis (25b).

(25) a. Dualist analysis: John wants [a lollipop]DP .

b. Monist analysis: John wants [PRO PHAV E a lollipop]P P .

There is considerable syntactic evidence favoring the dualist structure.

First, the history of English undercuts the original motivation for the controlled
PP (recall (17) above). The earliest attestations of want actually took a DP object,
with the meaning ‘lack’ (c1200). From ‘lack’ it drifted to ‘desire’; and started
taking infinitive complements (1706). (It’s not clear which of these two happened
first.) It was not until 1836 that we find directional PP’s and particles as in I want
in, I want in (OED example). It is anachronistic to cite the PP complements as the
basis for DP complements, when the PPs were a very late innovation that showed
up at least 500 years after the DPs. Also, these PPs were, and still are, rather
specialized for indicating implicit motion, as the OED notes. We cannot say *I
want in Austin to mean ‘I want to be in Austin.’ But the want+DP cases never
involve motion.
The want+PP pattern is found in other Germanic languages. This ‘go-deletion’,
as it is sometimes called in Swedish grammars, is independent of the want+DP
pattern. Like English want, Swedish vilja ‘want’ allows go-deletion but not DP
objects (26), while önska ‘want, wish’ allows DP objects but not ‘go-deletion’
(27). Both allow infinitives:

(26) a. Jag vill äta middag.

I want eat.INF dinner
‘I want to eat dinner.’
b. Jag vill hem / in i rummet.
I want home / into in room.DEF
‘I want (to go) home / into the room.’
c. *Jag vill en ny bil.
I want a new car
(‘I want a new car.’)

(27) a. Jag önskar att åka til Tyskland.

I wish to travel to Germany
b. Jag önskar en ny bil.
I wish a new car
c. *Jag önskar hem / in i rummet.
I wish home / into in room.DEF

284
So the want+DP pattern and the ‘go-deletion’ pattern do not correlate either his-
torically or across closely related languages, suggesting that the two patterns are
unrelated.
Secondly, want passivizes, suggesting it takes a DP direct object and not a PP:
The war was not wanted (by anyone). English sometimes allows ‘prepositional
passives’ like Mary was being stared at. But these are rather poor with controlled
PPs: *??The team was not wanted off (by anyone). This contrast is expected on
the dualist structure but not on the monist structure. Similarly, adjectivalization is
possible only for the DP taking verb: an unwanted war versus *an unwanted off of
team or *an unwanted into house.
Harley (2004, p. 264, footnote 8) notes another problem for the PP analysis: an
overt NP can replace PRO in the go-deletion type PP (28a), but not in the putative
PP structure posited for the DP complements (28b,c):2

(28) a. John wants Bill/PRO off the team.

b. *John wants Bill a beer.
c. *John wants [Bill PHAV E a beer]P P .

Next, if the apparent DP complements of want are really PPs, then they should
coordinate just as well with (uncontroversial) PPs as with other DPs (the latter
would be covert PPs). But coordination with PPs is almost impossible, as predicted
by the Dualist Analysis:

(29) a. I want [a vodka martini] and [a hot bath]. DP+DP

b. I want [out of these wet clothes] and [into a hot bath]. PP+PP
c. *I want [out of these wet clothes] and [a martini]. *PP+DP
d. *I want [a martini] and [out of these wet clothes]. *DP+PP

On the monist analysis all of the bracketed phrases in (29) are PPs, making it
mysterious that (29a,b) sound so much better than (29c,d).
English infinitival relative clauses allow pied piping of PPs (30a) or the filler-
less bare (or ‘simple infinitival’) type (30b), but disallow DP fillers (30c).

(30) a. a bench [on which]P P to sit

b. a bench to sit on
c. *a bench [which]DP to sit on

If the complement of want were a PP as claimed then it should be possible to

relativize it in infinitivals, but it is not:

(31) a. a reasonable type of bike to want for commuting

2
Harley (2004, p. 264, footnote 8) floats an idea for solving this problem, which will not be
discussed here for lack of space.

285
 b. *a reasonable type of bike [Phave which]P P to want for commuting
c. a reasonable sort of outcome to wish for
d. a reasonable sort of outcome [for which]P P to wish

On the putative PP structure shown in (31b), the contrast with (31d) is mysterious.
In contrast to PP complements, direct objects famously resist separation from
their verb by an adverb (32a,b). Once again, we find a clear contrast between DP
complements of want, and true PPs (32c-f):

(32) a. He nibbled quietly [on the carrot].

b. He nibbled (??quietly) [the carrot].
c. He wants desperately [out of his job].
d. He wants (??desperately) [a better job].

Yet another property distinguishing PPs from DPs is modification by right:

(33) a. So you bring this poor dog in from the rain,

Though he just wants right [back out]P P . (Metallica)
b. *He just wants right [a rapid exit]DP .

Covert HAVE was originally proposed to explain the scope of durative adverbials
(14) in terms of adjunction. But (33b) shows that putative PP constituent does not
allow modification normally permitted for PPs.
Harley (2004), citing McIntyre (2002), argued for the monist analysis on the
grounds that neither have (34a) nor want (34b) is a particle shift verb:

(34) a. He had { his jacket off / *off his jacket }.

b. The doctor wants { those clothes off / *off those clothes }.

The idea is that parallel constraints on the local syntactic environment of the two
verbs can be explained by positing a silent HAVE in both. In my personal judg-
ment, shifting is better in (34a) than (34b), and the former can be found on the
web, for what it’s worth: ‘A fox,’ he gloated to the housekeeper once he’d had off
his coat. (www). Moreover, with other particles the contrast is much sharper:

(35) a. He had { his jacket on / on his jacket }.

b. He wanted { his jacket on / *on his jacket }.

A check of the British National Corpus turned up many hits like (35a) for the string
[pers. pron.] had on [poss. pron.], but none for [pers. pron.] wanted on [poss.
pron.]. With regard to particle shift, the two verbs are not parallel after all.
Finally, want can coordinate and share its object with other transitive verbs,
as in The bear wanted, got, and ate it. (We know this is V-zero coordination and
not right node raising out of coordinated VPs because right node raising is not

286
possible with unstressed pronouns like it.) Such coordination is expected if all
these verbs select DP, but it is problematic if some select PP and others DP, since
the complement it cannot be both at once.
My arguments above focus specifically on the covert PP analysis from Harley
(2004). But in a broader sense these arguments are completely general. For exam-
ple, suppose we try to rescue the syntactic (monist) analysis by replacing the PP
with an outer DP shell.3 Hence the PP in (25b) would be replaced by [PRO HAVE
[a lollipop]DP  ]DP  (the numbers in DP1 and DP2 have no formal significance
and are for identification only). The idea would be that DP2 has the same category
label (namely DP) as any other DP, so on this view we would expect a lollipop to
have the same syntactic properties, regardless of whether it is the object of want or
in some other context such as The dog ate a lollipop— which is what I showed in
this section.
But the whole point of the syntactic analysis is that DP2 differs syntactically
from other DPs. If they are truly identical then the extra structure for the outer
DP2 shell, as well as PRO and silent HAVE, are merely graphical decorations
with no syntactic interpretation. On the other hand, if the syntactic representations
really differ then it should be possible to demonstrate that difference empirically.
The arguments above support the conclusion that there is no difference and that
therefore the syntactic analysis is wrong.

4 Argument three: count and mass nouns

Our last case study concerns Borer’s (2005a,b) particularly strong statement of
the monist theoretical perspective. Borer distinguishes two types of formatives,
listemes and f-morphs. Listemes, which are content words such as nouns, verbs,
and adjectives, have no grammatical features. As far as the grammar is concerned,
they are pure atoms, without subcategorization frames, argument structure, lambda
abstracts, part-of-speech category such as N or V, or minor category features such
as count versus mass noun. A listeme is associated only with an agrammatical
conceptual representation. Borer (2005a:11) submits that there is

no direct interface between the conceptual system and the grammar,

in that properties of concepts do not feed directly into any determina-
tion of grammatical properties. A substantive listeme is a unit of the
conceptual system, however organized and conceived, and its mean-
ing, part of an intricate web of layers, never directly interfaces with
the computational system.

In contrast, f-morphs, which are functional morphemes such as plural inflection,

determiners, numerals, and classifiers, do have grammatical features. For Borer,
the grammar (or ‘computation’) deals in rigid, categorical values, while the con-
ceptual system is highly malleable and subject to contextual factors. This leads her
3
This was suggested by a member of the audience at Wechsler (2008).

287
to the interesting prediction that in conflicts between the two, it is the concepts that
stretch to fit the exigencies of the grammatical construction. Let us consider her
illustration of this point, an analysis of the count/mass distinction.
By way of background, traditional grammars often distinguish count nouns
such as suggestion from mass nouns such as advice, with respect to whether they
allow plurals (suggestions; *advices), indefinite articles (a suggestion; *an advice), what?! really?!
Apparently, confirmed
quantification by too much (??too much suggestion; too much advice), and one- by Google... They say
anaphora (John gave me (some) suggestions and Mary gave me one too.; *John "give some advice" and
not "give an advice"...
gave me (some) advice and Mary gave me one too). However, it has long been
noted that nouns of one type can often be forced into the other type:

(36) ‘grinding’: count noun ⇒ mass noun

a. Jonas is eating a banana. (count)
b. There’s too much banana in this cake. (mass)

(37) ‘portioning’: mass noun ⇒ count noun

a. I drank too much beer last night. (mass)
b. Would you like a beer? (count)

In detailed lexicalist analyses, Copestake (1992) and Copestake and Briscoe (1995)
analyze this as systematic polysemy, in which a class of words productively alter-
nates between systematically related senses. Copestake and Briscoe (1995) gener-
ate the sense extensions with productive lexical rules: a ‘grinding rule’ converts a
count noun into a mass noun, while a ‘portioning rule’ applies in the opposite direc-
tion. This analysis is very detailed and sophisticated, combining Krifka’s (1989)
mereological account of nominal reference with Link’s (1983) treatment of plu-
rals, and carefully addressing the empirical question of the scope and productivity
of various sense extension rules: for example, whether there is a special ‘animal-
grinding’ rule deriving a mass noun referring to the meat or flesh of the animal
denoted by the corresponding count noun (e.g. too much chicken), or whether it
should be subsumed under a more general grinding rule. My sketch of this work
does not do it justice but it will suffice to illustrate the basic strategy and the form
of the grammatical theory under a lexicalist approach.
On Borer’s monist theory, words like banana and beer, like all listemes, are
grammatical atoms. So they cannot be distinguished by a count / mass lexical
feature; nor does Borer allow for lexical rules or coercion. All noun type listemes
denote masses. Instead of coercion, the f-morphs themselves impose structure on
these listeme-denoted masses: divider f-morphs (e.g. plural inflection) portion out
the mass into countable entities; counter f-morphs (several, two, there, etc.) count
out portioned entities; and some f-morphs (a(n), one, each, every) perform both
functions at once.
Since all listemes start out as masses, the grinding function is eliminated en-
tirely, with mass interpretations simply arising in the absence of a divider f-morph.
A seemingly ‘coerced’ phrase like three waters (‘three portions, e.g. glasses, of

288
water’) has the same grammatical analysis as three cats. It is the plural morpheme
itself that does the portioning, with pragmatics and world knowledge determining
that the appropriate portion of cat-mass is an individual cat, while the appropriate
portion of water-mass depends on context, e.g. a glass of water in one context, a
kind of water in another.
While the coercion of listemes between count and mass is fairly free, Borer ob-
serves that the addition of a divider morpheme makes an expression more resistant
to coercion. Hence There’s too much rabbit in this stew sounds much better than
*There’s too much rabbits in this stew. For Borer, *too much rabbits is ruled out by
the grammatical computation as a clash between the [±Divider] and [±Counter]
features of the f-morphs [plural] and much (Borer 2005a:104ff). Such examples
illustrate ‘the complete impossibility of coercibility, or type-shifting, whenever the
noun in question is marked by means of overt inflection.’ (Borer 2005a:105)
Borer’s argument goes as follows: on a lexical coercion account, if we can
coerce rabbit into a mass, then why can’t we coerce rabbits into a mass? Pluralia
tantum nouns provide Borer with a particularly striking evidence since ‘just like
regular plurals, they cannot be coerced into a mass context’ (Borer 2005a:105,
ex. 26b): *There’s too much scissors around this house. Thus the malleability
observed in rich conceptual representations should not be captured in grammatical
coercion rules, nor indeed in the grammar at all, because the grammar proper is
not susceptible to coercion. Borer’s theory explains this observation by keeping
the computational grammar radically insulated from the effects of such conceptual
representations.
Let us assess this argument. Assuming, as seems reasonable, that some words
are more semantically malleable than others, then in conflicts the malleable ones
will stretch more than the rigid ones, and a clash between two rigid ones will
sound worst of all. And ungrammaticality due to errors of ‘agreement’ between
formal grammatical features have a more pronounced quality than what results
from semantic incompatibility.
The question is whether this interaction between malleable and rigid semantics
justifies the radical separation that Borer advocates. As it turns out, f-morphs like
plural actually can be coerced into masses:

(38) How much refried beans / chopped nuts / scrambled eggs / mashed potatoes
/ mashed yams were consumed yesterday?

A listeme like chopped converts a plural like nuts into a mass, which therefore
accepts the mass quantifier much. Interestingly, chopping doesn’t seem to help ex-
amples like *too much chopped rabbits, perhaps because a single rabbit is large
enough to provide a reasonable amount of rabbit meat for a stew (cp. #too much
chopped nut). Examples like (38) show that the computation and conceptual repre-
sentations are interleaved in semantic composition: first the plural -s (an f-morph)
applies to the noun, indicating more than one unit; then chopped (a listeme) con-
verts it to a mass; and then too much (an f-morph) measures the amount of that

289
mass. This seems to be contradict Borer’s basic supposition that computational
grammar— in this case the [±Divider, ±Counter] feature system— is blind to
the rich conceptual properties of listemes.4
As for pluralia tantum nouns, a subset of them actually do appear with mass
quantifiers (Copestake (1992:98), Sag et al (2003:124–5)): cp. How much feces /
grits / collard greens / clothes are there?
So the radical separation between conceptual and grammatical systems seems
unwarranted. Still, the lexicon may not be the right place for all mass-count conver-
sion. For example, consider the ‘beverage portioning’ rule that allows us to order
two waters, but not to point out a puddle by saying *There’s a water on the floor.
Beverage portioning is not (only) lexical: one could order a Stoli and kiwi juice,
please, where Stoli and kiwi juice is a conjoined phrase produced by the syntax.
Let us posit that a concept has some structure, including preferred Individua-
tion Units (IUs). If the concept is a word meaning, then its structure ipso facto
becomes lexicosemantic structure. The phrase two beers draws its IUs either from
the conceptual representation of ‘beverage’ (the IU is roughly a serving, as in two
beers, please) or, since beer is found in many varieties, another IU is ‘kinds’ (They
serve two beers that I like). If a concept lacks any potential IU’s at all, then it
can never appear in count noun contexts: this may be the case for the concepts
denoted by the words evidence, furniture, and clothing. This approach captures the
positive aspects of Borer’s approach. But it crucially rejects the monist principle
of grammar-free words, since syntax and compositional semantics refer directly to
(the IUs within) the conceptual representation of words.
In addition, there is important evidence that count versus mass is also a for-
mal feature of at least some nouns: nouns preserve their count/mass-dependent
distribution even when they denote kinds (Krifka 1995).

(39) What do you value most in life?

a. Flattery. / Advice. / Evidence. / Fruit.
b. *Compliment. / *Suggestion / *Clue. / *Vegetable.

These NPs refer to the kinds or concepts themselves, not to particular specimens, so
there is no question of portioning by f-morphs. This would seem to require a lexical
feature, presumably a simple grammaticalization of the conceptual representation,
produced by a rule stating that a noun denoting an IU-less concept is marked as a
mass noun. The concept ‘flattery’ lacks IUs, so the word flattery is classified as a
mass noun and thus can appear in singular form without a specifier, as in (39a).
Summarizing, we saw from cases like too much chopped nuts that functional
and content morpheme meanings are interleaved in semantic composition. Some
aspects of the count/mass split should perhaps be pushed out of the lexicon proper
into pragmatic conceptual structure, but then noun syntax is crucially sensitive
4
If it turns out that such examples are not problematic for Borer’s theory, then I no longer know
what the theory actually predicts.

290
to that structure. Moreover, at least some nouns lexicalize grammatical features
reflecting their conceptual structures.

5 Conclusion: the autonomy of the lexicon

The old Generative Semantics idea of generating sublexical semantic structure
along with the compositional semantics of sentences has been revived in recent
years. But the attempts to support this approach have back-fired. First we saw
evidence that lexical argument structure is autonomous from the syntactic expres-
sion of it: deverbal nouns preserve the argument structure of the cognate verbs
(contrary to what is often claimed), but systematically differ in syntax. Our second
case study reached the tentative conclusion that word meaning is the smallest unit
for contextual polysemy. Sub-lexical semantic formatives, even those available for
some adverbial modification, are unavailable for forming idioms with surround-
ing words— a problem for monist approaches that treat such formatives as if they
were words. Finally, the mass/count distinction fails to support Borer’s notion of
a computational syntax hermetically sealed off from the vagaries of conceptual
knowledge. The available evidence still supports a dualist syntax comprising two
components: words— which are interfaces between conceptual representations and
grammatical subcategorization instructions— and the combinatory rules that abide
by those grammatical instructions.

References
Barker, Chris: 1995, Possessive Descriptions. CSLI Publications, Stanford.

Beavers, John, Elias Ponvert, and Stephen Wechsler: to appear, ‘Possession of a

Controlled Substantive: Light have and Verbs of Possession’, in Proceedings
of Semantics and Linguistic Theory (SALT) 18. University of Massachusetts,
Amherst, MA, March 21-23, 2008.

Borer, Hagit: 2003, ‘Exo-skeletal vs. endo-skeletal explanations’, in John Moore

and Maria Polinsky (eds.), The Nature of Explanation in Linguistic Theory.
CSLI and University of Chicago Press, Chicago.

Borer, Hagit: 2005, Structuring Sense, Volume I: In Name Only. Oxford University
Press, Oxford.

Borer, Hagit: 2005, Structuring Sense, Volume II: The Normal Course of Events.
Oxford University Press, Oxford.

Chomsky, Noam: 1970, ‘Remarks on Nominalization’, in R. Jacobs and P. Rosen-

baum (eds.), Readings in English Transformational Grammar, 184-221. Ginn
and Co., Waltham MA.

291
Copestake, Ann: 1992, The representation of lexical semantic information, PhD
dissertation, University of Sussex.

Copestake, Ann and Ted Briscoe: 2005, ‘Semi-productive Polysemy and Sense
Extension’, Journal of Semantics 12 (1), 15.

Copestake, Ann, Dan Flickinger, Carl Pollard, and Ivan A. Sag: 2005, ‘Minimal
Recursion Semantics: An Introduction’, Research on Language and Computa-
tion 3, 281–332.

Dowty, David: 1979, Word Meaning and Montague Grammar. Reidel, Dordrecht.

Dowty, David: 1989, ‘On the Semantic Content of the Notion of “Thematic Role”’
in Gennaro Chierchia, Barbara H. Partee, and Raymond Turner (eds.), Proper-
ties, Types and MeaningVolume II: Semantic Issues. Kluwer, Dordrecht, 69–129.

Egg, Marcus: 1999, ‘Deriving and Resolving Ambiguities in wieder-Sentences’,

in Proceedings of the 12th Amsterdam Colloquium.

Harley, Heidi: 2004, ‘Wanting, Having, and Getting: A Note on Fodor and Lepore
1998’, Linguistic Inquiry 35, 255–267.

Harley, Heidi and Rolf Noyer: 2000, ‘Formal versus encyclopedic properties of
vocabulary: Evidence from nominalizations’, in Bert Peeters (ed.), The Lexicon-
Encyclopedia Interface. Elsevier, Amsterdam.

Haspelmath, Martin: 1993, ‘More on the typology of inchoative/causative verb al-

ternations’, in Bernard Comrie and Maria Polinsky (eds.), Causatives and tran-
sitivity, 87-120. John Benjamins, Amsterdam.

Krifka, Manfred: 1989, ‘Nominal reference, temporal constitution and quantifica-

tion in event semantics’, in R. Bartsch, J. van Benthem, and P. von Emde Boas
(eds.), Semantics and Contextual Expression, 75115. Foris Publications, Dor-
drecht.

Lakoff, George: 1965, On the Nature of Syntactic Irregularity. PhD thesis, Harvard
University.

Link, Godehard: 1983, ‘The logical analysis of plurals and mass terms: A lattice-
theoretical approach’, in Bäerle, Schwarze, and von Stechow (eds.), Meaning,
Use and Interpretation of Language, 302–23. de Gruyter, Berlin.

McCawley, James D.: 1974, ‘On Identifying the Remains of Deceased Clauses’,
Language Research 9, 73–85.

McIntyre, Andrew: 2002, ‘The verb get and the syntactic semantics of VP’, Ms.,
University of Leipzig.

292
Marantz, Alec: 1997, ‘No Escape from Syntax: Don’t Try Morphological Analysis
in the Privacy of Your Own Lexicon’, U.Penn Working Papers in Linguistics 4.2,
201–25.
Partee, Barbara H.: 1999, ‘Weak NP’s in HAVE Sentences’, in J. Gerbrandy, M.
Marx, M. de Rijke, and Y. Venema (eds.), JFAK, a Liber Amicorum for Johan
van Benthen on the occasion of his 50th Birthday, 39–57. University of Ams-
terdam, Amsterdam.
Pesetsky, David: 1995, Zero Syntax: Experiencers and Cascades. MIT Press,
Cambridge.
Rappaport, Malka: 1983, ‘On the nature of derived nominals’, in M. Rappaport
B. Levin, and A. Zaenen (ed.), Papers in Lexical-Functional Grammar Indiana
University Linguistics Club, Bloomington, 113–42.
Richards, Norvin: 2001, ‘An Idiomatic Argument for Lexical Decomposition’, Lin-
guistic Inquiry 183–192.
Ross, John Robert: 1976, ‘To Have Have and to Not Have Have’, in M. A. Jazayery,
E. C. Polomé, and W. Winter (eds.), Linguistic and Literary Studies in Honor of
Archibald A. Hill, 263–270. Mouton, The Hague.
Sag, Ivan, Thomas Wasow, and Emily Bender: 2003, Syntactic Theory: A Formal
Introduction, 2nd edition. CSLI Publications, Stanford.
Smith, Carlota S.: 1970, ‘Jespersen’s “move and change” class and causative verbs
in English’, in M. A. Jazayery, E. C. Polomé, and W. Winter (eds.), Linguis-
tic and Literary Studies in Honor of Archibald A. Hill, 101-09. Mouton, The
Hague.
Smith, Carlota S.: 1972, ‘On Causative Verbs and Derived Nominals in English’,
Linguistic Inquiry 3 136-38.
Tham, Shiao Wei: 2006, ‘The Definiteness Effect in English Have Sentences’,
in P. Denis, E. McCready, A. Palmer, and B. Reese (eds.), TLS 8, 137–149.
Cascadilla Proceedings Project, Somerville.
Wechsler, Stephen: 2006, ‘Having a topic, wanting a focus’, Proceedings of 4th
Workshop on Discourse Structure, 498-506. University of Texas, Austin. March
3-5, 2006.
Wechsler, Stephen: 2007, ‘A Diachronic Account of English Deverbal Nominals’,
in Charles B. Chang and Hannah J. Haynie (eds.), Proceedings of the 26th
West Coast Conference on Formal Linguistics, 498-506. Cascadilla Proceed-
ings Project, Somerville.
Wechsler, Stephen: 2008, ‘Idioms, Light Verbs, and Lexical Decomposition’.
NORMS Workshop on Argument Structure, Lund University, February 6.

293