Partitive readings and verb transitivity in English...
Transcript of Partitive readings and verb transitivity in English...
Partitive readings and verb transitivity in English:
argument structure versus Derived Kind Predication∗
Andrew Weir
Generals Paper 1, University of Massachusetts Amherst
April 2012
1 Introduction and data
This paper discusses verbs with kind-type direct objects inEnglish, as in (1).
(1) a. Yesterday, I ate that bread you were telling me about. Hungry Ghost’s rosemary bread sure is
delicious.
b. Yesterday, I drank that beer you were telling me about – Wachusett Nut Brown Ale.
c. [Looking at a picture of a type of bird in a biology textbook]
I shot that bird while I was out hunting yesterday; the carcass is in my freezer.
d. [Looking at a picture of a car in a magazine]
I went for a test-drive yesterday: I drove that car.
e. [Looking at a picture of a Spirit ‘stealth’ bomber]
I flew that plane yesterday; flew a couple of sorties.
f. [In a liquor store]
I brought that beer [points] to the party last night – Wachusett Nut Brown Ale.
∗I would like to thank Seth Cable, Alice Harris, and Angelika Kratzer for their comments during the course of writing this GP,and Lyn Frazier for assistance in the construction of the judgments study, and useful discussion. All remaining errors are of coursemine.
1
Andrew Weir Partitive readings and verb transitivity in English
g. [Looking at a picture of a type of bird in a biology textbook]
I saw that bird yesterday when I was strolling through the woods!
In each of the sentences in (1), the objectthat NP is plausibly analyzed as referring to thekind of NP; the
kind of bread, beer, bird, car, etc. That’s what we’re talking about, or looking at a representative of. However,
what the sentences actually mean is that there is an affectedtoken of the type being referred to. So, I ate an
instance of that kind of bread, and so on. In the cases above where there is a stimulus such as a picture, the
token involved can be completely different from the stimulus. SoI flew that plane yesterdayneed not mean
that I flew thatspecificplane; it can mean that I flew some other plane that was of that type. I will refer to
this reading as the ‘instance-of’ or ‘existential’ reading.
Consider, however, the following sentences.
(2) a. ?Yesterday, I devoured that bread you were telling me about. Hungry Ghost’s rosemary bread
sure is delicious.
b. ?Yesterday, I guzzled that beer you were telling me about.You remember, Wachusett Nut Brown
Ale.
c. [Looking at a picture of a bird in a biology textbook]
?Yesterday, I euthanized that bird in the veterinary surgery.
d. [Looking at a picture of a Spirit ‘stealth’ bomber]
?I piloted that plane yesterday; flew a couple of sorties.
e. ?In one of my truck-driving runs last night, I transportedthat beer to the warehouse – Wachusett
Nut Brown Ale.
f. [Looking at a picture of a type of bird in a biology textbook]
?I noted that bird yesterday when I was strolling through thewoods.
In my judgment, the sentences in (2) are relatively marked, compared with the examples seen in (1). The
‘instance-of’ reading is somewhat harder to access, without considerable contextual support. Compare the
2
Andrew Weir Partitive readings and verb transitivity in English
two below examples. In the first example (3), a minimal context (just enough to plausibly allowthat horseto
have a kind reading) seems sufficient to licenseshoot that horse, buteuthanize that horseis (in my judgment)
degraded. Considerable context, as provided in (4), is required to make such a sentence seem natural.
(3) I look at a picture of a particular type of horse, and say:
a. I shot that horse (when I was out hunting yesterday.)
b. ?I euthanized that horse (when I was on duty in the veterinary surgery yesterday).
(4) a. We work for a animal euthanization mill. It’s a large facility that employs hundreds of
individuals. We both work in the ‘horse department’, that part of the mill that specializes in
the euthanization of horses. Every day, we are assigned a particular breed of horse to euthanize.
Yesterday, I was assigned ‘Clydesdale’, one of the hardest ones. Today, I’ve received ‘Falabella’,
one of the easiest ones to handle. You, however, have been assigned ‘Clydesdale.’ You try to
trade me assignments, but I protest:
b. Are you kidding? I euthanized that horse yesterday!1
I propose the hypothesis that the difference between the verbs which allow the ‘instance-of’ reading easily
and those which don’t is a difference of transitivity; verbssuch aseat, drink, shoot, driveare not obligatorily
transitive, while verbs such asdevour, euthanize, pilot, guzzleare obligatorily transitive. The judgments
presented above are based on intuitions, and as they stand are quite subtle. In order to test this hypothesis and
to give an empirical foundation to these intuitive judgments, a judgment study was conducted. In this paper,
I report the results of this judgment study, which supportedthe hypothesis. I will argue that the difference
between non-obligatorily transitive verbs and obligatorily transitive verbs is that the former syntactically
construct their object arguments, while the latter do not. If a verb syntactically constructs its object argument,
then it can allow for the intermediation of a partitive head,which I argue is cross-linguistically motivated,
and provides the semantics of the ‘instance-of’ reading. The relative markedness of kind-type direct objects
1This context and example are due in their entirety to Seth Cable; I have taken it verbatim from him and thank him for it.
3
Andrew Weir Partitive readings and verb transitivity in English
with obligatorily transitive verbs is due to the fact that anextra-syntactic process, similar to the processes
of semantic coercion discussed by Pylkkanen & McElree (2006), is required to achieve the ‘instance-of’
reading.
2 Kinds and the ‘instance-of’ reading
Recall one of the ‘instance-of’ readings discussed in (1) above.
(5) Yesterday, I ate that bread you were telling me about. Hungry Ghost’s rosemary bread sure is
delicious.
It is a matter of interest how we achieve the ‘instance-of’ reading here, ifthat breadunderlyingly refers to
a kind of bread. Various analyses have been proposed in the literature for achieving the reading which we
see in (5). For example, Carlson (1977) proposes that a lexical transformation transforms the verbs in these
episodic cases into verbs which take a kind-type argument and assert that some realization of the kind was
the theme of the event. For example, if the standard translation of drive is (6a), then the lexically transformed
version ofdrive when it combines with a kind argument is as in (6b), and a verb phrase likedrive that car
comes out as in (6c).2
(6) a. JdriveK = λx.λy.drive(x)(y)
b. Kind-taking variant of (a):λxk.λy.∃z.R(z, xk) & drive(z)(y)
(whereR(z, x) means ‘z is an object which is a realization of the kindx’)
c. Jdrive that carK = λy.∃z.R(z, thatCarkind) & drive(z)(y)
This means thatdrive that carcould have the meaning ‘drive a realization of that car-kind’. A similar
treatment is proposed by Chierchia (1998) for sentences such as the below.
2I provide translations of words and phrases of English into Ty2. Types: entitiese, eventualitiess, truth valuest. Variables:x, y, z entities (variables subscripted thus:xk stand for kinds),e events; types of other variables will be noted in the text. I workhere with an extensional semantics and so do not make use of a world type; the task of extending my analysis here to the intensionaldomain remains to be done.
4
Andrew Weir Partitive readings and verb transitivity in English
(7) (Chierchia’s (30))
a. That kind of animal is ruining my garden.
b. [Pointing at the picture of a lion in a biology book] That iswhat you saw this morning at the zoo.
These sentences plausibly mean that there are one or more instances of that kind of animal, and they are
ruining my garden; and that there are one or more instances ofthe lion-kind (by assumption, thethat in (7b)
is referring to the lion-kind) that you saw this morning at the zoo.
Chierchia argues that this comes about not via a lexical transformation of the predicate as in Carlson (1977),
but rather by a rule of Derived Kind Predication:
(8) (Chierchia’s (31c))
Derived Kind Predication (DKP):
If P applies to objects andk denotes a kind, then
P (k) = ∃x[∪k(x) ∧ P (x)]
(9) Definition of ‘∪’ (Chierchia’s (15), adapted3)
Let d be a kind [in our extensional terms, the maximal plural individual corresponding to a kind –
AW]. Then,
∪d = λx.x ≤ d (where ‘≤’ is the part-of relation).
(10) (Chierchia’s (31a, b), adapted)
a. Jthat kind of animal is ruining my gardenK = ∃x.∪thatKindOfAnimal(x) & ruinMyGarden(x)
‘There is a part of that kind of animal which is ruining my garden.’
b. Jyou saw that in the zooK = ∃x.∪that(x) & sawInTheZoo(you, x)
3I have removed the intensionality from Chierchia’s definition, which originally relativizes ‘∪’ to a situation. This is to maintainconsistency with my extensional stance throughout this paper.
5
Andrew Weir Partitive readings and verb transitivity in English
On Chierchia’s view, DKP is a freely available typeshiftingoperation, used when necessary to generate
the ‘instance-of’ readings of sentences with kind referents. Chierchia then uses this rule to generate the
correct reading of sentences such as (11a), arguing that bare plurals such aslions represent (or are ultimately
computed to represent, after the application of type-shifting operations)∩lions – the kind generated from the
plural propertyλx.lions(x).
(11) (Chierchia’s (32))
a. Lions are ruining my garden.
b. ruiningMyGarden(∩lions)
⇐⇒ (via DKP)∃x[∪∩lions(x) ∧ ruiningMyGarden(x)]
If the difference I report between non-obligatorily transitive and obligatorily transitive verbs is a genuine one,
it is slightly mysterious for an analysis like the one in Chierchia (1998), which uses a type-shifting rule to
allow an existential reading in these cases. The type-shifting rule should be able to apply regardless of the
verb which is used, with no difference in ‘markedness’. The mystery is somewhat less for an analysis like
Carlson (1977)’s; perhaps the relevant lexical rule only applies to non-obligatorily transitive verbs, and does
not apply to obligatorily transitive verbs. As we will see, the analysis I will present of the non-obligatorily
transitive cases hews more closely to a Carlsonian analysis, although I will recast it in terms of a structure-
building operation within the syntax, rather than the application of a lexical rule to the verbal predicate.
However, before I proceed to the analysis, I report the results of a judgments study, conducted in order to
confirm that the judgments reported above represent a genuine fact about the grammar of English.
6
Andrew Weir Partitive readings and verb transitivity in English
3 Judgments study
3.1 Methods and results
In Experiment 1, a set of sixteen sentences were prepared, containing the frame “. . . VERBed that NP . . . ”.
A set of eight non-obligatorily transitive verbs was used,4 paired with eight obligatorily transitive verbs with
similar meanings, namely:
(12) eat/devour, paint/depict, clean/sanitize, shoot/euthanize, shoot/murder, stab/impale, scratch/nick,
pull/lug
Sentences, then, were of the formI ate/devoured that bread for my lunch yesterday. Sentences were
constructed in such a fashion as to bring out episodic readings (with the use of time phrases such asyesterday),
rather than generic readings, as there may be other grammatical means available to access ‘instance-of’
readings in generic contexts (see Chierchia (1998) for discussion). Sentences were presented to subjects, and
on the same screen, subjects were asked ‘Can “that NP” mean “that kind of NP”?’, with the relevant noun
substituted for ‘NP’. Subjects had the choice of ‘Yes’, ‘No’, or ‘Not sure’. These sentences were presented
randomly and in a Latin square design; each subject saw only half of the intransitive verbs and half of the
transitives, and which verbs each subject saw was balanced between speakers.
Interspersed with these sentences were fillers, which askedirrelevant questions about pronominal binding;
and ‘catch’ questions, asking about the type-token ambiguity, prepared in order to ensure that the subject was
performing the task. Six of these ‘catch’ questions contained an NP which is not amenable to the ‘instance
of’ reading in any circumstances (e.g.I gave that student a stern lecture yesterday, wherethat student(for
unclear reasons; see e.g. Carlson (1977)) cannot have a ‘kind’ reading); two of these sentences were presented
in generic or habitual frames, and should (by assumption) beacceptable on the ‘kind’ interpretation (I drank
that coffee for years; we started growing that crop on the farm five seasons ago). In addition, before the
4Seven distinct verbs;shootwas used twice.
7
Andrew Weir Partitive readings and verb transitivity in English
experiment proper began, four sentences (two type-token, two binding fillers) were presented in a practice
session to familiarize the subjects with the procedure. A full list of sentences is presented in the Appendix.
The experiment was prepared on the Ibex Farm website, and a link to the experimental website was posted on
the author’s personal Facebook page with an exhortation to native speakers of English to take part. 39 people
responded. One was excluded after indicating him/herself as not a native speaker of English. A further 11
subjects were excluded after giving a ‘No’ response to the practice itemI drove that car for years. It was
assumed that such speakers were simply not prepared to give the kind reading to phrases of the typethat
NP, and an informal inspection of their responses to the experimental items suggests that this was indeed the
case; these subjects consistently gave low ratings (that is, answered ‘Not sure’ or ‘No’) at a higher rate than
other subjects. While this is a high rate of exclusion of subjects, if data from these subjects is included, the
direction of the overall results does not change.5 In addition, due to technical problems with the Ibex Farm
server, an additional five subjects had to be excluded from one of the counterbalanced lists in order to ensure
that the total number of responses to each of the two questionnaires was balanced. This was accomplished
simply by removing every third subject from the list of responses to that questionnaire. There were finally
22 subjects whose data was used. These participants had varied backgrounds. Data about education level
was not collected, but most had achieved undergraduate degrees (at least), and were native speakers of either
British or American English.
The results of this study are presented in Table 1. The following points are of note.
• In all cases where there is a difference in the percentage of ‘Yes’ responses between the two conditions,
that difference goes in the direction predicted by the hypothesis, that is, obligatorily transitive verbs
are degraded on ‘instance-of’ readings when compared to non-obligatorily transitive verbs.
• However, the distinction can be slight: compareeat that bread(91% giving a ‘Yes’ response to the
question ‘Can this mean ‘that kind of bread?’) withdevour that bread(73% giving a ‘Yes’ response,
5That is, the differences reported in the results section still obtain, although not to the same magnitude. This is true except for thepairsclean/sanitizeandscratch/nick: if no subjects are excluded, the percentage Yes (Not sure, No) responses are 31% (13%, 56%)for cleanversus 40% (13%, 45%) forsanitize; and 50% (13%, 37%) forscratchversus 55% (18%, 27%) fornick.
8
Andrew Weir Partitive readings and verb transitivity in English
Unergative verb Yes (Not sure, No) Transitive verb Yes (Not sure, No)eat (that bread) 91% (9%, 0%) devour 73% (27%, 0%)paint (that bird) 91% (9%, 0%) depict 82% (18%, 0%)clean (that instrument) 45% (9%, 45%) sanitize 45% (9%, 45%)shoot (that animal while hunting) 82% (9%, 9%) euthanize (that animal in the surgery) 9% (0%, 91%)shoot (this bird yesterday) 27% (18%, 55%) murder (this birdyesterday) 27% (18%, 55%)stab (that fish) 27% (18%, 55%) impale 18% (18%, 64%)scratch (that metal) 64% (9%, 27%) nick 45% (27%, 27%)pull (that car) 73% (9%, 18%) lug 9% (0%, 91%)
Table 1: Reponses for each verb in Experiment 1.n = 22.
with no ‘No’ responses). While this is indicative of a difference between the two conditions, it is not
on a scale which one would want to label categorical; that is,I devoured that bread yesterdayis clearly
not simply ungrammatical to English speakers.
• In cases where there is no difference between the two conditions, overall acceptability is degraded
across both conditions (e.g. the paircleanandsanitize, both with 45% ‘Yes’ responses). I propose that
this is due to difficulty accessing the kind reading of the NP independently. A noun phrase likethat
instrumentis much more plausibly analyzed as meaning ‘that (particular) instrument’ than ‘an instance
of that kind of instrument’. While this should not block the ‘instance of’ reading, it may be that in these
cases, which were presented without rich contexts, the ‘instance of’ interpretation was simply hard to
access for reasons unrelated to the type of verb, and this hasovershadowed any difference which the
verb may have had.
On the basis of the data presented in Table 1, I tentatively conclude that the hypothesis suggested by the
intuitive judgments is upheld; that is, while the availability of the ‘instance-of’ reading is not impossible
with strongly transitive verbs, it is marked, and relatively dispreferred to the ‘instance-of’ reading with non-
obligatorily transitive verbs.
Experiment 2 was initially designed as a ‘control’ to ensurethat the results of Experiment 1 were genuinely
due to an interaction between the verb and thethat NPphrase. The initial assumption was that if a phrase like
9
Andrew Weir Partitive readings and verb transitivity in English
Intransitive verb Mean rating Transitive verb Mean ratingeat 5.00 devour 3.27paint 4.64 depict 4.36clean 4.73 sanitize 4.09shoot 4.55 euthanize 4.64shoot 4.73 murder 3.82stab 3.91 impale 4.27scratch 4.27 nick 3.73Mean for intrans. 4.53 Mean for trans. 4.01
Table 2: Reponses for each verb with athat kind of NPobject.n = 22.
that kind of NP, with an overtkind ofphrase, could for some independent reason have an existential/‘instance-
of’ reading (see for example Wilkinson (1995)), then there should not be any incompatibility between
strongly transitive verbs and phrases likethat kind of NP. To test this, a second experiment was constructed,
in this case a straightforward grammaticality ratings study. Sentences similar to the stimuli in the first
experiment were constructed, using the same verbs, but in this case using athat kind of NPphrase rather
than athat NPphrase; for example,I {ate/devoured} that kind of pasta for my lunch yesterday. Subjects
were asked to rate each sentence on a scale from 1 to 5, where 5 was defined in the instructions as ‘a good
sentence of English, one that you could naturally imagine saying’ and 1 was defined as ‘does not sound like
a good sentence of English’. (These were labeled as ‘Bad’ and‘Good’ on the edges of the scale presented to
the subjects for each item.) Irrelevant fillers were also included, and subjects were given four practice items,
as before.
The same subjects were used as for Experiment 1, with the samesubjects excluded, as noted. Experiment
2 followed on from Experiment 1. The sentences were, again, counterbalanced using a Latin square design;
each subject saw only the verbs they had not seen in Experiment 1. So, for example, if a subject saweat –
depict – clean – euthanize. . .in Experiment 1, they sawdevour – paint – sanitize – shoot. . .in Experiment
2.
The results are given in Table 2. Again, if the excluded subjects are included in the analysis, the direction
of the data reported remains the same for each pair of verbs, with the exception thatshootandeuthanize
10
Andrew Weir Partitive readings and verb transitivity in English
are reversed in direction (the means with no subjects excluded are 4.31 versus 4.18 forshootandeuthanize,
respectively). We can see from the results that the ‘control’ condition was not in fact a control; in most cases,
that kind of NPis degraded in the object position of an obligatorily transitive verb. I do not present any
particular hypothesis for the contrary results obtained inthe case ofshoot/euthanizeandstab/impale; but, as
the means indicate, the general trend holds.
3.2 Discussion of results
The key point I take from the results of Experiment 2 is the following. It appears that, at least with respect
to the diagnostic pursued here,that NPandthat kind of NPbehave in the same way from the point of view
of the grammar. Both can receive an existential (‘instance-of’) reading in the object position of an optionally
transitive verb; but this reading is degraded when combinedwith an obligatorily transitive verb. I take from
this that there is no particular grammatical means for obtaining an existential reading for a phrase likethat
kind of NPwhich is not available with phrases likethat NP. Both of these phrases, I posit, denote the same
thing: they have a kind reading. The existential reading is derived by the grammar; and, based on the results
of Experiments 1 and 2, it is derived in a way which appears to be sensitive to the transitivity of the verb.
This distinction between non-obligatorily transitive andobligatorily transitive verbs presents a problem for
the view that existential/‘instance-of’ readings with kind-type DP arguments are delivered via the application
of a type-shifting rule such as Chierchia (1998)’s Derived Kind Predication. Such a rule applies to kind-type
nouns to deliver an object-type entity which can serve as argument to a predicate. In principle, such a rule
should not be sensitive to the type of the verb, and should apply freely in any case where a kind-denoting DP
is in an argumental position. There should not, on this view,be any distinction between verbs such aseat
anddevour, or shootandeuthanize, in terms of their ability to take kind-type direct objects;neither of these
types of verb should result in a more ‘marked’ structure thanthe other.
In what follows, I will propose that while a process like Derived Kind Predication is available to the grammar,
there is also another way to derive the existential reading.I will argue that there is a syntactically present
head, which may associate verbs likeeat, drink, shoot, drive(but not verbs likedevour, guzzle, euthanize,
11
Andrew Weir Partitive readings and verb transitivity in English
pilot) with their direct objects, and which can deliver the existential reading. I will propose that the two
types of verb under discussion systematically differ in howthey construct their argument structure. Verbs
of the sorteat, drink, shoot, drive, even if they appear transitive on the surface, are underlyingly unergative,
of type 〈s, t〉, but have their patient arguments ‘built up’ in the syntax through secondary predication. I will
refer to this type of verb as ‘basically unergative’ verbs. By contrast, verbs likedevour, euthanizetake their
patient arguments ‘directly’; they are underlyingly of type 〈e, st〉. I will refer to this type of verb as ‘basically
transitive’ verbs. (I assume, following Marantz (1984) andKratzer (1996), that external arguments (agents)
for both types of verb are introduced in the syntax by secondary predication, through a ‘little v’ head or a
Voice head.)
The fact that the ‘basically unergative’ verbs build up their direct objects in the syntax, but ‘basically
transitive’ verbs do not, is what allows ‘basically unergative’ verbs to be generally felicitous with kind-
type direct objects. Syntactic construction of direct object arguments allows for a silent partitive head to be
inserted between the verb and the object; the effect is that,with ‘basically unergative’ verbs, a sentence like
I ate that breadcan mean that I ate an instance of that kind of bread. By contrast, ‘basically transitive’ verbs
such asdevourdo not allow the insertion of such a partitive head; they taketheir direct object arguments
directly. As such, ‘basically transitive’ verbs cannot ‘structurally’ derive the existential reading with kind-
type direct object. Rather, in order to obtain an existential reading, a process like Chierchia’s rule of DKP
is required. I will argue that the need to resort to a process like DKP is marked, leading to the degradation
noted in the experimental study.
Firstly, I will defend the assertion that verbs likeeat, drink, drive, shootare basically unergative, placing my
proposal in the ‘constructivist’ mold of, for example, Borer (2005) and Kratzer (2003). Then I will proceed to
elaborate my proposal for a partitive head and account for the difference in the ability of ‘basically unergative’
and ‘basically transitive’ verbs to take kind-type arguments and deliver an ‘instance-of’ reading.
12
Andrew Weir Partitive readings and verb transitivity in English
4 Argument structure: built or altered?
In this paper I take a ‘constructivist’ approach to verb meanings: verbs are composed of basic primitives
(‘roots’, in the terms of Pesetsky (1995); ‘constants’, in the terms of Rappaport Hovav and Levin
(Rappaport Hovav & Levin (1998) et passim)). In my version ofthis proposal, such primitives may be
either of semantic type〈s, t〉 (in the case of ‘basically intransitive’ verbs, such aseat) or of type〈e, st〉 (in the
case of ‘basically transitive’ verbs such asdevour). In this constructivist conception, any argument structure
beyond the very limited structure verbs can be ‘born’ with (the two cases of verbs likeeat, taking only an
event argument, and verbs likedevour, taking a theme argument and event argument) is ‘built up’ via the
mediation of functional heads. Schematically, the two types of verb can be represented like this (at the level
of the verb phrase).
(13) a. Intransitiveeat(I’m eating, Have you eaten?, I ate already)VP
V
eatλe.eat(e)
b. Transitiveeat(I ate the sandwich)VP
λe.eat(e) & theme(theSandwich)(e)
V
eatλe.eat(e)
XPλe.theme(theSandwich)(e)
Xλx.λe.theme(x)(e)
DPtheSandwich
the sandwich(V and XP semantically combine via Event Identification (Kratzer 1996))
13
Andrew Weir Partitive readings and verb transitivity in English
(14) Purely transitivedevourVP
λe.devour(theSandwich)(e)
V
devourλx.λe.devour(x)(e)
DPtheSandwich
the sandwich
This approach to variance in argument structure captures the obligatory transitivity of a verb likedevour,
in comparison to optionally transitiveeat. Borer (2005) contains an extensive discussion of how argument
structure might be built up in this way, and what particular types of secondary predicates might be necessary;
Borer also discusses reasons to believe that the ‘internal’argument of a verb is syntactically constructed in
the way proposed here (sec. 2.4). There are many questions about how the thematic role ‘theme’, for direct
objects, is to be understood. Kratzer (2003) argues that a general thematic role ‘theme’ does not exist, but
doesn’t deny the possible existence of a set of more fine-grained thematic roles such as ‘affected patient’,
‘consumed patient’, ‘created patient’ etc. I will not conduct a full exploration of what possible secondary
predicates (both in the sense of the inventory of functionalheads in syntax, and in the sense of thematic
roles/predicates in the metalanguage) are available to combine basically unergative verbs with direct objects,
although clearly this will be an important topic for future research (and Borer (2005) contains extensive
discussion here). Concretely, in the current investigation, I will use the Davidsonian predicatetheme, with
the proviso that this is to be understood as a provisional usage pending further detailed investigation of the
nature of the thematic role at issue.
‘Constructivist’ views of argument structure alternations are not the only views possible. Other views of
argument structure changing (not necessarily ‘building’)operations are represented in the literature. For
example, Dowty (1978)’s analysis of verbs such as intransitive eat proposes that an argument structure
changing rule applies to a basically transitiveeat; the rule typeshifts a transitive verb likeJeatK =
λx.λy.eat(x, y) by existentially binding its object, i.e.λy.∃x.eat(x, y). Other views include, for example,
the program of Beth Levin and Malka Rappaport Hovav (see, forexample, Rappaport Hovav & Levin (1998),
14
Andrew Weir Partitive readings and verb transitivity in English
Levin & Rappaport Hovav (2011), and references cited therein). In their work, they propose that verb
‘constants’ consist of a basic frame describing an event which is a relation between various participants.
For example, (Rappaport Hovav & Levin 1998:114) gives the event structure for a verb likesweepas (15)
below.
(15) [ x ACT<SWEEP> y ]
The verbsweeplinks a participant x, who acts in a ‘sweeping’ manner, on a surface y. However, according
to Rappaport Hovav and Levin, Universal Grammar provides various other templates which can replace the
basic template of a verb, in a process they term Template Augmentation. Two examples are given below
of templates which Rappaport propose are provided by Universal Grammar, and the effects which Template
Augmentation has on the verbsweepin each case (taken from Rappaport Hovav & Levin (1998)).
(16) a. [ [ x ACT<MANNER> ] CAUSE [ BECOME [ Y< STATE > ] ] ]
John swept the floor clean.
b. [ [ x ACT<MANNER> y ] CAUSE [ BECOME [ z< PLACE > ] ] ]
John swept the crumbs onto the floor.
John swept the crumbs off the table.
Argument structure alternations are thereby accounted fornot in a ‘constructivist’ way, by ‘building’
argument structure in the syntax as I have proposed, but rather by lexical operations which alter the argument
structure.
Obviously the debate between ‘building’ and ‘altering’ argument structure is not one that I will be able to
resolve fully in the confines of this paper. I will note, however, that one interesting consequence of allowing
argument structure to be built up using intermediate functional projections is that, depending on the inventory
of functional heads posited, an argument introduced by these heads may end up standing in a fairly abstract
relation to the actual verb, if it stands in any relation at all. This is true, for example, of the resultative
15
Andrew Weir Partitive readings and verb transitivity in English
construction (as analyzed in a constructivist vein by Kratzer (2005)):
(17) a. I ran my shoes threadbare.
b. I drank the teapot dry. (Kratzer 2005)
c. I sang the baby asleep.
d. I talked him asleep.
e. I laughed myself happy again.
One does not run shoes, drink teapots, etc. Some ‘basically unergative’ verbs can also be used in a verb of
creation frame, as below:
(18) a. The mouse ate a hole in the wall.
b. I drove a hole in my tires.
c. I pulled a string out of the goo. (i.e. a string made out of goo, not, say, a woolen string that was
floating in it)
These sentences don’t entail that I ate a hole, I drove a hole,or I pulled a string, respectively (in the ‘normal’,
affected-patient sense we understand these verbs in). So a plausible hypothesis is that these ‘created patient’
arguments are being ‘built up’ in the syntax, via the mediation of some head with semantics more complicated
than simply assigning a thematic role ‘theme’. The important question on a constructivist approach then
becomes what functional heads are possible. Some that seem like good possibilities are possibly CAUSE
and BECOME heads, particularly in the analysis of the causative/inchoative alternation (von Stechow 1995,
1996, Folli & Harley 2004) and of verbs of creation, as above (von Stechow (2001), Beck & Johnson (2004)
on benefactives), and perhaps a head introducing the arguments of the double object construction (a HAVE
head, following Harley (2002), Beck & Johnson (2004), or an Appl head, following Pylkkanen (2008)). Here,
I will make a case for adding a partitive head to the list of heads which may participate in the construction of
argument structure. I will first lay out the semantics of sucha head, showing how the ‘instance-of’ reading
16
Andrew Weir Partitive readings and verb transitivity in English
can be generated if such a head is used; I will then argue that such a head is cross-linguistically motivated.
5 Predicates applying to objects and to kinds
5.1 Getting a kind denotation from the NP
I assume here that the grammar can assign a kind-type interpretation to the DPthat breadwhich denotes a
particular sub-kind of bread. For example, if we are talkingabout sourdough bread, thenthat breadcan refer
to the individual which is the fusion of all the sourdough bread in the world. Concretely, I follow Krifka
(1995) and Kratzer (2008) in this regard. I assume that common nouns likebreador car are born denoting
kinds, or more precisely, the nominal roots√
breadand√
car denote kinds. There are then various pieces
of nominal inflection, classifiers (in English silent, but overt in languages such as Chinese) which could
combine with this root and return an object-type reading or akind-type reading. (19) below is taken from
Kratzer (2008); (20) shows how the system derives a kind-type reading for a phrase likethat bread.
(19) (Kratzer’s (2), adapted)
a. J√
zebraK = zebrak (AW: the kind,Hippotigris)
b. JCLindK = λx.λy.kind(x) & individual(y) & y ≤ x
(takes a kind and returns the property of being a member of that kind)
c. JCLkindK = λx.λy.kind(x) & kind(y) & y ≤ x
(takes a kind and returns the property of being a subkind of that kind)
(20) a. DP
D
that
NP
CLkind
√bread
b. (i) J√
breadK = breadk (the kind)
(ii) JCLkindK = λx.λy.kind(x) & kind(y) & y ≤ x
17
Andrew Weir Partitive readings and verb transitivity in English
(iii) Function Application on (i) and (ii):λy.kind(breadk) & kind(y) & y ≤ breadk
(iv) JthatK = λP〈et〉.ιx.P (x)6
(v) Function Application on (iii) and (iv):ιx.kind(breadk) & kind(x) & x ≤ breadk
‘The contextually salient subkind of bread’
I presume that there are many different possible salient subkinds. Carlson (1977) discusses the possible
natural ways of ‘splitting up’ a kind into its subkinds (ch. 6sec. 2). One plausible way of ‘splitting up’
bread, for example, would be into the subkindssourdough, wheat, rye, 8-grain. . .. So the phrasethat bread
could refer to sourdough bread, for example. As notation forthis, I will useΣSourdough. This refers to
the kind ‘sourdough’, the fusion of all the sourdough bread in the world; it is of entity type.7 So given the
semantics proposed in (20), taken from Kratzer (2008), and given the appropriate deictic resolution ofthat,
thenthat breadcould be computed, in the end, to refer toΣSourdough.
5.2 Getting the ‘instance-of’ reading in object position
Given this, we need to account for how a sentence likeI ate that breadcan have a sensible reading. A
predicate likeeator drive, as Chierchia notes (1998:364), operates on objects, not kinds. I propose that such
kind-type arguments are shifted up to property type, via the(covert) insertion of Chierchia’s ‘∪’, which I take
to be freely available.
(21) Definition of ‘∪’ (repeated from (9))
Let d be a kind [in our extensional terms, the maximal plural individual corresponding to a kind –
AW]. Then,
∪d = λx.x ≤ d (where ‘≤’ is the part-of relation).
6I am ignoring the deictic contribution ofthat here and simply identifying it withthe.7It may actually be of individual concept type, that is, a function from worlds to entities. See Chierchia (1998) for discussion.
For simplicity, I am assuming a completely extensional semantics here.
18
Andrew Weir Partitive readings and verb transitivity in English
I further propose that there is a silent head which takes a property-type object, existentially quantifies over it,
and returns an event description asserting that the entity returned by the existential quantification is the theme
of the event. I notate this silent head as PART, for ‘partitive’. This may be a somewhat misleading name; the
head is not ‘partitive’ in a true sense, as it does not bring about reference to a part of an entity (say, a slice of
a cake), but rather to an instance of a kind. The ‘partitivity’ (in the sense of taking a part of a kind) is really
built into the definition of ‘∪’, above, rather than PART. I will call the head ‘partitive’,however, to bring the
terminology into line with such overt heads as Frenchde, referred to as partitives, and which (I will argue
below) perform the same task as PART. The denotation of PART is below.
(22) JPARTK = λP〈e,t〉.λe.∃x.P (x) & theme(x)(e)
How can this help us understand the ‘instance-of’ reading ofa sentence likeI ate that bread? I assume a tree
as below for this sentence. Note that, in order to fix a type mismatch between PART (seeking a property) and
that bread(denoting a kind), a ‘∪’ operator has been covertly inserted.
(23) VP
V
eat
PartP
PART∪ DP
that bread
(24) a. Jthat breadK = ΣSourdough
b. ‘∪’ on (a): λx.sourdough(x).
c. Function Application on PART, (b):λe.∃x.sourdough(x) & theme(x)(e)
d. JeatK = λe.eat(e)
e. Event Identification on (c), (d):λe.eat(e) & ∃x.sourdough(x) & theme(y)(e)
19
Andrew Weir Partitive readings and verb transitivity in English
(ΣSourdough is the kind, the fusion of all sourdough bread;sourdough is the property borne by any piece
of sourdough bread.∪ΣSourdough = sourdough = λx.sourdough(x).)
This is the ‘instance-of’ reading. Via the mediation of the PART head between a kind-type object and the
predicate, we derive the meaning that there was an instance of sourdough bread that was eaten, the correct
meaning for the relevant reading ofI ate that bread.
5.3 The ‘instance-of’ reading in other positions
We have, then, an analysis for the felicity of a sentence likeI ate that breadon the ‘instance-of’ reading.
I now turn to the relative markedness of a sentence likeI devoured that breadon the ‘instance-of’ reading.
On the analysis here, an obligatorily transitive verb likedevourdoes not syntactically construct its theme
argument. It takes it directly. On the constructivist view pursued here, that is what it means to be obligatorily
transitive. That is,devourcombines with a DP argument; it cannot combine with a phrase headed by PART.
If we left things as they currently are, then we would predictthat the ‘instance-of’ reading was simply
ungrammatical with obligatorily transitive verbs. PART cannot intervene to ‘mediate’ between the predicate
and the kind-type direct object to provide the ‘instance-of’ reading. We have seen, in the experimental study
reported in section 3, that this is not the case; an existential reading of kind-type DPs in the object position
of obligatorily transitive verbs is certainly possible. Weneed, then, to provide some way for the grammar
to derive the existential reading. However, the other fact we took out of the experimental results is that
‘instance-of’ readings in the object positions of obligatorily transitive verbs seem to be marked, in some
way; somewhat degraded with respect to the basically unergative verbs. Our analysis should capture this in
some way.
20
Andrew Weir Partitive readings and verb transitivity in English
I argue that, with obligatorily transitive verbs, a rule like Chierchia’s DKP does apply to derive the existential
reading. So, for example, we have the following derivation:
(25) a. JdevourK = λx.λe.devour(x)(e)
b. Jthat breadK = ΣSourdough
c. (a) and (b) cannot, or do not sensibly, combine via Function Application, so apply DKP:
Jdevour that breadK = λe.∃x.∪ΣSourdough(x) & devour(x)(e)
I argue that it is this process which is marked. I would analyze the application of a rule like DKP as being
an instance ofsemantic coercion, as discussed by Pylkkanen & McElree (2006). On this view, aprocess like
DKP is a process which can apply to make sense of a derivation which would otherwise be deviant.
I need to tread carefully here: I do not want to claim that all instances of type-shifting are marked. Type-
shifting is a device we may often want to have recourse to in natural language. For example, in the analysis I
have proposed above of the intransitive cases likeeat, I have argued for the free insertion of the type-shifting
‘∪’ operator, suggesting that we do not want this operation, atleast, to be ‘marked’. And Pylkkanen &
McElree (2006) argue that there is no evidence that the familiar type-shifting operation LIFT (Partee 1986),
lifting a type〈e〉 argument to a type〈et, t〉 argument, causes an increase in processing difficulty. The claim
should not be, then, that type-shifting in general is ‘marked’.
There are, however, certain cases discussed by Pylkkanen &McElree (2006) where type-shiftingdoesappear
to lead to an increase in processing difficulty. An example from Pylkkanen & McElree is cases of complement
coercion: a verb which normally takes an event complement (like transitivebegincombining withbuilding
the table) can take a nominal complement:
(26) a. The carpenter began building the table.
b. The carpenter began the table.
21
Andrew Weir Partitive readings and verb transitivity in English
Pylkkanen & McElree report that a sentence likethe carpenter began the tablehas an increased processing
cost (as measured by self-paced reading studies and eye-tracking studies) when compared to a sentence like
the carpenter built the table. They argue that the extra cost is due to the extra-syntacticprocessing required to
generate an interpretation; the DPthe tablein (26b) has to be coerced such that its meaning is something like
‘an event of building the table’, so that it can combine with averb likebegin. It is the shifting ofthe tableinto
denoting an ontologically different thing – not an entity, but an event – that Pylkkanen & McElree argue is
marked: ‘ontological mismatch [between a verb and complement – AW], such as those found in complement
coercion, [blocks] basic composition, and hence immediately mandates the on-line deployment of a costly
coercion process.’ (p. 571) By contrast, Pylkkanen & McElree argue that the more familiar noun phrase
shifting operations (as in Partee (1986)), which do not involve an ontological shift of this sort – shifting from,
say, an entity to a generalized quantifier over entities, rather than shifting from an entity to an event – do not
prompt a processing difficulty.
A similar process, I argue, is going on with sentences likeI devoured that (kind of) bread yesterday. An
attempt is being made to combine two constituents,devourandthat (kind of) bread. As one cannot devour
kinds –devouris an object-seeking predicate, not a kind-seeking predicate – it is reasonable to assume that
composition is blocked in this case. The structure needs to be coerced. The coercion process used can be
notated as something similar to Chierchia’s DKP; the verb phrasedevour that (kind of) breadis coerced
into meaning that there was an instance of that kind of bread which was devoured. The fact that this extra-
syntactic work has to be done is what results in the relative markedness of sentences likeI devoured that (kind
of) bread. They are not ungrammatical, just asThe carpenter began the tableis not ungrammatical; they are,
however, more marked, as shown by the experimental results reported here. I argue that this ‘markedness’ is
the reflex of extra processing difficulty occasioned by the need to resort to a non-structural operation in order
to derive the truth conditions. Further psycholinguistic work would need to be done to firmly support this
hypothesis; for example, self-paced reading or eye-tracking studies of sentences such asI devoured that kind
of bread. However, the theory I have put forward here seems to capturethe facts that existential readings are
available with ‘basically unergative’ verbs, and are possible, but marked, with ‘basically transitive’ verbs.
22
Andrew Weir Partitive readings and verb transitivity in English
6 Partitive heads: cross-linguistic evidence
I have argued that a partitive head can be used to introduce a theme argument. Up to now, the motivation
for this head has been largelyad hoc. Ideally, in a constructivist framework, we would not posita (silent)
argument structure building head without overt evidence for it in at least some languages. In this section,
I would like to present cross-linguistic evidence that a partitive head which is responsible for introducing
theme arguments may indeed be commonly attested.
6.1 Partitives seem to be VP-internal
Cross-linguistically, partitive case appears to often be aVP-internal case. This is the case, for example, in
Finnish (Karlsson 1984, Kiparsky 1998), where partitive case appears only on the objects of transitive verbs
and the subjects of what Kiparsky calls ‘presentational verbs’, as in (27) (Kiparsky’s (64c)).
(27) Laps-i-achild-PL-PART
leikki-iplay-3SG
piha-lla.courtyard-ADESS
‘There are children playing in the courtyard.’
Kiparsky argues that these ‘presentational verbs’ have VP-internal subjects, granting them partitive case. It is
of note also that Kiparsky gives a gloss suggesting that partitive case in Finnish can license the ‘instance-of’
reading discussed in this paper (Kiparsky’s (9); the surrounding prose notes ‘in [(28) ], ‘this rose’ is partitive
because it means “roses of this particular kind”’):
(28) Puutarhurigardener
istutt-iplant-PAST.3SG
kaikkialleeverywhere
tatathis-PART
ruusu-a.rose-PART
‘The gardener planted this rose everywhere.’
Similar facts concerning VP-internal assignation of partitive case appear to be true of Italian, in particular
a set of facts from Italian often adduced in support of the unaccusative hypothesis; the partitivene clitic
can resume objects of transitives and subjects of unaccusatives, but not subjects of transitives, argued by
23
Andrew Weir Partitive readings and verb transitivity in English
Belletti (1988) to be a reflex of VP-internal assignation of partitive case. The Russian Genitive of Negation is
another case which brings the ‘existential’ meaning observed in Finnish in (27), and which is available (in the
scope of negation) in the object position of transitives andthe subject position of unaccusatives and passives
((Pesetsky 1982:47); although see also Partee & Borschev (2004) for some skepticism and a suggestion that
the distribution may be sensitive to conditions other than unaccusativity alone).
This consistency across at least a subset of languages suggests that the existence of a ‘partitive’ head, which
can introduce VP-internal mass or kind-type direct objectsand existentially quantify over them, is cross-
linguistically motivated. A head which has an overt effect in other languages – case morphology in Finnish
and Russian, the creation of a particular type of clitic pronoun in Italian – may also covertly appear in English;
while silent, the effects of the partitive head are detectable due to its ability to license kind-type arguments in
direct object position.
6.2 Unaccusative and unergative subjects in English
Given the argument put forward above, we make a prediction for English. Just as there is a distinction
between kind-type arguments in the object position of ‘basically unergative’ verbs (where the argument
may be introduced by a partitive head) and in the object position of transitive verbs (where the argument
cannot be so introduced), we predict that there should be a distinction between the subject position of
unaccusative and unergative verbs. The facts from the languages surveyed above suggest that the subject
position of unaccusatives, or passives, may itself be derived from a syntactically constructed direct object;
that is, the subject of an unaccusative or passive verb may beintroduced by a partitive head. By contrast,
the subject argument of an unergative verb cannot be so introduced. We therefore expect kind-type DPs
to be acceptable in the subject position of unaccusative or passive verbs (as they can be introduced VP-
internally by a partitive head), but relatively marked in the subject position of unergative verbs (as they
cannot be introduced by a partitive head, and must rather receive their existential reading from the application
of a marked procedure such as DKP). While I have not conductedan experimental investigation of this
hypothesis, informal introspection and inquiry among native English speakers suggest that this hypothesis
24
Andrew Weir Partitive readings and verb transitivity in English
may be borne out. Speakers judged kind-type DPs to be relatively more marked in the subject position of
transitives or unergatives (the (b) examples below) than inthe subject position of unaccusatives or passives
(the (a) examples).
(29) a. (At 4 p.m. precisely) the very kind of bird we were talking about fell from the sky.
b. ?(At 4 p.m precisely) the very kind of bird we were talking about sang a sweet melody.
(30) a. (At 4 p.m. precisely) that very kind of student arrived/knocked on the door.8
b. ?(At 4 p.m. precisely) that very kind of student told a story/put his hand up.
(31) a. (At 4 p.m. precisely) the very kind of animal we were talking about was shot in the woods.
b. ?(At 4 p.m. precisely) the very kind of animal we were talking about howled in the woods.
(32) [We are talking about a species of frog, the Bururi Long-Fingered Frog.9]
a. The frog was last seen by scientists at 7.30 pm, December 14th, 1949.
b. ?The frog last attacked a human at 7.30 pm, December 14th, 1949.
Judgments varied in their strengths between items and between consultants, but all consultants agreed that
the a-sentences were good, and that there was a grammaticality contrast of some sort between the a-sentences
and the b-sentences.10 Note that this contrast cannot be explained by appealing to the notion that existential
closure occurs only at the vP level, as in Diesing (1992); on that analysis, the subjects of stage-level predicates
like howl or attack, originating within vP, should be able to receive an existential reading. But this appears
difficult. I argue that this is the same effect as is seen in theeat/devourcases. An argument that is introduced
as a theme, whether it ends up as a subject or an object, can be introduced by the partitive head and so can
8Knock on the dooris an unaccusative, at least in my judgment:There knocked on the door many yowling students.9(32a) was adapted fromhttp://www.sciencedaily.com/releases/2012/03/120327152850.htm. The
context given to consultants was similar to that article; that scientists have recently rediscovered a type of frog.10If anything, the reports were that the contrast feels in somecases stronger than mere ‘markedness’; particularly in thecase of
(32), some consultants reported that (32b) was simply ungrammatical in the context given, where we are making clear thatwe aretalking about a species of frog, while (32a) was fine.
25
Andrew Weir Partitive readings and verb transitivity in English
receive an ‘instance-of’ reading. An ergative subject cannot.
6.3 French de
There may also be overt evidence for a functional head which performs a job very close to the null partitive
head I have proposed in English: namely, Frenchde. This word can do a multiplicity of jobs in French; as
well as having a possessive function (la plume de ma tante‘the pen of my aunt’), it also acts as the so-called
‘partitive article’ in cases such as the below.
(33) DesDE.the.PL
garconsboys
jouaient.were.playing
‘Boys were playing./Some boys were playing.’
(34) J’aiI=have
misput
deDE
l’huilethe=oil
dansin
lathe
casserole.saucepan
‘I put (some) oil in the saucepan.’
There is a considerable body of work discussing the semantics of this partitive article in French (see, for
example, Kupferman (1977, 1994), Anscombre (1996), Ihsane(2005), and references cited therein) and I
will not attempt to recap the work here. I will, however, focus on an intriguing piece of data from Kupferman
(1977). Kupferman argues that there are two sorts ofdewhich appear before direct objects in French. One,
exemplified in (35), carries agenuine‘partitive’ meaning, as the glosses indicate. Note that crucially the
sentences in (35) arenot being considered on a instance-of-a-kind reading. That is,a noun phrase likece
gateau ‘that cake’ is, in these cases, being understood as meaning that very cake, not the type of cake in
question.
(35) a. J’aiI=have
mangeeaten
deDE
cethat
gateau.cake
‘I ate some of that cake, a bit of that cake.’
b. J’aiI=have
budrunk
deDE
cethat
vin.wine
‘I drank some of that wine, a bit of that wine.’
26
Andrew Weir Partitive readings and verb transitivity in English
c. J’aiI=have
pristaken
deDE
cethat
cognac.cognac
‘I took some of that cognac, a bit of that cognac.’
However,dealso appears with different verbs such asvoir ‘to see’. I do not give free translations for these
cases yet. The judgments are Kupferman’s.
(36) a. J’aiI=have
vuseen
deDE
cettethat
viande.meat
b. Ilhe
montreshows
deDE
cettethat
biere.beer
c. Ilhe
apportebrings
deDE
cethat
cognac.cognac
Kupferman argues, on the basis of various syntactic diagnostics, that thedes in (35) and (36) are not the
same. Thedeshown in (35) is only available with a restricted set of verbs, such asmanger‘to eat’ orprendre
‘to take’. Such constructions can be analogized to the (slightly marginal) English constructions of the form
eat of (the bread), drink of (the water), where a genuinely ‘partitive’ semantics is at play; that is, eat of the
breadmeans to eat a part of the bread. Thisof, like the partitive Frenchde, is restricted in the verbs it can
co-occur with (??read of a book, ??write of a dissertation, ??watch of a film).
Of greater interest to us is the semantic judgment reported by Kupferman for the second kind ofde: ‘. . . j’ai
vu de ce gateaucan be paraphrased asI saw that sort of cake, while j’ai mange de ce gateau[. . . ] can only
be paraphrased asI ate a bit of the cake in question.’ (p. 10)11 A reading ofvoir demeaning something like
what Kupferman reports is precisely what would be expected if a head with the semantics I have proposed
for PART were available in French asde:
11‘ j’ai vu de ce gateauest paraphrasable parj’ai vu de cette sorte de gateau, tandis quej’ai mange de ce gateau[. . . ] ne peut etreparaphrase que parj’ai mange un peu du gateau en question’; my translation.
27
Andrew Weir Partitive readings and verb transitivity in English
(37) a. JdeK = JPARTK = λP.λe.∃x.P (x) & theme(x)(e)
b. Jce gateauK = Jthat cakeK = ΣRedVelvet (for example)
c. Jj’ai vu de ce gateauK = λe.∃x.∪Jce gateauK(x) & see(e) & theme(x)(e) & agent(me)(e)
‘x is an instance of that kind of cake, and I sawx’.
A question arises about the syntactic distribution of this element. In Kupferman’s syntax, thedeof something
like voir de is internal to the nominal constituent. This would be inconsistent with the story I have told for
the proposed null PART head in English, which I have analyzedas a functional head external to the nominal
projection. I will not attempt to embark on a syntactic analysis of Frenchde here. It is possible thatde is
the morphological realization of partitive case granted bythe (null) head PART; in this case, PART would be
external to the nominal projection in French as in English, but the presence of PART would be signaled by a
piece of morphology which surfaces asdebefore the noun.
A second problem which has to be addressed for the hypothesisthat Frenchde is dependent on a morpheme
like VP-internal PART is thatde can, in some form, appear in the subject position of transitives and
unergatives, contrary to expectations.
(38) a. DesDE.LES
garconsboys
jouaient.were.playing
‘Boys were playing./Some boys were playing.’
b. DeDE
l’eauthe=water
boueusemuddy
recouvraitcovered
lathe
route.road
‘Muddy water covered the road.’ (Kupferman 1977:(24))
My defense would be that this is, again, not the samede. I will not attempt to analyze the semantics of thisde;
however, it does not appear to be the same form ofdeas is introducing the objects in the cases in (36). The
form in (38) may be a genuine article, something akin tosomein English. Luraghi (forthcoming) suggests
that this article may have been derived diachronically fromthe partitivede (in both French and Italian; in
a cross-linguistic survey, she argues that various languages show various forms which belong on a ‘cline of
28
Andrew Weir Partitive readings and verb transitivity in English
grammaticalization’ between a partitive case or preposition and an indefinite article). However, thede in
(38) no longer bears the partitive semantics and is rather simply an indefinite article. This may be what has
happened here. There is some evidence thatderetains the ‘instance-of’ meaning in the object positions of at
least some verbs, but this appears to be systematically impossible in subject position, as expected.12
(39) a. Ilhe
vagoes
apporterbring
deDE
cethat
bierebeer
(a(to
lathe
feteparty
cethis
soir).evening)
‘He’ll bring that kind of beer to the party this evening’
b. J’aiI=have
chassehunted
deDE
cetthat
animalanimal
(quand(when
j’etaisI=was
jeune)young)
‘I hunted that kind of game when I was young’ (consultant’s free translation)
c. %J’aiI=have
vuseen
deDE
cettethat
viande.meat
‘I have seen that (type of) meat.’
(40) a. *DeDE
cethat
vinwine
seREFL
vendsells
ici.here
(‘That (kind of) wine is sold here’; cp.on vend de ce vin ici‘that kind of wine is sold here’,
with ce vinin object position)
b. *DeDE
cetthat
animalanimal
ahas
hurlehowled
dansin
lethe
foret.forest
(‘That (kind of) animal howled in the forest.’)
c. *DeDE
cettethat
voiturecar
ahas
percutecrashed.into
una
murwall
(‘That (kind of) car crashed into a wall.’)
In the interests of full disclosure, it should be noted that my consultant rejected sentences like (39a) withvoir
de ‘seeDE’, even in a context like (41) below.
12I thank Jeremy Pasquereau for judgments here. The task presented to Jeremy is presented in Appendix C.
29
Andrew Weir Partitive readings and verb transitivity in English
(41) We are in a cafe. On the counter there is a particular type of cake with fruit and chocolate, a type
I haven’t seen before. I wonder if this type of cake is a specialty of the region. I know that you’ll
know, so I turn to you and say:
a. Tuyou
ashave
dejaalready
vuseen
deDE
cettethat
gateau?cake
(‘Have you seen that (kind of) cake before?’)
corrected to:tu as deja vu ce genre/type de gateau? ‘you have already seen that kind/type of
cake?’
This is contra the judgment contained in Kupferman (1977) for voir. My consultant also rejectedmontrer
de ‘to point outDE’, although Kupferman accepts it. It is possible that this difference between verbs reflects
a dialectal difference; I have no immediate explanation forwhy some verbs should allow thisde in object
position and others not. However, in subject position, it isalways rejected. We can see thatdephrases, on
the ‘instance-of’ reading, are felicitous in object position but not in subject position, as would be expected if
PART is strictly a VP-internal head.
We have seen that there is cross-linguistic evidence that partitives are linked with the ‘theme’ thematic role;
partitive case appears with theme arguments but not with agents, for example, and Frenchdeappears (on the
relevant reading) in the object position of transitives butnot the subject position. As such, there is evidence
to suggest that a PART head as a means of introducing a theme argument is motivated rather than being an
ad hocproposal.
7 Further work and conclusion
In this paper, I have proposed an analysis of some curious data concerning the availability of the ‘instance-of’
reading in the object position of ‘basically unergative’ verbs, but not in other positions. I argued that this is
due to the availability of a partitive head, which can introduce the theme argument of unergatives, but which
cannot appear in other configurations. Much remains to be said to flesh out the theory I propose here. One
area I have not addressed at all is interactions with bare plurals. ‘Instance-of’ readings are available, by my
30
Andrew Weir Partitive readings and verb transitivity in English
intuition easily, with bare plurals in any position.
(42) a. I ate pizzas (yesterday).
b. I devoured pizzas (yesterday).
c. Lions ruined my garden (yesterday).
I make no argument here about how to interpret bare plurals, but if we treat them as kind-denoting, as for
example Carlson (1977) and Chierchia (1998) do, then there is an incompatibility with the analysis put
forward in this paper. Sentences like (42b, c) should feel ‘marked’; special type-shifting devices (like DKP)
should be required to make them interpretable. Intuitivelya sentence like (42c) does not feel as marked as
?That kind of animal ruined my garden yesterday, however, and if this intuition is well-grounded, then either
my position is wrong – theme-introducing PART and marked DKPdo not represent the only way of deriving
an ‘instance-of’ reading of a kind argument – or bare pluralsdo not denote kinds (but rather, for example,
properties). I do not attempt to resolve this argument here,but simply flag the concern.
A further worry, raised (in a slightly different form) by Seth Cable (p.c.), concerns cases like (43) (which
context and example are, again, entirely due to Cable):
(43) a. Three days ago, you told me about Razzle bread and how good it was. You also gave me a
chocolate cake. Yesterday, I finally got my hands on some Razzle bread, and ate it with some
of the cake you gave me. We’re now talking about our meals yesterday, and I wish to tell you
about mine.
b. Yesterday, I ate that bread you were telling me about and that cake you gave me.
The concern is the conjunction. If we assume thatthat breadis introduced by PART, butthat cake(as an
entity) is introduced by something else (some other theme-introducing head, perhaps), the grammaticality of
(43b) poses a problem. We cannot generally conjoin two ‘different’ argument structures under the same verb
(as shown in (44)), which is what (43b) would amount to.
31
Andrew Weir Partitive readings and verb transitivity in English
(44) a. *I drank [[some beer] and [the teapot dry]].
b. *I sang [[a sweet song] and [the baby asleep]].
c. *I drove [[a fast car] and [a hole in my tires]].
d. *I bought [[a car] and [Mary a gift]].
Conjoining the ‘same’ argument structures, however, is good:
(45) a. I drank the teapot dry and the pint glass empty.
b. I sang the baby asleep and the neighbors awake.
c. I drove a hole in my tires and a rut in the road.
d. I bought Mary a gift and John a book.
One suggestion could be that, in (43b), the ‘instance-of’ reading ofthat breadis not being brought about by
PART, but by DKP, perhaps in the following way:
(46) VP
V
eat
ConjP
ThemePλe.∃x.∪ΣRazzle(x) & theme(x)(e)
(via DKP)
Themeλx.λe.theme(x)(e)
DPΣRazzle
that bread
and ThemePλe.theme(theCake)(e)
Themeλx.λe.theme(x)(e)
DPtheCake
that cake
32
Andrew Weir Partitive readings and verb transitivity in English
The semantics given above would predict thatI ate that bread and that cake(in the context in (43)), as it
would require DKP, should prompt the ‘markedness’ effect. Isuspect that confirmation of this would require
subtle psycholinguistic experimentation which is outsidethe scope of this paper: an interesting possibility
for future work, however, would be to test psycholinguisticresponses to stimuli such asI ate that cake and
that (kind of) bread. Would processing be slower for the kind DP (because DKP is forced) than for the
object DP? The analysis here would predict so. There is also aproblem with the precise semantics given in
(46): one event is asserted to have two themes (an instance ofRazzle Bread, and a cake). Whether this is an
insurmountable problem is a question I do not tackle here.
These problems aside, however, I hope to have made the case inthis paper for a (silent) partitive head in
English, which can introduce theme arguments of kind-type and existentially quantify over them. The fact
that such an interpretation seems to only be possible for syntactically constructed direct objects is linked
to the fact that partitive is, cross-linguistically, a VP-internal case; if a partitive head is one way in which
internal arguments of a verb can be introduced, then we derive the asymmetry between the possibility of an
‘instance-of’ reading of a syntactically constructed direct object, and the relative markedness of the ‘instance-
of’ reading with non-constructed direct objects and with ergative subjects.
33
Andrew Weir Partitive readings and verb transitivity in English
References
Anscombre, Jean-Claude. 1996. Partitif et localisation temporelle.Langue francaise109. 80–103.
Beck, Sigrid & Kyle Johnson. 2004. Double objects again.Linguistic Inquiry35(1). 97–124.
Belletti, Adriana. 1988. The case of unaccusatives.Linguistic Inquiry19(1). 1–34.
Borer, Hagit. 2005.The normal course of events, vol. 2 Structuring Sense. Oxford: Oxford University Press.
Carlson, Gregory N. 1977.Reference to kinds in English: University of Massachusetts Amherst dissertation.
Chierchia, Gennaro. 1998. Reference to kinds across languages.Natural Language Semantics6. 339–405.
Diesing, Molly. 1992. Bare plural subjects and the derivation of logical representations.Linguistic Inquiry
23(3). 353–80.
Dowty, David. 1978. Governed transformations as lexical rules in a Montague Grammar.Linguistic Inquiry
9(3). 393–426.
Folli, Raffaella & Heidi Harley. 2004. Flavours of v: Consuming results in Italian and English. In Roumyana
Slabakova & Paula Kempchinsky (eds.),Aspectual inquiries, 95–120. Dordrecht: Kluwer.
Harley, Heidi. 2002. Possession and the double object construction. In Pierre Pica (ed.),Linguistic variation
yearbook, vol. 2, 31–70. Amsterdam: John Benjamins.
Ihsane, Tabea. 2005. On the structure of Frenchdu/des‘of.the’ constituents. InGenerative Grammar in
Geneva, vol. 4, 195–225.
Karlsson, Fred. 1984.Finnische Grammatik. Hamburg: Helmut Buske Verlag.
Kiparsky, Paul. 1998. Partitive case and aspect. In Miriam Butt & Wilhelm Geuder (eds.),The projection of
arguments: lexical and compositional factors, Stanford: CSLI Publications.
34
Andrew Weir Partitive readings and verb transitivity in English
Kratzer, Angelika. 1996. Severing the external argument from its verb. In J. Rooryck & L. Zaring (eds.),
Phrase structure and the lexicon, Dordrecht: Kluwer.
Kratzer, Angelika. 2003. The event argument.http://semanticsarchive.net/Archive/
GU1NWM4Z/.
Kratzer, Angelika. 2005. Building resultatives. In C. Maeinborn & A. Wollstein-Leisten (eds.),Event
arguments in syntax, semantics and discourse, 177–212. Tubingen: Niemeyer.
Kratzer, Angelika. 2008. On the plurality of verbs. In Johannes Dolling, Tatjana Heyde-Zybatow & Martin
Schafer (eds.),Event structures in linguistic form and interpretation, 269–300. Berlin: Walter de Gruyter.
Krifka, Manfred. 1995. Common nouns: a contrastive analysis of Chinese and English. In Gregory N.
Carlson & Francis J. Pelletier (eds.),The generic book, 398–411. Chicago: University of Chicago Press.
Kupferman, Lucien. 1977. L’article partitif existe-t-il?Le francais moderne1–16.
Kupferman, Lucien. 1994. Du : un autre indefini ?Faits de Langue4. 195–203.
Levin, Beth & Malka Rappaport Hovav. 2011. Lexical conceptual structure. In K. von Heusinger,
C. Maienborn & Paul Portner (eds.),Semantics: an international handbook of natural language meaning
i, 418–38. Berlin: Mouton de Gruyter.
Luraghi, Silvia. forthcoming. Partitives and differential marking of core arguments: a cross-linguistic survey.
Ms., Universita di Pavia. Submitted.http://allegatifac.unipv.it/silvialuraghi/
Partitives.pdf.
Marantz, Alec. 1984.On the nature of grammatical relations. Cambridge, MA: MIT Press.
Partee, Barbara H. 1986. Noun phrase interpretation and type-shifting principles. In J. Groenendijk,
D. de Jong & M. Stokhof (eds.),Discourse representation theory and the theory of generalized quantifiers,
115–43. Dordrecht: Foris.
35
Andrew Weir Partitive readings and verb transitivity in English
Partee, Barbara H. & Vladimir Borschev. 2004. The semanticsof Russian Genitive of Negation: The nature
and role of Perspectival Structure. In Robert B. Young (ed.), Proceedings of SALT 14, 212–34. Ithaca, NY:
CLC Publications, Cornell University.
Pesetsky, David. 1982.Paths and categories: Massachusetts Institute of Technology dissertation.
Pesetsky, David. 1995.Zero syntax. Cambridge, MA: MIT Press.
Pylkkanen, Liina. 2008.Introducing arguments. Cambridge, MA: MIT Press.
Pylkkanen, Liina & Brian McElree. 2006. The syntax-semantic interface: On-line composition of sentence
meaning. In Matthew Traxler & Morton Gernsbacher (eds.),Handbook of psycholinguistics, chap. 14,
539–79. Elsevier 2nd edn.
Rappaport Hovav, Malka & Beth Levin. 1998. Building verb meanings. In Miriam Butt & Wilhelm Geuder
(eds.),The projection of arguments: lexical and compositional factors, Stanford: CSLI Publications.
von Stechow, Arnim. 1995. Lexical decomposition in syntax.In Urs Egli, Peter E. Pause, Christoph
Schwarze, Arnim von Stechow & Gotz Wienold (eds.),The lexicon in the organization of language, 81–
118. Amsterdam: John Benjamins.
von Stechow, Arnim. 1996. The different readings ofwieder ‘again’: A structural account.Journal of
Semantics13(2). 87–138.
von Stechow, Arnim. 2001. Temporally opaque arguments in verbs of creation. In C. Cecchetto, G. Chierchia
& M. Guasti (eds.),Semantic interfaces: Studies offered to Andrea Bonomi on the occasion of his sixtieth
birthday, 278–319. Stanford: CSLI Publications.
Wilkinson, Karina. 1995. The semantics of the common nounkind. In Gregory N. Carlson & Francis J.
Pelletier (eds.),The generic book, 383–97. Chicago: University of Chicago Press.
36
Andrew Weir Partitive readings and verb transitivity in English
A Experiment 1 stimuli
A.1 Test items
(1) a. I ate that bread for my lunch yesterday.
b. I devoured that bread for my lunch yesterday.
(2) a. The student painted that bird yesterday in his art class.
b. The student depicted that bird yesterday in his art class.
(3) a. The nurse cleaned that instrument yesterday in the operating theater.
b. The nurse sanitized that instrument yesterday in the operating theater.
(4) a. I shot that animal yesterday when I was out hunting.
b. I euthanized that animal yesterday in the surgery.
(5) a. You shot this bird yesterday! I saw you!
b. You murdered this bird yesterday! I saw you!
(6) a. The hunter stabbed this fish yesterday with a fish spear.Then he took his catch home and ate it.
b. The hunter impaled this fish yesterday with a fish spear. Then he took his catch home and ate it.
(7) a. I scratched that metal yesterday with my fingernail andleft a visible mark.
b. I nicked that metal yesterday with my fingernail and left a visible mark.
(8) a. A tow truck pulled that car all the way to Argentina.
b. A tow truck lugged that car all the way to Argentina.
37
Andrew Weir Partitive readings and verb transitivity in English
A.2 ‘Catch’ trials
(9) Expected to be bad on the ‘instance-of’ reading:
a. I gave that student a stern lecture yesterday.
b. The president visits that city frequently.
c. That country has a very high level of debt.
d. I could never vote for that political party.
e. That airport has a very inefficient air conditioning system.
f. That book scandalized the nation.
(10) Expected to be good on the ‘instance-of’ reading:
a. I drank that coffee for years.
b. We started growing that crop on the farm five seasons ago.
B Experiment 2 stimuli
(11) a. I ate that kind of pasta for my lunch yesterday.
b. I devoured that kind of pasta for my lunch yesterday.
(12) a. The artist painted that kind of animal yesterday in his studio.
b. The artist depicted that kind of animal yesterday in his studio.
(13) a. The technician cleaned that kind of instrument yesterday in the lab.
b. The technician sanitized that kind of instrument yesterday in the lab.
(14) a. Bob shot that kind of bird yesterday when he was out hunting.
b. Bob euthanized that kind of bird yesterday in the veterinary surgery.
38
Andrew Weir Partitive readings and verb transitivity in English
(15) a. The greedy hunters shot that kind of bird yesterday.
b. The greedy hunters murdered that kind of bird yesterday.
(16) a. The hunter stabbed that kind of deer yesterday with a spear.
b. The hunter impaled that kind of deer yesterday with a spear.
(17) a. John scratched that kind of glass yesterday with his fingernail.
b. John nicked that kind of glass yesterday with his fingernail.
(18) a. A tow truck pulled that kind of car down the highway yesterday.
b. A tow truck lugged that kind of car down the highway yesterday.
C French task
The task and replies are presented below.
“What I’d like you to do is:
1. Tell me if the sentences are grammatical or not. (or use finer gradations if you want, like OK/?/*)
2. If a sentence is grammatical, I’d like you to try to give a free translation of it into English.
3. If a sentence isn’t grammatical, I’d like you to suggest ifthere is a way of making it grammatical (for
example, have I used a wrong preposition, have I got a gender wrong, should I change the tense, have
I just spelt a word wrong, etc.)”
• Des garcons jouaient dans la rue.
Good: ‘Boys played in the street.’
• J’ai mis de l’huile dans la casserole.
Good: ‘I put some oil in the saucepan.’
39
Andrew Weir Partitive readings and verb transitivity in English
• J’ai mange de ce gateau.
Good: I ate some of this cake.
• J’ai bu de ce vin.
Good: I drank some of this wine.
• J’ai pris de ce cognac.
Good: I took some of this cognac.
• J’ai vu de cette viande.
Wrong: J’ai vu de la viande = I saw some meat; j’ai vu un peu de viande = I saw a little meat.
• Il montre de ce cognac.
Wrong: Il montre ce cognac = He is pointing out to this cognac.
• Il apporte de cette biere (a la fete ce soir).
?: I guess the sentence sounds better if, given the time adverbial you’ve used, you change the tense to
the future. Il apportera/Il va apporter de cette biere. . .
• J’ai chasse de cet animal.
?: For it to sound natural, I’d add a time adverbial like “quand j’etais jeune” = “when I was young”. I
hunted this kind of game as a child.
• J’ai conduit de cette voiture.
Wrong: J’ai conduit cette voiture = I drove this car.
• De ce vin se vend ici.
Wrong: Ce type/genre de vin se vend ici. They sell this kind ofwine here.
• De cette voiture a percute un mur. [‘DE that car crashed into a wall.’]13
Wrong: Cette voiture a percute un mur.
13Anything in square brackets is presented for information here and was not presented to the consultant.
40
Andrew Weir Partitive readings and verb transitivity in English
• De cet animal a crie dans la foret. [‘DE that animal shouted in the forest.’]
Wrong: ‘de cet animal’ is wrong, but also because animals ‘hurler’ (howl) not ‘crier’ (only for humans).
Follow up questions:
(1) Given this context, could you utter the sentence in below(does it improve at all?)
We are in a cafe. On the counter there is a particular type of cake with fruit and chocolate, a type I
haven’t seen before. I wonder if this type of cake is a specialty of the region. I know that you’ll know,
so I turn to you and say:
a. Tu as deja vu de ce gateau?
Answer: unequivocally no.
(2) Given this context, could you utter the sentence below (does it improve at all?)
We are looking at a car magazine and look at a picture of a car. Iwant to warn you not to buy the type
of car shown in the picture, because I think they are unsafe tohandle. I know this because:
a. De cette voiture a percute un mur hier.
Answer: unequivocally no.
(3) Are these sentences grammatical or not?14
a. J’ai chasse de cet animal quand j’etais jeune.
b. Je chassais de cet animal quand j’etais jeune.
c. J’ai chasse de cet animal hier.
d. Je chassais de cet animal hier.
e. J’ai chasse cet animal hier.
Answer: ‘Differences in aspect notwithstanding, I think all of these sentences are somewhat at
the edge but I consider them grammatical. I think here “animal” is used as a mass noun like
14j’ai chasseis the present perfect,je chassaisis the imperfect, of the verbchasser‘to hunt’.
41
Andrew Weir Partitive readings and verb transitivity in English
“gibier=game” and maybe that’s what might make someone raise an eyebrow. For instance, I can’t
see myself saying “J’ai chasse de l’animal. . . ” so the demonstrative seems to play a key role whereas
I can say both “J’ai bu du lait” and “J’ai bu de ce lait”.’
(4) Is the sentence below good in this context?
We are in a liquor store wondering what beer to buy to take to the party. I look at Hoegaarden, but
you tell me that I shouldn’t buy that, because you already know:
a. Marie va apporter de cette biere a la fete. (Donc, nous devrions choisir un autre type.)
[‘Marie is going to bringDE that beer to the party. So we should choose another type.’]
Answer: ‘Yes that’s one good context in which I would say it this way. My feeling is that, although it’s
grammatical and perfectly comprehensible, in some way it sounds a bit off and I think that someone
such as a French teacher or anyone who prides themselves in knowing how to speak “correct French”
would correct the speaker and say “Marie va apporter cette sorte de biere a la fete.”.’
42