No Cause for a Small Clause?
(Non-)Arguments for the Structure of Resultatives

Marcel den Dikken & Eric Hoekstra

J.-W. Zwart (ed) Minimalism and Kayne’s Asymmetry Hypothesis. Groninger Arbeiten zur Germanistischen Linguistik 37, 89-105.

Carrier and Randall (1992) (henceforth C&R) present two arguments for a ternary-branching analysis of resultative constructions, as in (1). (1a) instantiates what C&R refer to as the transitive resultative construction, featuring a transitive matrix verb, and (1b) instantiates the intransitive resultative construction:

(1) VP

a. water the tulips flat
b. run their Nikes threadbare

In both cases, the postverbal NP is T-marked by AP; on C&R’s assumptions, in the transitive resultative construction NP is also T-marked by V.
In addition, C&R provide six arguments against a binary small clause (SC) analysis, which in its archetypal form is illustrated in (2) below (where “SC” is a convenient shorthand for whatever the categorial status and internal structure of small clauses may turn out to be):

(2) VP

run NP AP
the tulips flat
their Nikes threadbare

The purpose of our reply is to call the ternary analysis into question, arguing that the case for a SC is actually much stronger than C&R suggest. To the extent that we succeed, we obviate the need for weakening the T-Criterion, forfeiting the restrictive view that a subject and a predicate consistently form a single constituent, and abandoning strict binarity. We adopt the framework of Chomsky (1986) in order to have the same frame of reference as C&R.

1. Two arguments from selection

C&R's major arguments for a ternary-branching approach to resultatives come from the realm of selectional restrictions imposed by the verb on (i) the result phrase and (ii) the postverbal NP. We will discuss and disqualify these in turn.

1.1. Selection of the resultative phrase by V

C&R (section 2.1.) observe, following Green (1972) and others, that there are semantic restrictions on result phrases. These restrictions, illustrated by the examples in (3) (from C&R's 25a-b), are found in both transitive and intransitive resultative constructions:

(3) a. The maid scrubbed the pot shiny/*shined/*shining
b. The tourists walked their feet sore/blistery/*blistered

A crucial assumption in C&R's account of these facts is that “the semantic features of a result predicate cannot percolate up to the next higher node the way categorial features can” (1992:184). This leads them to conclude that the facts in (3) show that the verb must be a sister of the resultative phrase, as in their ternary-branching analysis. The premise on which this argument is based is questionable, however, in view of data like the following, which show that the semantic features denoting negation (4) and aspect (5) can percolate quite readily.

(4) No boy's father saw anything
(5) a. She played the sonata (terminative aspect)
b. She played sonatas (durative aspect)

The facts in (5) are especially relevant in the light of C&R’s observation that the illformedness of (3) with -ed and -ing adjectives is due to “an aspectual clash between the meaning of resultatives and the meanings of -ed and -ing adjectives” (1992:184). If it is really aspect that is responsible for the data in (3), then the fact that aspectual properties do seem to percolate up, as shown in (5), undermines this argument of C&R’s for a ternary-branching analysis of resultatives.
But even if it should be true that semantic features do not percolate, the requirement that the verb be a sister of the resultative phrase is directly accommodated by a binary-branching SC analysis according to which SC's are projections of the predicate head, as proposed by Stowell (1981), Manzini (1983), Koopman and Sportiche (1991), among others.1

1.2. Selection of the postverbal NP by V

The verb in transitive resultative constructions seems to impose selectional restrictions on the postverbal NP, whereas in intransitive resultatives, no such selectional relationship is found. C&R illustrate this difference between transitive and intransitive constructions with examples of the following type:

(6) a. The bears frightened the hikers
b. * The bears frightened the campground
(7) a. The bears frightened the hikers speechless
b. * The bears frightened the campground empty
(8) a. The bears growled the hikers speechless
b. The bears growled the campground empty

C&R conclude from this that the verb T-marks the postverbal NP in a transitive resultative construction. This would seem to be impossible on a SC-analysis, in which the postverbal NP is not a sister of the verb. C&R argue that the selectional correspondence between (6) and (7) shows that the postverbal NP is a sister of V, regardless of the presence of an (overt) result predicate.
Our way of countering this argument is to show that it is really an illusion that the verb ever T-selects the postverbal NP in transitive constructions of the type in (6) and (7).2
Consider in some more detail the pair of examples in (7). What is striking about these sentences (and about all of C&R's transitive resultative pairs) is that the a- and b-examples do not form a truly minimal pair. It is not just the postverbal NP that is different, but the resultative phrases are conspicuously different as well. The adjective speechless used in (7a) must be predicated of an animate NP like the hikers, while empty in (7b) can only be felicitously combined with inanimate NPs like the campground. The animacy contrast between (7a) and (7b) may now be taken to be an immediate side-effect of the s-selectional relationship between frighten and the type of resultative phrases that it may be combined with. What must, on all accounts, be specified about the verb frighten is that it does not s-select resultative phrases like empty. This being said, the deviance of (7b) follows from the s-selectional relationship between V and the resultative phrase (cf. section 1.1), without anything being said about the postverbal NP. The verb s-selects the resultative phrase, and the resultative phrase in turn “selects” (is predicated of) the postverbal NP.
So far, we have concentrated on the examples in (7). C&R's main point, however, is that the same “selectional restrictions” show up in the simple transitives in (6), and that hence (6) and (7) must have the same structure. We agree. We will argue in the next section that (6) contains an empty resultative predicate heading the verb's SC complement. The account proposed for (7) can be extended to (6). The facts in (6) and (7) can thus be shown to be neutral with respect to the choice between a SC analysis or a ternary-branching account of the type proposed by C&R. We will present further evidence that does discriminate between the two analyses, and which turns out to favor our approach over C&R's.

2. Selectional restrictions on the postverbal NP as an argument for the SC analysis

Consider the data in (9)-(11) (cf. Hoekstra 1991):3

(9) a. Jan sloeg zijn broertje het ziekenhuis in
Jan hit his brother the hospital into
b. Jan sloeg zijn broertje
Jan hit his brother
(10) a. Jan sloeg het kopje stuk
Jan hit the cup broken
b. * Jan sloeg het kopje
Jan hit the cup
(11) a. Jan sloeg de bal weg
Jan hit the ball away
b. Jan sloeg de bal
Jan hit the ball

The examples in (9) seem to show that Dutch slaan is like the transitive verbs discussed by C&R, the selectional restrictions of the verb being identical in resultative (9a) and non-resultative constructions (9b). From C&R's perspective, however, the difference between (10a) and (10b) is unexpected. After all, if the verb slaan cannot select an inanimate object in a transitive construction like (10b), C&R would predict that this selectional property should carry over to the transitive resultative in (10a), but it does not.4 Furthermore, the data in (11) would seem to suggest that this verb does select inanimate objects. There appears to be conflicting evidence concerning the selectional restrictions that slaan imposes on its object. How can these facts be made sense of?
The key to a proper understanding of the difference between these sentences lies in recognizing that (11b), but not (9b) and (10b), has a resultative interpretation. In this respect (11b) is semantically similar to (11a). Following a suggestion of Hoekstra (1991), we submit that the resultative semantics of the VP is reflected in the syntactic structure of this construction. All resultatives involve SC-complementation. In the specific case of (11b), then, slaan takes a SC-complement with an empty predicate, the empty counterpart of the particle weg ‘away’ (compare 11a with 11b), as illustrated below:

(12) [VP [SC de bal [e] ] slaan ]

This directly accounts for the fact that (10b) could be used in the context of an Alice-in-Wonderland baseball match in which cups are used instead of balls.
Let us now turn to the contrast illustrated in (9b) and (10b). In keeping with the non-resultative interpretation of (9b), we propose that the verb slaan here selects a NP-complement rather than a SC-complement. The verb slaan imposes on its NP-complement the restriction that it be animate. Since this constraint is violated in (10b), this example is bad in a real-life context, though it could be used with an Alice-in-Wonderland interpretation according to which cups are live creatures that are being (non-resultatively) hit.
In resultative constructions, slaan does not take a NP-complement but a SC-complement. Hence it cannot require that the postverbal NP be animate. The grammaticality of (11b) is then accommodated, given that in this example the postverbal NP is not (in fact, cannot be) selected by the verb. The SC analysis of (11b) accounts for its resultative interpretation, and for the possibility of having an inanimate NP.
Let us now return to the data in (6-7), repeated below as (13-14):

(13) a. The bears frightened the hikers (=6a)
b. * The bears frightened the campground (=6b)
(14) a. The bears frightened the hikers speechless (=7a)
b. * The bears frightened the campground empty (=7b)

We have argued that the apparent selectional relationship between V and NP in (14) is really a side-effect of the s-selectional relationship between V and the resultative phrase. We can now do the same thing for the examples in (13), once we recognize that (13a), in spite of featuring no overt resultative phrase, has resultative semantics, just like (11b) from Dutch.
Consistently analyzing resultative constructions in terms of SC-complementation (following among others Hoekstra 1988, 1991, Den Dikken 1992, Sybesma 1992), we are then led to the conclusion that (13a) involves an empty resultative phrase functioning as the predicate of the verb's SC-complement.5 Just like we did in the case of (11b), we can attribute specific semantic properties to this empty resultative phrase.6 Specifically, we may say that this resultative predicate is like speechless in (14a) (and unlike empty in 14b) in taking inanimate subjects only. In this way, the parallelism between (13) and (14) can be made to follow.
Finally, the fact that frighten and growl apparently differ in the kinds of postverbal NPs they can “take” reduces to a lexical distinction between the types of resultative complements they select. That no deep dichotomy between transitive and intransitive verbs is at stake is evident from our earlier examples in (10), which show that the selectional restrictions that V imposes on its object in a truly non-resultative transitive construction do not in fact carry over to the corresponding resultative construction.
In sum, in none of the examples (6-8) is there a selectional relationship between the verb and the postverbal NP. The verb s-selects a specific type of resultative phrase, and this resultative phrase in its turn predicates of a specific type of subject.

3. The remaining arguments

3.1. Long-distance extraction

C&R (section 2.2.) observe that long-distance extraction of result phrases patterns with long-distance extraction of objects:

(15) a. ? How flat do you wonder whether they hammered the metal t b. ? Which boys do you wonder whether we should punish t

C&R (1992:185) conclude from this that resultative phrases are lexically governed by the verb. Notice that, in the face of facts of long-distance extraction of the subject of the resultative phrase, C&R (1992:204) conclude that “lexical government does not require T-marking”. But since being lexically governed by a head does not presuppose being an internal argument of that head, the facts of long-distance extraction do not argue for C&R's (1992:184) claim that “the result XP is an internal argument of the verb”. The data are fully compatible with the SC analyses which C&R discuss.7

3.2. An apparent asymmetry

C&R base three arguments on data from middle formation, adjectival passives and resultative nominals. In each case, transitive resultatives seem to behave differently from intransitives, as illustrated below:

(16) Middle Formation
a. New seedlings water flat easily
b. * Competition Nikes run threadbare easily
(17) Adjectival Passive Formation
a. the stomped-flat grapes
b. * the danced-thin soles
(18) Resultative Nominal Formation8
a. The watering of tulips flat is a criminal offense
b. * The talking of your confidant silly is a bad idea

C&R show that postverbal NPs of transitive resultatives behave like direct objects. The generalization they propose is that only NPs which are T-marked by V can undergo the three processes illustrated above. A SC analysis cannot make this generalization, since the postverbal NP is not T-marked by V. However, we call this generalization into question, basing ourselves on facts from Dutch.
Dutch facts fail to display any asymmetry between transitive and intransitive resultatives with respect to the three phenomena:

(19) Middle Formation9
a. ? Kleine plantjes gieten zo lekker/gemakkelijk plat
little plants water so nicely/easily flat
b. ? Goedkope schoenen lopen zo lekker/gemakkelijk scheef
cheap shoes run so nicely/easily threadbare
(20) Adjectival Passive Formation
a. de platgestampte druiven
the flat-stomped grapes
b. de dungedanste zolen
the thin-danced soles
(21) Resultative Nominal Formation
a. het platgieten van tulpen
the flat watering of tulips
b. het gekpraten van je vertrouweling
the silly talking of your confidant

The grammaticality of the (b)-sentences contradicts C&R's generalization that only arguments which are T-marked by V participate in the three processes. A comparison of Dutch and English suggests that the basic symmetry between transitive and intransitive resultatives is obscured in English by some language-particular factor.

3.3. Left-branch effects

C&R present empirical evidence testifying to the absence of left-branch effects under subextraction from the postverbal NP in (both types of) resultative constructions:

(22) a. the gang (that) you shot the leaders of t dead
b. the gang (that) I drank the leaders of t under the table

C&R construe this as an argument against a binary SC analysis, under which these (grammatical) sentences violate the Left Branch Subpart Condition (LBSC). Obviously, however, the LBSC should be reduced to independent principles of the theory. In Kayne's (1984) work, for example, the LBSC was a direct corollary of the overall Connectedness framework. However, C&R adopt Chomsky's Barriers theory, and in this theory, the LBSC no longer follows in the case of complement SCs. Since Chomsky assumes that L-marking can take place via Specifier-Head agreement (which obtains in SCs), the theory predicts that the subject of a SC governed by a lexical T-marker will be transparent for subextraction. The Barriers theory is therefore perfectly compatible with a SC treatment of resultatives. As the extraction facts do not decide between the two analyses, no argument can be built on them.

3.4. Predication and Control

As Simpson (1983) first pointed out, resultatives cannot be predicated of D-structure subjects of S. This is illustrated below (cf. C&R's 107):

(23) a. * The tenors sang hoarse
b. * Joggers often run sick
c. * The teacher talked blue in the face
d. * The tourists walked ragged

A binary SC analysis would assign examples like these a structure of the type in (24) (cf. C&R's 118):

(24) [IP the tenors [VP sang [SC PRO hoarse]]]

C&R (1992:220) claim that, in order to accommodate the deviance of (23), the SC analysis would need “an extra stipulation to the effect that a subject NP cannot control a PRO subject of a resultative SC complement”, adding in a footnote that the illformedness of (23) would in fact follow from the PRO-theorem “if the verb were to govern into the SC” (fn. 47). As a matter of fact, PRO is effectively governed by V on virtually all current definitions of government. Hence the binary SC analysis of resultatives encounters no problems in the domain of examples like (23). Quite the contrary: “Simpson's Law” is one of the strongest arguments for a SC analysis of resultative constructions (cf. also Hoekstra 1988).
Interestingly, C&R's own analysis of Simpson's Law forces them to abandon the VP-internal subject hypothesis (defended by Fukui and Speas 1986, Kuroda 1988, Sportiche 1988, Koopman and Sportiche 1991, Burton and Grimshaw 1992, McNally 1992, and others). C&R argue that the appropriate condition on predication is mutual m-command. The subject is base-generated as a daughter of S, as in the example below:

(25) [S the tenors [VP sang hoarse]]

Since the tenors and hoarse are not in a configuration of mutual m-command, the impossibility of a predicative relationship between them follows. If, however, the VP in (25) contained a trace of the subject NP the tenors, this trace and the predicative AP would stand in a configuration of mutual m-command, and predication of the subject would be possible. Thus Simpson's Law is incorporated at the cost of forfeiting the well-motivated VP-internal subject hypothesis.
C&R do not have any conceptual motivation for making this move. The VP-internal subject hypothesis, on the other hand, has a conceptual reason for excluding the option of base-generating the subject as a daughter of S: it requires that all T-roles of a head be discharged in the domain of the projection of the head. Empirical and conceptual evidence thus indicates that the VP-internal subject hypothesis must be maintained. The binary SC analysis allows us to do this and, moreover, directly accounts for Simpson's Law.

4. Concluding Remarks

Ternary branching is less restrictive than binary branching. The burden of proof in matters of syntactic tree structures hence a priori resides with those who invoke ternary branching. C&R take up this challenge, but the arguments they present do not hold water and some of them may in fact swing the empirical pendulum in the direction of a SC approach. Furthermore, the SC Analysis is conceptually superior in its restrictive use of tree structures and of Theta theory.10 The SC approach does not embody a research strategy which has been unsuccessful in the past. Some of the major breakthroughs in the past decade consist in proposing larger (but not X'-theoretically different) tree structures: consider, for instance, the Unaccusativity Hypothesis, the introduction of I' and IP instead of S, of C' and CP instead of S', the idea that IP itself must be split into TP, AgrP and NegP, and the DP-Hypothesis.
It is perhaps appropriate here to point out that the SC Analysis per se is not on trial in C&R's article. Rather, C&R argue against implementing the SC Analysis in the particular domain of resultatives. Although in section 7 C&R cursorily review the possibilities of extending their ternary-branching analysis to all constructions which have been analyzed in terms of SC complementation (including the paradigm case of consider constructions), they do not push their approach to its logical limit.11 Striving to abolish SCs across the board would in fact seem fairly counterproductive, for SCs, as projections of a head, are trivially allowed by X'-theory. A more fruitful tack to take is therefore to try and exploit their virtues to the full.


We would like to extend our gratitude to Teun Hoekstra and Jan-Wouter Zwart for comments on previous versions of this reply.

1Analyses in which SC's are functional projections (Hornstein & Lightfoot 1987, Den Dikken 1987, Raposo & Uriagereka 1990, Chomsky 1993, among others) could still derive the data in (3) from the assumption that there is head-head agreement (cf. also Chomsky 1986) between the functional head of the SC and the lexical head of the SC predicate, through which the semantic features of the result predicate are visible on the functional head of SC (cf. also Grimshaw’s 1991 “extended projection”).

2There are other ways of countering it as well. One would be to contest the claim that all lexical selection involves T-marking (C&R 1992:188: “selection can only operate over arguments”). The phenomenon of lexically selected adverbs, as in John worded the letter *(carefully), may well show that C&R's view is overly restrictive; for recent discussion of lexically selected adverbs, cf. Rizzi (1990:chapter 3), Frampton (1991).
Another option would be to show that a thematic relationship between the verb and the subject of its SC complement can be rendered compatible with restrictive views on T-role assignment after all. This tack might be worth taking in the context of the minimalist framework (Chomsky 1993). Given an analysis of SC’s of the type in (i) (an adjunction structure; cf. Manzini 1983), combined with a slight modification of Chomsky’s definition of complement domain (in which “containment” should replace “domination”), it is possible to construe the SC-subject as a member of the internal domain of the verb.

(i) [V' V [XP NP [XP ... X ... ]]]

It then becomes possible for the verb to assign a T-role to NP, given that “elements of the internal domain [of A] are typically internal arguments of A” (Chomsky 1993:12). We will not pursue this line of argument any further here.

3It has been pointed out to us that the English verb to slug appears to work almost the same way:

(i) a. John slugged his brother into the hospital
b. John slugged his brother
(ii) a. John slugged the cup to pieces
b. * John slugged the cup
(iii) a. John slugged the ball away
b. John slugged the ball

4As a way out of this problem, C&R might perhaps want to claim that slaan is an intransitive verb in its use in (10), as intransitive verbs do not impose selectional restrictions on the postverbal NP (cf. (8) above). The verb slaan would then be ambiguous with respect to its transitivity, much like the case of English water discussed by C&R (1992:186-187). Parallel to water, Dutch slaan can also occur in resultative constructions in which it is clear that there is no selectional relationship between the verb and the postverbal NP; cf. (i). However, while water passes the intransitivity test of being able to occur without an object (cf. iia), slaan does not, (iib) being unacceptable (excluding contexts which generally render intransitive usage of otherwise obligatorily transitive verbs possible). An account of (9)-(11) along the same lines as C&R's approach to water thus seems unavailable.

(i) a. The gardener watered the tulips flat / his sneakers soggy
b. Jan sloeg zijn broer het ziekenhuis in / zijn vingers bont
Jan hit his brother the hospital into / his fingers black
en blauw
and blue
(ii) a. The gardener watered (for hours)
b. * Jan sloeg (urenlang)
Jan hit (for hours)

5This gains interesting support from the fact that the Dutch counterpart of frighten is syntactically periphrastic: bang maken (‘make frightened’).

6The verb entertains a close relationship with the empty SC predicate, which is presumably incorporated into the verb. This would be a normal instance of incorporation into V (Baker 1988) if the SC, the complement of V, is a projection of the resultative predicate, as we have assumed. The morphologically complex nature of frighten, like sadden, might indicate that such an analysis is on the right track; cf. also the previous note.

7C&R only present evidence that is consistent with the ternary-branching analysis but does not favor it over binary branching; they apparently believe that it does, though, since they write that the facts that they present lead to the conclusion that “the result phrase must therefore be an argument of the verb. This requires that it be a sister of the verb for both transitive and intransitive resultatives, as claimed by the Ternary Analysis, but not by the Binary SC or the Hybrid SC Analyses” (1992:184).

8C&R (1992:203) note that resultative nominals “sound better either with an indefinite object or with a context that induces a ‘habitual’ reading, or both”. All grammatical examples involving transitive resultatives (their 74a) meet this requirement . The ungrammatical cases of nominalization based on intransitive resultatives (their 74b) conspicuously do not. The contrast between transitive and intransitive nominals may hence be seriously affected by the aspectual difference between the two example sets. Once C&R's aspectual requirement is complied with, intransitive nominals are practically acceptable:

(i) The talking of ignorant people into voting for the right
candidate is an essential task of campaign-leaders

9Strikingly, to our ears the transitive resultative construction in (19a) actually sounds slightly worse than the intransitive one.

10Since C&R assume that in intransitive resultative constructions the verb takes the postverbal NP as its complement without assigning a T-role to it, they are committed to the view that complements of a head need not be arguments of that head.

11Conversely, SC champions do not always carry over the SC account of “classical” cases involving verbs like consider to resultatives. Thus, Stowell (1983) suggests that consider constructions should be distinguished from what he calls “quasi-causatives” (i.e. resultatives), though he does not present any empirical motivation for this view.

Baker, Mark. 1988. Incorporation. A Theory of Grammatical Function Changing. Chicago: Chicago University Press.
Burton, Strang, and Jane Grimshaw. 1992. Coordination and VP-Internal Subjects. Linguistic Inquiry 23:305-313.
Carrier, Jill, and Janet Randall. 1992. The Argument Structure and Syntactic Structure of Resultatives. Linguistic Inquiry 23:173-234.
Chomsky, Noam. 1986. Barriers. Cambridge, Mass.: MIT Press.
Chomsky, Noam. 1991. Some Notes on Economy of Derivation and Representation. In Principles and Parameters in Comparative Grammar, ed. Robert Freidin, 417-454. Cambridge, Mass.: MIT Press.
Chomsky, Noam. 1993. A Minimalist Program for Linguistic Theory. In The View from Building 20, eds. Kenneth Hale and Samuel Jay Keyser, 1-52. Cambridge, Mass.: MIT Press.
Dikken, Marcel den. 1987. Secundaire predicatie en de analyse van small clauses. GLOT 10:1-28.
Dikken, Marcel den. 1992. Particles. Doctoral dissertation, University of Leiden.
Frampton, John. 1991. Relativized Minimality: A Review. The Linguistic Review 8:1-46.
Fukui, Naoki, and Margaret Speas. 1986. Specifiers and Projections. MIT Working Papers in Linguistics 8, 128-172. Department of Linguistics and Philosophy, MIT, Cambridge, Mass.
Green, Georgia. 1972. Some Observations on the Syntax and Semantics of Instrumental Verbs. In Papers from the 8th Regional Meeting. Chicago Linguistic Society, University of Chicago, Illinois.
Grimshaw, Jane. 1991. Extended Projection. Ms., Brandeis University.
Hoekstra, Teun. 1988. Small Clause Results. Lingua 74:101-139.
Hoekstra, Teun. 1991. Small Clauses Everywhere. Ms., University of Leiden.
Hornstein, Norbert, and David Lightfoot. 1987. Predication and PRO. Language 63:23-52.
Kayne, Richard. 1984. Connectedness and Binary Branching. Dordrecht: Foris.
Koopman, Hilda, and Dominique Sportiche. 1991. The Position of Subjects. Lingua 85:211-258.
Kuroda, S.-Y. 1988. Whether We Agree or Not. Linguisticae Investigationes 12:1-47.
Manzini, Rita. 1983. Restructuring and Reanalysis. Doctoral dissertation, MIT, Cambridge, Mass.
McNally, Louise. 1992. VP Coordination and the VP-Internal Subject Hypothesis. Linguistic Inquiry 23:336-341.
Raposo, Eduardo, and Juan Uriagereka. 1990. Long-Distance Case Assignment. Linguistic Inquiry 21:505-537.
Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press.
Simpson, Jane. 1983. Resultatives. In Papers in Lexical-Functional Grammar, ed. L. Levin, M. Rappaport, and A. Zaenen, 143-157. Bloomington: Indiana University Linguistics Club.
Sportiche, Dominique. 1988. A Theory of Floating Quantifiers and Its Corollaries for Constituent Structure. Linguistic Inquiry 19:33-60.
Stowell, Tim. 1981. Origins of Phrase Structure. Doctoral disertation, MIT, Cambridge, Mass.
Stowell, Tim. 1983. Subjects across Categories. The Linguistic Review 2:285-312.
Sybesma, Rint. 1992. Causatives and Accomplishments: The Case of Chinese Ba. Doctoral dissertation, University of Leiden.

(Den Dikken) (Hoekstra)
Vakgroep Taalkunde Afdeling Dialectologie
Vrije Universiteit P.J. Meertens Instituut
De Boelelaan 1105 Koninklijke Nederlandse Akademie van
1081 HV Amsterdam Wetenschappen
Nederland Keizersgracht 569-571
1017 DR Amsterdam Nederland