object positions

60
KYLE JOHNSON OBJECT POSITIONS* Nominals precede other verbal complements in English. This paper argues against accounts of this fact based on an adjacency requirement on Case assignment, and proposes instead that (I) main verbs in English move out of the VP they head, and (2) nominal complements are forced by the Case Filter to "Object Shift" either to Specifier of VP or to the Specifier of a functional projection that contains VP. 1. INTRODUCTION With the abandonment of accounts for the linear order of complements that hold the idiosyncracies of the selecting heads responsible, many linguists have sought to implicate a syntactic force. (Stowell (1981) and Chomsky (1981, sec. 2.3) are representative.) One effect of this force in English is to position nominal complements to verbs before those of other categorial status, yielding paradigms like those in (1). (1)a. Gary introduced Mittie to Sam. b. *Gary introduced to Sam Mittie. c. Gary told Sam to leave. d. *Gary told to leave Sam. There are two aspects to the effect. First, the force involved singles out nominals: its signature is the ability to distinguish NPs from all other phrases. Second, the effect involves reserving the initial position among complements for nominals. This paper attempts to provide the underlying reasons for the second of these aspects to the problem - to give an account for the "NP First" property of complements. One influential account of the paradigm in (1) relies on the CASEFILTER, which does the work of distinguishing NPs from other phrases (cf. Chom- sky 1981, p. 49). * This paper has benefitted from exposure to audiences at UT-Austin, UC-Irvine, University of Massachusetts-Amherst, Universit6 de Gen~ve, Cornell University and Brandeis Univer- sity, and from the participants of seminars at UCLA and the University of Wisconsin- Madison. Thanks are due also to Marcel den Dikken, Mtirvet Eng, Teun Hoekstra, Osvaldo Jaeggli, Hilda Koopman, Mark Lencho, Anoop Mahajan, Elaine McNulty, Sigri~Sur Sigur- j6nsd6ttir, Dominique Sportiche, Sten Vikner and Dag Wold. Natural Language and Linguistic Theory 9: 577-636, 1991. © 1991 Kluwer Academic Publishers. Printed in the Netherlands.

Upload: kyle-johnson

Post on 06-Jul-2016

213 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Object positions

KYLE JOHNSON

O B J E C T P O S I T I O N S *

Nominals precede other verbal complements in English. This paper argues against accounts of this fact based on an adjacency requirement on Case assignment, and proposes instead that (I) main verbs in English move out of the VP they head, and (2) nominal complements are forced by the Case Filter to "Object Shift" either to Specifier of VP or to the Specifier of a functional projection that contains VP.

1. INTRODUCTION

With the abandonment of accounts for the linear order of complements that hold the idiosyncracies of the selecting heads responsible, many linguists have sought to implicate a syntactic force. (Stowell (1981) and Chomsky (1981, sec. 2.3) are representative.) One effect of this force in English is to position nominal complements to verbs before those of other categorial status, yielding paradigms like those in (1).

(1)a. Gary introduced Mittie to Sam.

b. *Gary introduced to Sam Mittie.

c. Gary told Sam to leave.

d. *Gary told to leave Sam.

There are two aspects to the effect. First, the force involved singles out nominals: its signature is the ability to distinguish NPs from all other phrases. Second, the effect involves reserving the initial position among complements for nominals. This paper attempts to provide the underlying reasons for the second of these aspects to the problem - to give an account for the "NP First" property of complements.

One influential account of the paradigm in (1) relies on the CASE FILTER, which does the work of distinguishing NPs from other phrases (cf. Chom- sky 1981, p. 49).

* This paper has benefitted from exposure to audiences at UT-Austin, UC-Irvine, University of Massachusetts-Amherst, Universit6 de Gen~ve, Cornell University and Brandeis Univer- sity, and from the participants of seminars at UCLA and the University of Wisconsin- Madison. Thanks are due also to Marcel den Dikken, Mtirvet Eng, Teun Hoekstra, Osvaldo Jaeggli, Hilda Koopman, Mark Lencho, Anoop Mahajan, Elaine McNulty, Sigri~Sur Sigur- j6nsd6ttir, Dominique Sportiche, Sten Vikner and Dag Wold.

Natural Language and Linguistic Theory 9: 577-636, 1991. © 1991 Kluwer Academic Publishers. Printed in the Netherlands.

Page 2: Object positions

578 KYLE JOHNSON

(2) Case Filter *NP, if NP has phonetic content and has no Case.

That NPs must precede other complements can now be restated as the fact that Case-marked positions precede others, and the force involved can be sought in the conditions under which Case is assigned. Pursuing this line of thinking, Stowell (1981) and Chomsky (1981, p. 94ff) exploit the fact that the first position among the verbal complements in English is also the one closest to the Case-assigning verb and suggest, therefore, that Case is assigned to adjacent positions. This "adjacency condition" on Case assignment in conjunction with the Case Filter guarantees that NPs will precede other complements in English.

There is a caveat to this scheme made necessary by the existence of constructions where a verb appears to have two nominal complements. In certain of these cases, the two nominals form a "double object" construc- tion, whose peculiarities we return to in Section 6. In other cases, a verb has two nominal complements, typically one in the Accusative and the other in the Dative or Genitive Case. In these constructions, found in various Germanic languages other than English, the accusative Case- marked NP usually precedes the other.l Thus, these constructions encour- age viewing the adjacency condition as one on certain Cases but not others. It is conventional to restrict the adjacency condition to structural Cases, exploiting a distinction between structural and lexical Cases (cf. Babby (1980), Chomsky (1981, p. 170) and Zaenen, Maling and Thrfiins- son (1985, pp. 464-467)). In the Germanic languages (putting aside certain problematic constructions in Icelandic), only nominative and accusative are structural.

Even with the caveat, however, I believe this account for the NP-first arrangement of complements is wrong. Although I will not challenge the claim that conditions on Case assignment are involved, there are many problems (chronicled in the following section) for the hypothesis that adjacency is that condition. Instead, I will suggest that NPs come before

1 There are systematic exceptions to this generalization in Icelandic. For example, certain classes of verbs have nominal complements in the dative or genitive case which bear all the properties that other Germanic languages normally reserve for nominals in the accusative case. Thus, these NPs can undergo A-movement to "'subject" position (as in the passive and raising constructions), they must be adjacent to the selecting verb (modulo the effects of Verb Movement) and they are able to undergo "Object Shift" (see Section 5). These facts indicate that there is no direct relationship between "'structural Case," as it is used here (see below), and morphological case. See Zaenen, Maling and Thrfiinsson (1985) for extensive discussion. Rather, for the purposes of this paper, we may understand "structural Case" to have a closer affinity to the OBJECT relation defended in Zaenen, Maling and Thrfiinsson.

Page 3: Object positions

OBJECT POSITIONS 579

o ther complements because they move to a posit ion preceding V ' ; Section

3 sketches how this hypothesis resolves the difficulties for the adjacency

theory. In this respect , my suggestion bears a resemblance to a recent

proposa l made for Du tch by K o o p m a n (1989), and more distantly to

Postal (1974).

The remainder of the paper is concerned with finding the cause for the

obiect 's movemen t . A n examinat ion of particle construct ions in Section 4

leads me to equate the object 's m o v e m e n t in English with a type of short

scrambling found most plainly in Scandinavian languages. Section 5 houses

my proposa l for the condit ion on Case assignment that guides this short

scrambling, and therefore gives rise to the NP-first effect. Finally, Section

six extends the system to double object constructions.

2. PROBLEMS FOR ADJACENCY

If there is an adjacency condit ion on the assignment of structural Case,

then its effects should be evident for bo th nominat ive and accusative

Case. But it is difficult to detect any such condit ion on the assignment of

nominat ive Case. If C o m p assigns nominat ive Case, then the grammatical-

ity of (3) is unexpected; if Infl assigns it, then the examples in (4) should be ungrammatical . 2

(3)a, I knew that p robably Gary had left.

b. I wonder whether in fact G a r y will leave.

2 It is sometimes suggested that the ungrammaticality of examples like (i) is due to an adjacency condition on Nominative Case assignment.

(i)a. *Will probably Gary leave? b. *Has in fact Sam left?

But I believe that the evidence implicates a constraint on the verb's movement to Comp in these examples, rather than a condition on Case assignment. To begin with, it is difficult to find any other difference between these examples and those in (3) that could be responsible for the contrast in grammaticality; certainly there appears to be no difference with respect to Case assignment. Further, Lasnik (forthcoming) (see also Shlonsky (1987)) shows that the NP in (ii) is dependent on be for Case,

(ii)a. There was a woman in the garden.

b. There will be a unicorn at the party.

And yet in these constructions as well, an adverb cannot immediately follow an auxiliary in Comp, cf. (iii).

(iii)a. *Was in fact there a woman in the garden?

b. *Will probably there be a unicorn at the party?

Page 4: Object positions

5 8 0 K Y L E J O H N S O N

(4)a.

b.

Gary probably has left.

Gary in fact will leave.

In both cases, an adverb interrupts adjacency between the nominative Case-marked NP and the supposed Case marker. Restricting the adjacency condition to accusative case would reduce the hypothesis to vacuity; to preserve the adjacency condition therefore requires maintaining that ad- jacency is preserved in (3)/(4) and giving an account for its invisibility. Such an account has not been found,, and therefore this stands as our first problem for the adjacency condition.

A second problem concerns the paradigm in (5)/(6).

(5)a.

b.

C.

*Mikey visited quietly his parents.

*Betsy sung loudly the anthem.

*Chris hit quickly the dog.

(6)a.

b.

C.

d.

e .

Chris walked quickly down the street.

Mikey talked slowly to Gary.

Betsy spoke loudly with everyone.

Sam said suddenly that we must all leave.

Gary tried diligently to leave.

Here the NP/non-NP asymmetry found in the order-of-complements prob- lem is repeated: only NPs do not tolerate non-complements to intervene between them and the verb. On Stowell's and Chomsky's account, the parallelism is captured with the adjacency condition on Case assignment. But the grammaticality of (6) raises an independent problem that threatens this account of the paradigm. The problem involves the PROJECTION PRIN- CIPLE, which requires that only complements be sisters to the verb (cf. Chomsky (1981, p. 38) and Jackendoff (1977, pp. 55-72)). On standard conceptions of phrase markers, this entails that complements must precede non-complements. Thus, the ungrammaticality of (5) follows from the PROJECTION PRINCIPLE and the grammaticality of (6) calls for an explana- tion.

One conceivable account of (6) involves invoking a movement rule that relocates the complements to the right. This would give to (6a), for example, a representation like that in (7).

Page 5: Object positions

O B J E C T P O S I T I O N S 581

(7) IP

NP r

Chris I VP

i Z, V' down the street

V' AdvP

V ppi quickly

I walked t

But there are numerous reasons for believing that this is not the correct analysis of (6). First, it leaves mysterious why this option cannot be exploited for the examples in (5), and does not relate the NP/non-NP asymmetry found in the order-of-complements problem with that found in (5)/(6).

Further, it predicts that there should be no difference between (6) and (8).

(8)a. ?*Mittie went with it quickly to her.

b. ?*Gary said to leave to her.

c. ?*Sam talked about it calmly to her.

These examples improve if the rightmost PP has the "heaviness" charac- teristic of rightward movement in these contexts (cf. Ross (1967), Roche- mont (1978) and Culicover and Rochemont (1990)), as in (9).

(9)a. Mittie went with it quickly to the appropriate office in the administration building.

b. Gary said to leave to every one of the presidential candidates.

Page 6: Object positions

582 I~YLE JOHNSON

(9)c. Sam talked about it calmly to the woman with the painted blouse.

If the contrast between (8) and (9) indicates that PPs with pronominal complements cannot move rightwards, then we should conclude that no such movement is involved in (6).

Additionally, a rightward movement account cannot extend to cases like those in (10).

(10)a. Sam spoke loudly to everyone in falsetto.

b. Betsy walked quickly down the street naked.

The standard hypothesis (see Jackendoff (1977, sec. 4.3), for example) is that in falsetto is adjoined to V' in (10a). It is also commonly assumed that adjunction of a maximal projection must be to a maximal projection (cf. Baltin (1982, p. 2)). Hence, to everyone cannot have moved rightward in (10a). Similarly, in (10b) down the street cannot have adjoined right- ward, as naked is most likely adjoined to VP (cf. McNulty (1988, Chapter 2)).

Pesetsky (1989) discusses examples like those in (6) and also argues against a rightward movement account; I reproduce two of his arguments here. Citing Avery Andrews, Pesetsky assumes that adverbs modify their sisters and notes that sentences like (6) but with two adverbs (as in (11)) have the meaning corresponding to (12) and not (13).

(11) John knocked intentionally twice on the door.

(12) . . . [v, intentionally [v, twice [v' on the door]]]

(13) . . . [[[v, ti] intentionally]v, twice]v, [on the door]/]v,

Thus intentionally modifies the constituent containing twice, not the other way round, and the hierarchical arrangement is therefore as in (12). Second, Pesetsky argues on the basis of extraction facts that the comple- ments in (6) have not been rightward moved. Because extraction from a moved phrase is blocked in English (see the "Raising Principle" of Wexler and Culicover (1981, Chapter 5)), we should expect movement from the complements in (6) to yield ungrammaticality. But as Pesetsky notes, extraction is permitted in these contexts, as shown by (14).

(14)a.

b.

C.

d.

Which girli did Betsy talk quickly about ti ?

Whoi must you speak loudly to ti ?

This is whati Bill has looked recently at t~.

Which departmenti has Sam talked repeatedly to t,-?

Page 7: Object positions

O B J E C T P O S I T I O N S 583

(14)e. Which childi did you look diligently for t/?

f. Who, did you bake it quietly for 5?

g. Mary's the one who~ Bill relied stupidly on tg.

For these reasons, then, ! abandon an account for (6) based on right- ward movement. Some other means of circumventing the violation of the PROJECTION PRINCIPLE that these examples invoke must be found. And this solution will replace the account of the contrasts in (6)/(5) based on the Adjacency condition. Further, this solution must make the same distinction between NPs and other phrases that is found in the order-of- complements problem: it bears the signature of the force responsible for bringing NPs before other complements. Hence, unless we are to abandon explaining this similarity of the two paradigms, we should look for a different solution to the order-of-complements problem as well.

The third problem for the adjacency requirement that I catalogue here concerns its application in contexts where movement to an A ' position has applied. In these situations the Case Filter applies so as to guarantee that the trace left by A ' -Movement is in a Case-marked position. Thus the Case Filter needs to be rendered in a somewhat more sophisticated form; we may take it to be as in (15). 3

(15) If o- = (o~1,... , an) is an A-Chain of a phonetically overt NP, then al is the head of cr and is in the unique Case-marked position of o-.

(15) incorporates other necessary properties of A-Chains; see Chomsky

3 This version of the Case Filter distinguishes overt NPs from others, along the lines originally proposed by Vergnaud. But there are several difficulties for this approach; for example, it does not extend to the movement of null operators nor to predicate nominals (cf. Chomsky (1981, p. 223 note 18; 1986b, p. 95,208 note 39)). In response to these problems, Chomsky (1981, p. 333ff; 1986b, pp. 94-95) suggests deriving the Case Filter from the Theta Criterion with a condition that would make the theta-role some A-chain bears "visible" only if that A-chain contains a Case-marked position. Among other consequences, this "Visibility Hypothesis" would make the Case Filter unresponsive to the phonetically overt/silent distinc- tion - a harmless change, from the standpoint of the proposals made here. In its purest form, however, the Visibility Hypothesis causes the Case Filter to be blind to the NP/non- NP distinction, and this threatens its use in explaining the NP-first order of complements (see Stowell (t981) for a method of combining these ideas). One conclusion we shall reach is .that, in fact, the Case Filter is not what is responsible for singling out NPs to be comple- ment-first. Nonetheless it is unclear to me that my use of the Case Filter is compatible with the Visibility Hypothesis. One outcome of Section 5 is that only NPs are in structurally Case- marked positions. Thus, either the strategy of meeting the Visibility requirement in Stowell can be extended to the framework here, or the Visibility Hypothesis in its simplest form is wrong (cf. Epstein (1990)).

Page 8: Object positions

584 KYLE JOHNSON

(1981, p. 175ff, 333-339; 1986b, p. 136ff) for discussion and Section 4 for an application.

Interestingly, there is evidence that the Adjacency requirement on Case assignment is relaxed when the trace of A ' -Movement is involved. The contrasts in (16), discussed by Epstein (1991, Chapter 4), show that an overt NP cannot be separated from the Case assigning verb by an adverb, but that the trace of A ' -Movement can be.

(16)a. ??I've believed for a long time now Gary to be a fool.

b. *Mary believes sincerely Gary to be a fool. c. Who have you believed for a long time now to be a fool?

d. Who does Mary believe sincerely to be a fool?

If Case is assigned under adjacency, then how can the differences in (16) be accommodated?

3. A SOLUTION

I will explore an account for the NP First property that does not exploit an adjacency condition on Case assignment. The solution rests on the

three proposals in (17).

(17)a. Specifiers of XP precede X'.

b. Verbs always move out of the VP they head.

c. Accusative Case-marked NPs move to Specifier of VP.

I will concentrate on the last two of these. I know of no way to derive (17a); but note that something of its nature is probably needed indepen- dently to account for the fact that with overwhelming frequency the Speci- fier of IP (=surface subject) comes to the left of I' in the world's languages

(see Greenberg (1963)). That verbs move out of the VP they head is suggested by certain

coordination facts. 4 Larson (1988, p. 345, note 11) points out that because coordination normally involves constituents, examples such as those in (18) suggest that the bracketed strings form phrases.

(18)a. Gary put [the book on the table] and [the lamp on the sofa].

b. Chris ate [the meat slowly] but [the vegetables quickly].

4 See also the discussion of Stylistic Inversion in Culicover and Rochemont (1990, Chapter 3), where it is argued that in examples like Into the r o o m nude wa lked John the constituent into the r o o m nude is a VP out of which walked has moved.

Page 9: Object positions

OBJECT POSITIONS 585

c. Sam talked [to Mittie yesterday] and [to Betsy the day before].

The bracketed strings in examples such as these are identical to VPs without a verb. Following Larson, therefore, ! will analyze these construc- tions as VPs that have undergone across-the-board Verb Raising, giving to (18b), for example, the representation in (19).

(19) Chris ate/ [vP [vP ti the meat slowly] and [vP ti the vegetables quickly]].

Additional support for (17b) is found in section 4. One question concerning (17b) for which I will not provide an answer

is where the verb moves to and why it moves there. Although I believe plausible answers to these questions can be found, I will not address this aspect to the problem directly and assume that there is a functional head, /x, to which the verb must adjoin (see Jaeggli and Hyams (1989) and Pesetsky (1989) for some discussion).

The assumptions in (17) derive the "NP First" property: if NPs must move to the Specifier of VP and if this position precedes V', then NPs will necessarily precede all other complements. Moreover, (17) does not suffer from the problems listed for the adjacency condition in the previous section.

Consider first the contrast between (5) and (6), which illustrates the generalization that adverbs may intervene between a verb and its comple- ment only if the complement is not an NP. If verbs move out of their VP, then examples like those in (6) may be given the parse in (20) (=(6b)).

(20)

# vP

v~ ~ v'

talked slowly V'

Vi PP

I / ' x t to Gary

Assume that adverbs, like adjectives, may precede as well as follow the X' they are adjoined to, and (20) emerges as an immediate consequence

Page 10: Object positions

5 8 6 K Y L E J O H N S O N

of Verb Raising. Note that the Projection Principle is preserved in (20): the complement to Gary is a sister to its Theta-marker, the trace of talked.

A representation like that in (20) is not possible however when the complement is a (structural Case-marked) NP. In this situation, the NP must move to the Specifier of VP, giving to examples like those in (5) a representation like that in (21) (=(5c)).

#

hit

• . . ] ,L'

VP

NPi V'

the dog quickly V'

Vi NPi

I I t t

(21)

The examples in (5), then, violate the requirement that NPs move to Specifier of VP.

The problematic difference between traces of A'-Movement and overt NPs with respect to Case assignment also does not arise under (17). Thus, examples (16a) and (16b) are ungrammatical for the same reason that the examples in (5) are: the structural Case-marked NP fails to move to Specifier of VP, giving to (16b) the illicit representation in (22).

(22)

i i I t

believe

• . • I t '

VP

V'

sincerely V'

i i IP

t NP r

Gary to be a fool

Page 11: Object positions

O B J E C T P O S I T I O N S 587

But examples (16c) and (16d), where Wh-Movement has occurred, may receive a representation like that in (23).

(23) . . - # '

# VP

vk g v'

believe t' sincerely V'

vk Ip

t NP i r

t to be a fool

Here the structural Case-marked NP has moved to Specifier of VP, in compliance with (17c), before moving to the A' position it is found in at S-structure. Hence, the contrasts in (16) are a straightforward consequence of the proposal in (17).

There is a difficulty for this account of (16), however. If the derivation represented by (23) is possible, we should expect the subject of an infinitive following a believe-type verb to be able to move past the adverb and remain there. But this is not always possible, as the ungrammaticality of the following examples illustrates.

(24)a. *Gary believes Mikey sincerely to be intelligent.

b. *Sam considers Mittie strongly to have been misled.

In other contexts, however, sentences with this structure do arise, as the examples in (25) from Kayne (1984b, p. 114) and Postal (1974, p. 146) indicate.

(25)a.

b.

C.

I've believed Gary for a long time now to be a fool.

Jane proved Bob, unfortunately, to be a werewolf.

I have found Bob recently to be morose.

Page 12: Object positions

588 KYLE J O H N S O N

I must assume that there is some S-structure constraint that prevents NPs from remaining in Specifier of VP when they have moved past certain adverbs. I will not be able to take up this problem here. s

Consider finally the contrast between nominative and accusative Case with respect to adjacency between Case assigner and Case bearer. For the sake of concreteness, let us suppose that nominative Case is assigned by finite Comp - the following scenario can be straightforwardly transposed to a system under which finite Infl assigns nominative Case. The contrast at issue, then, is as in (26).

(26)a. Mary said that probably Gary had left.

b. *Gary told probably the story.

The requirement that accusative Case-marked NPs move to Specifier of VP already accounts for why (26b) does not receive a parse where proba- bly is adjoined to V'. But what prevents probably from being adjoined to VP, as in (27), but allows it to be adjoined to IP, as in (28)?

(27) . . . / . t '

VP

probably VP

NP i V'

the story Vk NP i

I h t t

#

v~ ~t

I told

5 The contrast between (16b,d) and (24) is strongly reminiscent of the account given to the French and Italian cases of exceptional Case-marked infinitives in Kayne (1984a, Chapter 5). Further, as an N L L T reviewer reminds, Postal (1974, p. 290 note 2) points out that (24) contrasts with passive sentences such as Mikey was believed sincerely to be intelligent. This is also true of a certain class of epistemic verbs in French and Italian (cf. Rizzi (1981, pp. 148-152)). Thus, insofar as the Romance facts parallel the English phenomena involving adverbs, a common account should be found. The situation is made more complex by the existence in English of verbs that pattern like the class of verbs discussed in Rizzi even when adverbs are absent (cf. Postal (1974, pp. 297-318)).

Page 13: Object positions

O B J E C T P O S I T I O N S 589

(28) . . . c e

C IP

that probably IP

NP I'

Gary had left

On the account that makes use of Case adjacency, (27) is blocked by the requirement that the story be adjacent to told. But this leaves mysterious why (28) is not similarly blocked.

One possibility is that the contrast between (27) and (28) is due to independent constraints on adjunction in English. Note that while there are numerous examples suggesting that it is possible to adjoin a phrase to the left of IP at S-structure (those in (29) are representative, see Baltin (1982, pp. 16-22)), there are no such examples with respect to leftward adjunction to VP, as in (30).

(29)a. She said that on the floor Betsy had put it.

b. ?Sam remembered that only his problems they'd talked about.

(30)a. ?*She said that Betsy had on the floor put it.

b. *Sam remembered that they'd only his problems talked about.

I will assume, therefore, that the contrast between (27) and (28) has nothing to do with Case assignment, but follows entirely from the prin- ciples governing adjunction structures.

The "NP First" property of the order-of-complements problem follows from (17), then, without invoking the problems that an Adjacency con- dition account does. 6 In the next section, we examine a special case of

6 As Gert Webelhuth points out, there is one case where an adjacency condition's effects are not captured by (17); this is the ungrammaticality of examples like (i).

(i)a. *I desire for tomorrow Betsy to arrive.

b. *It's possible for on Monday Liz to leave.

I have nothing but a speculative reply to this problem. Imagine that the representation for infinitives like those in (i) is as in (ii).

Page 14: Object positions

590 KYLE JOHNSON

the "NP Firs t" proper ty , which r ecommends the approach to the p rob lem

we have sketched here and sheds light on the cause of (17c).

4. THE PARTICLE CONSTRUCTION

The "NP Firs t" p roper ty admits of apparen t except ions in the PARTICLE

CONSTRUCTION, one of the few places in English syntax where it appears

that individual words do not map onto single syntactic pos i t ions . : Thus,

the verbs look up, throw out, and dust o f f are b roken up by their object

in the following examples.

(31)a. Mikey looked the reference up.

b. Betsy threw the bicycle out.

c. Chris dusted the counte r off.

There are three clear senses in which these particle verbs, as I shall

refer to them, are single lexical items. First, they are able to undergo

morphologica l processes thought to apply only to verbs in the lexicon, s

Hence , part icle verbs can be used to form no uns with -ing, adjectives with

-ed and, as G u 6 r o n (1987) points out (see also Bol inger (1971, p. 180)),

middles. 9

(32)a. Mikey 's looking up of the reference is a t rying affair.

(ii) . . . [eP for [,p NP [,, to VPll]

(See Emonds (1985, Chapter 7) for arguments in favor of (ii). The account offered here, however, crucially differs from Emonds in that parsing clauses as projections of prepositions must be confined to these cases. Therefore, in so far as Emonds' arguments lead to assigning this representation more liberally, his conclusions must be seen as an argument against the proposals of this paper.) If (ii) is correct, then the grammaticality of (i) will follow from whatever constraint prevents leftward adjunction to the complement of a preposition. 7 See Bolinger (1971) and Fraser (1976) for attempts to find criteria that distinguish particle constructions from the rest. s I follow Baker (1985, pp. 8-9) in treating nominalizations in ing as formed pre-D-structure, and Wasow (1977) in giving a similar treatment to the adjectival ed. (The arguments in Levin and Rappaport (1986) that adjectival ed is derived syntactically from the passive participle are weakened by the existence of forms based on unaccusatives: the fallen toy, the wilted leaf, etc. I don't find the argument in Levin and Rappaport (ibid note 36) that these forms are based on past participles convincing.) The middle is most likely lexically formed in English for the reasons given in Roberts (1987, pp. 190-195) who attributes them to Jaeggli; but see Roberts (ibid pp. 207-227) for an alternative). 9 Note that we would not wish to resort to claiming that these examples are produced by some form of syntactic particle construction formation applying to adjectives or nouns derived from the root verbs because this would leave accidental that these constructions arise only with the appropriate deverbal nouns or adjectives.

Page 15: Object positions

OBJECT POSITIONS 591

(32)b.

c.

(33)a.

b.

c.

(34)a.

b.

(35)a.

b.

The calling out of his name is heart-wrenching.

Their pointing out that we should leave was timely.

The table remained dusted off.

Their relationship seemed broken up.

The dance seemed called off.

the dusted off table

a looked up number

Bridges blow up easily.

His car breaks down easily.

This distinguishes particle verbs from idioms, compounds and other phra-

sal words. Second, as Gu6ron (1987, p. 9) and le Roux (1988, p. 31) note (see

also Fraser (1976, p. 8)), the selectional requirements of particle verbs are not derived from the selectional requirements of their parts, as the

examples in (36) indicate.

(36)a. We can't make out whether he is lying or not.

b. Fill in whether you are married or not.

Third, as Fraser (1976, p. 3) notes, Gapping treats the verb + particle as a single verb; thus the particle gaps with the (rest of the) verb and cannot be stranded, as the contrasts in (37) illustrate.

(37)a. Gary looked up Sam's number, and Mittie, my number.

b. *Gary looked up Sam's number, and Mittie, up my number.

c. *Gary looked Sam's number up, and Mittie, my number up.

If Stillings (1975) is correct in concluding that only verbs, either singly or in series, can be Gapped, then, as Selkirk (1982, p. 28) argues, these facts support treating the particle as part of the verb.l°

An indication that particle verbs are single terms at S-structure comes

10 Note that these facts cannot be accomodated by accounts of the particle construction that are based on an Incorporation of the particle into the verb (see note 12). For although such an account will explain the contrast between (37a,b) if Incorporation is obligatory (see Hankamer (1973, p. 35) for evidence that Incorporation feeds Gapping), it still leaves (37c) unexplained.

Page 16: Object positions

5 9 2 KYLE J O H N S O N

from the s e r v e s .

(38)a.

grammaticality of examples like (38), as Anoop Mahajan ob-

Betsy looked up the address quickly and the phone number slowly.

b. Liz turned out the porch light today and the living room light yesterday.

As in (18), these examples involve coordinating VPs that are missing their heads• If the account for (18) is extended to (38), then look up and turn

out are, like other verbs, able to move as singletons to/x. There is a correlate to this fact found with the "separable prefix" verbs

in standard Dutch• These verbs are able to undergo "Verb Raising" as a unit, as in (39)• 11

(39) . . . omdat Carol hem tv kon opbellen.

• . . because Carol him could up-ring.

"Verb Raising" in standard Dutch is a process that moves certain verbs to the right of a higher verb (see Evers (1975))• In (39), then, the moved V ° must include the particle.

If by these criteria we have correctly concluded that particle verbs are single lexical items, then the constructions in (31) suggest that the theory that pairs words with syntactic positions needs to be enriched• One obvious site for this enrichment is lexical insertion. We might imagine that the procedure by which words are matched with syntactic positions, guided, as it is, by subcategorization and selectional information, could be rigged in such a way that bimorphemic words are able to project to different syntactic positions• This is essentially the tack taken by Kayne (1984b), who argues that the structure underlying (31a), for example, involves a "small clause" (see Stowell (1983)) headed by the particle, as in (40). 12

la Similar facts are reported for Afrikaans in le Roux (1988, sec. 2.2) who suggests that unlike Dutch the particle cannot normally be stranded under Verb Raising. 12 I do not consider here theories of the particle construction that involve a rule moving the particle rightwards, as in Fraser (1976). If the operation he proposes is an instance of Head Movement, then it does not conform to the conditions on such operations that I have presupposed. I will also not review the possibility that particles Incorporate into the verb, a process evident in Dutch (cf. van Riemsdijk (1982, p. 54), Emonds (1976, p. 82) and den Dikken (1990a)). These proposals do not permit an account of the characteristic paradigm that relates it to the NP/non-NP distinction found in the order-of-complements problem (see Section 5). For critique of the proposals in Simpson (1983) and Stowell (1981), see le Roux (1988, p. 54-66, 223-229).

Page 17: Object positions

O B J E C T P O S I T I O N S 5 9 3

(40) IP

NP I'

Mikey I VP

V t

V PP

looked NP P'

the reference up

Gu6ron (1987) suggests a similar account, differing mainly in the linear order that the particle and its complement take.

A different approach would preserve the one-to-one mapping of syntac- tic positions and lexical items that current theories of lexical insertion enforce, and seek a means of breaking up words after D-Structure. The system sketched in the previous section provides a straightforward method of giving an account of particle verbs along the second line. In this section, I will argue that such an account gives a better explanation of the phenom- ena attendant with particle constructions than does one based on (40).

4.1. The Characteristic Paradigm

Particle constructions enter into a "characteristic paradigm," illustrated by the examples below.

(41)a.

b.

C.

d.

e .

f.

Mikey looked the reference up.

Mikey looked up the reference.

Betsy threw the bicycle out.

Betsy threw out the bicycle.

Brent dusted the counter off.

Brent dusted off the counter.

Page 18: Object positions

594 K,ZLE JOHNSON

(42)a. Mikey looked it up.

b. *Mikcy looked up it.

c. Betsy threw it out.

b. *Betsy threw out it.

e. Brent dusted it off.

b. *Brent dusted off it.

The object of the particle verb may show up on either side of the particle, unless this object is a pronoun, in which case it must precede the particle. There are a variety of ways in which the restriction on pronouns may be weakened (see, for example, Bolinger (1971, pp. 39-41)). If the pronoun is focussed or stressed, as in (43a), if it is conjoined with another pronoun, as in (43b), or if the pronoun has deictic force, as when (43c) is uttered while pointing, the order "particle-pronoun" is permitted.

(43)a. Betsy threw out THEM!

b. Mikey looked up him and her.

c. Brent dusted off that.

Interestingly, the alternation in (41) is restricted to NPs. As Kayne emphasizes, prepositional (44) or clausal (45) complements of particle verbs must follow the particle.

(44)a.

b.

C.

d.

(45)a.

b.

C.

d.

Further, particle,

(46)a.

b.

C.

Mikey teamed up with the women.

*Mikey teamed with the women up.

Betsy narrowed in on the problem.

*Betsy narrowed on the problem in.

Mikey pointed out that Gary had left.

*Mikey pointed that Gary had left out.

Betsy tried out loudly singing the national anthem.

*Betsy tried loudly singing the national anthem out.

non-arguments of the particle verb may not come before the as in (46).

Mikey slips up all the time.

*Mikey slips all the time up.

Betsy figured out the problem carefully.

Page 19: Object positions

OBJECT POSITIONS 595

(46)d. *Betsy figured carefully out the problem. (compare: "Betsy figured the problem out carefully.")

e. Brent pushed down the pin with a hammer.

f. *Brent pushed with a hammer down the pin. (compare: "Brent pushed the pin down with a hammer.")

There is one situation where a non-argument of the particle verb may precede the particle: when the NP is the subject of an infinitival comple- ment, as in (47).

(47)a. Mikey made out George to be a liar.

b. Mikey made George out to be a liar.

Here, as in (41) and (42), the pre-particle position becomes obligatory if the NP is a pronoun.

(48)a. *Mikey made out him to be a liar.

b. Mikey made him out to be a liar

And, as before, this restriction can be overcome in a variety of ways, as (49), parallel to (43), illustrates.

(49)a. Mikey made out THEM to be liars!

b. Mikey made out her and him to be liars.

c. Mikey made out that to be false.

These observations can be summarized with the following description.

(50) Noun Phrases dependent on the particle verb for accusative Case may appear on either side of the particle, unless the Noun Phrase is a simple pronoun, in which case it must appear preceding the particle.

(50) makes explicit that the word order alternation of the characteristic paradigm is related to Case: it is just those terms that are Case-marked by the particle verb that undergo the alternation. Further, the same NP/non-NP distinction found in the order-of-complements problem is reproduced here: only NPs undergo the alternation. We should find an account that explains this parallelism between the two phenomena.

We begin by examining the accounts of Kayne (1984b) and Gu6ron (1987), which base-generate the verb and particle in different positions. After raising certain problems for these accounts, we turn to an account

Page 20: Object positions

596 KYLE JOHNSON

of the construction that makes use of the mechanisms introduced in the

previous section.

4.2. V + PP Accounts

Although Kayne and Gu6ron both separate verb and particle with lexical insertion, they place the argument that undergoes the characteristic para- digm in a different D-structure position. Kayne suggests that this argument is a D-structure subject of the particle phrase, as in (40) above, while Gudron argues that it lies in complement position at D-structure.

On Kayne's account, the characteristic paradigm is the result of an operation like "Heavy NP Shift" that adjoins the moved phrase to the right of V'. Because Heavy NP Shift may only affect phrases with focal stress, pronouns are typically prevented from being moved; in this way the contrasts in (42) are accounted for. Heavy NP Shift is made obligatory on Kayne's account when the particle phrase contains in its subject position a phrase of a category typically banned from subject position; for this reason only NPs may stand in the pre-particle position and the contrasts in (44)-(46) follow.

On Gu6ron's account, the characteristic paradigm stems from a topicali- zation-like operation that moves the affected phrase and adjoins it to the left of the particle phrase, as (51) illustrates.

NP

I Mikey

(51) IP

r

I VP

I V'

V PP

looked NPI PP

the reference right P'

P NPi

I I up t

Page 21: Object positions

O B J E C T P O S I T I O N S 597

This operation is obligatory when an unstressed pronoun is involved be- cause "A pronoun must be contiguous to a lexical case assigner" (p. 7), and Gu6ron argues that in particle constructions, the particle is unable to assign Case. Thus a pronoun must move to a position adjacent to the Case assigning verb. That the characteristic paradigm is restricted to NPs (i.e., the contrast in (44)-(46)) is not addressed in Gu6ron's paper.

An inadequacy of Gu6ron's account is the mechanism used to drive the characteristic paradigm. If topicalization is involved, then there would seem to be no account for the restriction to NPs; topicalization in other contexts is free to move phrases of any category. Further, it introduces problems to allow topicalization to move pronouns without special demon- strative force. But on Gu6ron's account, pronouns are always topicalized in particle constructions thus rendering the asymmetry below mysterious.

(52)a. Mikey looked it up.

b. *Up it, Mikey ran., or: *It, Mikey likes.

Somewhat less problematic is this respect are accounts of the character- istic paradigm which adopt Gu6ron's (1987) D-structure representation of particle constructions, but use A-movement to bring the object before the particle (cf. Gu6ron (1990), Gu6ron and Hoekstra (1990), den Dikken (1990b) and Koopman (1990)). On these accounts, the object can be moved to Specifier of the particle phrase under A-Movement. This avoids the problematic claim that weak pronouns must topicalize, but it still leaves unexplained why only NPs can be moved. A-movement does not appear to be restricted to NPs, as the grammaticality of examples like That Betsy left seems to bother everyone indicates.

Another difficulty for these accounts is that a distinguishing character- istic of the particle construction is the unmovability of the particle (cf. Palmer (1968, p. 186) and Emonds (1972, p. 554)). The following are ungrammatical. 13

(53)a. *Up which number did you look?

b. *Which number up did you look?

c. *Up did you look the number.

13 Similar facts are reported for particle verbs in Dutch and Afrikaans (cf. van Riemsdijk (1982); le Roux (1988, pp, 45-47)). Cases such as Down they sat and Out jumped the cat! (cf. Palmer (1968, p. 186), Emonds (1985, p. 260)) are exceptional and appear to he restricted to intransitive predicates. Also exceptional are Down on which shelf did they put the books? and Out to whom did they send a schedule? cited in den Dikken (1990b, p. 27).

Page 22: Object positions

598 K Y L E J O H N S O N

d. *Out which mountain did you point?

e. *Out did you point the mountain.

This fact is unexpected on these accounts; there is no obvious reason why particle phrases should differ from prepositional phrases in this respect.

If Heavy NP Shift derives the characteristic paradigm, as Kayne's ac- count claims, then the absence of the typical "heaviness" of the moved phrase in particle constructions is puzzling; that is, contrasts like the following are unexpected if Heavy NP Shift is involved in both.

(54)a. *Mikey found sad the child. (compare: Mikey found the child sad.)

b. Mikey found out the information.

Similarly, as Gu6ron points out, if Heavy NP Shift is responsible for the particle + X P variant, then it is expected that XP should be an island for extraction (cf. Wexler and Culicover (1981, Chapter 5)). But this is not the case, as the contrast below demonstrates.

(55)a. *What did you find unbelievable descriptions of t ?

b. What did you look up descriptions of t ?

Kayne accounts for these difficulties by suggesting that the "heaviness" restriction on Heavy NP Shift is sensitive to the relative "heaviness" of the material over which the phrase is moved, and by formulating a process that allows the particle to theta-mark the position into which a Heavy NP Shifted phrase is brought. (This proposal presupposes that the ungram- maticality of (55a) stems from the Condition on Extraction Domains (see Huang (1982, sec. 6.4)).) Further, Kayne keys the definition of "heavi- ness" to the ability to undergo the process of theta-marking underlying (55b) in order to account for the correspondence of these two properties. Nonetheless, Kayne's account fails to link these two properties with the other defining properties of particle constructions, most notably that the verb + particle combination behaves like a single lexical item. Thus, the asymmetries in (54) and (55) do not emerge as a consequence of what defines a particle verb, as would be desirable.

The NP/non-NP distinction evident in the characteristic paradigm is captured on Kayne's account by reference to conditions on what may form the subject of a small clause. Kayne suggests that all phrases but NPs must shift to the right because only NPs may be in subject position; thus the ungrammaticality of (44) and (45) is due to a prohibition on non-NPs in subject position.

Page 23: Object positions

O B J E C T P O S I T I O N S 599

There is an empirical difficulty with this conjecture, however. Although generally true that non-NPs cannot stand in subject position, it is possible for a clausal gerund to be in this position, as (56) demonstrates.

(56)a.

b.

C.

She remembers Gary telling the story being uncomfortable.

He made Sam liking chocolate be the climax of his story.

She let Chris being there make the point.

As Bolinger (1971) observes, however, clausal gerunds cannot appear to the left of particles; hence the contrasts in (57).

(57)a.

b.

C.

d.

They put their studying off.

*They put them studying off.

He gave his smoking up.

*He gave him smoking up. (based on Bolinger (1971, p. 127))

It does not seem correct, then, to place the burden of capturing the NP/non-NP distinction on the conditions governing what can stand in subject position.

Beyond these various particular difficulties, all these accounts more generally fail to give a satisfactory account of the NP/non-NP distinction that the characteristic paradigm displays, both on empirical grounds and in the sense that this distinction is not related to the parallel one found in the order-of-complements problem.

4.3. A Verb Raising Account

Let us now examine an account that preserves a theory of lexical insertion whereby the mapping between lexical items and syntactic positions is one- to-one. On this approach, the verb and particle are inserted at D-structure in a single position and separated syntactically. Thus, the D-Structure representations of particle verbs will be (58).

Page 24: Object positions

600

(58)

KYLE JOHNSON

CP

C'

C IP

NP r

Mikey I P-P

/l'

# VP

V'

V NP

V up the reference

look

W e have seen ev idence tha t the par t i c le verb m a y move to /x (cf. (38)

above) ; thus f rom (58) may be de r ived (59).

(59)

• . : ~ f f

#

V #

look up

VP

V'

v NP

t the reference

Page 25: Object positions

O B J E C T P O S I T I O N S 601

But Head Movement also appears able to move the verbal portion alone. This is suggested by the grammaticality of sentences in Dutch where "Verb Raising" strands the particle, as in (60).

(60) . . . omdat Carol hem op kon bellen

• . . because Carol him up could ring.

Assuming English to have the same range of options, (58) may therefore also yield (61).

(61) . . . #'

# VP

V ~ V'

look V NP

t up the reference

Hence, let us take the syntactic force responsible for separating the particle from the particle verb to be Head Movement. 14

The derivations corresponding to (59) and (61) make it appear that the separation of the particle from the particle verb is optional at S-structure• But this is incorrect• The contrast in (62) shows not only that there is something forcing weak pronouns to precede the particle but also that the particle is necessarily separated from the verb in these circumstances.

(62)a. I looked it up.

b. *I looked up it.

If movement to /x can optionally involve the particle, then what forces the separation in cases like (62)?

A similar question arises with respect to Dutch data• Although the

14 This is consistent with Roberts ' (1991) treatment of Head Movement, if there is no selection between particle and verbal stem, as seems likely.

Page 26: Object positions

602 KYLE J O H N S O N

particle is moved optionally by "Verb Raising," when the verb moves into Comp (as in (63)) the particle is necessarily stranded.t5

(63)a. Carol belt hem op. Carol rang him up

b. *Carol opbelt hem.

What distinguishes these cases? One approach to this problem would be to assume that whether or not

the particle can be moved under Head Movement is contingent on proper- ties of the position moved to. In particular, following a suggestion of Hilda Koopman's, let us suppose that when Head Movement adjoins a verb to inflectional morphology the moved verb cannot contain a parti- cle. 16 This might follow from whatever restriction it is that forces bound morphology to attach to the verbal portion of a particle verb rather than to the entire verb + particle. (That is, looked up contrasts with *look uped.) Now if this hypothesis is combined with the assumption that inflec- tional morphology must be adjoined to its stem at S-structure (cf. Lasnik (1981)), then Head Movement will be forced to separate the verbal portion from the particle when the adjunction site is a bound morpheme.

In Dutch, then, the verb's movement to Comp will necessarily involve stranding the particle because the verb moves to the positions holding tense and agreement morphology as a preliminary to its movement to Comp. Under "Verb Raising," on the other hand, the particle is not obligatorily stranded plausibly because this operation does not move the verb into a position holding inflectional morphology. (See Evers (1975), Haegeman and van Riemsdijk (1986), Haegeman (1988) and Baker (1988a) for discussion.) Instead, it is possible that the verb inflects as an infinitive through some other means: prior to D-structure, after "Verb Raising" or, on accounts of "Verb Raising" that involve Reanalysis, as part of Reanalysis.

To explain the English data, we must assume that verbs not only move to/~, which presumably hosts no morphology, but also to a higher position that does contain morphology. In keeping with the structure of clauses proposed in Chomsky (1989) and Belletti (1990), let us take this higher position to host either tense morphology (in simple tenses) or the affix that forms a participle (in the complex tenses) or the infinitive. With these

15 Let us adopt the account of Verb Second that involves movement of the verb to Comp; see den Besten (1983) and Koster (1975). 16 This is independently suggested by an N L L T reviewer.

Page 27: Object positions

O B J E C T P O S I T I O N S 603

assumpt ions , then , the S-s t ruc tures c o r r e s p o n d i n g to (59) and (61) a re as in (64) and (65). 17

(64) . . . T '

V T t t'

I look tt

V #

t up

VP

V'

V NP

t the reference

, ° ° W ~ (65)

/ k /< . /.t T /.t'

V tt tt VP

¢ V' look

V NP

t up the reference

17 I assume that there is a null morpheme which forms the infinitive in English and forces the particle's separation.

Page 28: Object positions

604 KYLE JOHNSON

Because the affix in T requires the verbal portion of the particle verb as stem, the particle must be stranded either in /x or in its D-structure position.

Note that to preserve the account given in section 3 for the fact that an

accusative Case-marked NP is adjacent to a verb, we must assume that no adverb can adjoin to projections of p,.lS This very probably follows

from the strict locality conditions that connect an adverb to the constituent

it modifies. (For various proposals, see Jackendoff (1972, Chapter 3),

McConnell-Ginet (1982, pp. 167-182), Higginbotham (1985, pp. 562-565)

and Sportiche (1988a, pp. 428-432).) Most analyses argue that adverbs

modify their sisters. If this requirement also brings with it the mandate that a term must be meaningful for it to be modified (the proposals in

McConnell-Ginet and Higginbotham come closest to having this result), then /x cannot be modified as it has no semantic content.

If it is Head Movement to T that breaks up a particle verb, there must

be an independent process responsible for bringing the object NP between

verb and particle since this procedure leaves the linear order of verb,

particle and object NP unchanged. Further, this process must have the

properties that give rise to the idiosyncrasies of the characteristic para- digm; it must yield (50).

It 's significant that the characteristic paradigm precisely mimics the propert ies of a phenomena discovered in the Scandinavian languages

known as Object Shift. t9 I adopt the account of this phenomena offered

by Holmberg (1986, Chapter 6) and elaborated in Vikner (1990, Chapter

4). Object Shift is the name given to a case of A-Movement that relocates

a structurally Case-marked NP just when the verb assigning Case to that

NP has been moved. Thus, the object may move past sentence negation

when the Case assigning verb has, but not otherwise, as the contrast in the following Icelandic examples illustrates.

18 In particular, we must abandon Pollock's (1989, pp. 376-383) suggestion that various adverbs are positioned between T (his Agr) and VP in English. 19 There are certain cases where the restriction to NPs found in the characteristic paradigm is lifted, as in (i) (cf. Kayne (1984b, p. 107); and it's significant that Object Shift's restriction to NPs seems to be lifted in parallel contexts, cf. (ii) from Zaenen, Thrfiinsson and Maling (1985, note 9).

(i) ?I bet they pick under the bed out as their new hiding place. (ii) l~g taldi undir rdminu f barnaskap mfnum vera gdSan felustaS.

I believed under the bed in my foolishness to be a good hiding-place.

This strongly suggests identifying the two processes.

Page 29: Object positions

O B J E C T POSITIONS 605

(66)a . . . . a~5 J6n keypti b6kina ekki. • . . that John bought book-the not • . . that John didn't buy the book.

b. * . . . a~5 J6n hefur bdkina ekki lesi~5. • . . that John has book-the not read • . . that John hasn't read the book.

That this is an instance of A-Movement is supported by its failure to license parasitic gaps or invoke Weak Crossover violations (see Holmberg (1986, pp. 173-174) for discussion).

Holmberg suggests that the fact that Object Shift creates A-chains ex- plains its dependency on verb movement; and in particular, that the ungrammaticality of (66b) stems from the Case Filter. He proposes to extend the definition of "Case assigner" so that it includes not only the lexical term itself, but all the traces left by moving that term as well. Thus, not only may keypti assign Case in (66a), but any of its traces may too. Holmberg formulates this idea so that it extends to other situations where a phonetically null term is apparently assigning Case (cf. Holmberg (1986, pp. 176-184)). I will frame Holmberg's suggestion somewhat differ- ently although I believe his particular version is compatible with the proposals made here; let us therefore adopt (67).

(67) Structural Case may be assigned anywhere from D-structure to S-structure inclusive.

With this extension to the conditions on Case assignment, the contrast in (66) derives from that part of the Case Filter that requires the highest position in the A-chain to be the sole Case-marked one (cf. (15) on page 583). Thus in (66a) for example, keypti assigns Case at S-structure to the position that b6kina moves into and this forms an A-chain consonant with (15). But in (66b), because lesi~5 assigns Case to a position to its right, b6kina must have moved to an A-position from a Case-marked position, in violation of (15).

Interestingly, there are two properties of Object Shift that do not follow from its status as A-Movement. First, it is only able to affect NPs: only NPs are able to move to the position preceding negation, as the ungram- matically of the following sentences exemplifies.

( 6 8 ) a . *Jeg betalte for den ikke. (Danish)

I paid for it not

b. Jeg betalte ikke for den.

Page 30: Object positions

606 K Y L E J O H N S O N

(69)a. *J6n reyndi [a6 raka sig] ekki. (Icelandic)

John tried to shave sel f not

(John didn't try to shave himself)

b. Jdn reyn6i ekki [a6 raka sig].

Crucially, then, Object Shift makes the distinction between NPs and other phrases that we have seen arise in both the order-of-complements problem and the characteristic paradigm of the particle construction: it bears the signature of the force we are in search of.

Second, as in the characteristic paradigm, Object Shift makes a distinc- tion between "weak" pronouns and full NPs. It is optional when the structurally Case-marked NP is full, but obligatory when the NP is a pronoun, as the contrast between (70) and (71) demonstrates•

(70)a . . . . a6 Jdn keypti bdkina ekki.

• . . that John bought book-the not

b . . . . a6 Jdn keypti ekki bdkina.

• . . that John bought not book-the

• . . that John didn't buy the book.

(71)a. * . . . a6 Jdn keypti ekki hann.

• . . that John bought not it

b . . . . a6 Jdn keypti hann ekki.

• . . that John didn't buy it.

And just as in the characteristic paradigm, the effect is suspended if the pronoun is conjoined or given special emphasis.

(72)a . . . . a6 Jdn keypti ekki HANN.

b . . . . a6 Jdn heimsdtti ekki hana og hann

• . . that John visited not her and him

• . . that John didn't visit her and him.

These idiosyncrasies presumably arise as a result of particularities of the environment in which these cases of A-Movement occur• I have little to offer in the way of deriving these properties (see Section 7 for some discussion), so let's continue to use "Object Shift" as a label for those instances of A-Movement that are constrained in these ways.

The parallelism between the characteristic paradigm and the facts in Scandinavian just reviewed will be explained if it's Object Shift that is responsible for putting the "object" between the verbal portion of the

Page 31: Object positions

O B J E C T P O S I T I O N S 607

particle verb and the particle. Indeed, this is precisely what is expected if the (null) hypothesis that Object Shift occurs in English is combined with the previous section's conclusion that the verb in English always moves out of the VP it heads. As Holmberg shows, Object Shift is contingent on Verb Movement for reasons having to do with the Case Filter. Thus, if structural Case is assigned after the verb has moved, then the object must move to Specifier of VP - between the verb and its particle - as in (73) .2o (For reasons discussed in the following section, only the derivation where the particle is stranded in its D-structure position is considered here.)

z0 An apparent problem for this account arises when the extraction asymmetry illustrated below is considered.

(i) What/ did Chris look up stories about ti?

(ii) *Whati did Chris look stories about ri up?

In these examples, the object can be in Specifier of VP, on the account proposed here, and it is therefore mysterious why both don't trigger a Generalized Left Branch Condition violation. This is a quite general problem, as all nominal complements occupy a Specifier position under the present framework.

I suggest the following solution. Assume that Chomsky's (1986a) account of the Gen- eralized Left Branch Condition as a product of Subjacency is correct. Subjacency is violated when movement occurs past more than one Barrier; and a Barrier is a maximal projection that is either not "L-marked" or immediately dominates a Barrier. Chomsky defines "L- marks" as in (iii); let us change it to the very similar (iv).

(iii) Where a is a lexical category, c~ L-marks fi iff 6 agrees with the head of .r that is theta-governed by c~. c~ theta-governs 6 iff c~ theta-marks 8, and ~, 6 are sisters.

(Chomsky (1986a, p. 24))

(iv) cr L-marks 6 iff c~ is a lexical category and theta-governs 8. cr theta-governs 6 iff c~ theta-marks and governs 6.

With this change, consider the representation that (ii) receives, shown in (vi).

(vi) *Whatl . . . lookk [vP [Ne stories about ti ] [v' [ta. up] ]]?

In this case, look does not L-mark the stories-NP because it does not theta-mark that NP. Instead the particle verb from which look moves, look up, assigns the relevant theta-role. But look up does not L-mark the stories-NP either because it does not govern it. Hence, both NP and VP are barriers, and movement of who past these phrases violates Subjacency. In (i), on the other hand, the entire particle verb look up moves to a position that governs, and therefore L-marks, the stories-NP.

Page 32: Object positions

6 0 8 K Y L E J O H N S O N

(73) . . . #'

v~ tt N P ~ V'

I /% look the reference V NPi

tj up t

If accusative Case is assigned before the verb moves, then the object remains in its D-structure position. This second option, as (71) shows, is not available for weak pronouns. Hence, the central properties of the characteristic paradigm emerge straightforwardly, if the verb and particle are base-generated together and separated by Verb Movement.

Further, this account gives a direct explanation for Palmer's observation that the particle and verb cannot be separated by A'-Movement, as the examples in (53) show. These examples would have to involve movement of the entire VP which the verb has vacated, giving to (53a), for example, a representation like that in (74).

(74) [vP [v te up] which number]j did you looki tj?

But representations such as these presumably violate the ECP: the trace left by raising look is not properly governed at S-structure o r L F . 2~ Hence, none of the difficulties facing analyses of the particle construction that give the particle and verb different D-structure positions arise on the account advocated h e r e . 22

21 We must assume that traces left by Head Movement are gamma-marked at S-structure in the scheme proposed by Lasnik and Saito (1984), or that VP Reconstruct ion at LF is not sufficient to license these cases. 22 There are some puzzles for the account I am offering, however. I see no straightforward account, for example, of Kayne 's discovery that expletives and idiom chunks resist combining with particle verbs in E C M constructions, as in (i) (see Kayne (1984b, p. 114)).

(i) *They're trying to make out advantage to have been taken of them.

*They're trying to make advantage out to have been taken of them.

*They're trying to make out there to be no solution to this problem.

*They're trying to make there out to be no solution to this problem.

(It 's worth noting in this context that Object Shift of expletives out of ECM infinitives is

Page 33: Object positions

O B J E C T P O S I T I O N S 609

5 . C O N D I T I O N S G O V E R N I N G S T R U C T U R A L C A S E

Let us now reconsider the order-of-complements problem discussed in Section 3. This phenomena, as well as the characteristic paradigm of the particle construction, both make a distinction between NPs and other categories. I have suggested that Object Shift is responsible for the distinc- tion in the particle construction; thus to account for this shared property, it is natural to extend Object Shift's responsibility to the distinction in the order-of-complements problem as well. This amounts to the hypothesis that it's Object Shift that underlies (17c) of Section 3: it's Object Shift that is responsible for moving structurally Case-marked NPs into the Specifier of VP thereby causing NPs to precede all other complements.

The difficulty with this conjecture is that Object Shift applies optionally in the Scandinavian paradigms, but the movement of NPs to the Specifier of VP is obligatory. This suggests that there is a factor independent of Object Shift insuring that nominals move (at least) to Specifier of VP. That is, the two aspects to the order-of-complements problem identified in the Introduction - the force that singles out nominals and the force that guarantees their initial position - appear to have separate causes.

perfectly grammatical in Swedish. (Cf. Holmberg, 1986, p. 175)). These judgements do seem sensitive to vague semantic factors however; compare (ii) with (i).

(ii) ??I made out advantage to be taken of them.

??I made advantage out to be taken of them.

?I made out there to be a unicorn in the garden.

I made there out to be a unicorn in the garden.

Also left mysterious are contrasts like (iii), from Kayne (1984b, pp. l l l f f ) .

(iii)a. Who are they trying to make out a liar? b. *?What kind of a liar are they trying to make Bill out?

Kayne's account relies on giving these a D-structure like: . . . Iv,make [ [NP a liar] out]], which I believe to be effectively rebutted in Gu~ron (1987; 1990). Den Dikken's (1990a, pp. 27-28) suggestion that particles are not "Lexical," and therefore that (iiib) violates the ECP because the trace in it is not L-marked, might be extendable to the account offered here. Nonetheless, it seems to predict that What did you make John out to like? should be a Subjacency violation; and the trace in 1 made Sam out t to be intelligent, would have to be exempt from the ECP, Further, the contrast between (iiia) and (iiib) on den Dikken's account derives from Incorporating the particle into the verb; but embracing this proposal makes an account for the characteristic paradigm obscure.

Finally, I do not see how the account of particle constructions advanced here can be extended to the varying restrictions on particle constructions in Swedish and Danish (Ice- landic and Norwegian pattern like English). More particularly, I do not see a straightforward way of linking the word order restrictions that emerge in the particle constructions of these languages with the obviously related word order restrictions on the causative construction. (See Taraldsen (1983) and Vikner (1987) for discussion, and Bolinger (i971, pp. 70-72, 79- 80) for examples in English.)

Page 34: Object positions

6 1 0 K Y L E J O H N S O N

To identify this second factor, let us examine the contexts where Object Shift is applicable• Object Shift may move a nominal to the left of negation in Icelandic, and in this context it is optional, as in the following examples.

(75)a . . . . a~5 J6n keypti b6kina ekki.

• . . that John bought book-the not

b . . . . a~5 keypti ekki b6kina.

• . . that bought not book-the

It may also apply optionally to move nominals to the left of a particle, as shown in the previous section. 23 The only case where Object Shift is obligatory is when it moves a nominal to Specifier of VP. This is true not only for the English cases that have preoccupied our discussion, but for Scandinavian examples as well. In Scandinavian, just as in English, com- plements to verbs are arranged in such a way that nominals precede other categories. Thus Icelandic, for example, hosts paradigms parallel to those found in English.

(76)a. I~g skrifa6i Haraldi um ritger0ina hans.

I wrote Harald about thesis his

b. *t~g skrifaN um ritger~Sina hans Haraldi.

c. t~g lofaN Haraldi at5 raka mig.

I promised Harald to shave myself

d. *l~g lofa~Si a6 raka mig Haraldi.

Let 's begin by confining our attention to a comparison of these cases of obligatory movement and the optional movement past negation in Ice- landic. Thus, what we search for is a reason for the obligatoriness of the movement to Specifier of VP, and the optionality of movement of NPs past this position•

Let us explore the possibility that it is the Case Filter that forces move- ment to Specifier of VP. This can be achieved if we exploit Chomsky's (1989) proposal that structural Case is assigned not by the verb, but instead by a functional head immediately above VP which he identifies with Object Agreement• Let's adopt this suggestion, and assume that /x

23 I wil l a ssume tha t the ob l iga to ry m o v e m e n t of weak p ronouns has a d i f ferent source.

Page 35: Object positions

OBJECT POSITIONS 611

is the assigner of structural Case, leaving open the question of whether /x is Object Agreement . Further, let us assume that structural Case is as- signed under government, as is usual, and that government is defined as in (77). 24

(77) o~ governs ,c i f f a c-commands ~- and no more than one projec- tion includes 7 and excludes oz.

c - c o m m a n d s ~ iff every X ~, n > 0, that contains a contains /3, and a does not dominate/3

This has the consequence that structural Case will be assigned no lower than the Specifier of VP, for this is the only position governed by/x. Note that complements within V' are not governed, hence not Case-marked, by /z since two projections, V' and VP, exclude/x. This, then, will guaran-

tee that Object Shift applies so as to move NP complements to Specifier of VP, for only then will they meet the requirements of the Case Filter.

The optionality of Object Shift beyond Specifier of VP (in the Scandi- navian languages) now follows from the premise that Case may be assigned at any point in the derivation of S-structure from D-structure. Different from earlier assumptions, however, it is /x, not the verb, that assigns structural Case. Thus, e i ther /x may assign Case before it is moved, and the object remains in Specifier of VP; o r / x may assign Case after it and the verb adjoined to it have moved to a position preceding negation, in which case the object also moves to a position preceding negation.

Let 's now reconsider the role Object Shift plays in deriving the charac- teristic paradigm of the particle construction. In these cases, as we have seen, movement of the Case-marked NP between the particle and the rest of the verb is optional. There are two circumstances to consider: one where the particle has been stranded in its D-structure position, and one where Head Movement brings the particle t o /x and strands it there. The first of these yields a representation like that in (78) after the verb has Incorporated into T.

24 This departs from the definition of government advanced in Chomsky (1986a), though not materially for the cases under discussion. (77) is a strengthened version of the definition defended in Johnson (1988), and is very close to the definitions in Sportiche (1988) and Culicover and Rochemont (1990, Chapter 1). That government should be defined in terms of c-command rather than m-command is argued in Culicover and Rochemont (1990, Chapter 1), and would follow if Rizzi's (1990, pp. 40-44) conclusion that proper head government is defined on c-command is generalized.

Page 36: Object positions

612 KYLE JOHNSON

(78)

T UP

/,t T /z'

V /~ ed VP

look t' V'

V NP

t up the reference

In this situation, either the Specifier of VP or the Specifier of /xP will be a Case-marked position, depending on whether /x assigns Case before or after its movement to T. Thus the reference must move to one of these two positions, and will in either case precede the particle.

When the particle has moved to Ix, the representation in (79) arises.

(79)

T UP

V T #'

look ed #

V #

t' up

VP

V'

V NP

I /" , t the reference

Page 37: Object positions

O B J E C T P O S I T I O N S 613

Note that unlike (78), /x does not move to T in (79). Instead, because T does not tolerate look up as a stem, Head Movement is forced to move only look, stranding up and /x both. As a consequence, the only Case- marked position in this representation is the Specifier of VP; and therefore the reference will move to this position.

Finally, consider the situation that obtains when a particle verb's com- plement is a weak pronoun. As the facts from Scandinavian have indi- cated, Object Shift obligatorily moves weak pronouns as close to the verb as possible. In English, this means that Object Shift must apply so as to move weak pronouns to Specifier of /xP. Thus, the only representation consonant with this requirement is one like (80).

( 8 0 ) . . . T'

V #

I look

# T NPi #'

f ed it I

t '

VP

?Z, t V NPi

t up t

Note that the particle has been stranded in its D-structure position so that /x can be brought into T. This is necessary if the chain formed by moving it into Specifier o f /x P is to meet that requirement on A-chains which allows only the topmost position to be Case-marked.

In this way, then, the characteristic paradigm of the particle construc- tion, as well as the NP-first order of complements, derives from combining Object Shift with the hypothesis that /x assigns structural Case. Most importantly, if Object Shift is taken to underlie both these phenomena, then an explanation emerges for why the same NP/non-NP distinction is

Page 38: Object positions

614 KYLE JOHNSON

found in each. Object Shift's mark is found in several other situations as well; we turn to one of these in the next section.

6. THE DOUBLE OBJECT CONSTRUCTION

There are situations, licensed by a lexically idiosyncratic class of verbs, of which give, bake, send, and envy are representative, where two NPs follow the verb. These double object constructions pose another apparent counterexample to the "NP First" generalization that Sections 3 and 5 attempt to explain. This section sketches an analysis of this construction that is in line with the results of the previous sections and accounts for its apparent violation of the "NP First" generalization.

Of the many interesting peculiarities of the construction, one is that there appears to be an asymmetry with respect to the relations of the two NPs. In particular, as Barss and Lasnik (1986) have discussed, the second NP appears to be asymmetrically in the scope of the first. Thus, as (81) illustrates, there is a difference between the two NPs with respect to various scope sensitive phenomena.

(81)a. Gary showed

b. *Gary showed

c. Gary showed

d. *Gary showed

e. *Gary showed

f. Gary showed

(82)g. Gary showed

the boys each other's parents.

each other's parents the boys.

[every girl]/hers parent.

hers parent [every girl]/.

[the devil]; Brenti's parents.

Brenti's parents [the devil]s.

nobody anything.

h. *Gary showed anybody nothing.

This contrast also appears in other constructions; compare (81) with (82).

a. Gary showed the boys to each other's parents.

b. ??Gary showed

c. Gary showed

d. *Gary showed

e. *Gary showed

f. Gary showed

each other's parents to the boys.

[every girl]i to heri parent.

heri parent to [every girl],-.

[the devil]s to Brent~'s parents.

Brenti's parents to [the devil]~.

Page 39: Object positions

OBJECT POSITIONS 615

(82)g. Gary showed nothing to anybody.

h. *Gary showed anything to nobody.

The ungrammaticality of the examples in (82) is expected; in each case, the PP blocks c-command of the first NP by the second. In so far as scope is determined by S-structure c-command relations, then, the contrasts in (82) are understandable. The contrasts in (81), however, suggest that there is hidden structure to the VP that prevents the second NP from c- commanding the first.

There is a class of analyses of the double object construction which claims that the two NPs of the construction form a "clause" of which the first is in a subject-like relation to the second; recent examples are Kayne (1984a, Chapter 7), Herslund (1986) and Larson (1988). If under these analyses the construction is given a representation like that in (83), then the asymmetric c-command relations can be explained.

(83) . . . [vp V [TP NP' [?, Np2]]]

NP 1 in (83) c-commands NP 2, but if c-command is sensitive to ?', then NP 2 does not c-command NP 1.

A number of phenomena recommend assigning (81) the hidden struc- ture in (83); I will review several here. First, as Kayne (1984a) argues, if verbs of this class select small clauses as in (83), then certain facts concern- ing nominals built on these verbs become understandable. In particular, like other verbs that select small clauses, these verbs cannot head nominals that have identical argument structure; thus, the ungrammaticality of the examples in (84) parallels that of the examples in (85).

(84)a. *the gift of Gary (of) the book

b. *Gary's gift (of) the book

c. *envy of Sam (of) his job

d. *Sam's envy (of) his job (with the relevant meaning)

(85)a. *the belief of Mittie intelligent

b. *Mary's belief intelligent

c. *a consideration of Chris unhappy

d. *Chris's consideration unhappy

Second, in certain languages where anaphors have a "subject orien- tation," the first NP of the double object construction behaves as a "sub- ject." For example, in Danish, as Herslund (1986) shows, the reflexive sig, normally only able to accept NPs in the Specifier of NP or IP as

Page 40: Object positions

616 KYLE J O H N S O N

antecedent, can be bound by the first NP of the double object construction. Thus, contrasts such as the following obtain.

(86)a. *Jeg fortalte drengeni om sin,- bamse.

I told boy-the about sells teddybear b. ?De gav hams sins bekomst.

They gave him sells what-he-deserved

(Herslund (1986), p. 135)

Finally, as Vikner (1990, sec. 4.3) argues, the two NPs of the double object construction move as a unit under Object Shift. Thus, in Danish, where Object Shift applies only when the object is a pronoun, Object Shift may apply to double object constructions in such a way that both NPs are moved. The example in (87) is representative. 25

(87) Peter viste [Te hende den] jo.

Peter showed her it indeed

It's also possible for the first of these NPs to move independently; the second NP, however, does not have this freedom.

(88)a. Peter viste hende jo bogen.

Peter showed her indeed book-the

b. *Peter viste den jo Marie.

Peter showed it indeed Mary

These contrasts receive a straightforward explanation, if the two objects form a phrase that Object Shift may move.

In fact, this last paradigm points to a more particular version of the approach to the double object construction that (83) expresses. Recall that characteristic of Object Shift is its ability to distinguish NPs from other phrases. Therefore, that Object Shift is able to move both NPs of the double object construction suggests that the phrase these two NPs make-up is itself an NP. There is evidence supporting this conclusion that comes from the range of meanings that double object constructions are limited to. As Green (1974) shows, a constant element to the meaning of double object constructions in English is a possession relation that holds

25 All examples are from Vikner (1990, Sec 4.3).

Page 41: Object positions

O B J E C T P O S I T I O N S 617

between the first and second NP. 26 Consider by way of illustration the contrasts that obtain when the double object construction is compared to the other word order characteristic of the relevant verbs.

(89)a. Mittie taught a lesson to Gary.

b. Mittie taught Gary a lesson.

c. Sam sent a telegram to New York/Gary.

d. Sam sent *New York/Gary a telegram.

e. Betsy threw a ball to first base/Gary.

f. Betsy threw *first base/Gary a ball.

g. Liz baked a cake for Mikey.

h. Liz baked Mikey a cake.

i. Gary spared Betsy the ordeal.

One difference between (89a) and (89b) is that (89b) entails that Gary learned the lesson, whereas (89a) does not. Similarly, the difference be- tween the pairs (89c,e) and (89d,f) is that in the second but not the first does the agent of sending or throwing intend that the telegram or ball be possessed. Thus, presumably, the ungrammaticality of Sam sent New York a telegram and Betsy threw first base a ball follows from the inability of New York andfirst base to have the relevant possession relation. Likewise, note that the range of relations that Mikey may bear in (89g) is limited to just the possessor relation in (89h): thus, as Green notes, (89h) has a paraphrase something like "Liz intended Mikey to have a cake by baking it." Finally, in (89i) it is a possession relation between Betsy and the ordeal that is being prevented.

Interestingly, it is just this range of meanings that arises with respect to the "possession" relation between a genitive NP and the remainder of the NP it is contained in: compare (89) to the parallel examples in (90).

(90)a. Gary's lesson

b. Gary's telegram

c. *New York's telegram

26 See also Oehrle (1976), Kayne (1984a) and Herslund (1986) where this fact is discussed. The relation can be subtly involved, as Green (1974, pp. 98-103; 110-133) discusses. Thus, for example, in Mary denied John his dessert there is no possession relation entailed, rather there is one denied; and in Mary sent Bill a letter, a possession relation is merely intended. Further, the relation can be abstract: in Sam showed Sally the picture, Bets), played Liz a song, etc. it is a percept or experience that is possessed.

Page 42: Object positions

618 KYLE JOHNSON

(90)d. Gary 's ball

e. *first base 's ball

f. Mikey's cake

g. Betsy's ordeal

This parallelism won' t be accidental if the first NP in the double object

construction is part of an NP in the same way that the genitive NPs in (90) are.

There are a number of conceivable ways of rendering this idea; let me

very tentatively offer the following proposal. Adopting a " D P Hypo- thesis"-type analysis of nominals, 27 suppose that the double object con-

struction has a representation like that in (91).

(91) . . . VP

V DP

give DP D'

D' D

D NP

I I a chiltl

DP

tile book

Take r to be the Head of the DP (henceforth the "clausal DP" ) whose

complement is the second nominal and whose Specifier is the first nominal

of the double object construction. Further, assume that r assigns Case to its complement nominal, accounting for the two Cases that this construc-

tion hosts. The representation in (91) is consistent with all the facts listed above,

and moreover gives an account for the particularities of the effect of Object Shift in the double object construction. The grammaticality of (88a) follows if hende is dependent on /x for structural Case. When p~

27 See Abney (1987), Speas (1986), Fukui (1986) among others. I believe that the proposal that follows can be translated without significant change to a framework that gives a more traditional representation to nominals.

Page 43: Object positions

O B J E C T P O S I T I O N S 619

(containing the incorporated verb) moves to a position to the left of the adverb, accusative Case can be assigned to the position into which hende moves. The grammaticality of (87) also follows if (91) is correct. If the clausal DP in (87), containing both hende and den, is also dependent on /x for Case, then it too will be a possible Object Shifter. Finally, the ungrammaticality of (88b) follows if the DP containing den is dependent on ~" for Case; it would then not be able to undergo Object Shift indepen- dently without violating the well-formedness conditions on A-chains. 28

There are many questions raised by an analysis of double object con- structions like this one. One concerns the relative immobility of the clausal DP - a problem for all "small clause" accounts of the double object construction. The clausal DP cannot undergo A'-movement, as in (92), nor A-movement to subject position, as in (93).

(92)a. *Which woman a gift did you send?

c. *The boy a cake, you baked!

(93)a. *Mittie a gift was sent.

b. *Liz a cake was baked.

Why is Object Shift the only operation able to move the clausal DP? Another concerns the Case assigning properties of the construction.

The first DP we may assume receives its Case from /x, for it has the "object'Mike properties characteristic of being assigned structural Case (e.g. it loses Case under Passive). The second DP we have assumed receives Case from z. But what of the clausal DP itself? I f /x assigns its Case to the first DP, then how does the clausal DP satisfy the Case Filter?

I would like to explore a common solution to these problems that relates them to an idea of Baker's (1988b, pp. 106-120; 340-349) (see also Baker, Johnson and Roberts (1989, pp. 234-241)). Baker argues that in certain contexts Noun Incorporation stands in place of Case assignment. In parti- cular, he argues that the Case Filter can be satisfied by adjoining the head of a nominal to the verb that theta-marks it. Let's suppose that it is this alternative method of satisfying the Case Filter that licenses the occurrence of the clausal DP in the double object construction.

Consider now the derivations that this assumption will give to a double object construction. The head of the clausal DP will Incorporate into the verb, yielding a representation like that in (94).

as See Vikner (1990, Sec. 4.3.3) for a different account of the ungrammaticality of (88a) based on Relativized Minimality.

Page 44: Object positions

620 K Y L E J O H N S O N

(94) ° ° ° ]..~I

# VP

V'

V

D V

I r

DP /N DP D'

T /N # T

/N V /~

J give

# VP

c i~ v' G a r y / ~ / ~

D~ V i~ o' I I / N

t the book

give Gary Du DP

t the book One S-structure that can be formed from (94) arises by stranding z in the verb's D-structure position, moving give to T and Object Shifting the first DP into Specifier of VP or p.P, as in (95).

( 9 5 ) . ~ ~

Page 45: Object positions

OBJECT POSITIONS 621

Whether Gary is moved to Specifier of VP or /xP will depend on whether

/x assigns Case before or after it moves; we consider here only the deri- vation where Case is assigned to Specifier of VP. Interestingly, it appears that when z Incorporates, it must be stranded in the verb's D-structure position, as in (95), for reasons that we turn to at the end of this section.

Note that this hypothesis provides an explanation for the relative immo- bility of the clausal DP. Because the clausal DP contains the empty category left by Incorporating z, it cannot move to a position where this

trace is not antecedent governed by z and must therefore remain within the c-command domain of the position into which ~- has Incorporated. 29

Thus, it cannot be moved outside the VP to an A'-position, as in (92), or an A-position, as in (93), because this brings it beyond the scope of the Incorporated z triggering an ECP violation.

Let us now turn to cases where the clausal DP undergoes Object Shift. This is possible, as the Danish (87) illustrates, but should be blocked along the lines just discussed, if z remains Incorporated in the verb's D-structure position. As noted above, the clausal DP's ability to Object Shift indicates

that it is dependent on /x for Case. It therefore appears that under Object Shift z does not Incorporate, but that instead the clausal DP is Case- marked by /x. But if /x assigns Case to the clausal DP, then it would appear that the first nominal (the one in the Specifier of the clausal DP) is left without Case.

I suggest that this is precisely what happens, and that to satisfy the Case Filter the first DP undergoes cliticization. This would explain Vikner's (1990, sec. 4.3) observation (see also Holmberg (1986, pp. 206-207)) that the clausal DP may Object Shift (in Mainland Scandinavian) only if the nominal in its Specifier position is a weak pronoun. Thus the grammatical- ity of (87), repeated here as (96), contrasts with the examples in (97).

(96) Peter viste [oe hende den] jo.

Peter showed her it indeed

(97)a. *Peter viste hende bogen jo.

Peter showed her book-the indeed

b. *Peter viste Marie bogen jo.

Peter showed Mary book-the indeed

29 We must assume, as is implicit in the work of Chomsky (1989), that any member of the chain into which some head has Incorporated can act as an antecedent governor for that head. Further, assume that antecedent government requires c-command (contra Chomsky (1986a)); see Rizzi (1990, pp. 6, 24-27) for discussion.

Page 46: Object positions

622 KYLE J O H N S O N

(97)c. *Peter viste Marie den jo.

Peter showed Mary it indeed

If cliticization is a form of Incorporation, as Kayne (1989a) argues, then this should be an alternative method of satisfying the Case Filter, along the lines described by Baker. And if cliticization is restricted to weak pronouns, then the ungrammaticality of (97b,c) will follow from the Case Filter.

Note that as (97a) illustrates, not only must the first nominal in the double object construction be a weak pronoun but so also must the second nominal. I have no suggestion as to the cause of this restriction. 3°

Although these suggested refinements of the "clausal" analysis of dou- ble object constructions are very tentative, if they are in the right direction, they support the account of particle constructions offered in the previous section. Support comes from the behavior of certain particle verbs that invoke the double object construction. In cases such as these, a paradigm reminiscent of (96) and (97) emerges. 3~ The first object must appear before the particle, as the paradigm in (98) illustrates; and both objects (i.e., the clausal DP) may appear before the particle only when both are weak pronouns, as the paradigm in (99) shows.

(98)a. Sam handed Mittie down the tools.

b. Gary poured Betsy out the milk.

c. Sam handed her down the tools.

d. Gary poured her out the milk.

e. *Sam handed down Mittie the tools.

f. *Gary poured out Betsy the milk.

g. *Sam handed down Mittie them.

h. *Gary poured out Betsy some.

i. *Sam handed down her the tools.

j. *Gary poured out her the milk.

30 It's worth noting in this connection that Object Shift is only possible for weak pronouns in Danish and the remaining Mainland Scandinavian languages (see Section 7). Thus, what- ever causes this restriction might extend to the cases at hand. Icelandic data are problematic, as here the clausal DP may Object Shift even when the first nominal is full (see Vikner 1990, sec. 4.3.3.2, and 4.3.3.3 for an extension to verb-final Germanic languages.) 31 See Poutsma (1928, p. 421), Bolinger (1971, pp. 168 169) and Emonds (1976, pp. 82- 86), where these facts are introduced. The dialectal variation discussed in Emonds might find a parallel variation with respect to structures like (103).

Page 47: Object positions

OBJECT POSITIONS 623

(98)k. *Sam handed down her them.

1. *Gary poured out her some.

(99)a. ?Sam handed her them down.

b. Gary poured me some out.

c. *Sam handed Mittie them down.

d. ?*Gary poured Betsy some out.

e. *Sam handed Mittie the tools down.

f. *Gary poured Betsy some milk out.

g. *Sam handed her the tools down.

h. *Gary poured her some milk out.

Now if the account of the particle construction is combined with that of the double object construction, then an example like (98a) will receive the S-structure representation in (100).

(100) . . . #'

VP

I handed

i Mittie

V,

V DP

v iP; ~r ~) down t

/ t

DT

D k DP p / ' , , t the tools

The ungrammaticality of (98i-1) follows from the requirement that weak pronouns Object Shift (see the discussion surrounding (71)). The ungram- maticality of (98e-h) is mysterious at this point - we return to these cases directly.

The paradigm in (99), however, is fully expected. The paradigm that

Page 48: Object positions

624 K Y L E J O H N S O N

Vikner describes for the application of Object Shift to the double object construction is reproduced precisely here. In particular, the clausal DP may Object Shift only if both nominals within it are weak pronouns, as shown in (97) above. Thus, (99a,b) are grammatical because both nomin- als in the clausal DP are weak pronouns licensing Object Shift and yielding a representation like that in (101).

(101) ...t~'

poured

VP

me D DP V, out t

I I "g some t

(For simplicity's sake I have not represented the necessary cliticization of me.) The remaining examples in (99) are ungrammatical because the clausal DP contains at least one nominal that is not a pronoun, and Object Shift is therefore blocked.

Let's now return to consider the ungrammaticality of (98e-h). These examples receive an S-structure representation like that in (102) (=(98e)).

(102) / . ~

T

hand

/%

°/% /% V i V'

t i down Mittie V/ DP

t il~ D'

t D DP

~" the tools

Page 49: Object positions

O B J E C T P O S I T I O N S 625

(I have suppressed Incorporation of ~- here.) (102) conforms to the con- ditions on Case assignment we have adopted, and should therefore be grammatical. Note in particular that Mittie is governed by/x, and should as a consequence be in a Case-marked position. Nonetheless, something blocks this representation.

It's worth noting that, as Kayne (1984a, sec. 3.2) points out, the same fact reappears in (other) situations where a particle verb combines with a small clause. Thus, parallel to (98e-h), examples such as (103) are ungrammatical.

(103) *Betsy made out Gary a liar.

Just as with double object constructions, the particle must separate the two nominals in these examples:

(104) Betsy made Gary out a liar.

This appears to be a fact peculiar to small clauses; it is optional when the particle verb combines with an infinitival clause, as in (105).

(105)a. Betsy made out Gary to be a liar.

b. Betsy made Gary out to be a liar.

Thus, these facts strengthen the view that the double object construction involves a small clause. Further, they suggest that it is some property of small clauses that is relevant.

This generalization can be described under the present system along the following lines. I have argued that when the clausal DP of the double object construction does not Object Shift, its head Incorporates into the verb. Stowell (forthcoming) makes a similar suggestion for small clauses in general. He argues that their heads Incorporate into the selecting verb at LF; but given the present proposals, his idea can be transposed to the claim that the Incorporation occurs at S-structure. Thus, what distin- guishes the infinitives in (105), which allow the particle to fall on either side of the Case-marked nominal, from double object constructions and small clauses, where the particle must follow the nominal, is whether an Incorporation is involved. This correlation follows if there is something that allows only the stem portion of the verb that hosts the Incorporation to move to/x. Therefore, if in (106) only V* can move, the facts outlined above will follow. 32

32 This constraint on structures hosting Incorporation will also derive the fact, observed by den Dikken and Mulder (1991, p. 74), that the clausal DP in a double object construction is not able to move to Specifier of IP even when the verb hosting Incorporation of ~ is moved to Comp.

Page 50: Object positions

626 KYLE JOHNSON

(lo6) • • . ~ r

# VP

V'

V XP

X i V X'

V* prt X i

I t

I do not know what derives this descriptive generalization, but it might find independent support in another context. When particle verbs combine

with a certain class of modifiers, the order V N P pr t is forced. Thus,

(107b,d) are awkward.

(107)a. Gary looked the fact right up.

b . *Gary looked right up the fact.

c. Betsy mopped the floor all up.

d. *Betsy mopped all up the floor•

The modifiers that have this effect are those that bear a close semantic

connection to the particle port ion of the verb. If this semantic connection is modification, then it is reasonable to assume that the modifiers are adjoined to the particle verb or, perhaps, to the particle itself. In these

examples too, then, the particle verb will receive a representat ion similar to (106), where right or all stands in place of Xi. If so, the particles in (107) will be anchored in their D-structure positions, and the Case Filter will force the nominals to migrate leftward. 33

33 The weak ungrammaticality of John sent Bob right offa package, (cf. Carlson and Roeper (1980, p. 150)) might follow if no more than one term can Incorporate into the particle verb. den Dikken's (1990a, p. 1) discovery that cases of locative inversion involve a similar constraint, as the ungrammaticality of To Bob was sent right off a package indicates, will

Page 51: Object positions

OBJECT POSITIONS 627

7. CONCLUSION

In the preceding sections, I've argued that the "NP-First" order to comple- ments of verbs is due in part to the conditions under which structural Case is assigned. Case is assigned to the Specifiers of VP or/zP, and nominal complements are therefore forced to move into one of these positions -- causing them to precede all other complements. Thus, as in the Chomsky/ Stowell approach to these phenomena, Case Theory is responsible for forcing Case-seeking terms to precede other complements. But unlike these accounts, it is not Case Theory that distinguishes nominals from the others. This is due to Object Shift, whose signature is the ability to affect only nominals.

If Object Shift is to be assimilated into the rule-flee format of the Principles and Parameters framework, then an explanation for its restric- tion to nominals should be discovered. Holmberg (1986, pp. 207-213) makes one proposal in this direction. He suggests that a distinction be made between morphological Case (m-Case) and abstract Case, and then devises a Case Theory that permits only m-Case marked terms to Object Shift. Because only nominals bear morphological Case in the relevant languages, this will prevent Object Shift from applying to other categories. Holmberg's evidence comes primarily from typological considerations. It correctly accounts for the fact that Object Shift is possible only of pronouns in the mainland Scandinavian languages, but can affect both pronouns and full nominals in Icelandic. This follows from the differing manifestations of m-Case in these languages: in mainland Scandinavian, only pronouns show morphological Case, whereas both full nominals and pronouns show morphological Case in Icelandic.

Although Holmberg's proposal is compelling, it cannot be adopted here for it is inconsistent with the use I have made of Object Shift. If Object Shift moves full nominals to Specifier of VP or/~P in English (as well as the mainland Scandinavian languages), then its application cannot be restricted just to terms bearing m-Case. Some alternative account for the typology of Object Shift and its restriction to nominals must be found. This task falls well beyond my present abilities, but let me nonetheless sketch a different approach to the restriction to nominals that Object Shift shows.

Case Theoretic methods of singling out nominals I believe rest on the

follow if in these cases too the nominal complement must move to a V' external position to receive Case.

Page 52: Object positions

628 KYLE J O H N S O N

observation that it is typical only of nominals to bear case (in Indo- European). Thus, "Case-marked position" is equated with "surface posi- tion of nominal" largely because of the privileged correspondence that appears to hold between morphological case and nominals. But nominals appear to have a exclusive relationship with another grammatical system in Indo-European: agreement. 34 We might exploit this observation along the same lines that Case Theory exploits the observation concerning morphological case. In particular, imagine that there is a restriction on agreement morphemes that forces them to share the relevant agreement features only with a nominal. Such a restriction would allow only nominals into the positions which agreement morphemes target.

In the cases at hand, this idea might be formalized along the following lines. First, let us adopt Chomsky's suggestion that/x is (abstract) Object Agreement. Second, adopt Kayne's (1989b, p. 87) suggestion that agree- ment holds between Agr and a term that Agr governs. This gives the cases we have been examining a representation like (108), before movement to T.

(108) TP

T'

AgrO'

AgrO VP

/ L / L V/ agrO V t

I V~ . . ,

I t

34 This is valid in so far as the cases where verbs appear to agree with clauses are actually instances of default agreement, or more likely, agreement with an expletive or trace linked with the clause. See Koster (1978), Stowell (1981) and Safir (1985) for discussion.

Page 53: Object positions

OBJECT POSITIONS 629

Now if government is defined in terms of c-command, as we have assumed, then AgrO will govern the Specifier of VP if it does not move to T, and the Specifier of AgrOP if it does. Thus, these positions will be reserved for nominals, if the relationship with agreement is restricted in the way proposed above.

Many questions remain.

A P P E N D I X

This appendix examines a class of problems that involve manner adverbs. If Head Movement is responsible for separating the verb from its comple- ment in (109a) and the particle from the remainder of the verb in (109b), then we should expect the two strategies to combine so as to separate a verb from its particle across an adverb.

(109)a. Gary talked quietly to her.

b. Sam looked the number up.

But as Bolinger (1971, p. 147) notes, except for the small class of adverbs discussed in connection with (107), such cases are uniformly ungrammati- cal in English, as the examples in (110) illustrate.

(110)a. *Betsy looked carefully up the number.

b. *Betsy looked the number carefully up.

c. *Mittie threw quickly out the trash.

d. *Mittie threw the trash quickly out.

e. *Sam dusted rapidly off the table.

f. *Sam dusted the table rapidly off.

These facts can be related to another paradigm that crucially involves the presence of an adverb. Recall that examples such as (111) may receive a parse like that in (112).

(111)a.

b.

e.

Gary put the book quickly on the table.

Sam told the story quietly to Mittie.

Betsy sent the letter secretly to Liz.

Page 54: Object positions

630 K Y L E J O H N S O N

(112) • . ° [ . . ~ t

# VP

v'

Adv V'

/

t

\ ~ PP

Interestingly, these examples contrast with a class of other superficially similar cases of which (113) are representative. 35

(113)a. *Sam told Mittie quietly the story.

b. *Betsy sent Liz secretly the letter.

c. *Gary believes Mikey sincerely to be intelligent.

d. *Sam considers Mittie strongly to have been misled.

e. *Betsy believes Liz sincerely intelligent.

f. *Sam considers Mittie strongly misled.

What distinguishes these examples from those in (111) is that the NP that precedes the adverb in (113) is not an argument of the root verb, whereas this NP in (111) is. In (113), the root verb has a small clause for comple- ment, whose "subject" has Object Shifted to the left of the adverb.

One factor that distinguishes (111) from (113), therefore, is the status of the empty categories left by Object Shift. In (111), these empty categor- ies are in theta-governed positions, whereas in (113), Object Shift leaves an empty category in the non-theta-governed "subject" position of the small clause. A similar property distinguishes the ungrammatical examples where a particle verb moves past an adverb (as in (110)) from examples where a morphologically simple verb moves past an adverb (as in (109a)). A comparison of these two cases shows that the empty category left by moving the entire verb, as in (109a), is in the head position of a comple-

3s The adverb is to be understood as modifying the root verb in each of these examples.

Page 55: Object positions

OBJECT POSITIONS 631

merit, whereas the empty category created by moving a portion of the verb, as in (110), is not in the head position of a complement, but is instead inside that head.

The Empty Category Principle (ECP) is able to distinguish empty categ- ories in theta-governed positions from those that aren't; therefore I suggest blaming the ECP for the ungrammaticality of the examples in (110) and (113). We can take the ECP to be as in (114).

(114) The Empty Category Principle A non-pronominal empty category must be properly governed. o~ properly governs 6 iff:

a. ~ theta governs 6, or

b. oz antecedent governs 6.

antecedent governs 6 iff c~ and 6 are coindexed and no barrier for 6 excludes oz.

Consider first (110). Because it's the presence of the adverb that causes the ungrammaticality of this example, it's a reasonable conjecture that the additional V' invoked by the adverb is responsible for the ECP violation. Let us therefore revise Chomsky's (1986a) definition of Barriers, modi- fying a suggestion in Sportiche (1988b), so that X' projections are admitted to the set of possible Barriers as in (115).

(115) o~ n, n > 0, is a blocking category for r i f f ~ includes r and is not theta-governed. • is a barrier for r i f f q~ includes r and: (i) is a blocking catgory for r , or

(ii) d) immediately dominates a blocking category for r.

oz theta-governs 6 iff a theta-marks and governs 6.

Consider now the parse that (110) receives. (I ignore throughout this section the irrelevant movement to T.)

(116) . . . [p.' [/z Vi [ ll [vP [v .... adverb [v,* [vti part]]]]]

By (115), V'* in (116) is not theta-governed as it is not governed by/x. Thus, both V'* and V'** will be Barriers for ti, and antecedent govern- ment of ti by Vi will fail. Because ti is also not theta-governed, it violates the ECP.

This contrasts with the situation in (109a), which receives a parse like that in (117).

(117) . . . [., [. v, [ 11 [,,p [v .... adverb [v,* [vti] PP]]]]

Page 56: Object positions

632 KYLE JOHNSON

As in (116), both V '* and V'** are Barriers for ti in (117) and therefore ti cannot be directly antecedent governed by V~. However, note that Vi does antecedent govern VP. Hence, if we adopt Lasnik and Saito's (1984, p. 248) conclusion that a term, a, is antecedent governed by /3 if the phrase c~ heads is antecedent governed by /3, then ti in (117) will be antecedent governed by Vi because it heads VP. The difference between (116) and (117), then, derives from whether the trace left by moving the verb constitutes the head of the V P or is buried within that head.

In parallel fashion, the contrasts between (111) and (113) can be de- rived. The examples in (111) receive a parse like that in (118).

(118) . . . {,, [~ V, [/xl] [vP NP~ [v, adverb [v, tk t~ ]]]l •

Both empty categories in (118) are properly governed: t~ is properly governed because it is the head of the antecedent-governed VP, and ts is theta-governed by t~. Sentences with this shape are therefore grammatical.

The ungrammatical examples in (113) receive a parse like that in (119).

(119) . . . [~, [~ Vk[P,]] [vpNPi [v' adverb [v, tk [sc t~ [...]]]]1].

As before, t~ in these cases is properly governed because it is the head of the antecedent-governed VP. But t/- in (119) is not properly governed. It is not theta-governed as in (118) because it is in the subject position of the small clause; nor will it be antecedent governed by NPi, as both V's in (119) are Barriers for t;. Hence, these examples are ungrammatical for the same reason that the examples in (110) are: they produce an ECP violation.

R E F E R E N C E S

Abney, Steven: 1987, The English Noun Phrase in Its Sentential Aspect. Ph.D. Dissertation, MIT, Cambridge.

Babby, Leonard: 1980, "The Syntax of Surface Case Marking', in Cornell Working Papers in Linguistics, Number 1, 1-32.

Baker, Mark: 1985, 'Syntactic Affixation and English Gerunds', in Jeffrey Goldberg et al. (eds.), Proceedings of the West Coast Conference on Formal Linguistics, pp. 1-11. Stanford Linguistics Association.

- - : 1988a, 'Against Reanalysis of Heads', in McGill Working Papers in Linguistics, Special Issue on Comparative Germanic Syntax, Denise Fekete et al. (eds.), pp. 35-60, Montr6al: McGill University.

Page 57: Object positions

OBJECT POSITIONS 633

- - : 1988b, Incorporation: A Theory of Grammatical Function Changing, University of Chicago Press, Chicago.

Baker, Mark, Kyle Johnson, and Ian Roberts: 1989, 'Passive Arguments Raised', Linguistic Inquiry 20, 219-251.

Baltin, Mark: 1982, 'A Landing Site Theory of Movement Rules', Linguistic Inquiry 13, 1- 38.

Barss, Andrew and Howard Lasnik: 1986, ~A Note on Anaphora and Double Objects', Linguistic Inquiry 17, 347-354.

Belletti, Adriana: 1990, 'Generalized Verb-Movement', ms., Universit6 de Gen6ve. Besten, Hans den: 1983, 'On the Interaction of Root Transformations and Lexical Deletive

Rules', in W. Abraham (ed.), On the Formal Syntax of the Westgermania, John Benjamins Publishing Company, Amsterdam.

Bolinger, Dwight: 1971, The Phrasal Verb in English, Harvard University Press, Cambridge. Carlson, Greg and Tom Roeper: 1980, 'Morphology and Sub-categorization: Case and the

Unmarked Complex Verb', in Teun Hoekstra et al. (eds.), Lexical Grammar, pp. 123- 64, Foris Publications, Dordrecht.

Chomsky, Noam: 1981, Lectures on Government and Binding, Foris Publications, Dordrecht. - - : 1986a, Barriers, MIT Press, Cambridge. - - : 1986b, Knowledge of Language, Praeger Publishers, New York. - - : 1989, ~Some Notes on Economy of Derivation and Representation', in MIT Working

Papers in Linguistics 10, MIT, Cambridge. Culicover, Peter W. and Michael Rochemont: 1990, English Focus Constructions and the

Theory of Grammar, Cambridge University Press, Cambridge. Dikken, Marcel den: 1990a, ~The Structure of English Complex Particle Constructions', in

R. Bok-Bennema et al. (eds.), Linguistics in the Netherlands, pp. 23-32, Foris Publications, Dordrecht.

- - : 1990b, 'Particles and the Dative Alternation', ms., University of Leiden, Dikken, Marcel den and Ren6 Mulder: 1991, ~Double Object Scrambling', in Jonathan

David Bobaljik et al. (ed.), Papers from the Third Student Conference in Linguistics, pp. 67-82, MIT, Cambridge.

Emonds, Joseph E.: 1972, 'Evidence That Indirect Object Movement is a Structure Preserv- ing Rule', Foundations of Language 8, 546-561.

- - : 1976, A Transformational Approach to English Syntax, Academic Press, New York. : 1985, A Unified Theol y of Syntactic Categories, Foris Publications, Dordrecht.

Epstein, Samuel: 1990, ~Differentiation and Reduction in Syntactic Theory: A Case Study', Natural Language and Linguistic Theory 8, 313-324.

- - : 1991, Traces and Their Antecedents, Oxford University Press. Evers, A.: 1975, The Transformational Cycle in Dutch and German, Indiana University

Linguistics Club, Bloomington. Fraser, Bruce: 1976, The Verb-particle Combination in English, Academic Press, New York. Fukui, Naoki: 1986, A Theory of Category Projection and Its Application, Ph.D. Dissertation,

MIT, Cambridge. Green, Georgia M.: 1974, Semantics and Syntactic Regularity, Indiana University Press,

Bloomington. Greenberg, Joseph: 1963, 'Some Universals of Grammar with Particular Reference to the

Order of Meaningful Elements', in Joseph Greenberg (ed.), Universals of Language, pp. 73-113, MIT, Cambridge.

Gu6ron, Jacqueline: 1987, 'Clause Union and the Verb-particle Construction in English', Paper Presented at NELS 17.

Page 58: Object positions

634 KYLE JOHNSON

- - : 1990, 'Particles, Prepositions, and Verbs', in J. Mascar6 et al. (eds.), Grammar in Progress, Foris Publications, Dordrecht.

Gu6ron, Jacqueline and Teun Hoekstra: 1990, 'Tense, Particles and Causatives', Paper Presented at Time Conference, MIT, Cambridge.

Haegeman, Liliane: 1988, 'Verb Projection Raising and the Multidimensional Analysis: Some Empirical Problems', Linguistic Inquiry 19, 671-684.

Haegeman, Liliane and Henk van Riemsdijk: 1986, "Verb Projection Raising, Scope, and the Typology of Verb Movement Rules', Linguistic Inquiry 17, 417-466.

Hankamer, Jorge: 1973, 'Unacceptable Ambiguity', Linguistic Inquiry 4, 17-68. Herslund, Michael: 1986, 'The Double Object Constructon in Danish', in Lars Hellan et al.

(eds.), Topics in Scandinavian Syntax, Reidel, Dordrecht. Higginbotham, James: 1985, 'On Semantics', Linguistic Inquiry 16, 547-594. Holmberg, Anders: 1986, Word Order and Syntactic Features, Stockholm. Huang, James: 1982, Logical Relations in Chinese and the Theory of Grammar, Ph.D.

Dissertation, MIT, Cambridge. Jackendoff, Ray: 1972, Semantic lnterTretation in Generative Grammar, MIT Press, Cam-

bridge. - - : 1977, X ' Syntax, MIT Press, Cambridge. Jaeggli, Osvaldo and Nina Hyams: 1989, 'On the Independence and Interdependence of

Syntactic and Morphological Properties: English Aspectual come and go', ms., USC and UCLA.

Johnson, Kyle: 1988, 'Clausal Gerunds, the ECP and Government', Linguistic Inquiry 19, 583-609.

Kayne, Richard: 1984a, Connectedness and Binary Branching, Foris Publications, Dord- recht.

- - : 1984b, 'Principles of Particle Constructions', in Jacqueline Gu6ron et al. (eds.), Grammatical Representation, Foris Publications, Dordrecht.

- - : 1989a, 'Null Subjects and Clitic Climbing', in Osvaldo Jaeggli et al. (eds.), The Null Subject Parameter, pp. 239-262, Kluwer Academic Publishers, Dordrecht.

- - : 1989b, 'Facets of Romance Past Participle Agreement', in Paolo Benincfi (ed.), Dialect Variation and the Theory of Grammar, pp. 85-104, Foris Publications, Dord- recht.

Koopman, Hilda: 1989, 'The Structure of VP in Dutch', Paper Presented at the Summer Syntax Workshop, University of California, Irvine.

- - : 1990, 'The Syntactic Structure of the Verb Particle Construction', ms., UCLA. Koster, Jan: 1975, 'Dutch as an SOV Language', Linguistic Analysis 1, 111-136. - - : 1978, 'Why Subject Sentences Don't Exist', in Samuel J. Keyser (ed.), Recent

Transformational Studies in European Languages, Linguistic Inquiry Monograph, no. 3, MIT Press, Cambridge.

Larson, Richard: 1988, 'On the Double Object Construction', Linguistic Inquiry 19, 335- 392.

Lasnik, Howard: 1981, 'Restricting the Theory of Transformations', in Nobert Hornstein et al. (eds.), Explanation in Linguistics, Longmans, London.

Lasnik, Howard: forthcoming, 'Case and Expletives: Notes Towards a Parametric Account', Linguistic Inquiry.

Lasnik, Howard and Mamoru Saito: 1984, 'On the Nature of Proper Government', Linguistic Inquiry 15,235-290.

le Roux, Cecile: 1988, On the Interface of Morphology and Syntax: Evidence from Verb-parti- cle Combinations in Afrikaans, Ph.D. Dissertation, Stellenbosch Papers in Linguistics, No. 18, University of Stellenbosch, Stellenbosch.

Levin, Beth and Malka Rappaport: 1986, 'The Formation of Adjectival Passives', Linguistic Inquiry 17, 623-661.

Page 59: Object positions

OBJECT POSITIONS 635

McConnell-Ginet, Sally: 1982, 'Adverbs and Logical Form', Language 58, 144-184. McNulty, Elaine: 1988, The Syntax of Adjunct Predicates, PhD Dissertation, University of

Connecticut, Storrs. Oehrle, Richard: 1976, The Grammatical Status of the English Dative Alternation, Ph.D.

Dissertation, MIT, Cambridge. Palmer, F. R.: 1968, A Linguistic Study of the English Verb, University of Miami Press. Pesetsky, David: 1989, 'Language-particular Processses and the Earliness Principle', ms.,

MIT, Cambridge. Pollock, Jean-Yves: 1989, 'Verb Movement, UG and the Structure of IP', Linguistic Inquiry

20, 365-424. Postal, Paul: 1974, On Raising, MIT Press, Cambridge. Poutsma, H.: 1928, A Grammar of Late Modern English. Part I, Groningen. Riemsdijk, Henk van: 1982, A Case Study in Syntactic Markedness, 2nd ed., Foris Publi-

cations, Dordrecht. Rizzi, Luigi: 1981, 'Nominative Marking in Italian Infinitives and the Nominative Island

Constraint', in Frank Heny (ed.), Binding and Filtering, pp. 129-158, MIT Press, Cam- bridge.

- - : 1990, Relativized Minimality, MIT Press, Cambridge. Roberts, Ian: 1987, The Representation of Implicit and Dethematized Subjects, Foris Publi-

cations, Dordrecht. - - : 1991, 'Excorporation and Minimality', Linguistic lnqui O, 22, 209-218. Rochemont, Michael: 1978, A Theory of Stylistic Rules in English, Ph.D. Dissertation,

University of Massachusetts, Amherst. Ross, John: 1967, Constraints on Variables in Syntax, Ph.D. Dissertation, MIT, Cambridge. Safir, Kenneth: 1985, Syntactic Chains, Cambridge University Press, Cambridge. Selkirk, Elisabeth O.: 1982, The Syntax of Words', MIT Press, Cambridge. Shlonsky, Ur: 1987, Null and Displaced Subjects, Ph.D. Dissertation, MIT, Cambridge. Simpson, Jane: 1983, 'Discontinuous Verbs and the Interaction of Morphology and Syntax',

in Michael Barlow et al. (eds.), Proceedings of the West Coast Conference on Formal Linguistics, 2, Stanford Linguistics Association.

Speas, Margaret: 1986, Adjanctions and Projections in Syntax, Ph.D. Dissertation, MIT, Cambridge.

Sportiche, Dominique: 1988a, 'A Theory of Floating Quantifiers and Its Corollaries for Constituent Structure', Linguistic Inquiry 19, 425-449.

- - : 1988b, 'Conditions on Silent Categories', ms., UCLA. Stillings, Justine: 1975, 'Gapping in English and Variable Types', Linguistic Analysis 1,247-

274. Stowell, Timothy: 1981, Origins of Phrase Structure. Ph.D. Dissertation, MIT, Cambridge. - - : 1983, 'Subjects Across Categories', The Linguistic Review 2,285-312. - - : forthcoming, 'Small Clause Restructuring', in Robert Freidin (ed.), Principles and

Parameters in Comparative Grammar, Robert Freidin (ed.), MIT Press, Cambridge. Taraldsen, Tarald: 1983, Parametric Variation in Phrase Structure: A Case Study, Ph.D.

Dissertation, University of TromsO. Vikner, Sten: 1987, 'Case Assignment Differences Between Danish and Swedish', in R.

Allan et al. (ed.), Proceedings of the Seventh Conference of Scandinavian Studies in Great Britain, pp. 262-81, University College, London.

- - : 1990, Verb Movement and the Licensing of NP-positions in the Germanic Languages, Ph.D. Dissertation, Universit6 de Gen~ve.

Wasow, Tom: 1977, 'Transformations and the Lexicon', in Peter Culicover et al. (eds.), Formal Syntax, pp. 327-60, Academic Press, New York.

Wexler, Kenneth and Peter Culicover: 1981, Formal Principles of Language Acquisition, MIT Press, Cambridge.

Page 60: Object positions

636 KYLE JOHNSON

Zaenen, Annie, Joan Maling and Hoskuldur Thr~iinsson: 1985, 'Case and Grammatical Functions: The Icelandic Passive', Natural Language and Linguistic Theory 3, 441-483.

Received 26 November 1990 Revised 30 May 1991

Linguistics Department 1168 Van Hise Hall 1220 Linden Drive University of Wisconsin-Madison Madison, WI 53706 USA kbj @WISCMACC.bimet