wh-dependencies without movementling.umd.edu/~bradl/publications/wh sans movements.pdf · mation on...

43
WH-DEPENDENCIES WITHOUT MOVEMENT BRADLEY LARSON [email protected] UNIVERSITY OF MARYLAND 1. Introduction In this paper I argue that some wh-question constructions in which wh-phrase find themselves in left-peripheral positions are not in fact derived via movement trans- formations. As is well known, wh-question dependencies (and a-bar dependencies in general) display certain constraints and abilities. If we take these constraints and abilities to be indicative of wh-movement dependencies, then wh-questions that fail to display these ought not be considered dependencies of the same type. In particular, I argue that these non-movement wh-dependencies are found in coordinated wh-questions (CWH) questions like in (1). (1) What and when did Iris eat? In (1), the first word (‘what’) does not seem to show the signature of having moved from the complement position of the verb. I argue for a what-you-see-is-what-you-get analysis of this initial wh-word: it is only syntactically present in its overt position. Instead of by means of a syntactic representation of the wh-word in the pre- movement position, I argue that the interpretation of the relevant wh-words is guided through semantic reconstruction. 1

Upload: others

Post on 27-Mar-2020

22 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

WH-DEPENDENCIES WITHOUT MOVEMENT

BRADLEY LARSON

[email protected] UNIVERSITY OF MARYLAND

1. Introduction

In this paper I argue that some wh-question constructions in which wh-phrase find

themselves in left-peripheral positions are not in fact derived via movement trans-

formations. As is well known, wh-question dependencies (and a-bar dependencies in

general) display certain constraints and abilities. If we take these constraints and

abilities to be indicative of wh-movement dependencies, then wh-questions that fail

to display these ought not be considered dependencies of the same type.

In particular, I argue that these non-movement wh-dependencies are found in

coordinated wh-questions (CWH) questions like in (1).

(1) What and when did Iris eat?

In (1), the first word (‘what’) does not seem to show the signature of having moved

from the complement position of the verb. I argue for a what-you-see-is-what-you-get

analysis of this initial wh-word: it is only syntactically present in its overt position.

Instead of by means of a syntactic representation of the wh-word in the pre-

movement position, I argue that the interpretation of the relevant wh-words is guided

through semantic reconstruction.

1

Page 2: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

This will be made clearer and more explicit later. First some background infor-

mation on how such non-movement wh-questions are to be derived.

2. History

In this section I rehash a few of the differences between current and former mod-

els of generative syntax, focussing on interpretation and structure building. These

differences will provide the necessary background to deriving wh-questions without

relying on movement.

2.1. Interpretation. In the extended standard theory of generative grammar (EST)

(Chomsky 1965) wh-movement transformations were of a quite different type than

what is the state of the art today. First, there was a representational level of ‘deep

structute’ off of which thematic interpretation was read (2).

(2) [Ivy [fixed what]]

In the above representation, ‘what’ is interpreted as the internal argument of ‘fixed’

(the fixee) and ‘Ivy’ the external (the fixer). This representation would feed a trans-

formation into that would move the wh-word from its deeply embedded position to

a less embedded position (3).

(3) [what [Ivy fixed]

At this level of representation (‘surface structure’) such things as topic-comment

interpretations were determined. Note that since the thematic interpretation was by

this point already set in stone, there was no need to represent the wh-word in its

base position.

2

Page 3: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

This model of wh-movement is of course no longer in fashion. Nowadays it is

common practice to work under a model in which all semantic information is read

off the final derived representation, not just topic-comment information but also

thematic relations (Chomsky 1993,1995).1 In order for this to possible, it is necessary

that wh-words be somehow represented in their base positions. That is, roughly the

same representation as found in (2) feeds a wh-movement operation that results in

(4).2

(4) [what [Ivy [fixed what]]]

From this representation, the thematic relationship between verb and internal argu-

ment is preserved throughout the derivation and any information relevant to focus or

presupposition is also extant. Further, the lower position will not be pronounced and

as such the overt word order will be something like (5) (See Nunes 2001 for details).

(5) ‘what’ > ‘Ivy’ > ‘fixed’.

The descriptive results of the two approaches are identical in this constrained case.

The takeaway point is that in the current theory the correct interpretation of the

moved wh-word seems to require some sort of maintenance of its pre-movement

representation. The EST did not require this.

What I will show is that the current theory can accommodate the correct the-

matic interpretation of a wh-word without it having a syntactic representation in

the position where it would acquire that thematic interpretation.1 Of course not everyone believes only this. Some (Uriagereka 1999 and others) take semanticinformation to be read off the syntactic derivation multiple times throughout the derivation, notjust at the end.2I will assume the copy theory of movement (Chomsky 1995) here (though others would surelysuffice).

3

Page 4: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

2.2. Structure Building. Prior to the early nineties, structure building operations

necessarily preceded movement operations. Either phrase structure rules (Chomsky

1957) or later X-bar formulae (Chomsky 1970, Jackendoff 1977) would create a deep

structure like in (6)

(6)

Ivyfixed what

Transformations would then apply to the fully formed phrase structure, resulting

in a representation like that in (7) which was in turn transferred to the levels PF

and LF.

(7)

what ...

Ivyfixed what

The differences currently is that there are no longer the representational levels

of deep and surface structure. As such, structure building and movement trans-

formations can intermingle derivationally. In order to capture the parity between

structure building and movement transformations, the concept of merge was intro-

duced (Chomsky 1995). Merge, represented in (8), takes two syntactic elements and

combines them such that they create a new element that is also subject to merge.

(8) Merge(x,y) ⇒ {x,y}x

4

Page 5: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

Merge can take two entirely unrelated objects and combine them as shown above,

external merge. It can also merge an element that is embedded within another to that

self same element. That is, given a structure like that in (9), it is possible to merge

the embedded ‘y’ to the ‘q’-phrase that encompasses it, thus creating something like

(10).

(9) {q,{z,{x,yi}x}x}q

(10) {yi,{q,{z,{x,yi}x}x}q}q

This gives us exactly the effect of movement, yet still in terms of generalized structure

building.

The important point here is that the introduction of the a wh-word into its surface

position is now no longer a different type of operation from introducing it as in the

internal argument of a verb. So long as the more deeply embedded structure of the

derivation is such that it licenses the merger of the wh-word, the derivation ought to

work.

3. Deriving Syntactic Gaps

Imagine a language in which wh-question formation works like this: A wh-word

moves to the front of a clause and leaves behind an X (11). All other aspects of this

language are identical to English.

(11) a. Ivy ate something

b. What did Ivy eat X?

5

Page 6: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

In this reality, deriving (11b) could involve a movement transformation that leaves

behind the X, much like English wh-movement transformations leave behind gaps.

It would be prudent, however, to determine whether there were any other way to

derive the question construction. With the X there is novel entity that needs to be

accounted for. Perhaps the derivation does not involve movement. For example, by

merging the X in low as the complement to the verb and merging the wh-word higher

up in the structure, one could derive the same word order and interpretation. This is

of course logically possible and it is either a licit derivation or not. So much should

be obvious.

But back to reality and wh-movement does not leave behind an X but rather

negative space, a gap. This being this case, and given the history of the field, it is

perhaps more understandable why the non-movement possibility was not explored

as a non-mutually exclusive alternative for wh-movement. But given the state of the

field now with respect to structure building, there is no excuse not to explore such

an analysis.

In this paper, such an exploration is carried out with respect to wh-question forma-

tion. In the subsection below, I offer control constructions as an analogous instance

of gap derivation in which both tacks have been taken.

3.1. Control. With respect to control constructions (12), two means of gap creation

exist in the literature.One way is to code into the grammar a means to avoid the

overt expression of the base position of a moved element (Bowers 1973, OÕNeil 1995,

and Hornstein 1999 among others) (13).

(12) Ivan wanted [gap] to leave.

6

Page 7: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(13) Ivani wanted Ivani to leave.

Another way is to posit the introduction of null elements into the derivation. This

is followed by the introduction of the relevant antecedent in the form of the matrix

subject in this case (See Chomsky 1981, Landau 2001 and others) (14).

(14) Ivani wanted PROi to leave

Ignoring the arguments for adopting either of the above proposals, the distinction

is important with an eye towards wh-movement. Again in wh-movement, there is a

gap that is interpreted as related to an antecedent (15).

(15) What did Ivan fix [gap]

Most theorizing of such constructions has taken the movement-thoery tack. That is,

the wh-word moves from the lower position into the higher one (16).

(16) Whati did Ivan fix whati?

But an equally logically possible approach is that of the PRO theory: The antecedent

to is base generated in a different position from the gap. This is essentially the

approach taken here (sans the postulation of a PRO analogue). The derivation of

wh-questions may be possible via direct base generation of the wh-phrase into the

spec,CP position (17).3

3There is another logical possibility for deriving the gap in (): there is simply nothing there syn-tactically. There is no subject represented in the embedded clauses (17).

(17) Ivan wanted to leave

There are to be sure many reasons to doubt this as a means of deriving control sentences, butit remains a logical, if deeply flawed, possibility. When explored with respect to wh-movementhowever, the derivation of a gap without movement does not seem so far fetched.

7

Page 8: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(18) Whati did Ivan fix?

In the next section I show how this might be possible without falling afoul of syntactic

restrictions.

3.2. Syntactic Leeway. As noted earlier, there is no longer any difference in current

theory between merging a wh-word into a spec,CP position as its first entry into the

derivation and merging a wh-word into that position after having merge a copy of

it elsewhere. They are both instances of merge, well-formedness is a concern for

another part of the grammar.

This being the case, it ought to be possible to merge a wh-word for the first time

into the spec,CP position of the following structure:

(19) CP

what ⇒merge C

C TP

Ivy T

T V

fix

Given a few other operations, the structure above could easily produce the sentence

in (20) below:

(20) What did Ivy fix?

The structure may be ruled out at the interfaces or earlier, but the merge itself is

surely licit. It only differs from the traditional conception of wh-movement in that8

Page 9: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

there is no unpronounced copy of the wh-word elsewhere in the structure. Further,

the presumed Markovian nature of the derivation should make this information in-

accessible when merging in the wh-phrase.

However, there is something that prior to the introduction of the wh-word that

potentially marks the sentence in (20) as unacceptable independent of the wh-word.

The verb ‘fix’ in English obligatorily requires an internal argument. That is, (21) is

unacceptable:

(21) *Ivy fixed.

Some verbs in English work differently and do not require an internal argument, like

‘eat’:

(22) Ivy ate.

This being the case, it should be possible to derive a sentence whereby the wh-word

enters late so long as the verb optionally takes an internal argument. In other words,

a sentence like (23) ought to be derivable in the way that (24) suggests. In this

sentence, the subcategorization frame of the verb is respected.

(23) What did Ivy eat?

9

Page 10: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(24) CP

what ⇒merge C

C TP

Ivy T

T V

eat

Another potential roadblock arises that might preclude the above representation

from being syntactically well-formed. Under most theories of phrase structure, the

wh-word ‘what’ would be taken to be a noun phrase of sorts and thus require abstract

case. Case assigners in contemporary theory are finite T-heads as well as most verbs.

The wh-word in question is not in a relevant position in the structure above to receive

case from either of these sources. In the following subsection I will argue that these

concerns turn out not to be relevant. ex-situ wh-words do not require case.

3.3. What case? As noted above, even if a wh-word can be first-merged into the

spec,CP position in such a way that might be in many ways licit syntactically, it

would still be missing abstract case. As such, the case-less wh-word ought to mark

the derivation as being ill-formed.

In this subsection I argue that this qualm is misguided and that wh-words that do

not reside in-situ do not require abstract case.4

4Note that not all wh-words ought to be considered noun phrases. Wh-words like ‘why’, ‘where’,and ‘when’ are hard to justify as being noun phrases. So if it turned out to be the case that somewh-words did in fact require abstract case, non-movement wh-questions should still be derivable aslong as the wh-word in question is one of these

10

Page 11: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

It seems fairly clear that in-situ wh-words (by which I mean wh-words that are not

found in spec,CP) do indeed require case. They pattern exactly like regular nouns

in their distribution with respect to case:

(25) a. Which teacher said {who/Iris} seemed to be the best?

b. *Which teacher said it seemed {who/Iris} to be the best?

It is also almost unavoidable that licit ex-situ wh-words (wh-words in spec,CP) stem

from a position in which abstract case was acquirable. The wh-word in (26) could

have gotten case from being in the spec,TP position, the wh-word in (26) could have

gotten case from being the complement to the VP.

(26) a. Whoi [TP whoi likes Autumn]?

b. Whati does Ivy [V P like whati]?

It is not possible to determine whether these wh-words required case. It could have

been the case that they were in case positions and that is what in part makes the

derivations legitimate. It is equally possible given the facts that the wh-words simply

never needed case to begin with and being in a case position at some stage of the

derivation was irrelevant.

11

Page 12: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

So it seems that to be rather difficult to determine whether ex-situ wh-words

require abstract case. If they remain in-situ, they seem to require it, but if they

move it is hard to say.5

To test whether the spec,CP wh-words require case would call for an instance

where case is not assigned to the position from which wh-movement stems from.

Case is not assigned by tenseless T-heads to their specifiers, as evinced in (27).

(27) It is certain Ivan to win

Perhaps if wh-movement stemmed from the spec,TP position in the above sentence,

the sentence would be acceptable due to wh-words not requiring case. This turns out

not to be the case, as seen in (28). This does not necessarily show that the moving

wh-word requires case as I will argue below.

(28) *Who is it certain to be at the party?

First, Lasnik and Frieden 1981 argue that it is crucially wh-traces that require case,

not the wh-word itself. They note that in the case theory developed in Chomsky

1980 (following inspiration from Vergnaud) lexically realized NPs require case. They

take the overt wh-operator in (28) to work identically to an overt relative operator

like in (29). The sentences are both unacceptable, presumably due to the fact that

the wh-word does not receive case.5It may seem more elegant to say that if a wh-word in one position requires case then wh-words inany position require case. But it could be the case the in-situ wh-words in multiple wh-questionsin English deal with abstract case in a manner different than that of spec,CP wh-words, Along thelines of Marantz 1991. In that system accusative case is assigned due to the fact that a higher nounphrase received nominative case. The lower noun phrase is ‘dependent’ on the case of the uppernoun phrase and it is assigned in a different way. it could be that the in-situ wh-words case isdependent on there being a wh-word in spec,CP, one that doesn’t get case itself.

12

Page 13: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(29) *The man who it is certain to be at the party...

However, when the relative operator is made covert, the sentence is still bad (30). It

cannot be the case that there is a constraint on the overt operator, relative or wh-.

Rather, they argue that the trace of the movement is what requires case.

(30) The man it is certain to be at the party...

The approach to wh-questions suggested here eschews all traces and their modern

analogues. Since there are no wh-traces to acquire or fail to acquire case, the fact

that the wh-word does not acquire case is orthogonal to the issue.

The case against the wh-word requiring case can also be made in a different way.

First, the sentence above is quite reminiscent of super raising constructions in Eng-

lish, which are ungrammatical for reasons not solely stemming from case considera-

tions:

(31) *Ivan seems that it is certain to be at the party.

Setting aside this for a moment, there is another construction in which case is not

assigned to the position from which wh-movement stems from. Take (32). It seems

that the embedded subject, Ivan, receives case neither from the tenseless head nor

from the matrix verb.

(32) *Ivy assured Iris Ivan to be at the party.

If there is wh-movement from Ivan’s position we could avoid the resemblance to super

raising. When this is done, the sentence is surprisingly acceptable (Postal, 1974).

(33) Who did Ivy assure Iris to be at the party?

13

Page 14: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

Kayne (1980) takes the above sentence to suggest that the wh-word can receive case

from the matrix verb when it moves into the embedded spec,CP. But there is just

as much evidence suggesting that the wh-word does not require case at all. Matrix

verb to subordinate spec,CP is a rather unexpected relationship in which to assign

case.

Recall our super raising-like construction from above in () (repeated here as (34)).

If its ungrammaticality stems solely from a lack of case, it ought to be rescuable by

getting case from a verb like ‘assure’. This turns out not to be so as seen in (35).

(34) *Who is it certain to be at the party?

(35) *Who did Ivy assure Iris that it is certain to be at the party?

This suggests that the ungrammaticality of (35) is not due to lack of case on the overt

wh-word or one of its traces, but rather something else entirely. For some reason it

is not licit to wh-move a subject across an expletive in a position the subject could

have moved to.6

The results of this study allow for a conception of the grammar in which ex-

situ wh-words do not require case.7 This is an admittedly unintuitive result, but

consistent with the data. Assuming case to be irrelevant to these wh-words allows

us to insert a wh-word late into a non-case position without pain of Case-related

ungrammaticality.

6The construction in () is also not an island to all wh-movement. As seen below in (i), extractionof a complement from the most embedded position is licit.

() What did Jake assure Jim that Joey is certain to win?7Another, seemingly indistinguishable, logical possibility is that all moving wh-words get case inthe spec,CP position.

14

Page 15: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

3.4. Summary. What we have seen in this section is that inserting a wh-word into

a derivation late does not run afoul of anything syntactically. It is possible to derive

the lower gap from which the wh-word only apparently moves from. Further, it seems

that moving wh-words do not require abstract case in the relevant ways to mar the

derivation. That is, inserting the wh-word initially into spec,CP will not cause the

derivation to be ruled out due to lack of case on the wh-word. Case is simply not

required in that instance.

If this conception of wh-‘movement’ is not ruled out syntactically, it ought to

be considered whether such representations are ruled out semantically. This is the

subject of the following section.

4. Deriving Gaps Semantically

Assuming it to be possible to derive a wh-question both without movement and

without violating any narrowly syntactic strictures, it behooves us to determine

whether the derived representations are legitimate semantically.

4.1. Presupposition and event variable restriction. It is certainly the case that

certain apparent gaps are licit semantically. Take for instance optionally transitive

verbs like ‘eat’. This verb can take an internal argument (36) and it can also be

followed by a gap (37).

(36) Ivy ate something

(37) Ivy ate

15

Page 16: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

The example in (37) suggests that the lack of internal argument is licit semantically.

This contrasts with obligatorily transitive verbs like ‘fix’. Seen below, such verbs

require internal arguments and are unacceptable without them:

(38) Ivy fixed something

(39) *Ivy fixed

It seems to be the case that verbs like ‘eat’ allow for the presupposition of an internal

argument. The sentence ‘Ivy ate’ presupposes that Ivy ate something. In other

words, sentence () entails sentence ().8 This is something that words like ‘fix’ do not

seem to do.

How then to account for the varying acceptability ratings between to the two

sentences lacking internal arguments? One potential solution is to adopt a neo-

Davidsonian approach to semantic composition and attribute the difference to the

lexical semantics or pragmatics of the given words. That is, both sentence () and ()

are grammatical with respect the semantics yet () is licit because of some proprietary

aspect of the verb.

Under a neo-Davidsonian approach, the sentences in () and () would have the

rough logical forms in (40) and (41) respectively:

(40) [[Ivy ate]] = ∃e{eat(e) & Agent(Ivy, e)}

‘there exists an event of eating whose agent was Ivy’

8There may be some slight difference between the two sentences. Some speakers take ‘Ivy ate’ toonly suggest that she ate a meal or its equivalent instead of any given thing.

16

Page 17: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(41) [[Ivy fixed]] = ∃e{fixed(e) & Agent(Ivy, e)}

‘there exists an event of fixing whose agent was Ivy’

This approach holds that both of the above logical forms are perfectly licit. It is

simply some extra-semantic notion that makes it such that a fixing event without an

explicit patient is odd, whereas an eating event without an explicit theme is fine.

This seems like the wrong idea however. It is not the case that the sentence

represented in (40) is not interpreted as having an internal argument. It is rather

the case that the internal argument is implicit or presupposed. We then would

like to somehow code this distinction into the semantics. Many researchers have

previously done exactly that: code presuppositional material into the restrictor of

quantifiers (see Berman 1991, Krifka 1993, and Herburger 2000 as proponents of such

an approach). Following these authors, I posit that instead of the logical form in

(40), a sentence like ‘Ivy ate’ has a logical form like that in (42).

(42) [[Ivy ate]] = ∃x[∃e: Theme(x,e)]{eat(e) & Agent(Ivy, e)}

‘there exists an x and there exists an event e whose theme is x such that the

event was an eating and its agent was Ivy.

Following Heim (1982) and Kamp (1984), I take indefinites to be variables that are

unselectively bound by existential operators. This existential operator binds into

the restrictor of the event variable in this case. The event variable is then restricted

to only those events that have something as their theme, which is what we want

with respect to verbs like ‘eat’. This is not the case with ‘fix’. There is no lexically

determined variable in the restrictor of the event variable in that case. This is to say

that the gap after ‘fixed’ is not licensed semantically, but the gap after ‘eat’ is.

17

Page 18: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

4.2. Wh-questions and event variable restriction. We saw in the previous sub-

section that the optional gap after verbs like ‘eat’ can be licensed by existentially

bound variable in the restrictor of an event quantifier. In this subsection I will adapt

that idea to a representation of wh-movement and show that wh-questions can be

formed without movement and yet be licit semantically.

The representation involves an existential quantifier that binds a variable. This

variable could presumably be bound by some other sort of operator as well, including

a wh-operator.9 As we saw in the section, wh-words can be introduced late and high

into the structure and as such it should be possible for that this wh-word can function

as the wh-operator that binds into the restrictor.

That is, the sentence in (43) can be represented at logical form as in (44).

(43) What did Ivy eat?

(44) [[What did Ivy eat]] = ?x[∃e: Theme(x,e)]{eat(e) & Agent(Ivy, e)}

‘for what x does there exist an event e whose theme is x such that the event

was an eating and its agent was Ivy.

The same cannot be said about ‘fix’. Seen below, without movement of the wh-word

(thus leaving behind a variable to bind) the wh-operation has nothing to bind. I

take such vacuous binding to be illicit semantically.

9Note that Chomsky 1982 argues for a similar configuration in parasitic gap constructions. Thewh-phrase that moves in the matrix clause can serve as the operator to a variable that the parasiticgap consists of. This is in principle no different that what I propose here: a wh-operator can binda variable that is not derived from movement from that position.

18

Page 19: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(45) [[What did Ivy fix]] = ?x∃e{fix(e) & Agent(Ivy, e)}

‘for what x does there exist an event e such that the event was an eating

and its agent was Ivy.

This contrasts with full wh-movement which would create the licit representation

in (46). The trace of the wh-word would be represented as the relevant variable,

though not necessarily in the restrictor, but in the nuclear scope.

(46) [[What did Ivy fix]] = ?x∃e{fix(e) & Agent(Ivy, e) & Theme(x,e)}

‘for what x does there exist an event e such that the event was an eating

and its agent was Ivy and its theme was x.

Thus is it not the case that sentences of the form ‘What did Ivy fix?’ are necessarily

going to be illicit. In fact, there is no discernible difference between the movement-

derived version of any wh-question and the non-movement version. The mean the

same things and are phonetically realized identically. In the next section, I explore

a potential means of contrasting the two.

4.3. Summary. In the previous two sections we have seen that it is in principle

possible to derive wh-questions without resorting to movement. The only constraint

seems to be that for this to be possible, it must be that the wh-word that is undergo-

ing the movement comes from a position where it would not necessarily be required

to arise overtly in a declarative sentence. That is, wh-questions without movement

are possible when the wh-word could have stemmed from the object position of a

verb like ‘eat’, but not a verb like ‘fix’.

19

Page 20: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

5. Empirical differences

In this section, I will address questions as to whether or not this logically and

theoretically possible novel derivation of wh-questions has any merit empirically. It

will be seen that answering this question is not as simple might be expected.

As we have already seen, there is little to no difference between the posited non-

movment derivations of wh-questions and the traditionally type. They have very

similar semantics (so long as the verb is of the right type) and they are string iden-

tical. Any differences that arise are going to stem from syntax internal signatures of

movement.

5.1. A false lead. An obvious signature of wh-movement is that of island sensitivity.

If it is truly the case that object-wh dependencies for verbs like ‘eat’ do not require

movement, it should be the case that we fail to see the effects of island constraints

in those instances. That is, a sentence like (47) ought to be more acceptable than a

sentence like (48).

(47) *What did Ivy meet a man who ate?

(48) *What did Ivy meet a man who fixed?

It should be apparent to the reader that there is in fact no asymmetry in the accept-

ability of the above sentences. They are both bad.10

10It is interesting to note that this approach to islands and null arguments makes the correctpredictions for single-wh questions though not for English. Italian (and other null subject languages)not only allows full null subjects but also seems not to be subject to island constraints when thegap is in a subject position of a finite clause (Rizzi, 1982). The analysis here seems applicable tosuch cases, but in a way that is not clear to me yet.

20

Page 21: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

This lack of differential acceptability however does not necessarily rule out the

possibility of movementless wh-questions. From a parsing perspective, it would be

incredibly odd for the sentence in (47) to be judged acceptable.

When the parser comes across the initial wh-word in (47), it posits a gap to follow.

I assume that the default dependency that the parser predicts is one of movement.

It is surely the the more common dependency for wh-words. At this point in the

parse, it is unknown whether the wh-word ‘what’ will form a dependency with a gap

in the subject or object position. Since it is the case that in English non-movement

wh-questions can only form as dependencies with object gaps of specific verbs, it

makes sense for the parse to assume that such a circumscribed set of conditions is

less likely and thus to be ignored. Positing movement is going to be the choice that

leaves more options open for successful parses in English.

So the parser assumes a movement dependency between the wh-word in (47) and

the upcoming gap. When the gap is reached, it turns out that the movement predic-

tion cannot have been correct because movement out of the island was not possible.11

But what to do at this point. It has turned out by the time the gap is parsed out

that movement could not have worked. The intended meaning of the utterance has

surely been discerned and there there is no motivation to carry out a costly re-parse.

As such, the sentences are simply judged unacceptable.

11It is likely not to be the case that the movement prediction will be shown to be false any earlierthan the gap site. Any island for movement that is parsed does not necessarily need to be thesource of the potential movement. That is, the string in () does not necessarily have to end withan island violation.

(i) What did Ivy make the claim that Iris...

It could instead end up being entirely licit:

(ii) What did Ivy make [the claim that Iris ate] for?21

Page 22: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

In order to truly test whether non-movment wh-questions are empirically viable,

it would be necessary to test sentences in which the impossibility of a movement

dependency is made clear very early on in the parse. In the next section, I investigate

such sentences.

5.2. What and When? At the outset of the this article I claimed that sentences

of a the sort in (49) involve non-movement wh-question formation.

(49) What and when did Ivy eat?

For such sentence the initial wh-word ‘what’ does not arrive at the left periphery via

movement, but rather base generation. The second wh-word ‘when’ arrives there via

normal wh-movement.

The impetus for this analysis comes from many places. First, movement to the

sentence initial position would require wh-movement in English to be freer that

previously expected. Second, the parsing issues raised earlier suggest that this con-

figuration is one that could allow a non-movement parse. Third, the first wh-word

does not display the empirical signature of wh-movement.

5.3. Strange Movement. English is not a multiple wh-movement language. Unlike

some languages, English does not allow sentences like in (50).

(50) What when did Ivy eat?

It would then be quite unexpected that both of the wh-words in the sentences at

hand (49) would involve both of the wh-words having moved into that position. Were

that the case it would be a novelty in English.

22

Page 23: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

Further, the type of movement that would be required would be novel as well.

Movement is taken necessarily to target the root of trees (Chomsky 1995).12 As

such the derived position of the wh-word should c-command its trace. Yet in these

instances, the wh-words do not seem to c-command their traces (51).

(51)

&did

Ivytj

eat ti

whati &

& whenj

There are presumably other problems with positing this sort of movement. This

is not supposed to be an exhaustive demonstration. Rather, it should be sufficient

to show that assuming movement for both of the wh-words in the relevant sentences

requires one to posit novel types of movement for English.

5.4. Parsing hints. As suggested above, the parser needs to be tipped off as to

the applicability of a non-movement dependency early on so as not to default to

a movement prediction. In these sentences such a tip-off exists. When parsing,

the initial wh-word is reached and the default movement prediction is activated.

However, once the immediately adjacent word ‘and’ is met it becomes clear that a

movement prediction cannot work. This wh-word is coordinated and thus cannot

c-command anything but the other coordinand (52).

12Other options such as ‘tucking-in’ Richards 1999 notwithstandings.

23

Page 24: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(52) CP

& ...

what &

& ...

It cannot c-command any potential gap. Given this deduction, the parser should not

posit a movement dependency. As we have seen though, the parser has the option

of positing a non-movement dependency. Revising this initial prediction should not

be that costly as only two words have yet been parsed.

A prediction is thus made. At the point when the string ‘what and’ is parsed,

it should only be possible for optionally transitive verbs like ‘eat’ to host the gap.

We determined earlier that only those verbs could allow non-movment wh-questions.

Since those verbs are the only ones that can handle such a dependency, it follows

that only they should be able to licit in a sentence that begins with ‘what and’. This

prediction is borne out (see Gracanin-Yuksek 2007 among others):

(53) What and when did Ivy eat?

(54) *What and when did Ivy fix?

This is a stark contrast compared to which verbs can usually host wh dependency

gaps. That is, any verb that takes a internal argument can host such a gap generally:

(55) What did you eat?

(56) What did you fix?

(57) What did you want?

24

Page 25: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(58) *What did you sleep?

But only optionally transitive verbs can host an internal argument gap in ‘what

and’ questions. This is suggestive of a odd constraint on wh-dependecies if not a

different sort of dependency all together.

5.5. Empirical signature of movement. I take syntactic movement to display

a signature driven by two things: island constraints and syntactic reconstruction

effects. In this subsection, I investigat whether the initial wh-word displays any of

these effects.

5.5.1. Islands. We saw above that the relevant island effects are impossible to discern

in regular, singular wh-movement cases. Unfortunately it turns out that a different

problem arises here.

It should be possible to construct a sentence like (59) and determine whether the

first wh-word is subject to the island constraint that threatens from below. If it has

not actually moved into that position, it ought not be subject to that island and

thus not be judged as bad as (60).

(59) *What and when did Ivy meet the man that ate?

(60) *What and when did Ivy meet the man that fixed?

The sentences are equally bad. It is near impossible to determine whether the island

violation was effected by the first wh-word or the second. Even when there is no

potential island violation, the coordinated wh-words must be interpreted as stemming

from the same clause:

25

Page 26: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(61) What and when did Ivy say that Ivan ate?

The above question can only be interpreted as inquiring about what Ivan and when

Ivan ate. It does not have an interpretation concerning what Ivan ate and when Ivy

said that Ivan ate. Because of this, it will rarely be possible to test a sentence where

the first wh-word is interpreted into an island and the second wh-word not.

The only islands that we are allowed to test are left-branch islands like the one

exemplified in (62).

(62) *Whose did Ivan steal book at the party?

(63) *How many did Iris meet people in Rome?

(64) *How tall did Ivy kiss a soldier on the cheek?

If the wh-phrases in the three sentences above we the first conjunct of a fronted

coordinated wh-question. The island effects ought to be ameliorated because, by

hypothesis, there has been no movement out of the island.13 This turns out to

indeed be the case, the sentences below show ameliorated island effects:

(65) Whose and when did Ivan steal books at the party?

(66) How many and when did Iris meet people in Rome?

(67) ??How tall and when did Ivy kiss a soldier on the cheek?

13The sentences must be altered such that the first coordinated wh-phrase is necessary forgrammaticality. That is, the example in () should not count against the hypothesis because () isnot acceptable either.

() *Whose and when did Ivan shake hand?

() *When did Ivan shake hand?

26

Page 27: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

Since the island effects seem to be obviated. There is one fewer reason to assume

that the first wh-word of these sentences arrives in its overt position via movement.

5.5.2. Reconstruction. Another effect of movement is that it is often possible to re-

construct the moved element in to its base position. That is, for a sentence like in

(68). It seems that the wh-phrase can be syntactically realized in the direct object

position so as to license the reciprocal.

(68) Which pictures of each other did the women like <which picture of each

other>?

Reconstruction is commonly take to arise because the wh-phrase in the left periphery

left a copy of itself behind when it moved from the verb-complement position. If the

wh-phase had simply been base-generated high, there would be no copy of it in the

lower position and reconstruction should not be possible.

If the first wh-phrase of the sentences we have been investigating shows reconstruc-

tion effects, that would be evidence of it having moved from a lower base position.

If not, then it is evidence against such an analysis.

Contrasting with the sentence in (68) above, the sentences below are less accept-

able. The lack of abject unacceptability is presumably due to the fact that meaning

is relatively clear. This suggests that the first wh-phrase is unable to reconstruct.14

(69) *?Which books about each other and when did the women read?

(70) *?Which book about himself and when did Ivan read?

14Sentences with wh-phrases of similar complexity are still acceptable:

() Which books about cats and when did the women read?

27

Page 28: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

When reconstruction makes for an ungrammatical sentence like in (71), we predict

that the analogous ‘what and’ sentence to be more acceptable. This seems to be the

case in (72).

(71) *Which book about Ivani did hei read?

(72) Which book about Ivani and when did hei read?

5.6. Summary. I have argued for three things in this section. First, given current

theory and basic facts about the number of wh-words that can move per clause in

English, movement does not seem to be possible for both of the wh-words in the

relevant sentences like in (73).

(73) What and when did Ivan sing?

Second, given that non-movement wh-questions have already be argued to be deriv-

able, the particular type of sentence discussed here uniquely guide the parser to posit

a non-movement dependency.

Third, the first wh-word in these sentences fails to show the signature of traditional

movement: island-sensitivity and reconstruction effects.

All three of these taken together (movement is not likely, non-movement is possible,

no movement is detected) I posit a (preliminary) representation for sentence (73)

below in (74).

28

Page 29: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(74)

&did

Ivytj eatwhati &

& whenj

In the following section I will summarize pervious accounts of this type of sentence

and discuss their shortcomings. In the section following that, I will offer a detailed

derivation of the above type of sentence and give a semantic representation for them.

6. Previous Accounts

In this section I outline two strains of previous accounts and argue that they are

both empirically inadequate. There has already been a fair amount of ilk spilled on

this type of sentence and the criticisms here are not exhaustive. In fact, the criticisms

here hinge mostly upon the novel island and reconstruction data presented herein.

The main approaches fall neatly into two groups: Bi-clausal analyses and mono-

clausal analyses. There are of course sub-varieties of such accounts.15 These are

different in interesting ways, but not in ways that are relevant to the criticisms

presented here.

6.1. Mono-clausal Accounts. Mono-clausal analyses posit that the relevant sen-

tence types are derived by movement of two wh-words into the left periphery of a

single clause (Zoerner 1995, Kazenin 2001, Skrabalova 2006, Gribanova 2008, and

Zhang 2007 among others). The result of this movement is either a coordinated

15See Gracanin-Yuksek 2007 and Citko and Gracanin-Yuksek 2010 for a hybrid, bi- and mono-clausal approach.)

29

Page 30: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

wh-phrase (75) or an instance of two specifiers of C separated by a spurious ‘and’

(76).16

(75) [CP [XP whati and whenj ] [C did Ivy eat ti tj ]]

(76) [CP whati [CP and [CP whenj [C did Ivy eat ti tj ]]]]

The relevant fact here is that both wh-movements are entirely typical. There is

nothing elided in either case nor any other reason to suspect that any of the usual

movement constraints ought not still apply.

The most obvious criticism is that languages like English do not allow multiple

overt wh-movement. Whatever disallows such dual movement generally would have

to be obviated here somehow. This would require that this obviation be proprietary

to the particular construction at hand: something to be avoided if at all possible.

The second problem with this type of approach is that, as we have seen in the

previous section, the first wh-phrase of coordinated wh-phrase sentence does not

show any island effects as shown in the sentence repeated in (77).

(77) How many and when did Iris meet people?

(cf. *How many did Iris meet people?)

This is unexpected under any mono-clausal account. The movement of the string-

initial wh-phrase is not theorized to be any different in and of itself than movement

generally. Again, there could be some proprietary property to these constructions

16It should be noted that the representation in () is very similar to what is argued for here. Thecrucial difference is that there string-initial wh-phrase has not movement into its surface position.

30

Page 31: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

that obviates the island constraints, but nothing of the sort has yet be proposed and

if the lack of constraints can be captured in a generalizable way, so much the better.

There are other problems with a mono-clausal account. See (People) for details.

6.2. Bi-clausal Accounts. Bi-clausal accounts posit that there are in fact two sep-

arate clauses, each involving wh-movement (Browne 1972, Banreti 1992, Whitman

2002 and others). The leftward clause subsequently undergoes IP-deletion. In other

words, such sentence are an example of backward sluicing. Such an account is

sketched in (78) below.

(78) [whati [IP did Ivan eat ti ]] and [whenj [did Ivan eat tj ]]?

This approach happens to make the opposite prediction with respect to the island

observance of the first wh-phrase. IP-ellipsis has long been know to obviate island

effects (Ross, 1969 among others) including the ones relevant for our discussion here.

As seen below, left branch islands are seemingly obviated by traditional sluicing:17

(79) I know that Ivan stole a book, but I can’t remember [CP whosei [IP Ivan

stole ti book]]

(80) I know that Iris met some people, but I can’t remember [CP [how many]i [IP

Iris meet ti people]]

17It does not seem that sluicing obviates all left branch islands. Shown below in (), certain degreequestion phrases cannot seem to avoid island constraints (See Kennedy and Merchant 2000 fordetails):

() ?*I know that Ivy kissed a soldier, but I don’t know [CP [how tall]i [IP Ivy kissed a ti soldier]]

This counts as an asymmetry between the facts of traditional sluicing and backward sluicing. Butthe backward sluicing example is somewhat degraded:

() ?How tall and when did Ivy kiss a soldier.

31

Page 32: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

However, this approach has other problems. Kazanin 2001 and others point out

that the un-elided versions of the examples would have different interpretations from

the elided versions. That is, if (81) is the elided version of (82), it is a mystery why

they have different interpretations.

(81) What and when did Ivan eat?

(82) What did Ivan eat and when did Ivan eat?

The sentence in (81) has an interpretation that more closely resembles that of (83)

where there is an overt pronoun in the second clause that is co-indexed with the first

clause’s wh-phrase:

(83) Whati did Ivan eat and when did Ivan eat iti?

However, if backwards sluicing is forced on the above sentence, the result is ungram-

matical:

(84) *What and when did Ivan eat it?

(viz. *Whati did Ivan eat ti and when did Ivan eat it?)

There are, again, other criticisms of such accounts but the one above is sufficient for

serious doubts. Traditional sluicing does not run into the same problems as backward

sluicing does here.

Yet whether or not the above argument holds is beside the point with respect to one

final problem that both approaches succumb to. In both types of previous accounts

there is movement of the intial wh-phrase. As such we expect to find reconstruction

32

Page 33: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

effects of the sort discussed earlier. Regular movement allows for reconstruction

effects and sluicing does not obviate them (85).

(85) I know that the women read some books about each other, I just can’t

remember which books about each other.

Both types of previous accounts predict reconstruction effects, but there do not seem

to be any reconstruction effects in the sentences at hand. The approach posited here

is the only one that predicts this.

In the next section, I offer a derivation of the relevant sentences that avoids the

shortcoming of the previous accounts.

7. Deriving Gaps

In the previous sections I have argued that certain long-distance wh-depedencies

are dot derived via movement but rather base generation (first merge) of a wh-phrase

into the left-periphery. In addition to this, I have presented argumentation that

current accounts of the construction are inadequate. In this section I got through

an example derivation of the type I am discussing. Following this I offer a logical

form of the resulting syntactic representation. I conclude with a discussion of the

cross-linguistics viability of this approach.

7.1. Sample Derivation. In this section I offer a simplified derivation of the sen-

tence in (86).

(86) What and when did Ivy eat?

33

Page 34: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

The initial step in the derivation is the merger of a suitable verb into the derivation.

That is, an optionally transitive verb is merged not with an internal argument, but

rather with a small v (87).

(87) Merge(read,v) ⇒ v

v read

There is no subcategorization clash that arises from failing to merge an internal

argument with the verb and the derivation can continue as it would otherwise.

The next crucial step in the derivation is the introduction of the first wh-word in

its low, base-generated position (88).18

(88) v

whenj v

v read

The derivation will continue in unexciting fashion until the C-head is merged and

the impetus for wh-movement of ‘when’ is introduced. The only extant wh-word in

the derivation will then move to the Spec,CP position (89).19

18The precise structural location of this wh-word is irrelevant for our purposes.19The particular featural motivation for this movement is also irrelevant here.

34

Page 35: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(89) CP+wh

C+wh

C+wh ...

... v

whenj v

v read

At this point the derivation has obeyed all syntactic strictures and ought to be

well formed, all else being equal. For such a structure, as has been argued earlier, it

is possible for the late (though not post-spellout) introduction of another wh-word.

7.2. Introduction of the second wh-word. Following Hornstein (2009) and Horn-

stein and Nunes (2008) in particular, there is a point in the derivation of a wh-

question in which the moved wh-word has not fully Merged with the matrix C-head.

These researchers break down Merge into two sub-operations: concatenate and label.

Concatenate first combines two syntactic objects into an unlabeled set. This set

does not act as (nor is it really) a constituent. As such the members of the set are

liable to undergo further concatenation operations. After concatentation, the set can

(adjuncts may, arguments must) undergo Label which takes one of the members of the

set and labels it with respect to another object that has been concatenated with. This

decomposition allows for a principled means of accounting for adjunctival behavior in

a bare phrase structure system (see Hornstein 2009 with respect to Chomsky 1995)

as well as free relative constructions (Hornstein and Nunes 2008). For want of space,

35

Page 36: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

assume this to be a possible means of construing merge (though there are other

means of accounting for the following steps in the derivation).

Taking this to be a possibility, there is a stage in the derivation of the above

sentence at which the wh-C concatenation has not yet been labeled and wh-word is

still open to further concatenation (90).

(90)

whenj C+wh

C+wh ...

... v

tj v

v read

At this point, another wh-word can concatenate with the moved wh-word.20 The

result, before labeling, would be something like that in (91).

20Larson (2010) argues that this sort of merger of like categories is what underlies coordination.The lexical item ‘and’ is an epiphenomenal grammatical formative under this view.

36

Page 37: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(91)

whati whenj C+wh

C+wh ...

... v

tj v

v read

As argued by Larson (2011) the the extension condition is irrelevant to the labeling

operation. As such, the concatenated relations sketched above can label in any

order. For concreteness let us say that the two wh-words undergo labeling first. The

arguement wh-word ‘what’ has adjoined to the adjunct wh-word ‘when’ and thus the

latter should project under labeling. This is shown in (92).

(92)whenj C+wh

C+wh ...

... v

tj v

v read

whati whenj

Following this, the head of the coordinated wh-phrase ‘when’ labels with the C-

head, as shown in (93).21

21The same effect can be achieved by adopting a sideward movement (Nunes 2001) analysis of themoving wh-word whereby the moving wh-word moves first to a different derivational workspace.There, the second wh-word adjoins to it before this new object merges into spec,CP.

37

Page 38: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

(93) CP+wh

whenj C+wh

C+wh ...

... v

tj v

v read

whati whenj

The above is the final representation of the sentence in (). IN what follows I offer

a simple logical form for the above representation.

7.3. Logical Form. As noted earlier, verbs like ‘eat’ presuppose internal objects

that are need not be represented explicitly in the syntax. Such presuppositions can

be semantically captured by introducing the presupposed elements into the restrictor

of the event variable. If the event is restricted to only those event in which there was

a relevant internal argument, we can capture the inference that there is a theme in

sentences like (94). The logical form of such a sentence would then look something

like (95).

(94) Ivan read.

(95) [[Ivan read]] =

∃x[∃e: Theme(x,e)]{read(e) & Agent(Ivan, e)}

‘there exists an x and there exists an event e whose theme is x such that the

event was an reading and its agent was Ivan.’

38

Page 39: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

A similar thing can be done with respect to the sentence type at hand. A sentence

like (96) can have the logical form in (97).

(96) What and when did Ivan read?

(97) [[What and when did Ivan read]] =

?x?t[∃e: Theme(x,e)]{read(e) & Agent(Ivan, e) & Time(t,e)}

‘for what x and for what t does there exist an event e whose theme is x such

that the event was an reading and its agent was Ivan and it took place at

time t.’

The argument question operator binds into the restrictor of the event quantifier and

the time question operator binds into the nuclear scope of the event quantifier. The

specifics of such an account are again not very important. The crucial idea is that

the argument question operator can bind a position that it did not move from in the

syntactic derivation. Instead, it can bind into the event quantifier’s restrictor.

7.4. Cross-linguistic Predictions. The data in this article so far has come entirely

from English. This was a fitting starting point given the difference in English between

obligatorily transitive verbs and optionally transitive ones. It seems to be the case

that in English it is possible to infer covert direct objects of particular verbs (like

‘eat’) but not others (like ‘fix’). This is by no means a deep fact about English.

Some exceedingly similar verbs have different transitivity requirements. With ‘eat’

it is possible to infer the direct object, with ‘devour’ it is not (98).

(98) *Iris devoured

39

Page 40: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

There are many other pairs of the same sort (‘finish’ and ‘bake’ versus ‘complete’ and

’fry’). The point is that we do not this different to be reflective of a deep repercussion

of the lexical semantics of the verb. That is, there should not be something about

the act of baking that differs from the act of frying that allows one to presume the

theme.

English could have been different in such a way that all the optionally transitive

verbs were obligatorily transitive and vice verse. Equally possible would be that

English verbs were all optionally transitive. Were this the case, we would expect

all direct objects to be able to be represented by the first wh-phrase in the relevant

sentences. We are able to test this hypothesis with Hungarian, a language in which

all verbs are optionally transitive.

As shown in (99) below, the verb repair does not require an internal argument

(data from Liptak 2011 and an informant).

(99) Javítottál

repaired.2sgmeg

PV(You repaired.)

(100) Hol

wherejavítottál

repaired.2sgmeg

PV(Where did you repair?)

Given that the internal argument of the ‘repair’ in Hungarian seems to be be pre-

supposed, it ought to be possible to create a ‘what and’ from such a verb, contrary

to the facts in English. This is the case (101).

(101) Mit

whatjavítottál

repaired.2sgés

andhol

wherejavítottál

repaired.2sgmeg?

PV(What and when did you repair?)

40

Page 41: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

The same ought to hold in language where the subject of finite clauses is not

obligatorily overt. In English, this is obviously not possible (102). Furthermore, it is

not possible for the relevant types of sentences to have the first wh-phrase be related

to the subject position (103).

(102) *Will sleep

(103) *Who and when will sleep?

However, in a language like Russian, it is possible (given the correct context) for a

finite clause to be missing a subject (data from Gribanova 2008) (104).22

(104) Budet

will.3sgspat.

sleep.infHe will sleep.

(105) Gde

wherebudet

will.3SGspat?

sleep.infWhere will he sleep?

I then predict that Russian ought to allow subjects of finite verbs to serve as the

first wh-phrase of the relevant sentence types. This turns out the be the case as seen

below in example (106).

(106) Kto

whoi

andgde

wherebudet

will.3sgspat?

sleep.inf(who and where will sleep?)

22I am not adopting the traditional little pro coding of subject drop. Instead I follow Alexiadou andAnagnostopoulou 1999 and posit that the agreement marking on the finite verb suffices to satisfythe EPP in these cases.

41

Page 42: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

Given the cross-linguistic suggestions above, it seems that there is a generalization to

be made. As long as the argument is optional generally, the first wh-word in can be

related to that missing argument. The presentation in the previous sections predicts

this. So long as the inferred argument can be seen as a presupposition that can be

coded in the restrictor of an event variable.23

7.5. Summary. In this section I have shown that the particular approach to wh-

questions presented here are plausible. They can be derived with minimal extra

assumptions and they can also be given a relatively straightforward semantic repre-

sentation.

8. Conclusion

In this article I have argued essentially two things. First, I have argued that is

theoretically possible to derive a wh-question that has the appearance of movement

but did not involve any actual movement from a base-position in the syntax. Wh-

phrases can optionally be introduced into derivations at the spec,CP position. Ex-

situ wh-phrases were argued to not require abstract case and argued to have the

ability to bind variables that arise only in the semantics.

Second, I argued that a certain type of sentence allows us to unearth derivations in

which ex-situ wh-phrases do not move into their positions. Such sentences (beginning

with something like ‘what and’) have been difficult to capture in the past and I

suggest that a non-movement account works for reasons theoretical and empirical.

The first wh-word in such sentences does not seem to be able to arrive in its overt23It is possible that not all covert arguments can or should be captured in this way. Thus, it couldbe the case that a language has a null argument possibility that does not correlate with allowing thecorresponding wh-phrase as the first conjunct. The converse however ought to hold. Any wh-phrasein the first conjunct ought to correspond to a position that is optionally left empty.

42

Page 43: WH-DEPENDENCIES WITHOUT MOVEMENTling.umd.edu/~bradl/publications/Wh sans movements.pdf · mation on how such non-movement wh-questions are to be derived. 2. History In this section

position by traditional means of movement. Further, it does not show any of the

telltale signs of having moved at all.

All of this is made possible by the radical re-conceptualization of the syntax that

the minimalist program carries with it.

9. Works Cited

43