Long-Distance Wh-Movement and Minimalism

Aaron F. Kaplan

Second Qualifying Paper

March 31, 2005

1 Introduction

Long wh-movement is a phenomenon by which an element appears to move directly from

an A- or A-position to a higher A-destination, ignoring weak island constr aints and bypass-

ing the intermediate landing sites that are characteristic of successive-cyclic wh-movement.

Cinque (1990) gives some examples of the contrast between long and successive-cyclic wh-

movements, illustrating the long-distance variety’s immunity to weak island eﬀects (the (a)

examples show successful long movement, and the (b) examples show failure of long move-

ment because the moved element requires successive-cyclic movement):

(1) Extraposition Island (= Cinque (1990:2), ex. (7))

a. To whom is it time to speak t?

b. *How is it time to behave t?

(2) Factive Island (= Cinque (1990:2), ex. (6))

a. To whom do you regret that you could not speak t?

b. *How do you regret that you behaved t?

These data indicate that some sort of island-defying movement is available in certain in-

stances and/or to certain elements. This paper focuses on the properties of long-movement

itself, leaving the issue of which elements may participate in this phenomenon for future

research. For present purposes, it is su ﬃcient to note that Cinque (1990) and Chung (1994,

1998) notice that only certain DPs are eligible to participate in a long-movement operation.

Cinque argues that the moved element in a long-movement construction must be “referen-

tial.” In Cinque’s terminology, this is a DP “that refer[s] to speciﬁc members of a preestab-

lished set” (Cinque 1990:8). It is D-linked (Pesetsky 1987), and, unlike non-referential DPs,

it can enter into coreference relationships. Chung points out some deﬁciencies with this pro-

posal, but it is a suﬃcient working analysis for now. I return to this issue in the conclusion.

Further, Cinque’s formal notions of antecedent government and binding have quite dif-

ferent incarnations in the Minimalist Program (Chomksy 1999). For example, antecedent

government, whose role is to enforce strict locality in successive-cyclic movement, is su-

perseded by the Phase Impenetrability Condition. Antecedent government is consequently

discarded in Minimalism. With the theoretical advances of Minimalist syntax, Cinque’s ideas

merit new scrutiny and revision. I discuss this topic in more detail in Section 7 and, with

respect to Minimalism and long movement generally, immediately below.

As I show in Section 2, there is strong evidence that long wh-movement does indeed bypass

the typical stopping-oﬀ points that characterize ordinary wh-movement. A DP that under-

goes long wh-movement appears to move directly from its A-position to its ﬁnal destination,

potentially moving across several clauses (see (3)). An operation like this is incompatible

with Minimalist assumptions. Since CPs, at the very least, are phases, a probe cannot locate

a goal that is embedded multiple clauses below the probe’s phase. This would violate the

Phase Impenetrability Condition, as (3) makes clear. We seem to have a choice to make:

either we deny the facts of long wh-movement, or we abandon the notion of phases that is

central to the Minimalist Program. This paper essentially takes the former approach by de-

veloping an analysis of long wh-movement in Chamorro (Chung 1998) that does not involve

movement across any long distance and is compatible with the Minimalist framework. The

analysis involves base-generation of a DP in what appears to be the ﬁ nal destination of the

movement operation, coupled with a short (i.e., non-long-distance) movement in the lower

clause from which movement appears to have occurred.

While the analysis below argues that there is in fact no unusually long movement involved in long-distance constructions,

I continue to call the operations involved “long movement” or “long-distance movement” for terminological consistency.

(3) CP

2 Wh-Movement in Chamorro

Chamorro is particularly interesting with respect to wh-movement because of its highly

detailed morphology, which provides a visible record of the path of A-movement.

This

morphology is a clear diagnostic for long-distance constructions b ecause each movement

operation is signalled morphologically. Where this morphology is lacking, the moved element

must have combined what would be a series of successive movements into a single operation.

Before discussing this morphology, I outline the facts of wh-movement in Chamorro.

Wh-movement in Chamorro has all the properties of wh-movement in other languages.

Chung (1998) shows that wh-constructions in Chamorro “contain a syntactic dependency

between a displaced constituent. . . and a gap” (p. 208), and that this dependency holds

across an unbounded distance and exhibits island and crossover eﬀects. Finally, the displaced

constituent surfaces in an

A-position.

Consider ﬁrst the constituent question in (4). Like all constitu ent questions in Chamorro,

this one has an interrogative phrase (what I call below a “wh-phrase”) at its left edge. Chung

identiﬁes this position as the speciﬁer of C

, a conclusion conﬁrmed in part by the presence of

Other languages with similar record-keeping morphology include Irish (McCloskey 1990), Moore (Ha¨ık 1990), and Palauan

(Georgeopolous 1985, 1991a,b).

Virtually all of the analysis in this section comes from Chung (1998) chapters 6–8.

an overt complementizer to the right of the wh-phrase in this example. t marks the position

in which the wh-phrase is expected to appear in declarative constructions; it marks the gap.

(4) Ginin

from

hayi

who?

Comp

un-konni’

agr-take

the

neni

baby

t ?

“From whom did you take away the baby?” (Chung 1998:209)

The wh-phrase and the gap form an A-dependency. The position of the gap must be

empty, as (5) sh ows. In (5b), the overt material in the position of the gap in (5a) causes un-

grammaticality. The displaced element must also meet the semantic and syntactic selectional

requirements imposed on the gap.

(5) a. Hayi

who?

ma’a’˜nao-mu

WH[obl].afraid-agr

t ?

“Who are you afraid of?” (Chung 1998:210)

b. *Hayi

who?

ma’a’˜nao

agr.afraid

hao

you

Obl

guiya?

him

(Who are you afraid of him?) (Chung 1998:210)

This dependency may occur across an unbounded distance. For example, in (6) several

clauses intervene between the initial displaced element and gap. This freedom is constrained

by the normal array of island and strong crossover eﬀects. (7) shows attempted extraction

out of a relative clause (7a) and out of an embedded question (7b). (8) shows that a pronoun

that c-commands the gap may not be coindexed with the gap.

(6) Manu

which?

isla

island

Comp

masangani

agr.Pass-say.to

hao

you

man-ansias

agr-anxious

siha

they

p¨ara

Fut

uma-muv

agr-move

siha

themselves

gu¨atu

over.there

t ?

“Which island were you told that they are eager to move to?” (Chung 1998:211)

(7) a. *Hayi

who?

siha

famagu’un

children

un-rispeta

agr-respect

[¨adyu

that

the

palao’an

woman

[ni

Comp

fuma’na’gui

WH[nom].teach

t]]?

(Which children do you respect the woman who taught t ?) (Chung 1998:211–212)

b. *Taim¨anu

how?

in-tingu’

agr-know

[hayi

who?

si˜na

can

chumo’gui

WH[nom].do

the

che’chu’

job

t ]?

(how do you (pl) know [who can do the job t ]?) (Chung 1998:212)

(8) Hayi

who?

malago’-˜na

WH[obl].want-agr

pro

j,∗i

Comp

un-na’facho’chu’

WH[obj].agr-make.work

“Who

does he

j,∗i

want you to hire t

?” (Chung 1998:212)

Finally, wh-movement creates the full range of expected constructions in Chamorro. Con-

stituent questions were illustrated in (4)–(6) above. (9) shows relative clauses, embedded

questions, and clefts, each of which exhibit the same array of properties outlined above.

(9) a. P¨ara

Fut

bai u-sugun

agr-drive

gu¨atu

over.there

[[¨anai

Comp

p¨ara

Fut

u-gupu

agr-ﬂy

si Maria

Maria

t ] na

lug´at].

place

“I’m driving to the place that Maria is ﬂying to.” (Chung 1998:214)

b. Un-tungu’

agr-know

[taotao

person

[O ni

Comp

not

interes´ao

agr.interested

yu’

p¨ara

Fut

bai u-tungu’

WH[obj].agr-know

t]].

“You know the people who I’m not interested in knowing.” (Chung 1998:218)

c. Esta

already

alas otchu

eight.o’clock

Comp

man-maigu’

agr-sleep

the

famagu’un

children

“It was eight o’clock when the children fell asleep.” (Chung 1998:228)

Nothing discussed so far is unusual. These wh-constructions manifest the expected prop-

erties. In the next two sections I discuss the morphology that accompanies wh-movement in

Chamorro. This morphology is a diagnostic for the presence of wh-movement, and therefore

an indicator of successive-cyclic movement.

2.1 Operator-C Agreement

In Chamorro, unlike in English for example, there is direct morphological evidence of the

path a moved item takes during its move. Two agreement relationships are associated with

wh-movement. The ﬁrst is what Chung (1998) calls Operator-C Agreement.

The normal complementizer morphology (that which appears when no wh-movement oc-

curs) is illustrated in the following examples. Complementizers normally alternate according

to various clausal properties: ﬁnite/nonﬁnite; interrogative/noninterrogative; root/embedded.

In non-ﬁnite clauses (and ﬁnite non-interrogative root clauses), the complementizer is null:

(10) Mu-mal¨agu’

agr-want

yu’

lokkui’

also

[[

]

Comp

chum¨agi

Inﬁn.try

mama’].

chew.betelnut

“I too came to want to try chewing the betelnut.” (Chung 1998:223)

In ﬁnite non-interrogative non-root clauses, the complementizer is either na or null:

(11) [

]

Comp

Ta-tungu’

agr-know

[na

Comp

gu¨aha

agr.exist

man-mafa˜nagu

WH[nom].agr-born

Comp

man-mo’na

WH[nom].agr-front

than

hita].

“We know that there were some born earlier than us.” (Chung 1998:223)

In ﬁn ite interrogative clauses, both root and non-root, the complementizer is kao. In the

root variety, it may also be null (see the bracketed clause):

(12) P¨ara

Fut

u-li’i’

agr-see

[kao

magahit

agr.true

Comp

u-fan-osgi]

agr-AP-obey

“(So) I could see whether it was true that he would obey” (Chung 1998:224)

Complementizer morphology in non-wh-movement constructions is summarized in (13).

(13) C Morphology in Non-wh-movement Clauses

Type of Clause C Morphology

[–ﬁnite] —

[+ﬁnite,–q,+root] —

[+ﬁnite,–q,–root] na/—

[+ﬁnite,+q,+root] kao/—

[+ﬁnite,+q,–root] kao

When C’s speciﬁer position is ﬁlled, this morphology yields to a diﬀerent system that

reﬂects properties of C’s speciﬁer (DP vs. PP; locativity; null vs. non-null). With a null

locative DP, the complementizer is realized as ¨anai (“O” is the null relative operator):

(14) Pues

dumimu

agr.kneel

[guihi

there

[O ¨anai

Comp

gaigi

agr.be

si tata-˜na

father-agr

yan

and

si nana-˜na

mother-agr

t]].

“So they (du) knelt there where his father and mother were.” (Chung 1998:226)

When this same DP is overt, the complementizer becomes na in the Guam dialect and

nai/ni in the Saipan dialect:

(15) Esta

already

alas otchu

eight.o’clock

Comp

man-maigu’

agr-sleep

the

famagu’un

children

“It was eight o’clock when the children fell asleep.” (Chung 1998:228)

With non-DP speciﬁers, C is again realized as na (Guam), or nai/ni (Saipan).

(16) Ginin

from

hayi

who?

Comp

un-risibi

agr-receive

katta

letter

“From whom did you receive a letter?” (Chung 1998:227)

The Operator-C Agreement morphology is summarized in (17).

(17) C Morphology in Wh-movement

Type of Speciﬁer C Morphology

[+N,–locat] —

[+N,+locat,+O] ¨anai

[+N,+locat,–O] na (Guam), nai/ni (Saipan)

[–N] na (Guam), nai/ni (Saipan)

In addition to the patterns summarized in (13) and (17), postnominal relative clauses have

their own Operator-C Agreement morphology. In such cases C reﬂects both the agreement

morphology and the morphology of the “linker” morpheme, which accompanies modiﬁers

in Western Austronesian languages (see Chung (1998) for details). Operator-C Agreement

surfaces according to (19) in these sentences. Some examples are given in (18).

(18) a. Adyik

choose

[un

problema

problem

[O ni

Comp

impottanti

WH[nom].agr.important

t p¨ara

hagu

you

yan

and

the

famagu’on-mu]].

children-agr

“Choose a problem that is important to you and your children.” (Chung 1998:233)

b. Adyugui’

there.is

the

chinina

shirt

[O ni

Comp

malago’-mu

WH[obl].want-agr

t]].

“There’s the shirt that you wanted.” (Chung 1998:233)

(19) Realizations of C in Postnominal Relative Clauses

Type of Speciﬁer C Morphology

[+O,–locat] ni

[+O,+locat] ¨anai (Guam), ni (Saipan)

Chung (1998) analyzes the morphological alternations in (14)–(16) as an instance of

speciﬁer-head agreement. As is apparent from the above examples, this kind of agreement

is a feature of wh-movement. (14) shows the agreement in a relative clause, (15) shows it in

a cleft sentence, and (16) shows it in a question. When wh-movement occurs across more

than one clause, whether by successive-cyclic movement or long-distance movement, only

the highest C—the one whose speciﬁer is the ﬁnal destin ation of movement—enters into this

agreement relationship. The lower complementizers show their normal, non-movement mor-

phology. An example is shown in (20), where complementizers are underlined. The highest

C shows the morphology from (14), and its speciﬁer is ﬁlled with the same null relative

operator from (14). The lower complementizer, with a trace in its speciﬁer, is realized as na,

the form used for a ﬁnite non-interrogative non-root clause (see (11)). If it agreed with its

speciﬁer, we would expect to see the same morphology that appears on the higher C.

(20) Taya’

agr.not.exist

kasamentu

marriage

ma-susedi

WH[nom].agr.Pass-experience

[guihi

there

ha’ani

day

[O ¨anai

Comp

hinasso-tta

WH[obj].think-agr

[t na

Comp

um-¨asagua

agr-marry

the

dos

two

t]]].

“No marriage sacrament occurred on the day when we thought they were married.”

(Chung 1998:229)

2.2 Wh-Agreement

Chung calls the second morphological phenomenon Wh-Agreement. Verbs along the path

of wh-movement acquire special inﬂections, reﬂecting the grammatical function of the gap left

by movement.

When this inﬂection is overt, it replaces the normal subject-verb agreement

morphology. In glosses below, WH[ ] (with case indicated in the square brackets) signals the

presence of Wh-Agreement morphology in one form or another.

When the predicate is transitive and realis, -um- is the inﬂectional inﬁx for a nominative

gap:

(21) Hayi

who?

chum¨atgi-n

WH[nom].laugh.at-L

m¨amaisa

self.Prog

gui’

him

t ?

“Who was laughing at himself?” (Chung 1998:237)

With an objective gap,

the predicate is optionally nominalized.

If the predicate is

transitive, then the inﬁx -in- must accompany nominalization. (22) illustrates both the

inﬁx and nominalization associated with objective agreement. (23) contrasts two questions

that are identical except that the ﬁrst one realizes Wh-Agreement via nominalization (and

inﬁxation) while the second one does not. Notice the diﬀerent morphology of the verbs

glossed as “say” and the oblique marker in the ﬁrst example.

(22) Hafa

what?

kin

ann´ono’-mu

WH[obj].eat.Prog-agr

t ?

“What are you eating?” (Chung 1998:237)

(23) a. Hafa

what?

si Maria

Maria

sinangane-n˜na

WH[obj2].say.to-agr

Obl

Joaquin

t ?

“What did Maria tell Joaquin?” (Chung 1998:242)

b. Hafa

what?

si Maria

Maria

ha-sangani

say.to-agr

si Joaquin

Joaquin

t ?

“What did Maria tell Joaquin?” (Chung 1998:242)

Finally, with an oblique gap, nominalization is required. The inﬁx -in- optionally appears

More accurately, in clauses that are not the lowest CP along the p ath of movement, Wh-Agreement reﬂects the grammatical

function of the CP out of which the moved item has moved.

Chung (1998) identiﬁes two objective cases, “object” (a direct object’s case) and “object2” (“the [c]ase of the oblique object

of a verb of transfer” (p. 237)). While the Wh-Agreement morphology is the same for both cases, I follow Chung in glossing

the agreement as either [obj] or [obj2] as appropriate.

A nominalized predicate exhibits possessor-noun agreement rather than the usual subject-verb agreement, “and its direct

object occurs in the oblique morphological case, not the unmarked morphological case” (Chung 1998:242).

Chung includes a Wh-Agreement gloss in (23b). I have removed this part of the gloss to emphasize that objective Wh-

Agreement is optional, and this example fails to make use of that option.

when the predicate is unaccusative. For example, the predicate in (24) is nominalized:

(24) Hayi

who?

mahalang-mu

WH[obl].lonely-agr

t ?

“Who are you lonely for?” (Chung 1998:238)

The agreement morphology is summarized in the table in (25).

(25) Wh-Agreement Morphology (from Chung 1994:8)

Case of Gap Agreement Realization

Nominative Replace any ergative agreement with -um-.

Obj, Obj2 Optionally nominalize. If the nominalized

[+V]

is transitive, insert -in-.

Oblique Nominalize. If the nominalized [+V]

unaccusative, optionally insert -in-.

Crucially, this agreement is required only on the lowest verb in a wh-const ruction. For

higher verbs along the path of movement, agreement is optional when the moved element is a

referential DP. In the basic case, this optionality is an all-or-nothing eﬀect: either agreement

appears on all of the non-lowest verbs, or it appears on none of them. (But see Section 6

for some apparent counterexamples to the all-or-nothing eﬀect.) Agreement is required on

all verbs along the path of movement for certain DPs, roughly corresp onding to Cinque’s

(1990) “non-referential” DPs as noted at the outset of this paper.

(26) shows a wh-question in which the moved item is a referential DP. Agreement appears

on the lowest verb, u-ma-fa’maolik, but not on the higher verb, mal¨agu’.

(26) Hafa

what?

patti

part

Loc

atumobit

car

mal¨agu’

agr.want

hao

you

[u-ma-fa’maolik

WH[nom].agr-Pass-ﬁx

t ]?

“Which part of the car do you want to be ﬁxed?” (Chung 1998:248)

In contrast, when a non-referential DP moves, agreement must appear on every verb:

(27) Hafa

what?

malago’-˜na

WH[obl].want-agr

si Magdalena

Magdalena

[t p¨ara

Fut

ta-chuli’

WH[obj].agr-bring

t]?

“What does Magdalena want us to bring?” (Chung 1998:249)

The two verbs in (27) show diﬀerent agreement morphology. As a ﬁrst approximation,

Wh-Agreement reﬂects the case of the closest lower gap. For the lowest verb in (27), the

agreement morphology is dictated by the case of the moved item’s A-p osition. Higher verbs

show agreement based on the trace in the speciﬁer of the immediately subjacent CP. The

higher verb in (27) takes its cues from the trace in the speciﬁer position of the embedded CP.

Although I argue for a diﬀerent conclusion in Section 4.2, moved items appear to acquire

new case features with each movement operation, and Wh-Agreement reﬂects this new case.

To be more precise, the morphology of the higher verb reﬂects the case assigned to the

immediately lower CP in whose speciﬁer the trace sits. Chung (1998) argues that the C

head of this CP inherits the case assigned to the CP, and it passes this feature to its speciﬁer.

Thus while the intermediate trace in (27) may be a trace of something that was originally

assigned objective case, the moved item acquires oblique case when it moves to the speciﬁer

of the lower CP. This is because the lower CP is itself oblique, and it assigns this oblique

case to its speciﬁer. The Wh-Agreement on the higher clause is oblique, r eﬂecting the case

assigned to the intermediate trace by the embedded C. This line of reasoning is discarded

below, but it is useful to think in these terms for the time being.

In Chung’s (1998) analysis, Wh-Agreement is agreement between T (In Chung’s terms I)

and a lower trace, and this agreement is manifested on the VP complement of T. For a trace

to trigger agreement, it must be free within this T’s maximal projection. This means that

only the highest trace within TP can trigger agreement because all lower traces are bound by

higher traces in TP. For example, in (28), t

binds t

, so t

may not trigger Wh-Agreement

on the highest TP. Likewise, t

binds t

, so t

cannot trigger agreement either. However, t

is free within the highest TP, so it triggers agreement.

(28)

Wh-Agreement provides a morphological indication that (at least some) wh-movement

in Chamorro is successive-cyclic. Wh-Agreement appears on a verb precisely because some

wh-moved item stops, at least temporarily, in the immediately lower clause. When wh-

movement does not bring any moved wh-phrase into a clause, no agreement appears on the

immediately higher verb. If all wh-movement in Chamorro were successive-cyclic, sentences

like those in (26) would be ungrammatical. A clause that appears to be along the path of

wh-movement fails to trigger the requisite agreement morphology on the higher verb.

Constructions like those in (26) are the focus of this paper. The data and the theoretical

assumptions of Minimalism appear to be incompatible. The DP H afa na patti gi atumobit

“which part of the car” in (26) cannot move from the right edge of the sentence (where its

trace is the object of the embedded verb) to the left edge of the sentence (i.e., the speciﬁer

of the higher CP) via successive-cyclic movement because such movement would trigger Wh-

Agreement on the highest verb: The wh-phrase would have to stop in the speciﬁer of the

lower CP, triggering agreement on the higher verb. On the other hand, since the highest C

is separated from its goal’s trace by at least one phase boundary (the CP that heads the

embedded clause), the movement operation that creates the question cannot occur in a single

step: the probe cannot locate the goal’s A-position.

In the following sections I resolve these contradictions. These are the central facts that

motivate my analysis: In (26), the lowest clause—the one containing the A-position trace—

shows signs of wh-movement. There is a gap, and Wh-Agreement appears. And in the higher

clause, a wh-phrase appears in C’s speciﬁer, triggering Operator-C Agreement.

As outlined in the introduction, I propose that the wh-phrase is base-generated in the

speciﬁer of the highest C. In the lower clause, an operator (whose identity is explored below)

moves from the position of the trace to the speciﬁer position of the embedded CP, triggering

Wh-Agreement in this clause. Before developing this analysis in greater detail, I discuss

successive-cyclic movement, from which the necessary theoretical constructs are drawn.

2.3 Successive-Cyclic Movement

Consider again the example in (27), repeated here:

(29) Hafa

what?

malago’-˜na

WH[obl].want-agr

si Magdalena

Magdalena

[t p¨ara

Fut

ta-chuli’

WH[obj].agr-bring

t ]?

“What does Magdalena want us to bring?” (Chung 1998:249)

The Wh-Agreement pattern, in the form of nominalization of both verbs, indicates that

successive-cyclic movement occurs here. The moved item passes through the lower CP’s

speciﬁer. How is this construction produced in the Minimalist Program? Notice ﬁrst the

two CPs, which must be phases. Th ere are other phases in this construction ( notably the

vPs), but they are not crucial to the current discussion (v’s import as a phase boundary is

addressed in S ection 4). I therefore make the simplifying assumption that only two phases

are involved in this construction. The phase that becomes the higher clause has the structure

in (30) (abstracting away from the nominalization of both verbs in (29): Chung does not

give the pre-nominalization forms of these verbs, so I show nominalization here even though

it hasn’t yet been triggered). This phase is incomplete: A verbal complement is missing,

and this argument will be supplied by the phase constituting the lower clause.

(30) CP

C TP

malago’-˜na si Magdalena

(want Magdalena)

Likewise, the phase that constitutes the lower clause has this structure before wh-movement

(again showing premature nominalization):

(31) CP

C TP

p¨ara

(Fut)

ta-chuli’

(bring)

hafa

(what?)

The DP hafa must move to the speciﬁer of the CP in (31). In the Minimalist Program,

this means that this DP must be an available goal for some probe. To be an active goal, it

must have an uninterpretable [uQ] feature.

Similarly, the probe has uninterpretable features

The [Q] feature indicates that its host heads a constituent question, or, in semantic terms, that its host is an interrogative

operator (see Section 5.1). Such a semantic property is interpretable on C (because C can be an interrogative operator) but

not on DP (because DP cannot be an interrogative operator).

that must be checked, and the DP checks these features. At least one of these features must

be strong so that in situ feature checking is not an option. Since hafa moves to C’s speciﬁer,

C must be the probe. The relevant feature to be checked is the [uWH*] feature on C.

Hafa carries a [iWH] feature that checks C’s feature, and the strength of [uWH*] compels

movement.

So far, we have the structure in (32):

(32) CP

hafa



iWH



[uWH*]——–

p¨ara

ta-chuli’

At this point, Wh-Agreement between T in this clause and the newly created trace can

occur. (Since hafa is in C’s speciﬁer, the trace is free within TP.) This results in nominaliza-

tion on the verb. See Section 4.2 for a discussion of the formal properties of this agreement

relation in Minimalist terms.

As t he derivation continues, the CP-phase in (32) is in turn merged as an argument of

the higher verb. Hafa moves from its position in (32) t o the speciﬁer position of the higher

Asterisks indicate strong features.

The [WH] feature is interpretable on DP and uninterpretable on C because this feature signals semantic properties that

are appropriate for a DP but not for C. It distinguishes wh-phrases from ordinary DPs in that a wh-phrase is a semantically

appropriate restrictive clause f or a constituent question (see section 5.1), for example. Since C’s do not participate in wh-

constructions in the same way as DPs, the [WH] feature is uninterpretable on C; That its presence signals the semantic viability

of its host’s participation in wh-movement is meaningless on C.

CP. Once again, this movement is triggered by a strong feature on the matrix C that is

checked by hafa. As in the lower clause, the appropriate feature is [uWH*] on C. Further,

since this is hafa’s surface position, its [uQ] feature must be checked here. For this reason,

the matrix C must possess a [iQ] feature. This feature marks the clause as interrogative and

is interpretable (see fn. 8). (33) shows the surface structure of (29).

(33) CP

hafa



iWH

uQ—–





uWH*——–



malago’-˜na si Magdalena t p¨ara ta-chuilil’ t

I adopt this system, which is consistent with standard Minimalist analyses of wh-movement,

in the analysis of long movement below. While there are important diﬀerences between

successive-cyclic and long-distance derivations, I assume that the same feature system is at

work. Wh-phrases possess the [iWH] and [uQ] features. Complementizers that attract wh-

phrases to their speciﬁers have a [uWH*] feature, and the C that heads a question has an

additional [iQ] feature. I now turn to the analysis of long-distance Wh-movement, starting

with the lowest clause of these constructions.

3 The Syntax of Long Movement

3.1 The Lower Clause

Two facts suggest that movement occurs in the lower clause in (26), repeated below as

(34). First, there is a gap in this clause. Second, Wh-Agreement morp hology appears in this

clause.

Apparently, then, this clause has the pre-movement structure in (35).

(34) Hafa

what?

patti

part

Loc

atumobit

car

mal¨agu’

agr.want

hao

you

[u-ma-fa’maolik

WH[nom].agr-Pass-ﬁx

t ]?

“Which part of the car do you want to be ﬁxed?” (Chung 1998:248)

(35)

C TP

T VP

V DP

(35) abstracts away from certain details of the construction, such as the presence of vP.

is t he verbal complement, and it occupies the position of trace in (34). This DP then

moves to the speciﬁer of the lower C, creating this structure:

(36)

C TP

T VP

With this movement, Wh-Agreement on the verb is triggered. I suggest that this move-

ment is identical to the successive-cyclic movement described above. DP

possesses a [i WH]

feature that checks C’s [uWH*] feature.

needs an uninterpretable feature to be an active goal. It is not clear what this feature

is. [Q] is not an option because it is reserved for the head of a question. If [Q] were assigned to

Chung (1998) does not tie the appearance of Wh-Agreement morphology to movement directly. Rather, it is triggered by

variables that are A -bound. Nonetheless, since the most obvious way to produce an A-bound variable is through movement, it

is reasonable to take Wh-Agreement as a symptom of movement until evidence to the contrary appears.

the resumptive pronoun, this feature would have to appear on intermediate C’s, too, because

the resumptive pronoun doesn’t move to the CP that heads the wh-construction. Second,

it will become clear in Section 4.1 that positing [Q] on intermediate C’s causes problems for

an analysis of Operator-C agreement which ties this morphology to [Q].

Another feature, perhaps [Op] or [resumptive], could be adopted. Alternatively, we could

give up on the idea that goals need uninterpretable features to be active. This is compatible

with what I have said so far in that the [Q] feature could be removed from the overt wh-

phrases with [WH] r emaining to motivate movement.

I will not resolve this issue here,

but for concreten ess I assume a feature [uR(esumptive)] (perhaps a placeholder for a better

feature) that makes the resumptive pronoun active. An interpretable version of this feature

appears on C to check the resumptive pronoun’s feature, just like [Q] in successive-cyclic

movement. In (37), the [WH] and [R] features are added to the structure in (35).

(37)



uWH*



T VP

V DP



iWH



What exactly is the moved element? Given that it lacks phonological content, there seem

to be two plausible choices: It is either a trace or a null pronominal. I assume (partly for

reasons to be discussed below) that it is a (resumptive) pronoun. The resumptive nature

of this element is clear from the semantics: The overt wh-phrase in the matrix C’s speciﬁer

position is interpreted as if it were assigned the θ-role of the null element in the lower clause.

In other words, the overt phrase binds the null resumptive pronoun.

In the context of the

Or we could leave the analysis as it is with the understanding that the [uQ] feature is not responsible for wh-phrases being

active goals.

Although I will not discuss the point in any detail, it should be noted that the properties of the null DP generally match

the typical characteristics of resumptive pronouns as described by McCloskey (2005).

rest of the language, this is a reasonable conclusion. Null pronouns that are A-bar-bound

are well attested in Chamorro (in topicalization, for example; see Chung (1998)).

Returning to (37), when the probe C searches for some goal to satisfy its [uWH*] feature,

it ﬁnds the resumptive pronoun, and the strength of C’s [uWH*] feature causes movement

of the DP. This is identical to the ﬁrst step in the successive-cyclic wh-movement in (32) on

p. 15 above. If the next higher C lacks a [uWH*] feature, th e resumptive pronoun will not

move any farther. If the next higher C possesses this feature, it can attract the resumptive

pronoun to its speciﬁer, just as in normal successive-cyclic wh-movement. This happens in

some of the variations on the long-movement pattern described in Section 6. This is how

the surface appearance of the lower clause in (34) is derived. The resumptive pronoun’s

movement triggers Wh-Agreement on the embedded verb.

If the resumptive pronoun engages in successive-cyclic movement, it can, in principle, end

up in the speciﬁer of the matrix CP, just as if it were a full DP u ndergoing wh-movement.

However, this conﬁguration would be semantically unviable. The resumptive pronoun must

acquire its semantics from some other element in the construction. I argue in Section 5.1

that this is accomplished via binding by an overt wh-phrase higher in the structure. If the

resumptive pronoun moves all the way to the speciﬁer of the matrix CP, it cannot be bound

by a higher wh-phrase. (Nor would there be an acceptable place in which to merge this

overt wh-phrase: There would be no element to check its [uQ] feature.) Such a conﬁguration

is therefore semantically ill-formed, and the resumpt ive pronoun must stop in a position

lower than the speciﬁer of the matrix CP. In other words, the only semantically acceptable

structure is one in which a non-matrix C possesses a [iR] feature that can halt the resumptive

pronoun’s movement. For the d uration of this paper, until Section 6, I ignore the possibility

that the resumptive pronoun can move successive-cyclically. The rest of the analysis focusses

on constructions in which the resumptive pronoun stops after the ﬁrst movement.

3.2 The Higher Clause

The wh-phrase Hafa na patti gi atumobit “which part of the car” from (34) is base-

generated in the speciﬁer of the highest CP. This means that this DP satisﬁes (some of)

the selectional criteria of the root C and is therefore merged in C’s speciﬁer to fulﬁll these

criteria. The ﬁrst thing to consider, then, is the nature of C’s selectional criteria.

Since (34) is a question, the root C has the [iQ] feature. Further, since this a constituent

question, C should have the familiar [uWH*] feature. Something must be merged into

C’s speciﬁer position to check this feature. In successive-cyclic movement , this is achieved

through movement of a wh-phrase from the existing structure. This is not possible in long-

movement constructions. The only wh-phrase is the null pronoun from the previous section.

The pronoun’s [uR] feature has been checked, so it is inactive (keeping in mind the above

caveat concerning active goals). The root C has no choice but to accept a base-generated

wh-phrase in its speciﬁer.

Before the wh-phrase is merged, the higher clause has this structure:

(38) CP



uWH*



The DP Hafa na patti gi atumobit “which part of the car” is merged into C’s speciﬁer. C’s

[uWH*] feature is checked, and the DP’s [uQ] feature is checked. Descriptively speaking, no

Wh-Agreement is triggered because no trace is produced.

Operator-C agreement proceeds

as normal, though, and the appropriate agreement morphology appears in the construction.

That the wh-phrase was not moved is irrelevant to Operator-C agreement.

That there is no other candidate to satisfy the root C’s requirements is seen most clearly

when other clauses intervene between the lowest clause from Section 3.1 and the root clause.

It is useful to think in these terms for the time being, but the idea that traces trigger Wh-Agreement is rejected below.

Consider the schematic structure from (3), repeated and modiﬁed here as (39):

(39) CP

is the root CP discussed in this section. CP

is the clause containing the resumptive

pronoun. The intermediate CPs play no role in this analysis. If CP

’s head checks the

resumptive pronoun’s [uR] feature, these clauses do not participate in the long-movement

derivation. The root C is separated from resumptive pronoun in the lowest clause by at

least two phase boundaries. With th is DP well out of reach, the highest C has no choice

but to accept a ﬁrst-merged wh-phrase. In addition to the fact that the resumptive pronoun

is u navailable because its [uR] feature has been checked, examples like these show that the

resumptive pronoun can be buried too deep for the matrix C to access it.

Still ignoring the possibility of successive-cyclic movement of the resumptive pronoun, CP

and CP

are normal non-interrogative CPs. Lacking the relevant features, the complemen-

tizers in these clauses do not attract anything to their speciﬁers. These clauses behave just

like non-interrogative embedded clauses, even though they are in the middle of a question

construction. The crucial operations occur in the highest and lowest clauses.

Finally, DPs in the speciﬁer of CP characteristically take the default morphological case

(S. Chung, p.c.), so there is no need to pass the case of the gap (the case assigned to the

null resumptive pronoun) up to this DP. The morp hology of a DP in a speciﬁer of CP is

not dependent on the case of its A-position, even in instances of successive-cyclic movement.

Since the default morphology appears, no mechanism is need ed to connect the overt DP in

long movement with the case assigned to the resumptive pronoun. And as it is not formally

associated with a θ-role, the overt DP does not need to acquire abstract case.

In canonical long movement, Wh-Agreement appears only in the lowest clause because

wh-movement only occurs only here (via mechanisms to be developed below). Operator-C

agreement appears on the highest C because the wh- phrase appears in this C’s speciﬁer

position. The agreement facts are accounted for.

Non-question instances of wh-constructions require further comment because they lack a

[Q] feature. Consider the sentence in (40), which contains a relative clause.

(40) [Adyu

that

the

[O ma’a’˜nao

agr.afraid

yu’

[na

Comp

u-b¨aba

WH[obj].agr-open

t ]]], g´ofdangkulu.

agr.very.big

“That thing which I was afraid to open, it was very big.” (Chung 1998:248)

Of interest here are the bracketed constructions. The relative clause O ma’a’˜nao yu’ [n a

u-b¨aba t ] “which I was afraid to open” is an adjunct to the DP Adyu i “that thing.” This

relative clause contains an embedded clause, indicated by the innermost square brackets. As

the gloss shows, Wh-Agreement is present on this clause, but it is absent from the higher

clause in the relative clause. This is evidence that (40) is a case of long movement.

The lowest clause in the relative clause—the one with Wh-Agreement—behaves exactly

as described in Section 3.1. A null resumptive pronoun originates in the position of the trace,

and it moves to the speciﬁer position of the complementizer na to satisfy featural criteria.

Following Chung (1998), I assume a null relative operator in the sp eciﬁer position of the

relative clause’s matrix CP. This operator is marked as O in (40). As in English relative

clauses, this operator forms a dependency with the gap and acquires its semantic content

from the phrase to which the relative clause is adjoined (the DP adyu i “the thing”).

Since this is a long-movement construction, the operator O must be merged directly into

na’s speciﬁer position with no movement (because the overt DP is not the moved element

in long-movement operations). As a complementizer in a relative clause, na cannot possess

a [Q] feature, but does possess the same [uWH*] feature that was employed above. The

operator O is merged to check this feature. It must therefore have its own [WH] feature, but

I remain agnostic about whether or not this feature is interpretable. If it is uninterpretable,

we can account for why the operator does not participate in further movement, but since

it is buried within a DP, it is unclear whether such an explanation is necessary. If DPs are

phases, the operator is unavoidably inaccessible to further probes.

Despite the absence of a [Q] feature on the higher C, relative clauses behave exactly like

constituent questions with respect to long-distance movement. One further point is worth

noting. The null operator O checks a complementizer’s [uWH*] feature and receives its

semantic content from another DP. In this respect it is just like the resumptive pronoun

postulated above. It is conceivable that these two entities are actually the same, but at the

very least we have an interesting pattern. In three constructions (long-distance movement,

relative clauses, and topicalization), we have a null element that forms a dependency with

another DP. One might formulate a generalization concerning the (absence of) phonological

content of anaphoric elements in Chamorro. However, reﬂexive pronouns are not null:

(41) Ha-bira

agr-turn

gui’

herself

si Santa

Santa

Maria

and.then

ha-fana’

agr-face

the

liga.

wall

“The Virgin Mary would turn and face the wall.” (Chung 1998:36)

4 The Agreement Relations in Minimalist Terms

In this section I develop Minimalist analyses of the two agreement relations, Operator-C

Agreement and Wh-Agreement.

4.1 Operator-C Agreement

Chung’s rule for Operator-C agreement is (from Chung (1998:230)):

(42) Operator-C Agreement Rule (holds at s-structure)

and an Associate that is both its sp eciﬁer and an operator must have compatible

values for [N], [O], and [locat].

The details of this rule need not concern us here. Crucially, Operator-C agreement is a

relation between C and the element in its speciﬁer. This element must not be a variable:

Operator-C Agreement holds between C and the highest item in an A-chain (i.e. a full DP).

Clearly [N], [O], and [locat] are the relevant features in this relation. These correspond to

the features of C’s speciﬁer that determine C’s shape, as described in (17). [+N] indicates a

DP, [+O] indicates a phonetically null operator, and [+locat] indicates a locative operator.

If the highest C along the path of A-movement contains uninterpretable versions of these

features, we can explain their inﬂu ence on C’s morphology by assuming that the wh-phrase

moved into its speciﬁer checks these features. C is realized accordingly.

But we must be careful to ensure that only the highest CP possesses these features.

The easiest solution is to bundle these features with the feature that renders the wh-phrase

inactive when it reaches this CP. Consequently, [iQ] must be linked to the [N], [O], and

[locat] features. In constituent questions, when a wh-phrase reaches the speciﬁer of a CP

whose head has the [iQ] feature, the wh-phrase’s [uQ] feature is checked, and the phrase is

rendered inactive. In addition to checking C’s [WH] feature, this phrase also checks C’s [N],

[O], and [locat] features. Since [Q] cannot appear on C without [N], [O], and [locat], and

vice versa, the Operator-C agreement features will only appear on the C in whose speciﬁer

the wh-phrase appears at s-structure.

Operator-C Agreement is easily modeled in Minimalist terms. Because a unique local

agreement relations already exists between a wh-phrase and a C with a [Q] feature, additional

agreement relations can be posited in a way that capitalizes on this special association. Wh-

Agreement is much harder to accommodate.

4.2 Wh-Agreement

If we take the description of Wh-Agreement assumed above at face value, we must devise

a system in which C can acquire case an d transfer this case to its speciﬁer. DPs must be

As mentioned above, the fact that Operator-C agreement only appears on the highest clause in a long-movement construction

is one motivation for not using [Q] on the resumptive pronoun.

able to discard one case feature and acquire a new one. In addition to checking a subject’s

[NOM] feature, T must have its own additional case features that can be checked only by

a wh-element. Some unappealing aspects of this approach may already be apparent to the

reader; I will argue that this is in fact the wrong way to proceed, and the way in which

Wh-Agreement was characterized above is mistaken.

Within the Minimalist Program, Wh-Agreement is best understood as a morphological

phenomenon rather than a purely syntactic one. That is, Wh-Agreement d oes not moti-

vate the postulation of more features that trigger the morphology described in Section 2.2.

Instead, independently necessary featural requirements conspire to produce unique arrange-

ments among the participants in A-system case-checking and A-system wh-movement. These

unique conﬁgurations are signaled by Wh-Agreement. Such an approach has been proposed

by Watanabe (1996), although I depart from that analysis in important places.

The morphology itself is syntactically insigniﬁcant (as suggested by Chung (1998)) in that

it drafts existing morphology in the manifestation of the special featural conﬁgurations. For

example, verbal nominalization is simply a way to signal oblique Wh-Agreement. It does

not reﬂect any syntactically or semantically meaningful metamorphosis of the verb.

I ﬁrst discuss approaches that attempt to derive Wh-Agreement directly from a feature-

checking operation. These eﬀorts are u nappealing on both conceptual and empirical grounds.

I then turn to a more indirect method in the vein of Watanabe (1996) that frames Wh-

Agreement as a by-product of other syntactic processes.

4.2.1 Wh-Agreement as Feature Checking

The most obvious way to implement Wh-Agreement in a Minimalist framework is by

postulating a case-checking relationship that accompanies wh-movement. In this section I

sketch an attempt along these lines, but it proves conceptually unappealing.

Chung’s (1998) agreement rule is a reasonable starting point. Using T here instead of I,

Chung’s rule is formalized as in (43) (from Chung (1998:257)):

(43) Wh-Agreement (holds at s-structure)

and an A-bound trace that is free within T

’s minimal m-command domain must

have compatible values for [Case].

Chung (1998) provides careful argumentation to the eﬀect that T is the element that

participates in the agreement relation with the trace. I adopt this conclusion for the purposes

of the current discussion. The following arrangement correctly predicts the occurrence or

non-occurrence of Wh-Agreement in all clauses:

Following Chung (1998), the case assigned to CP is inherited by its head, C

. When a

wh-element moves to C’s speciﬁer, it acquires C’s case. T in the immediately higher clause

has a weak uninterpretable case feature bundled with a weak [WH] feature. Consequently

only a wh-element can check the case feature. To this end, T locates the wh-element in C’s

speciﬁer. The wh-phrase may then continue its successive-cyclic movement.

Only C with a [WH] feature may select this Wh-Agreement T with an additional case

feature. Further, only the Wh-Agreement T may (and in fact must) select v with a [WH]

feature. (vP is a phase boundary, so a [WH] feature on v is necessary independently. An

A-moved DP stops in each speciﬁer of vP in addition to each speciﬁer of CP.)

These selectional requirements are essential. If any C can select the Wh-Agreement T,

then Wh-Agreement should be possible on any clause that contains a CP whose speciﬁer

is ﬁlled by a wh-element. For example, in (the Chamorro equivalent of) I don’t know who

the police arrested, the matrix verb know should have the option of showing Wh-Agreement.

This does not happen, so the Wh-Agreement T must not be available in non-wh-CPs.

Also, we cannot require C with a [WH] feature to select the Wh-Agreement T. The long-

movement construction in (44) shows this most clearly. The matrix C has a [WH] feature,

but the T that it selects must not need Wh-Agreement case-checking. The only elements

that could check T’s case are the resumptive pronoun, which may be embedded many clauses

lower, and the overt DP, which in inaccessible to T because it is in C’s speciﬁer position. If

T required Wh-Agreement, certain long-movement constructions would be impossible.

(44) C

. . . RP

. . .

How, then, do we ensure that the special T appears in every clause along the path of wh-

movement? If T without a Wh-Agreement case feature appears in a clause headed by a C

that drives wh-movement, no Wh-Agreement will appear. Wh-Agreement would incorrectly

be predicted to be optional in all cases. We might expect to ﬁnd an alternating pattern of

agreement where one clause shows agreement, the next does not, and the next does, etc.

Consider the diagram in (45). We’ve already conclud ed that this C, which has a [WH]

feature, can optionally select the Wh-Agreement T. The goal now is to force C’s hand in

this case by ensuring that only the Wh-Agreement T will yield a well-formed structure. In

order for the movement shown to be successful, v must have a [WH] feature. If only the

Wh-Agreement T may select a v with a [WH] feature, then only the Wh-Agreement T may

appear in (45). Without a Wh-Agreement case feature, T cannot select a v with a [WH]

feature, and the derivation will crash. The result is that while C is not required to select th e

special T, only through the selection of this T will the derivation succeed. This is why it is

necessary to restrict T’s selectional criteria.

(45)

The stipulative nature of the selectional requirements is unappealing. It would be prefer-

able if the distribution of Wh-Agreement could be accounted for with the free distribut ion of

case and [WH] feature on the relevant heads. Ideally, the derivation would succeed in only

and all the cases in which the resulting morphological pattern is attested.

By themselves, the selectional stipulations do not condemn the analysis. But there is

another drawback. If Wh-Agreement reﬂects C’s case, C must be able to change its speciﬁer’s

case feature. How such an arrangement can be produced in Minimalist terms is diﬃcult to

imagine. Simple feature-checking isn’t enough: The case feature on C’s speciﬁer has already

been checked in the A-system and has no inﬂuence on Wh-Agreement.

This is a more serious problem than the selectional requirements because it challenges the

nature of feature checking in Minimalism. We might gloss over this diﬃculty and assume

that a solution is possible, but with the compounding issue of the selectional stipulations,

the wiser move is to question the whole approach.

There are a couple additional drawbacks to the feature-checking analysis. First, C must

have case features that it can transfer to DP. It is not clear what it means for C to have

case, nor is it clear why it might need case. Also, T must have more than one case feature.

It must participate in the nominative case-checking of a subject as well as case-checking for

Wh-Agreement. Such duplication is to be avoided: There are no other instances of a head

participating in two completely unrelated case-checking operations likes this.

One might attempt an analysis that takes DP out of the Wh-Agreement process, es-

tablishing a feature-checking relationship between C and T. v would need to serve as an

intermediary so the checking process could cross the vP and CP phase boundaries. C would

check v’s case, and v would check T’s case. No feature-changing power is needed. How-

ever, the other drawbacks are ampliﬁed here. v now needs case, just like C and T, making

Alternatively, we might give the DP a second case feature that matches C’s case and triggers Wh-Agreement. Aside from

the conceptual awkwardness of claiming a DP has more than one case feature, it would be necessary, in principle, to assign DP

a diﬀerent case feature for each A-movement operation it undergoes. Wh-Agreement may vary from clause to clause within

the same construction, so we cannot be sure the same “extra” case feature will work for each instance of Wh-Agreement in an

A-construction. Clearly such an analysis is to be avoided on conceptual grounds.

three non-nominal heads that require case. The case-checking duplication problem is com-

pounded, too. To see this, consider two scenarios. In the ﬁrst, C is not a verbal complement.

C must receive case from some head and then, for Wh-Agreement, check case on v. (v checks

accusative case on another element.) Case checking for Wh-Agreement must therefore be

independent of normal case checking because in this instance v does not check accusative

case on C. In the second scenario, where C is a verbal complement, C has two case-checking

relationships with v, both of which we know are necessary from the ﬁrst scenario but which

are undeniably redundant here.

Many aspects of the feature-checking analysis are unappealing. This is not an analysis

that is generally sound except for some residual issues. Instead, the core aspects of the

analysis present the greatest problems. I conclude that an approach along these lines is

technically feasible, but conceptually inadequate. Another analysis th at does not have these

conceptual problems is available, rendering the analysis sketched here inadequate. At least

as far as Minimalism is concerned,

Wh-Agreement is best analyzed in other terms.

4.2.2 Wh-Agreement as Morphology

Watanabe (1996) argues that Wh-Agreement is simply a morphological side-eﬀect of other

feature-checking relations. As such, it serves as a signal for certain featural conﬁgurations

but is not itself triggered by any part icular featural requirement. This is the system I adopt

here with some modiﬁcation.

Watanabe develops a system in which case absorption—the elimination of uninterpretable

case features before spell-out—is mediated through the Agr-s and Agr-o heads. In the

subject-verb agreement system, T adjoins to Agr-s, and the subject moves to the speciﬁer

of Agr-sP. We thus have the conﬁguration in (46).

An analysis in the spirit of GPSG or HPSG may have more success. For example, in GPSG terms, an XP/DP may signal

that the displaced DP is missing by reﬂecting the case of the gap or the case of the clause containing the gap.

(46) Agr-sP

[NOM]

Agr-s

[NOM]

Agr-s

Provided the case features on T and DP match, Agr-s may absorb one of these case

features so that it is invisible at LF. The other case feature must be dispensed with some

other way. Normally, Agr-s incorporates DP’s [NOM] feature, and the [NOM] feature on T

is dealt with by movement of the T/Agr-s complex to C. This movement eliminates T’s case

feature, yielding (47).

(47) CP

C Agr-s

T Agr-s

Agr-sP

Agr-s

. . . t

. . .

In subject extraction, DP then moves to C’s speciﬁer:

(48) CP

C Agr-s

T Agr-s

Agr-sP

Agr-s

. . . t

. . .

This creates a circle of feature-checking relationships with DP, C, and T/Agr-s. DP and

T/Agr-s are involved in a case-checking relationship, T/Agr-s and C have their own case-

checking relationship, and DP and C are the central players in wh-movement. Alternatively,

we might think of this in terms of a “double relationship” between DP and T/Agr-s. Within

the Agr-s projection, these elements are involved in A-system feature-checking, and within

the CP projection, when T/Agr-s occupy C, they are involved in A-feature-checking.

Subject Wh-Agreement is simply a reﬂection of this double feature-checking relationship

between DP and T/Agr-s. In the case of non-subject extraction, this peculiar relation-

ship does not exist (because the extracted DP’s case is not checked by T/Agr-s), and Wh-

Agreement reﬂects this in the form of diﬀerent morphology. We thus have a two-way contrast

between constructions with the double-agreement arrangement between DP and T/Agr-s and

constructions without this special conﬁguration. The morphophonological realization of this

dichotomy is arbitrary: the language simply coopts existing verbal morphology.

This analysis is suﬃcient for languages like Palauan (Georgeopolous 1985, 1991a,b) which

have a two-way Wh-Agreement system, separating subject from non-subject extraction. But

Chamorro has a three-way contrast among subject extraction, object extraction, and oblique

extraction. Before developing an account of the object/oblique distinction, it is worth bring-

ing Watanabe’s system in line with current syntactic theory.

In particular, Agr-s and Agr-o are unnecessary. Nominative case-checking occurs within

TP. The subject DP and T check each other’s cases directly, and the [NOM] features dis-

appear with no further operations. T-to-C movement is motivated on other grounds. The

basic premise behind Watanabe’s argument remains intact though. DP and T are still in

the double feature-checking relationship: case-checking within TP, wh-movement within CP.

Wh-Agreement reﬂects the presence or absence of this conﬁguration. (49) replaces (48).

(49) CP

C T

. . . t

. . .

The contrast between object and oblique extraction is not yet motivated. Perhaps this

distinction is of the same type as the subject/non-subject dichotomy. That is, perhaps object

or oblique extraction creates a feature-checking conﬁguration that the other does not. What

might this conﬁguration be?

In the previous section v was given a more prominent role because of its projection’s sta-

tus as a phase boundary. Rather than moving from CP to CP, a wh-moved item must move

from CP to vP to CP, etc. Verbal objects have their accusative case features checked within

vP. They therefore have a relationship with v that is the exact analogue of the relationship

between T and DP in subject extraction. In object extraction, v checks a DP’s [ACC] fea-

ture, and then this DP moves to v’s speciﬁer and checks v’s [uWH] feature. I suggest that

Chamorro’s objective Wh-Agreement morphology is just a reﬂection of this. Such a conclu-

sion is almost required by the analysis of subject extraction: If “double feature-checking”

triggers special agreement in one instance, it would be odd if other similar conﬁgurations

were barred from triggering their own special agreement morphologies. Oblique extraction is

the result of no such special feature-checking conﬁguration. In this light, Chamorro simply

opts to signal more unique conﬁgurations than Palauan. Both languages are sensitive to the

featural relationship between T and extracted subjects, and Chamorro is also sensitive to

the featural relationship between v and extracted objects.

Oblique Wh-Agreement morphology is just the elsewhere condition, arising when neither

of the special featural conﬁ gurations holds. This means that all DPs that trigger oblique

Wh-Agreement (i.e. any DP that is not a subject, direct object, or oblique object of a verb

of transfer) must not have their case features checked by either T or v. Pylkk¨anen (1997)

provides a way of ensuring this. In her system, only “core” arguments (generally, direct

objects) originate within VP. External arguments (one type of non-core argument) originate

in vP (VoiceP for her). This is exactly what I assumed above, so nothing new must be said for

these arguments. Other non-core arguments are introduced by one of six functional heads:

high applicative, low recipient applicative, low source applicative, root-selecting CAUSE,

verb-selecting CAUSE, and phase-selecting CAUSE (Pylkk¨anen 1997:15). The details of

these heads need not concern us here. What is important is that Pylkk¨anen provides a way

to introduce certain arguments independently of VP and vP. If an argument is licensed by

one of these six, it will trigger oblique Wh-Agreement under extraction because it will not

enter into one of the special featural relationships described above.

For an analysis like this to hold, it must be shown that arguments that trigger objec-

tive morphology are Pylkk¨anen’s core arguments, and those that trigger oblique morphology

are Pylkk¨anen’s non-core (but also non-external) arguments. Objective Wh-Agreement is

triggered by direct objects and “the oblique object of a verb of transfer” (Chung 1998:237).

Clearly, direct objects fall within Pylkk¨anen’s conception of core arguments; they’re pro-

totypical examples of such arguments. If we take Chung’s description seriously, we can

conclude that oblique objects of verbs of transfer are also core arguments. While oblique

DPs are in general non-core, it is reasonable to assume that the oblique object of a verb of

transfer is a core argument: The verb itself requires this DP to ﬁll the θ-role of recipient

or source. DPs in this case are distinct from recipients and sources that are introduced by

Pylkk¨anen’s applicative heads in that oblique objects of verbs of transfer are directly licensed

by the verb of transfer because of this verb’s semantic properties. Verbs of transfer require

recipients and/or sources independently of other functional heads, so the DPs that ﬁll these

θ-roles for verbs of transfer are distinct from DPs that ﬁll these θ-roles for other verbs.

DPs that trigger oblique Wh-Agreement are “oblique complements of intransitive pred-

icates, instruments, and subcategorized comitatives” (Chung 1998:238). None of these

categories overlaps, in whole or in part, with Pylkk¨anen’s core arguments. Consequently,

Pylkk¨anen’s system would introduce these arguments through nonverbal functional heads.

Further, these DPs are non-subjects, so their case is not checked by T. We thus have precisely

the situation we want: These DPs will not enter into the special featural conﬁgurations th at

trigger subject and object Wh-Agreement. They therefore trigger oblique Wh-Agreement,

which is the morphology that appears when no special featural conﬁguration holds.

Watanabe does not elaborate on the mechanisms that translate the double-feature-checking

relationships into morphology, and this is a recalcitrant p roblem for both his analysis and the

current proposal. The intuition behind Watanabe’s proposal is clear, but its formal imple-

mentation is less than straightforward. The syntax seems to require memory: For subjective

Wh-Agreement, the syntax must know that the DP that checked T’s [NOM] feature is the

same DP that checked C’s [WH] feature. Simply examining the features on T and C won’t

help: It will always be the case that these two features are checked. The important point

for Watanabe is that the same DP checked both features.

One solution reassigns the duty of checking [NOM] to ﬁnite C.

With this change, the

features t hat trigger Wh-Agreement are consolidated onto two heads: [NOM] and [WH] on C,

[ACC] and [WH] on v. Now, case-checking coupled with [WH] has a morphological exponent

that varies with the case assigned, either nominative or accusative. Checking of [WH], when

not “linked” with any case-checking operation, has its own morphological exponent, oblique

Wh-Agreement. This solution is speculative at best, but it at least provides a means to

formally understand Watanabe’s ideas. Notice that it also allows us to dispense with some

assumptions from the above analysis, notably that T moves to C.

Under this approach, Wh-Agreement is not a reﬂection of case in any way. Subject

agreement doesn’t appear because the moved item is nominative. It appears because the

moved item has a unique relationship with T. Wh-Agreement is sensitive to which head

checks the moved item’s case but not to the speciﬁc case feature involved. Agreement

Thanks to J. McCloskey (p.c.) for suggesting this approach.

morphology varies with case but is not triggered by case.

5 Semantic Interpretation and Its Consequences

5.1 Semantic Interpretation

How does the current analysis square the requirements of the semantic component of the

grammar? The syntax must supply the semantics with an interpretable structure. While

I do not present a full-ﬂedged semantic analysis here, I suggest that the long-movement

structures proposed above are compatible with existing semantic ideas.

Heim & Kratzer (1998) develop a system for assigning a denotation to relative clauses.

The crucial part of their system is Predicate Abstraction, which is responsible for turning

the otherwise (semantically) ordinary relative clause into something that can be integrated

with th e NP to which it is attached. This system can be straightforwardly applied to relative

clauses in Chamorro as long as the resumptive pronoun is accommodated.

Perhaps the most straightforward remedy is to assign the resumptive pronoun an empty

semantic value. This is not an unusual move in the Heim and Kratzer system. For example,

the complementizer that does not aﬀect the denotation of the construction it is a member

of. To extend this property to the resumptive pronoun, we simply need to adopt the rule in

(50). (For typographical reasons, “CP

” is used to stand in for the structure in (51).)

(50) [[CP

]] = [[C

]]

(51) CP

= CP

This rule ensures that the resu mptive pronoun will not aﬀect the semantic interpretation

of the construction. The binding relation between the null operator O and the resumptive

pronoun ensures that O and the pronoun’s trace will be coindexed, mimicking the situation in

standard relative clauses. The computation of the denotation proceeds exactly as described

by Heim and Kratzer, and Predicate Abstraction will ensure that the appropriate value is

assigned to the trace. The relative clause can then be integrated with the larger sentence.

The system in Heim & Kratzer (1998) is not well suited for questions on conceptual

grounds. It is unclear exactly what the denotation of a question should be, and Heim and

Kratzer oﬀer no speculation. Consequently, I adopt the analysis used in Chung, Ladusaw,

& McCloskey (1995; henceforth CLM).

CLM, making use of Berman (1991), argue that three syntactic elements are required

for a well-formed constituent question: an interrogative operator, a restrictive clause, and a

nuclear scope. These elements must meet certain syntactic and semantic requirements. The

restrictive clause is the displaced constituent in C’s speciﬁer position. It “must syntactically

bind a position within the TP complement of C

” (CLM:244). The restrictive clause must

contain or be a wh-phrase, and it “must contribute to semantic interpretation just as if it

were sitting in the syntactically bound position” (CLM:244).

The role of the interrogative operator, which is associated with the [Q] feature, is ﬁlled

by C. It semantically binds the wh-phrase in the restrictive clause.

The nuclear scope is the TP complement of C. It supplies a propositional function whose

domain is deﬁned by the restrictive clause. The interrogative operator forces an interpreta-

tion in terms of a set of propositions.

What are the parts of this system in the analysis developed here? Clearly, the interrogative

operator must be the higher C because only this C has a [Q] feature. This means that the

restrictive clause must be the wh-phrase in this C’s speciﬁer, and the nuclear scope is the

matrix TP. All of th e requirements mentioned above are met in this conﬁguration. The

wh-phrase syntactically binds the trace through the resumptive pronoun, just as if it headed

a conventional A-chain formed by successive-cyclic movement. The wh-phrase also behaves

“just as if it were sitting in the syntactically bound position.” Because of (50), the resump tive

pronoun does not aﬀect the derivation except to connect the overt DP to the trace.

The analysis of long-movement proposed here is comp atible with the semantic require-

ments of wh-constructions. Long-movement constructions are interpreted exactly like their

successive-cyclic counterparts. The only addition to the semantic component of the grammar

is the rule in (50), which is essentially an instruction to ignore the resumptive pronoun.

5.2 Codependencies Between the Higher and Lower Clauses

The analysis of long-distance constructions developed here involves syntactically distinct

and independent operations. In the higher clause, a wh-phrase is merged into the speciﬁer

of the matrix CP. In the lower clause, a resumptive pronoun is merged into an argument

position and undergoes A-movement. Since there is no syntactic requirement that one half of

this analysis be present for the other half to appear,

we might expect to ﬁnd constructions

that contain just a resumptive pronoun with no binder or a binder with no resumptive

pronoun. This would be an incorrect prediction, and I suggest that this is a semantic fact.

In order for Predicate Abstraction to apply successfully, the syntactic unit from which the

predicate is to be constructed must contain some syntactic object that can be construed as

a variable. “Traces” and resumptive pronouns serve this function prototypically. If there is

no such element, the assignment function introduced by Predicate Abstraction will not have

the necessary eﬀect. A useless denotation will result, one which includes lambda abstraction

of a variable that is not present. We can therefore conclude that the operator O in relative

clauses requires the resumptive pronoun in the lower clause when O is ﬁrst-merged into its

surface position. Without the resumptive pronoun, Predicate Abstraction will fail.

Likewise, in questions, the restrictive clause and interrogative operator cannot be inter-

preted without an appropriate nuclear scope. Without an TP that contains an unbound

variable, no semantic computation can be performed.

Perhaps a similar argument can be made for why the resumptive pronoun requires the

presence of a wh-phrase elsewhere in the construction. The resumptive pronoun seems to

To be more speciﬁc, there is no syntactic requirement that merger of the wh-phrase is permitted only if there is a resumptive

pronoun elsewhere in the construction. Likewise, no syntactic restriction permits the merger of a resumptive pronoun only where

a wh-phrase can bind it.

require a wh-binder, and this may be attributable to semantic or syntactic attributes of the

resumptive pronoun. Alternatively, when no suitable binder is present, the anaphor may

simply be a (non-resumptive) pronoun.

6 Variations on the Long-Movement Pattern

Much of the discussion so far has assumed that in long-distance movement, the resumptive

pronoun moves to the ﬁrst speciﬁer of CP and stops, and the wh-phrase merges directly into

its surface position. But the analysis in fact predicts other options.

The current analysis does not prevent the wh-phrase from merging in a non-matrix spec-

iﬁer of CP and moving successive-cyclically to its surface position until it is deactivated

by a [iQ] feature, triggering Wh-Agreement along the way. We would predict that such a

conﬁguration would involve, from the top of the structure down, one or more clauses with

Wh-Agreement (triggered by the wh-phrase), followed by one or more clauses with no agree-

ment, followed by one or more clauses with agreement (triggered by the resumptive pronoun).

In other words, we predict what appears to be long-movement followed by successive-cyclic

movement. This is exactly what we ﬁnd:

(52) Hayi

who?

malago’-mu

WH[obl].want-agr

[t p¨ara

Fut

u-ma’a’˜nao

agr-agraid

si Carmen

Carmen

[p¨ara

Fut

ali’e’-˜na

WH[obl].meet-agr

t ]]?

“Who do you want Carmen to be afraid to meet?” (Chung 1998:365)

This construction must be of the long-d istance variety. With no Wh-Agreement in the

middle clause, it appears that the moved item has skipped over a clause. Since this is not

possible, a resumptive pronoun must be in the lower clause, triggering agreement there,

followed by base-generation of the over wh-phrase in a higher position. In this respect, (52)

is just like a long-movement sentence. But the Wh-Agreement on the highest clause signals

the presence of a trace in the speciﬁer of the next lowest CP. The analysis developed here

permits Hayi to merge into this position, presumably to satisfy this intermediate C’s [uWH*]

feature, and then move into the matrix CP to check the same feature on the higher C.

Similarly, suppose the C that immediately dominates the resumptive pronoun’s A-position

has a [uWH*] feature but not a [iR] feature. Upon moving to this C’s speciﬁer, the resumptive

pronoun remains an active goal because its [uR] feature has not been checked . If the next

higher C has a [uWH*] feature, the resumptive pronoun may move again. In fact, it may

move successive-cyclically, triggering Wh-Agreement with each move, until it encounters a

[iR] feature. We therefore predict an analog of (52) where the higher clauses have no Wh-

Agreement but two or more lower clauses do. That is, we shou ld ﬁnd what appears to be

successive-cyclic movement followed by long-movement. This prediction is correct:

(53) Esti

this

pitsonas

person

comp

ma’a’˜nao

agr.afraid

yu’

[man-malagu’-˜niha

WH[obl].agr-want-agr

[t p¨ara

Fut

uma-kuentusi

WH[obj].agr-speak.to

t ]].

“It’s this person who I’m afraid they want to speak to.” (Chung 1998:365–366)

Again, this sort of structure is expected in the context of the current analysis. All that is

required is that the C into whose speciﬁer the resumptive pronoun ﬁrst moves not possess a

[iR] feature. In other words, the kind of C that appears in the middle of normal successive-

cyclic constructions is necessary.

Finally, the current analysis predicts that both the resumptive p ronoun and the overt

wh-phrase can m ove successive-cyclically in the same construction. There are no such ex-

amples in Chung (1994) or Chung (1998), but this may be an artifact of complexity. Such

a construction would require at least four clauses. Two lower clauses with Wh-Agreement

would show successive-cyclic movement of the resumptive pronoun, and Wh-Agreement in

the highest clause would reﬂect the movement of the wh-phrase. A fourth clause with no

agreement is needed between these two sets of clauses to show conclusively that the construc-

tion is not a normal successive-cyclic wh-construction. It may be diﬃcult to elicit reliable

judgments on constructions with this kind of embedding, so their absence is not surprising.

7 Conclusion

The analysis of long-distance wh-movement presented here reconciles the empirical facts

of Chamorro and the theoretical assumptions of Minimalism. Because no Wh-Agreement ap-

pears on intermediate verbs in long movement constru ctions, it seems as though a wh-phrase

moves directly from its original clause to its ﬁnal surface position without stopping in any

intermediate positions. An operation like this would ignore the Phase Impenetrability Con-

dition. Minimalism therefore predicts that long-distance movement should not be possible.

By separating what appears to be a long movement operation into two parts, th e conﬂict

is resolved. In fact, “long-distance movement” is no longer an accurate description of these

constructions. They involve normal successive-cyclic movement and the base-generation of

a wh-phrase in a higher position. No movement across long distances is required.

In addition, this analysis accounts for the fact that apparent long-distance movement is

exempt from island violations. Since little movement is involved, movement out of an island

is not an issue except where it concerns the null pronoun’s short movement. Any number of

islands may occur between the wh-phrase and the null pronoun, but no ungrammaticality

will result. Th is is because th e wh-phrase is base-generated in its surface position and thus

cannot incur any island violations.

Two issues have yet to be settled. The most important remaining issue is that of the

distinction between DPs that can participate in the long-movement construction and those

that can’t. Recall that Cinque (1990) identiﬁes the set of DPs that may undergo long-

movement as “referential.” In his system, referential DPs may bind, rather than antecedent

govern, pronouns and traces. Non-referential DPs, such as quantiﬁed DPs, must antecedent

govern traces and pronouns. Either binding or government is suﬃcient to license a trace or an

anaphoric pronoun. Because government is an inherently local relationship, it is unavailable

in long-movement constructions. Binding is required in these cases. Since only referential

DPs may be binders, only they may participate in long wh-movement.

Chung (1994) discusses some shortcomings of Cinque’s analysis. The notion of referen-

tiality is unsatisfyingly vague, and the data concerning which DPs may participate in long-

distance movement lead Cinque’s system to draw some puzzling conclusions about what

is a referential DP and what is not. For example, certain pairs of DPs in Chamorro that

contrast in their ability to participate in long movement do not diﬀer along the lines of ref-

erence, familiarity, or sp eciﬁcity, all semantic factors that might be imagined to contribute

to referentiality. Rather, familiarity seems to be the critical factor in these cases.

Whatever the right factor is, I assume that it is tangential to the present investigation.

Syntactic resources make long movement available, but other factors can inﬂuence the well-

formedness of th e resulting structures. Kluender (1998) provides evidence that cognitive

processing is (one of) these additional factors. The parsability of a wh-construction improves

as the extracted DP increases in “identiﬁability” (roughly, speciﬁcity or richness of descriptive

content). Perhaps long movement is parsable only if the overt wh-phrase meets some level

of speciﬁcity. (See also the references in Kluender (1998), as well as Gordon et al. (2001a,b),

for discussion of the kinds of DPs that can aﬀect processing in wh-constructions.)

Also, the mechanism that makes the resumptive pronoun an active goal has not been

identiﬁed. A [uResumptive] feature was used here, but we have seen no evidence that this is

indeed the correct feature. Perhaps, as already suggested, long-movement provides evidence

that goals do no need uninterpretable features to be active.

The analysis developed above makes long-movement constructions look strikingly similar

to phenomena like partial wh-movement in German, for example. In partial movement, an

A-chain is produced through two distinct but connected operations. An element moves from

an A-position into the A-system, moving successive-cyclically from one speciﬁer of CP to

the next. At some point, this element stops moving and another item is merged into the

next higher speciﬁer of CP. This second item continues moving and completes the A-chain.

The two halves of the partial movement operation are comparable to the two halves of the

long-movement construction: Some element moves from the A-system into the A-system,

but its movement is insuﬃcient to “complete” the derivation of a wh-construction. Another

element is merged in a higher position to complete the A-chain, binding the ﬁrst element.

In sum, the analysis of long-distance movement presented here removes much of the

oddness of these constructions. While it appears at ﬁrst that certain DPs can defy the

standard locality conditions of wh-movement, the phenomenon of long-distance movement

is in fact compatible with these restrictions.

References

Berman, Stephen (1991) On the Semantics and Logical Form of WH-Clauses. Ph.D. thesis,

University of Massachusetts, Amherst.

Chomksy, Noam (1999) Derivation by Phase. In MIT Occasional Papers in Linguistics, no.

18, Cambridge, MA: MIT Working Papers in Linguistics, Department of Linguistic and

Philosophy.

Chung, Sandra (1994) Wh-Agreement and “Referentiality” in Chamorro. L ingui stic Inquiry

25(1): 1–44.

Chung, Sandra (1998) The Design of Agreemen t: Evidence from Chamorro. Chicago and

London: The University of Chicago Press.

Chung, Sandra, William A. Ladusaw, & James McCloskey (1995) Sluicing and Logical Form.

Natural Language Semantics 3: 239–282.

Cinque, Guglielmo (1990) Types of

A-Dependencies. Cambridge, Mass.: MIT Press.

Georgeopolous, Carol (1985) Variables in Palauan Syntax 3: 59–94.

Georgeopolous, Carol (1991a) On A- and A’-Agreement. Lingua 85: 135–169.

Georgeopolous, Carol (1991b) Syntactic Agreement. Dordrecht: Kluwer Academic Publish-

ers.

Gordon, Peter C., Randall Hendrick, & Marcus Johnson (2001a) Memory Interference Dur-

ing Language Processing. Journal of Experimental Psychology: Learning, Memory and

Cognition 27: 1411–1423.

Gordon, Peter C., Randall Hendrick, & Aaron F. Kaplan (2001b) NP Types in Relative

Clauses: Corpus Results and Reading-Time Measures. CUNY Conference on Human Sen-

tence Processing, Philadelphia, PA.

Ha¨ık, Isabelle (1990) Anaphoric, Pronominal and Referential INFL 8: 5–47.

Heim, Irene & Angelika Kratzer (1998) Semantics in Generative Grammar. Malden, MA,

and Oxford, UK: Blackwell Publishers.

Kluender, Robert (1998) On the Distinction between Strong and Weak Islands: A Processing

Perspective. In Syntax and Semantics, Volume 29: The Limits of Syntax, Peter Culicover

& Louise McNally, eds., San Diego and London: Academic Press.

McCloskey, James (1990) Resumptive Pronouns, A-Binding, and Levels of Representation in

Irish. In Syntax and Semantics 23: The Syntax of the Modern Celtic Languages, Randall

Hendrick, ed., New York: Academic Press.

McCloskey, James (2005) Resumption, University of California, Santa Cruz.

Pesetsky, David (1987) Wh-in situ: Movement and Unselective Binding. In The Representa-

tion of (In)deﬁniteness, Eric Reuland & Alice ter Meulen, eds., 69–120, Cambridge, Mass.:

MIT Press.

Pylkk¨anen, Mariliina (1997) Introducing Arguments. Ph.D. thesis, University of Pittsburgh.

Watanabe, Akira (1996) Case Absorbtion and W H-Agreement. Dordrecht: Kluwer Academic

Publishers.