ArticlePDF Available

Resumptive pronouns, A'-binding, and levels of representation in Irish

January 2011

DOI:10.1075/lfab.5.02mcc

Authors:

James Mccloskey

University of California, Santa Cruz

Content uploaded by James Mccloskey

Content may be subject to copyright.

is is a contribution from Resumptive Pronouns at the Interfaces.

Edited by Alain Rouveret.

is electronic le may not be altered in any way.

e author(s) of this article is/are permitted to use this PDF le to generate printed copies to

be used by way of oprints, for their personal use only.

Permission is granted by the publishers to post this le on a closed server which is accessible

to members (students and sta) only of the author’s/s’ institute, it is not permitted to post

this PDF on the open internet.

For any other use of this material prior written permission should be obtained from the

publishers or through the Copyright Clearance Center (for USA: www.copyright.com).

Please contact rights@benjamins.nl or consult our website: www.benjamins.com

Tables of Contents, abstracts and guidelines are available at www.benjamins.com

John Benjamins Publishing Company

part i

Syntactic uniformity/diversity of resumption

Resumptive pronouns, Ā-binding,

and levels of representations in Irish

James McCloskey

University of California, Santa Cruz

At the empirical level, this paper aims to provide a reasonably systematic

overview of the distribution of resumptive pronouns in one language (Irish)

and of their behavior under certain tests. At the theoretical level, it argues that

resumptive pronouns are syntactic variables (locally Ā-bound elements) and

that there is a syntactically visible relation between the pronoun and its binder.

It follows that it must be possible for pronouns to be directly Ā-bar bound.

e paper also argues, however, that the pronominal character of resumptive

pronouns emerges in the so-called Highest Subject Restriction which is analyzed

as an anti-locality eect (reminiscent of Condition B of the Binding eory)

acting in the Ā-domain.

1. Introduction

is article has the relatively modest goal of contributing to our understanding of

resumptive pronoun structures – in Irish in particular and in a broad theoretical

perspective in general. Such structures have gured in a peripheral way in the dis-

cussion of many theoretical issues in recent years, discussion centered, for the most

part, on the related notions of “syntactic variable” and “Ā-binding” and on the many

questions that in turn demand a proper understanding of those notions. Resumptive

pronoun structures have gured in only a peripheral way in these debates, in part

because the most familiar and best-studied languages largely lack robust or productive

resumptive pronoun strategies. As a consequence, there is a dearth of the kind of data

needed to resolve the various questions of theory and analysis that arise concerning

resumptive pronouns.1

1. Important contributions have, of course, been made to the study of resumptive pronouns,

most notably by Peter Sells. See Sells (1984a, 1984b, 1987), as well as Zaenen et al. (1981),

Chomsky (1982), Chao and Sells (1983), Borer (1984), Jaeggli (1984), and Saﬁr (1986).

 James McCloskey

Initially, here, I attempt to narrow the data gap for Irish by providing a fairly

complete and systematic overview of the distribution of resumptive pronouns and of

their behavior under various tests. Beyond this overview, the article is concerned as

well with a number of related theoretical questions. e most basic of these is whether

or not resumptive pronouns are properly construed as syntactic variables. By syntactic

variable I understand the following: an element that is syntactically bound and whose

most immediate binder is an element in an Ā-position.

As pointed out recently by Lasnik and Stowell (1987) and by May (1989), almost

all recent discussions of weak crossover, of bound variable anaphora, and of the syntax

of quantication in general proceed from the assumption (explicit or implicit) that

direct Ā-binding of a pronoun is impossible. Since an important aim of all of this

work has been to assimilate the structures produced by syntactic WH-movement

and quantier raising (QR) at LF, a consequence of the assumption is that structures

such as (1) should he legal, whereas structures such as (2) should he illegal (where

no other binder, i.e. coindexed c-commanding element, intervenes between XPj and

the element it binds).

(1) [S′ XP [S… tj …]]

(2) [S′ XP [S … PROj …]]

Now, of course, to demonstrate that structures such as (2) are in fact well formed, it

is not enough simply to point to the existence of examples involving resumptive pro-

nouns, such as (3):

(3) an fear ar labhair mé lena mhac

the man  spoke I with-his son

“the man that I spoke to his son”

One might argue, for instance, that the pronoun in (3) is not syntactically bound at

all. It is perfectly possible to set up the mechanism of semantic interpretation so that

there is no syntactically visible relation between the “resumptive” pronoun and any

other syntactic position. Alternatively, one might argue that the pronoun is indeed

syntactically bound, but that it is coindexed with the head NP of the relative clause

structure, not with an operator within COMP. One would have to worry then about

whether or not the head position of relative clause structures should he interpreted

as an Ā-position. It is hardly obvious what the correct answer to this last question is.2

. See Saﬁr (1986) for an ingenious argument (based on the interaction between pied piping

and the licensing of parasitic gaps) that suggests strongly that binding by a relative head would

not count as Ā-binding.

Resumptive pronouns, Ā-binding, and levels of representation 

So the question of whether or not resumptive pronouns are (in some construc-

tions and in some languages at least) Ā-bound is tied intimately with the question

of what element it is exactly that functions as antecedent to a resumptive pronoun.

e problem of identifying with any precision or condence the antecedent of a

resumptive pronoun (the head NP or some element in COMP, for instance) is a subtle

one, and it is not immediately obvious what kind of empirical evidence bears on this

issue. One of the ways in which the Irish data are interesting is that they provide a

relatively precise and robust way of arguing about the issue.

ere is another way in which arguing about the status of (3) is less than

straightforward. Since the proposals of Chomsky (1982), it is clear that one cannot

simply ask, for a given language, whether or not resumptive pronouns in that lan-

guage are or are not Ā-bound. It is perfectly possible that resumptive pronouns might

be Ā-bound at one level of representation (say, at LF), but not bound at all at other

levels of representation (at S-structure. for instance). Teasing out the consequences of

the dierent assumptions one might make in this regard is, again, a delicate matter;

but Irish also has the merit of providing data that bear on the question in a rather

direct way.

Finally, if we admit the existence in principle of Ā-bound pronouns, then cer-

tain obvious questions arise immediately for binding theory. Resumptive pronouns,

if they are taken to be directly Ā-bound, are pronominal variables, that is, they are

simultaneously pronouns and variables. Questions arise immediately about how we

should expect them to behave with respect to binding theory. Will their status as pro-

nominal variables be related to the way they behave with respect to the conditions of

binding theory?

. Background

Before addressing such questions, it is useful to review rst the background assump-

tions about the syntax of Irish that lie behind the discussion pursued throughout

this article.

.1 Clausal structure in Irish

Irish is VSO in its nite clauses. e basic facts of word order in nite clauses can be

illustrated with the example in (4).

(4) Bhéarfaidh mé an t-airgead do Chaaimhín i nDoire inniu

give- I the money to Kevin in Derry today

“I’ll give Kevin the money in Derry today”

 James McCloskey

at is, unmarked order follows the formula given in (5).3

(5) Verb < Subject < Object < Oblique Arguments < Adverbials

In nonnite clauses, however, VSO order is impossible; the subject is always initial in

such clauses, as illustrated in (6).

(6) Níar mhaith liom [Caoimhín a theacht abhaile].

I-wouldn’t-like Kevin come- home

‘I wouldn’t like Kevin to come home.’

Furthermore, subjects are also initial in small clauses in Irish (Chung and McCloskey

1987).

(7) a. Ba mhinic [Caoimhín ag caoineadh].

- oen Kevin cry-

“Kevin was oen crying”

b. agus [Caoimhín ag caoineadh]

and Kevin cry-

“while Kevin was crying”

An analysis of Irish clausal structure is developed in Chung and McCloskey (1987),

which seeks to account for these clause types and their interrelationships by assuming

that all three types (small clauses, nite clauses, and nonnite clauses) share important

aspects of phrase structure. In particular, small clauses are construed as constituents,

with the internal structure in (8).

(8) SC

NP XP

Chung and McCloskey (1987) argue that small clauses (in Irish, at least) are Ss, an

assumption that brings both advantages and embarrassments. Since the issue is not

crucial for present concerns, I for now use the abbreviation “SC” for small clauses and

leave the issue of categorial identity open.

Structures such as (8) are in turn seen as important subconstituents of full clauses,

both nite and nonnite. is is achieved by assuming the structure in (9) for both

nite and nonnite clauses.

. For further exempliﬁcation and discussion, see Stenson (1981: 40–47). McCloskey

(1983a: 10–12), and Ó Siadhail (1989: 23–26).

Resumptive pronouns, Ā-binding, and levels of representation 

(9)

INFL SC

NP XP

e SC-constituent of (9) has an independent distribution (as illustrated, for instance,

in [7]) but functions also to build full clauses (nite and nonnite) when a sister

to INFL, as in (9). On this view, the verb-initial order seen in nite clauses can he

accounted for by assuming that in the case where XP of (9) is headed by V0, V0 moves

to INFL if INFL is [+Finite]. For nite clauses, then, one will have the kind of deriva-

tions schematized in (10).

(10) S

INFL

[+FINITE]

NP VP

(YP)

To account for the subject-initial order of nonnite clauses, it is assumed that this kind

of head movement may not apply in nonnite clauses. I have no proposal to oer as to

why this movement should take place obligatorily in nite clauses but be inapplicable in

nonnite clauses. It should be noted, however, that similar eects (the nonapplication

of V-to-INFL movement in nonnite clauses) are well attested for other languages–

for Welsh, see Sproat (1985); for French, see Emonds (1978), Taraldsen (1983), Pollock

(1989), and Chomsky (1988); for Vata and Gbadi, see Koopman (1984).4

ere is another fact about the order of elements in nonnite clauses that should

be noted here. Direct objects are generally preposed to a position le of the verb and

to the right of the subject, as seen in (11).

. I assume here only V-to-INFL movement in the derivation of ﬁnite VSO order. One might

wonder if there is not further movement of the INFL-V complex to COMP, as in many studies

of the verb-second phenomenon. e absence of a pattern of complementarity between

preposed ﬁnite Vs and the appearance of lexical complementizers suggests that this further

movement does not take place in Irish. e matter is far beyond present concerns, however,

and is not discussed further here. See Guilfoyle (1988) for further discussion of the structure

of clauses in Irish. It is possible that heads other than V0 also undergo head movement to

INFL. is is arguably the appropriate analysis of ﬁnite copular clauses in Irish.

 James McCloskey

(11) a. Níor mhaith liom [iad Ciarán a eiceáil]

I-wouldn’t-like them Ciaran see-

“I would not like them to see Ciaran”

b. Níor mhaith liom [Ciarán a eiceáil]

I-wouldn’t-like Ciaran see-

“I wouldn’t like to see Ciaran”

c. Níor mhaith liom [Ciarán labhairt leat mar sin]

I-wouldn’t-like Ciaran speak- with-you like-that

“I wouldn’t like Ciaran to speak to you like that”

I assume here, following Chung and McCloskey (1987) and McCloskey and Sells

(1988), that this order reects application of a rule that moves a NP governed by V0

and le-adjoins it to the VP headed by V0. An example such as (11a) will then have the

kind of structure illustrated in (12).

(12) S

INFL

[–FINITE]

NP VP

NPj

V NP

iad

a eiceáil

Application of this rule is subject to dierent conditions in dierent dialects. In the

northern group of dialects, it applies in all nonnite clauses; in the southern group

of dialects, it applies only if the subject is phonologically empty (i.e. in control and

raising clauses). For detailed discussion of these matters, see McCloskey (1980) and

McCloskey and Sells (1988).

. Null arguments and agreement

Irish is a null-subject language. More precisely, Irish is a null-argument language in

the following sense: any argument position whose person and number features are

identied by agreement morphology on a lexical head can be null. Under most cir-

cumstances, the condition is actually stronger. Argument positions whose person and

number features are identied by agreement morphology on a lexical head must be

Resumptive pronouns, Ā-binding, and levels of representation 1

null. ese patterns can be seen in the case of subjects, when the verb inects for the

person and number of its subject.

(13) a. Cuirim Ø suas é

put--1 up it

“I put it up”

b. *Cuirim mé suas é

put--1 I up it

“I put it up”

e same patterns exactly hold for the objects of prepositions, which inect for the

person and number of their objects.

(14) a. Labhair leo Ø!

speak- with-3

“Speak to them!”

b. *Labhair leo iad!

speak- with-3 them

“Speak to them!”

Finally, similar patterns hold for nouns, which take agreement prexes that identify

the number and person of possessors:

(15) a. mo theach Ø

1 house

“my house”

b. *mo theach mé

1 house I/me

“my house”

ese “null” eects have been studied in detail in McCloskey and Hale (1984) and

in McCloskey (1986b). ose papers present evidence that the empty position in

structures such as (13)–(15) is occupied by a null pronominal – the element pro of

much recent work (see Chomsky, 1982; in particular). We will take pro to be subject

to an identication condition that requires (in this and similar languages) that it be

governed by AGR, taken to be a possible subcomponent of the feature complexes that

dene INFL0, P0, and N0.

All these assumptions are defended in some detail in McCloskey and Hale (1984)

and in McCloskey (1986b). For the case of subject-verb agreement, I assume that the

relevant AGR feature is a subpart of INFL, as sketched in (16).

 James McCloskey

(16)

INFL

[AGR: αF]

pro

αF

V0jV0

(YP)

A great deal of evidence is adduced in Chung and McCloskey (1987) that INFL

governs the subject position in structures such as (9). In a structure such as (16), then,

the AGR complex within INFL will govern, and hence license, the instance of PRO that

appears in subject position. e complex of INFL + AGR + fronted V0 will ultimately

be realized as a verb inected for tense and for the features of its subject.5

. Ā-binding in Irish

Before beginning a detailed analysis of resumptive pronoun structures, we should rst

consider certain facts about the syntactic realization of Ā-binding in Irish. One of the

most useful features of Irish syntax in this domain is that it provides more than one

diagnostic for identifying the presence or absence of a resumptive pronoun in a given

clause. Specically, clauses are introduced by dierent COMP-particles depending on

whether they contain a gap, a resumptive pronoun, or no syntactically bound element

at ail. Consider the latter case rst. In (17) an example is given of a structure in which

a clause is embedded as complement to a verb.

(17) Dúirt sé gur bhuail tú é

said he  struck you him

“He said that you struck him”

Notice that the embedded clause is introduced by the (past tense) complementizer gur.

Consider now, though, what is essentially the same clause with a gap in object position:

(18) an fear [S′ a bhuail tú __]

the man  struck you

“the man that you struck”

e clause now is introduced by the complementizer a.

Consider, nally, the case of a clause containing a resumptive pronoun rather

than a gap:

. See Andrews (1988) for a detailed set of proposals about how syntax and morphology

interact in this domain of Irish grammar.

Resumptive pronouns, Ā-binding, and levels of representation 

(19) an fear [S′ ar bhuail tú é]

the man  struck you him

“the man that you struck (him)”

In this case, the relative clause is introduced by the complementizer ar. So we have

here a three-way distinction of clauses that involve no Ā-binding at all, clauses that

involve an Ā-binding that terminates in a gap, and clauses (to anticipate a little) that

involve an instance of Ā-binding that terminates in a pronoun. Now the phonological

and morphological mechanisms that serve to mark the distinctions between the vari-

ous complementizers are very complex, since they depend on such factors as the tense

of the associated clause (past or nonpast), the presence of a negative particle, and the

presence of the copula.6 Furthermore, the distinction between the dierent comple-

mentizers is sometimes a distinction not in segmental phonology but rather in the

initial mutation induced on the following element (typically the verb).

We can nesse these morphophonological complexities here by indicating

the distinction, when necessary, with dierent subscripts on the glosses associated

with the dierent complementizers. Henceforth, the complementizer go/gur, which

appears in the absence of any Ā-binding, is glossed simply “COMP”; the comple-

mentizer a, which is associated with the binding of a gap (i.e. with the application of

WH-movement) is glossed “COMPt”; and the complementizer a/ar, which is associ-

ated with the appearance of a resumptive pronoun, is glossed “COMPpro.” e basic

patterns at issue, then, are as illustrated in (20). e structures schematized in (20)

are relative clause structures, but, as discussed later, these patterns are the same for all

constructions involving Ā-binding.

(20) a. [NP NP [S′ t [S … t …]]]

b. [NP NP [S′ pro [S … pro …]]]

It is also convenient at certain points to use the abbreviations aL for COMPt, and aN

for COMPpro. ese abbreviations are based on the fact that COMPt induces the leni-

tion mutation on a following verb, whereas COMPpro (in the present tense) induces

the nasalization mutation on the initial of a following verb.7 ese distinctions are

. For some discussion of the syntactic structures that form the basis for these complexities,

see Chung and McCloskey (1987).

. Not all dialects have the full range of distinctions documented here. In some southern

varieties, for instance, there is not a distinction between COMPpro and the default comple-

mentizer go. In such varieties, (19) appears as in (i):

(i) an fear gur bhuail tú é

the man  struck you him

“the man that you struck”

 James McCloskey

important for the clues that they provide about the proper analysis of resumptive pro-

noun structures. ey are also extremely useful in helping us detect the presence of

resumptive pronouns when their presence would not otherwise be obvious.

Recall from the previous section that Irish is a null-argument language, in the

sense that pronouns are frequently null. Specically, pronouns are obligatorily null

when governed by (and hence identied by) AGR. e evidence for the pronominal

status of the null positions identied by AGR is extremely strong (McCloskey & Hale

1984; McCloskey 1986b) and is independent of current concerns. It is important to

note from the outset, then, that with respect to the COMP-distribution phenomena

discussed here, there is no dierence whatsoever between the behavior of phonologi-

cally specied pronouns and phonologically null pronouns. Both demand the appear-

ance of COMPpro when functioning as resumptives. is is illustrated for two instance

of null pro (object of an inected preposition and possessor in a NP containing an

agreement prex) in (21).

(21) a. an fear ar labhair tú leis pro

the man pro spoke you with-3-

“the man that you spoke to (him)”

b. an bhean a raibh [NP a mac pro] breoite

the woman pro was 3- son ill

“the woman that her son was ill”

Put simply, the rules or principles that determine the distribution of the dierent

complementizers do not care whether pro in (20b) is phonologically empty or phono-

logically specied. is fact is illustrated as the discussion proceeds by citation, when-

ever possible, of a pair of examples – one involving an audible pronoun, the other

involving pro – wherever some aspect of the syntax or interpretation of pronouns is

characterized.

One further piece of information needs to be given at this point. It is a salient

feature of Ā-binding structures in Irish that they exhibit a classic “successive cyclic

eect” with respect to the distribution of complementizers. We have just seen that

the language prescribes the use of a distinctive complementizer for clauses that form

a domain in which WH-movement has applied [see (17)–(19)]. It is also crucial to

note that, in the case of “long” movement, the same complementizer must introduce

In the speech of some younger speakers, the distinction between COMPt and COMPpro

has been lost (at least in informal registers), so that COMPt (aL) is also used in resumptive

pronoun relatives:

(ii) an fear a bhí tú ag caint leis

the man t were you talk- to-him

“the man that you were talking to”

Resumptive pronouns, Ā-binding, and levels of representation 

every clause that contains the WH-trace but not its ultimate antecedent. e eect is

to create a chain of COMPt-particles between (for instance) the head and the gap in a

relative clause structure, as schematized in (22) and illustrated in (23).

(22) [NP NP [S′ aL … [S′ aL … [S′ aL … t …]]]]

(23) an rud a shíl mé a dúirt tú a dhéanfá

the thing t thought I t said you t do--2

“the thing that I thought you said you would do”

For a detailed discussion of this eect, see McCloskey (1979, 1985) and Chung and

McCloskey (1987). We return to the analysis of these facts below; for now, though,

two points should be stressed: (1) the obligatory appearance of the pattern schema-

tized in (22) is characteristic not just of relative clause structures but of all Ā-binding

constructions in the language (see McCloskey, 1979, 1985, for detailed discussion and

exemplication) and (2) the pattern of (22) is not characteristic of resumptive pro-

noun structures. e pattern classically found with resumptive pronoun binding is

one in which COMPpro appears only in the topmost COMP-position, but in which

the default complementizer gol/gur appears in all lower COMP-positions within the

relative clause:

(24) a. an rud ar dúirt sé go gcoinneodh sé ceilte é

the thing pro said he  keep- he hidden it

“the thing that he said he would keep (it) hidden”

b. radharc ar bhreá liom a bheith ábalta a mhaíomh

sight pro I-would-really-like be- able claim-

go bhfacas é

 see--1 it

“a sight that I would love to be able to claim that I saw (it)”

We will return to this later.

. e distribution of resumptive pronouns

One of the important points of data that I want to establish in this article is that the

pattern of distribution of resumptive pronouns in Irish is extremely broad. By this I

mean two things:

(25) Resumptive pronouns appear in every WH-construction.

(26) Within each WH-construction, resumptive pronouns can appear in every

clausal position but one.

 James McCloskey

e documentation of these distribution patterns is an important task, but it is also

one that makes for tedious reading. For that reason, I have assembled the relevant data

in the Appendix of this article. Here I simply summarize the principal observations

suggested by an examination of that data.

It has been customary, at least since Chomsky (1977), to classify certain construc-

tions as unied by the characteristic of being built around an instance of WH-binding

(Ā-binding). Chomsky (1977) identies the following eight constructions as being of

this type: (1) restrictive and nonrestrictive relative clauses, (2) constituent questions,

(3) cles, (4) topicalization structures, (5) comparative and equative clauses, (6) inni-

tival relatives, (7) purpose clauses, and (8) complex adjectival (“Tough Movement”)

constructions. It is a fact about Irish that the class of construction types in which

resumptive pronouns can appear is simply and exactly the class of WH-constructions

dened by Chomsky (1977).8

Consider now the second distributional question posed above: within a single

clause, what is the range of positions in which resumptive pronouns can appear?

Ihave argued at length elsewhere (McCloskey 1985) that the underlying rule governing

the distribution of gaps and resumptive pronouns in Irish is extremely straightfor-

ward – unless independent principles intervene, there is a free choice between use

of a resumptive pronoun and use of a gap (WH-trace). e independent principles

that most frequently intervene are subjacency and the ECP. ere are many positions

from which WH-movement is impossible (from within NP or PP, for instance), and

in such positions the appearance of a gap at S-structure is, of course, ungrammati-

cal. For detailed discussion of these constraints, see McCloskey (1979, 1985). But, in

the familiar way, resumptive pronouns can be used in the positions from which WH-

traces are excluded, and in these circumstances use of the resumptive pronoun will

have the appearance of being obligatory. Hence the contrasts in (27) and (28), which

stand here for many similar cases. For more detailed exemplication, see McCloskey

(1979, 1985).

(27) a. na dánta sin nach bhfuil ios againn cén

the poems  pro+ is knowledge at-us what

áit ar cumadh iad

place  were-composed them

“those poems that we do not know where they were composed”

. ere is no topicalization structure distinct from cles (unless one analyzes the “reduced

cles” discussed in McCloskey 1979: 90–92, in this way). Le dislocation is considered here

in due course.

Resumptive pronouns, Ā-binding, and levels of representation 

b. *na dánta sin nach bhfuil ios againn cén

the poems  t+ is knowledge at-us what

áit ar cumadh

place  were-composed

“those poems that we do not know where were composed”

(28) a. an fear a raibh mé ag caint leis pro

the man pro was I talk- with-3-

“the man that I was talking to him”

b. *an fear a bhí mé ag caint le

the man t was I talk- to

“the man that I was talking to”

e binding of resumptive pronouns is not constrained by subjacency or by the ECP;

WH-movement, however, is so constrained.9 From these two assumptions, the con-

trasts seen in (27) and in (28) follow. Every case that I am aware of in which the use

of a resumptive pronoun has the appearance of being obligatory yields to this sort of

explanation: application of WH-movement is blocked by subjacency, or by the ECP

(see Chung and McCloskey 1987), and the only possible way in which the structure

can surface at all is as a resumptive pronoun structure.

But in those circumstances in which WH-movement can apply to leave an

S-structure gap, the option of using a resumptive pronoun as a bound element (with

the concomitant choice of complementizer) is also available. is pattern of systematic

optionality is illustrated for all major clausal positions in the examples cited in the

second part of the Appendix.

ere is, in fact, just one position from which resumptive pronouns are excluded.

e highest subject of a clause cannot be occupied by a resumptive pronoun.

(29) a. *an fear a raibh sé breoite

the man pro be- he ill

“the man that (he) was ill”

b. *na daoine a rabhadar pro breoite

the people pro be--3 ill

“the people that (they) were ill”

. Irish displays a very conventional array of island eﬀects; the complex NP-constraint

(both cases), the WH-island condition and the adjunct island condition are all observed.

For discussion and exempliﬁcation, see McCloskey (1979, 1985). is fact suggests that it

would be wrong to postulate any principled connection between the productive availability of

resumptive pronouns in a language and the apparent inapplicability of subjacency (see Sells

1984a: 255, for this suggestion; and Engdahl 1985: 10, for critical commentary).

 James McCloskey

Note that this is a restriction that applies only to the highest subject of a WH-clause.

As we have already seen, resumptive pronouns appear freely in the subject position of

embedded clauses, nite and nonnite:

(30) a. an t-ór seo ar chreid corr-dhuine go raibh sé ann

this gold pro believed a few people  was it there

“this gold that a few people believed (it) was there”

b. cúpla muirear a bhféadfaí a rá go

a-few families pro one-couId say- 

rabhadar pro bocht

be--3 poor

“a few families that one could say (they) were poor”

I refer to this restriction from now on as the highest subject restriction (HSR). It gures

in an important way in argumentation at a later point, and it is discussed in greater

detail then. For now, let us merely note that the HSR determines the only clausal

position from which resumptive pronouns are excluded.10

. Resumptive pronouns as variables

Consider now the analytical implications of the distributional pattern established in

the previous section and in the Appendix. What we have established is that resump-

tive pronouns appear in all the WH-constructions in the language, and that resump-

tive pronouns can appear (alongside WH-traces) in every clausal position except that

excluded by the HSR. Given the breadth and regularity of the distributional pattern,

one clearly does not want to license resumptive pronouns by proposing rules that

are either particular to given constructions or particular to given clausal positions.

emost economical statement, and the one that does justice to the pattern we have

seen, is (31).

(31) Ā-chains in Irish can terminate in pronouns.

1. It should be pointed out, though, that there are no adjunct resumptive pronouns. at is,

there are no structures such as the following: the place that we bought a house there or the time

that we bought a house then. is could simply reﬂect a lexical gap (i.e. the absence of adjunct

pronouns of category NP; see Kuroda 1968; Bresnan & Grimshaw 1978), or it may require

some deeper explanation.

Resumptive pronouns, Ā-binding, and levels of representation 

ln fact, it is proposed in McCloskey and Sells (1988) that A-chains can also have pro-

nouns as their lowest elements; if that is correct, then (31) is in fact only one aspect of

a broader statement.

e chain-formation rule in (31) has the consequence, of course, that resumptive

pronouns are dened as syntactic variables, since, given (31), the closest binder for a

resumptive pronoun is some c-commanding element in an Ā-position. at being so,

resumptive pronouns should show the characteristic behavior of variables. In particu-

lar, they should be subject to Condition C of the binding theory, since variables are

among the elements subject to this condition. To be more specic, one should nd

strong crossover eects, such as illustrated in (32), for resumptive pronouns.

(32) *Whoj did you think that hej said that Mary would marry tj?

Sentence (32) is ungrammatical because it fails to meet the requirements of binding

theory. Condition C requires that names and variables be A-free (lack a c-commanding

antecedent in an argument position) in the domain of their operators (cf. Chomsky

1982). e object trace in (32) is coindexed with the pronoun he, which is in an argu-

ment position (subject position) and is also within the domain of the operator (whoj)

that binds tj. Condition C is therefore violated in (32).

e question to address at this point, then, is the following: do resumptive pro-

nouns in Irish exhibit strong crossover eects? e answer is that they do, but a certain

amount of work has to be done to establish the point. Simple translation of the English

example (32), substituting the resumptive pronoun for the trace, tells us nothing.

(33) Cé ar shíl tú gur dhúirt sé go bpósfadh

who pro thought you  said he  would-marry

Máire é?

Mary him

“Whoj did you think that hej said that Mary would marry tj?”

Sentence (33) is perfectly grammatical. But notice that it will always be possible to

analyze an example such as (33) in such a way that the lemost or highest pronoun is

the bound variable. e second pronoun is then not a variable but instead is simply

coindexed with the pronominal variable in a routine way analogous to English exam-

ples such as Whoj do you think tj said Mary would marry himj? Nothing can be deduced

from such examples. We must work a little harder to make the argument.

Specically, substituting an epithet for the c-commanding pronoun in examples

such as (33) eliminates the possibility of construing the c-commanding element

itself as a variable. For epithets cannot themselves function as resumptive pronouns

(i.e.asvariables).

 James McCloskey

Note rst that epithets do in fact give rise to strong crossover eects when they

c-command, and are coindexed with, R-expressions (variables or lexical NP).11

(34) a. e bastardj said hej would resign.

b. *e bastardj said Johnj would resign.

c. *Whoj did the bastardj say tj would resign?

Sentence (34b) is ungrammatical because the lexical NP John is c-commanded by, and

coindexed with, the epithet the bastard. Sentence (34c) is ungrammatical because the

embedded subject trace is A-bound by the epithet the bastard, giving rise to a Condi-

tion C (strong crossover) violation. Consider now the Irish examples in (35)–(37).

(35) Sin an fear ar dhúirt mé le tuismitheoirí an bhastaird

that the man pro said I with parents the bastard

gur cheart é a chaitheamh isteach i bpríosún.

+ right him throw- into in prison

“at is the man that I said to the bastard’s parents that he should be thrown

into prison”

(36) *Sin an fear ar dhúirt an bastard go maródh sé muid

that the man pro said the bastard  would-kill he us

“at is the man that the bastardj said hej would kill us”

(37) *Sin an fear ar dhúirt mé leis an bhastard gur

that the man pro said I with the bastard +

cheart é a chaitheamh isteach i bpríosún.

right him throw- into in prison

“at is the man that I said to the bastard that he should be thrown in prison”

11. e status of epithets with respect to binding theory has been discussed by Lasnik (1976,

1989), among others. Lasnik (1989) shows in particular that epithets simultaneously exhibit

the properties of R-expressions and of pronouns with respect to the binding conditions. is

is true of their behavior both as binding and as bound elements. In particular, when func-

tioning as c-commanding antecedents to lexical NPs, epithets pattern with pronouns in giving

rise to robust Condition C violations, even in languages (such as ai and Vietnamese) in

which lexical NPs give rise to only very weak or nonexistent Condition C eﬀects in the same

conﬁguration. e same contrasts are palpable in English:

(i) ??Johnj said that Johnj would come.

(ii) *e idiotj said that Johnj would come.

(iii) *Hej said that Johnj would come.

Epithets thus provide a particularly good way to test for Condition C eﬀects.

Resumptive pronouns, Ā-binding, and levels of representation 1

Sentence (35) is grammatical. is is because the epithet does not c-command the

resumptive pronoun and there is, therefore, no Condition C violation (the vari-

able is A-free, as required). Sentence (36) is ungrammatical. Here the epithet does

c-command the resumptive pronoun, and a strong crossover conguration results.

Sentence (37) is also ungrammatical. is is because in this case, too, the epithet

c-commands the resumptive, giving rise to a strong crossover conguration.12 ese

facts fall naturally into place if resumptive pronouns are in fact Ā-bound and are

therefore syntactic variables.

. e nature of the binding relation

e distributional arguments, and the arguments from binding theory, suggest that it

is correct to construe resumptive pronouns in Irish as Ā-bound variables. But this con-

clusion still leaves a number of issues unresolved. In particular, the question of what

precisely functions as antecedent to the resumptive pronoun remains to be settled.

Second, it remains unclear at what level of representation (D-structure, S-structure,

orLF) the binding is dened. It is the purpose of this section to try to establish answers

to these and related questions.

.1 e antecedent

Consider a relative clause such as (38a) and its schematic structure (38b).

(38) a. na míreanna a leanann séimhiú iad

the particles pro follow Lenition them

“the particles that Lenition follows (them)”

b. [NP NP [S′ pro [S … pro …]]]

1. is interpretation of (37) requires the assumption that subcategorized prepositions

(“case-marking” prepositions) do not “count” (to speak loosely) for the computation of

c-command conﬁgurations in Irish. is interpretation is required, in any case, for other

reasons (see Chung & McCloskey 1987: n. 26). One could perhaps conclude from this observa-

tion that phrase structure conﬁgurations at S-structure do not form the appropriate domain

over which to deﬁne the kind of command relations needed in the statement of anaphoric

(inthe broad sense) conditions. Alternatively, one might conclude with Stowell (1990) that the

so-called prepositions of Irish are in fact case-marking preﬁxes, such that the phrase leisan

bhastard of (37) will be a NP and the familiar deﬁnition of c-command will work in the

appropriate way. What is needed for our current argument is the observation itself, not the

analysis of the observation. Some speakers report a subtle diﬀerence between (36) and (37),

the latter being preferred. is diﬀerence (if real) could be related to the presence of the

governing preposition in (37). In general, judgments in this domain are surprisingly robust.

 James McCloskey

If one asks what precisely the antecedent is for the resumptive pronoun in (38), there

are two answers that suggest themselves, that is, two hypotheses that grow naturally

out of recent thinking on these and related matters.

One obvious possibility is that the head NP is the antecedent. is answer is com-

patible with the conclusions we have established so far, to the extent that it is plausible

or implausible to call the head position of a relative clause structure an Ā-position.

Corresponding hypotheses are available for the other WH-constructions in which

resumptive pronouns appear – the cleed phrase will be the antecedent in cles and

so on. For ease of exposition, I restrict attention here to the case of (restrictive) relative

clauses. e hypothesis being considered, however, can be extended, in mostly obvious

ways, to all the other WH-constructions.

e other possibility is to take the antecedent of the resumptive pronoun to be a

null operator within COMP. More specically, we can adopt the theory of the phrase

structure of clausal categories (in particular, S′) developed by Chomsky (1986a). is

proposal seeks to rationalize the analysis of the categories S and S′ within X-bar theory.

S′ is taken to be a projection of the zero-level category C0 and is taken to be a com-

pletely regular projection, in that C0 has a complement and a specier:

(39) CP

XP C1

C0S (=IP)

e specier position, XP in (39), is taken to be the landing site for WH-movement

and is, therefore, the locus of the various WH-operators that function as Ā-binders

at LF. A natural extension of this view is to propose that there are also null operators

in this position for resumptive pronoun structures, and that it is these operators that

function as antecedents to resumptive pronouns.

It is the purpose of this section to argue that the rst of these hypotheses is wrong

and that the second is right. To make this case, we have to consider in more detail the

nature of the HSR and also to look more closely at the system of principles that deter-

mines the distribution of complementizer particles.

We can begin with a consideration of the HSR. Recall that this is a constraint that

bars the appearance of a resumptive pronoun in a relative clause in the subject position

immediately subjacent to the head:

(40) a. *an fear a raibh sé breoite

the man pro was he ill

“the man that (he) was ill”

Resumptive pronouns, Ā-binding, and levels of representation 

b. *na daoine a rabhadar pro breoite

the people pro be--3 ill

“the people that (they) were ill”

ese facts reect a requirement that pronouns, if bound, be nonlocally bound; that is,

they have the character of a disjoint reference eect – a Condition B eect that applies

in the domain of Ā-binding rather than in the domain of A-binding. Like ConditionB,

this is a constraint that has the eect of rendering ungrammatical those structures

in which a pronoun (a pronominal variable, in this case) is too close to its anteced-

ent. e positions in which resumptive pronouns can legally appear within a clause

(direct object, prepositional object, possessor position within NP) are all separated

from their respective antecedents by the intervention of at least one more maximal

projection boundary (VP, PP, NP) than is true in the case of subjects. Recall also that

if the resumptive pronoun appears in embedded subject position, the eect disappears

and the structure is grammatical.

(41) an fear ar dhúirt mé go dtiocfadh sé

the man pro said I  would-come he

“the man that I said (he) would come”

is observation, too, suggests that syntactic distance plays a role in the HSR.

e disjoint reference character of the eect can also be seen in the fact that if one

arranges for the resumptive pronoun to be embedded within a subject NP, the eect

also disappears. is is true (unsurprisingly) for possessors within the subject:

(42) an ghirseach a bhfuil a máthair breoite

the girl pro is 3- mother ill

“the girl whose mother is ill”

More surprisingly, the same eect can be observed in the case of coordinate subjects:13

(43) a. duine ar bith a mbeadh sé féin

anyone pro would-be he+

agus Tom mór le-n a chéile

and Tom great with each other

“anybody that he and Tom would be very fond of one another”

b. daoine a rabhadar pro féin agus a gclann

people pro be--3 pro+ and their family

mhac ábalta ar iascach

sons- able on shing

“people that they and their sons were capable of shing”

1. For detailed discussion of the syntax of the kind of coordinate structures seen in (43b),

and of their interaction with agreement and identiﬁcation, see McCloskey (1986b).

 James McCloskey

For these reasons and others, a great deal of previous work on this topic (includ-

ing McCloskey 1983; Borer 1984) has been based on the idea that the HSR is to be

accounted for by extending the domain of application of Condition B in some appro-

priate way, so that it encompasses Ā-binding as well as A-binding. Recent work by

Aoun and Li (1989) also explores this idea, drawing on data concerning the distribu-

tion of pronouns bound by quantier expressions in Mandarin Chinese. ey propose

that bound pronouns (those linked to antecedents that at LF occupy Ā-positions) are

subject to a requirement of “Ā-disjointness” that governs the distribution of referential

pronouns. eir proposal, if modied slightly as in (44), will account for the data we

are concerned with here.

(44) A pronoun must be Ā-free in the least complete functional complex containing

the pronoun and a subject distinct from the pronoun.

e notion “complete functional complex” (CFC) is to be understood in the sense of

Chomsky (1986b: 169). e term subject in (44) refers to the NP-position of structures

such as (9) and (10). More precisely, a subject is a NP that is governed by INFL.

e basic examples in (40) will now be ungrammatical because the minimal CFC

that contains the pronoun and a subject distinct from the pronoun will be the higher

S in which the whole relative structure is embedded. e pronoun is not Ā-free in this

domain – whether one takes its antecedent to be the head NP or a null operator in

COMP, this antecedent will be contained within the higher S in violation of (44). For

the case of an embedded subject, as in (41), the minimal CFC containing the resump-

tive pronoun and a distinct subject will again be the next highest S of the relative

clauses structure. But the pronoun is Ā-free in this domain. Again, this conclusion is

independent of what one takes the immediate antecedent of the resumptive pronoun

to be. e next S up in a structure such as (41) contains neither the head NP, nor, on

the alternative conception, the null operator in Spec of COMP. For cases (42) and (43),

in which the resumptive pronoun in each instance is contained within a larger sub-

ject NP, the minimal CFC containing the pronoun and a distinct subject is simply the

minimal S containing the resumptive. But the pronoun is Ā-free in this domain, and

the structure is therefore expected to be grammatical, as is the case. Once again, this

conclusion holds no matter which hypothesis one endorses about the antecedent; both

the head NP and a null operator in Spec of COMP are external to S.

Principle (44) also accounts for some data we have not yet considered. Recall that

objects are preposed and le-adjoined to VP in nonnite clauses:

(45) a. Caithdh mé Ciarán a eiceáil

must I Ciaran see-

“I must see Ciaran”

b. [S NPsubj [VP NPobj [VP V tobj]]]

Resumptive pronouns, Ā-binding, and levels of representation 

Note that since the preposed object is in VP-adjoined position, it is in an Ā-position.

Ifthis is so, and if it is also correct, as we have argued, that Ā-chains can always ter-

minate in pronouns in Irish, then the question arises of why the structure illustrated

in (45) does not permit a resumptive pronoun in direct object position, as illustrated

in (46).

(46) a. *Caithdh mé Ciarán a eiceáil é

must I Ciaran see- him

“I must see Ciaran”

b. *Caithdh mé Ciaran dhá eiceáil pro.

must I Ciaran 3-+see-

“I must see Ciaran”

But notice that (44) provides a direct account for the ungrammaticality of (46). e

minimal CFC containing the pronoun and a subject distinct from it is the lower S;

but the pronoun is Ā-bound within this domain, and the structure therefore stands in

violation of (44).14

1. “Object” preposing also operates from the subject position of small-clause complements

to certain verbs (perception verbs, for instance; McCloskey & Sells 1988: 163–164). Peter Sells

has pointed out to me that this fact poses a potential problem for the view of object preposing

outlined in the text. If one had a coordinate subject in the small clause, one might expect a

resumptive pronoun within the coordinate subject to be bound by the preposed NP (on the

analogy of structures such as [43]). But such structures are ungrammatical:

(i) *Níor mhaith liom tú Eoghan a shamhlú é agus Ciarán ag imeacht

I-wouldn’t-like you Eoghan imagine him and Ciaran leave-

“I wouldn’t like you to imagine Ciaran and Eoghan leaving”

If (i) is to be declared ungrammatical by (44), the least CFC containing the resumptive

pronoun and a subject distinct from the pronoun must be the nonﬁnite clause headed by the

verb samhlaigh (“imagine”). is consequence, however, already follows given the deﬁnition

of subject associated with (44), namely, that it is the NP governed by INFL. e closest such

subject in (i) is the subject of the nonﬁnite clause. e nonﬁnite clause, then, deﬁnes the

domain in which the resumptive pronoun must be Ā-free. It is in fact Ā-bound in this domain,

and (i) is, as a consequence, ungrammatical.

Similar remarks apply in the case of copular clauses, as in (ii) and (iii). Sentence (iii)

represents the subclass of copular structures that are cles. As shown in (iii), it is possible for

the focus position of a cle to be occupied by a resumptive pronoun.

(ii) girseach ar de bhunadh Protastúnach í

girl pro +  of people Protestant her

“a girl who is of Protestant stock”

(iii) fear arbh é a bhí breoite

man pro +  him t was ill

“a man that it was him who was sick”

 James McCloskey

is interpretation of the HSR has certain empirical advantages over previous

proposals that have been made to account for the relevant data. Borer (1984) suggests

extending Condition B of the binding theory directly to Ā-binding. e idea is that

pronouns must be simply free (that is Ā-free and A-free) in their governing categories.

Subjects are taken to be unique in having two governors: INFL and COMP. Subjects

are thus distinct from all other clausal positions in having S′ rather than S as their gov-

erning category. Subject resumptive pronouns will always be bound within S′ and will

thereby always violate Condition B. Resumptive pronouns in other positions will also

be bound within S′, but this will not matter since their governing category is S and in

this category they are Ā-free.

is approach can be criticized on two levels. It is, to begin with, very unclear how

it will be possible to reconcile the idea that subject NPs are governed both by INFL and

by COMP with the minimality requirement on government. Yet much recent work

suggests that minimality is a crucial condition on government relations.

Second, it seems that this approach will have diculty in making the crucial dis-

tinction between (40) and (43). e government relations in which the rst NP of a

coordinate subject is involved seem to be exactly the same as those in which a simple

subject NP is involved. at is, it is just as true of sé féin in (43a) that it is governed

by INFL as it is of sé in (40a). is is argued in some detail in McCloskey (1986) on

the basis of evidence from case marking and identication of pro. e analysis in that

study suggests that there is no dierence in government relations between the subjects

of (40) and the subjects of (43). If this line of analysis is right, then it will be dicult to

make a distinction between the ungrammatical (40) and the grammatical (43) on the

basis of the notion of “governing category.”

ese observations also argue against the proposals made in Sells (1984b). ose

proposals are made within the context of a set of analytical assumptions very dier-

ent from those adopted here, but the basic idea can be translated and paraphrased as

in(47).15

ese examples will be grammatical to the extent that it is plausible to regard the resumptive

pronouns as not being subjects in the sense deﬁned in association with (44). In fact, this

position is not only plausible but also virtually demanded when one begins to look at the

syntax of the copula in a serious way. ere is a good deal of evidence that such structures

contain a null expletive that is governed by INFL (which is arguably the copula itself). is,

and not the resumptive pronoun, is the subject of the clause. Note that this conclusion is

already suggested by the fact that the pronouns in (ii) and (iii) are nonnominative. e

resumptive pronouns in (ii) and (iii), then, will have exactly the same status as any nonsubject

resumptive pronoun.

1. See Sells (1984b: 231–266), in particular, the case domain ﬁlter (p. 239).

Resumptive pronouns, Ā-binding, and levels of representation 

(47) An operator cannot bind a pronoun in a position assigned Case by that

occurrence of  that is most closely associated with the operator.

Examples such as (40) will then be ungrammatical because they each contain an

operator-INFL complex in which INFL assigns nominative case to a pronoun bound

by the operator. But Principle (47) still fails to make the necessary distinction between

the ungrammatical (40) and the grammatical (43), since (43) is not dierent from (40)

in any sense relevant to the function of the ller in (47).16

Whatever the specic or ultimate merits of the Ā-disjointness requirement as

stated in (44), this general approach to the HSR (in which it derives from a Condition-

B-like principle operating in the domain of Ā-binding) seems right. What I want to

consider next is the interaction between this principle and the rules or principles that

determine the distribution of complementizer particles. is interaction provides an

important clue about the exact nature and position of the antecedent in resumptive

pronoun structures.

Recall from our earlier discussion that long-distance relatives are subject to a

requirement that each COMP-position between a WH-trace and its antecedent binder

must be occupied by the complementizer a, in the pattern of (48).

(48) a. an rud a shíl mé a dhéanfainn

the thing t thought I t do--1

“the thing that I thought I would do”

b. [NP NP [S′ aL … [S′ aL … t …]]]

Recall too that this is not the pattern that is in general found in the case of long rela-

tives built around the binding of a resumptive pronoun. In this case, one nds the

distinctive COMPpro in the highest position, but the default complementizer in the

intermediate positions.

(49) a. an fear ar shíl mé go dtiocfadh sé

the man pro thought I  come- he

“the man that I thought (he) would come”

b. [NP NP [S′ aN … [S′ go … pro …]]]

Now this dierence in the two complementizer patterns is very natural, given certain

widely accepted assumptions. If WH-movement is constrained by subjacency and

1. Sells notes a similar diﬃculty involving the corresponding Hebrew data (1984b: 252–253),

but he suggests that the resumptive pronoun might be “intrusive” in the sense of Chao and

Sells (1983), that is, not truly Ā-bound. If this were so, however, quantiﬁed head NPs should

be impossible in such structures. But this is not the case, in Irish at least, as we have already

seen in (43a).

 James McCloskey

operates in a successive cyclic fashion, and if the appearance of the complementizer aL

is a formal mark of the application of WH-movement (in a sense soon made precise

here), then we expect the appearance of aL in every clause that has dened a domain of

application for WH-movement. But as we have already seen, the binding of a resump-

tive pronoun is not a relation constrained by subjacency, and there is, therefore, no

reason to expect that the distinctive complementizer COMPpro would appear at inter-

mediate COMP-positions.

But of more immediate concern is the fact that besides the pattern in (49), another

possibility exists for the case of long relatives with resumptive pronouns. is is the

pattern schematized in (50) and exemplied in (51)–(53):17

(50) [NP NP [S′ aN … [S′ aN … pro …]]]

(51) an talamh ar mheas mé a raibh an ceol

the land pro thought I pro was the music

ag teacht as pro

come- from-3-

“the land that I thought the music was coming from (it)”

(52) an méid den dán ar mheas sé a raibh feidhm

the amount of-the poem pro thought he pro was use

leis pro

with-3-

“as much of the poem as he thought was needed”

(53) an bhean a raibh mé ag súil a

the woman pro was I expect- pro

bhfaighinn uaithi pro arís é

get--1 from-3- again it

“the woman from whom I was expecting that I would get it again”

What we have in these structures is the appearance of the distinctive aN complemen-

tizer associated with the binding of a resumptive pronoun, at every COMP-position

between the pronoun and the head of the relative structure. is pattern is very

highly marked in comparison with the pattern in (49). Its marked character can be

seen in a number of ways. In the course of some ten years of searching for naturally

occurring examples of type (50) in speech or in writing, I have come across just the

three examples cited in (51)–(53). is contrasts with some y attested examples of

the pattern illustrated in (49). e marked character of the construction in (50) also

emerges in the degree of uncertainty and insecurity reported by native speakers in

1. Example (53) is cited in de Bhaldraithe (1956–1957). See also Ó Siadhail (1980: 90).

Resumptive pronouns, Ā-binding, and levels of representation 

their judgments about the relevant examples. Some speakers refuse to accept examples

such as (51)–(53); others report uncertainty about the status of these examples. More-

over, there is a great deal of uctuation from speaker to speaker and from session to

session in elicitation. It seems then, that the structure in (50) must have a place, albeit

a peripheral and highly marginal place, in the grammatical system of the language.

We return to the problem of accounting for the peripheral status of these structures

shortly. For the moment, we are concerned with two questions: rst, how the structures

can be licensed at all and, second, what implications the existence of such structures

has for the problem of identifying the immediate antecedent to a resumptive pronoun.

With regard to the second question, the crucial observation is this: the uncer-

tainty and insecurity that generally surround speakers’ reactions to (50) disappear

completely in the face of examples such as (54), in which the resumptive pronoun

appears in the subject position of the embedded clause. Examples such as those in

(51)–(53) are marginal but clearly acceptable to some speakers. Example (54), on the

other hand, is thoroughly and unambiguously ungrammatical.

(54) *an fear ar shíl mé a raibh sé breoite

the man pro thought I pro was he ill

“the man that I thought (he) was ill”

Consider the implications of this fact. Clearly we have here another reection of the

HSR. Viewed in isolation, or viewed locally, the conguration of the embedded clause

in (54) – a raibh sé breoite – is identical to the ungrammatical relative clause in the

core cases that illustrate the HSR, namely, the examples in (40). If we want to account

for these two facts by means of the same formal mechanism, then the explanation for

the ungrammaticality of (54) must be sought locally, that is, within the domain of the

embedded clause. If the Ā-disjointness requirement (or something like it) is to account

for the ungrammaticality of (54), then the subject resumptive pronoun must have an

antecedent that is “too close” to it within the embedded clause.

Now it is very dicult to see how this intuition could be given formal expression if

we make the assumption that the closest antecedent for the resumptive pronoun is the

head NP. In particular, it is very dicult to see how the robust contrast between (54)

and (55) could be accounted for in terms of the HSR.

(55) an fear ar shíl mé go raibh sé breoite

the man pro thought I  was he ill

“the man that I thought (he) was ill”

ere is no obvious sense in which the antecedent is “closer” to the resumptive pro-

noun in (54) than in (55), if we make the assumption that the antecedent is the head

NP. More specically, there seems to be no way for the Ā-disjointness requirement to

 James McCloskey

make any distinction between (54) and (55), if we take the position that the resumptive

pronoun is locally bound by the head NP.

But a simple and elegant connection is drawn among the ungrammaticality of (54),

the ungrammaticality of (40), and the grammaticality of (55) if we make the alterna-

tive assumption being considered here – namely, that the antecedent to the resumptive

pronoun is a null operator in Spec of COMP.

We can reason as follows. In the simplest cases, we have to deal with two kinds of

relative clauses:

(56) a. [NP NP [CP aL [… t …]]]

b. [NP NP [CP aN [… pro …]]]

ere is good reason to believe that both aL and aN are lexical complementizers; in

particular, there is rm evidence for the case of the movement type (56a) that the par-

ticle aL is not itself the moved phrase.18 What this suggests is that we are dealing with

structures such as (57).

(57)

[αF]j

[αF]

An Ā-chain is formed between the trace and the operator phrase in Spec of CP. It is

irrelevant for present concerns whether this chain is formed by way of movement, or

by way of base generation of the operator phrase and chain formation at S-structure.

Itseems reasonable to believe that items in the same chain must share feature speci-

cations; the morphological case of a moved WH-phrase, for instance, is normally that

assigned to the position occupied by the lowest trace it binds. If that is so, the crucial

1. See McCloskey (1979: 10–16); note, in particular, that aL can be used as a complemen-

tizer in a range of cases that clearly do not involve WH-movement:

(i) Is amhlaidh a bhí neart capaill ann

 so t was strength horse in-him

“It is a fact that he had the strength of a horse”

(ii) Sé rud a beireadh air i nDoire.

it-is thing t was-caught on-him in Derry

“e fact is he was captured in Derry”

Resumptive pronouns, Ā-binding, and levels of representation 1

features identifying the empty category under S as a WH-trace will also appear on

the NP in Spec of CP. ese same features will also be shared by C0 by way of the

mechanism of Spec-Head agreement discussed in Chomsky (1986a).19 e form of

resumptive pronoun relatives will be exactly analogous:

(58)

[αF]j

pro

[αF]j

[αF]

We thus have a way of making a connection between the character of the variable

(whether it is a gap or a pronoun) and the choice of complementizer particle in a rela-

tive structure. e connection is made by way of two mechanisms – feature sharing

among elements of a chain and Sped-Head agreement – that are of considerable gen-

erality and that have much independent justication. e nal element in the analysis

is the element that must minimally be there on any account, namely, a featural distinc-

tion between the two relevant complementizers, aL and aN.20

What are the relevant features? An obvious rst guess is that they are the features

[anaphoric] and [pronominal], rst introduced in Chomsky (1981, 1982) to distin-

guish among the various nominal categories relevant to the functioning of binding

theory. Within that system of denitions, we have the specications in (59).

(59) Pronouns:a.

b. WH-traces:

+pronominal

–anaphoric

–pronominal

–anaphoric

1. See Rizzi (1990) and Chung (1990) for extensions of the mechanism of Spec-Head agree-

ment to clausal projections.

. Note that this analysis succeeds in doing what McCloskey (1979) signally failed to do:

it accounts for complementizer selection in terms of a strictly local selectional mechanism

(Spec-Head agreement).

 James McCloskey

If we assume this featural distinction between WH-traces and pronominals, then the

features in question will eventually pass to the C0 position of the clause that contains

the binding operator (by way of the mechanisms just discussed), and the two comple-

mentizers can be distinguished thereby.21

is view of matters is a close formal expression of the intuition that the func-

tion of the complementizers aL and aN is to signal the presence of an operator of a

kind appropriate to bind either a trace or a resumptive pronoun. e function of the

complementizer go, on the other hand, is to signal the absence of a binder. e basic

facts of the HSR, as we have seen, follow from this conception since we will get HSR

eects just in case the complementizer aN signals the presence of an operator that is

“too close” (in the sense we have dened in terms of the Ā-disjointness requirement

in [144]) to a subject resumptive pronoun. When go signals the absence of such an

operator, there is no HSR eect. e contrast between (55), on the one hand, and

examples(54) and (40), on the other, thus falls out straightforwardly. e three struc-

tures are as in (60).

(60) a. [NP NP [CP Opj aN [… [CP go [V proj …]]]]] [= (55)]

b. [NP NP [CP Opj aN [… [CP Opj aN [V proj …]]]]] [= (54)]

c. [NP NP [CP Opj aN [V proj …]]] [= (40)]

is general line of analysis, then, accounts for the complete ungrammaticality of (54)

and links it to the ungrammaticality of (40) in a satisfying way.

Notice that what we are le with now is a mixed system of denitions in which

nominal categories such as anaphor, pronominal, and R-expression are dened inher-

ently in terms of the features [pronominal] and [anaphoric], but in which the notion

of variable is dened contextually. at is, a variable is an element locally bound by an

element in the Ā-position.22

Consider now the question of how the operators that bind resumptive pronouns

get to be in Spec of CP. Notice that they must be positioned there by means of base

generation. Or, at any rate, they cannot get to their ultimate position by way of move-

ment from the position they bind, since the relation between such operators and the

position they bind is not constrained by subjacency. We assume, then, that the operator

that binds resumptive pronouns is base generated in Spec of CP.

It is presumably the case that at least the topmost COMP-projection in a rela-

tive clause structure must contain a binding operator. is seems to be a general

requirement and must reect a principle whereby relative clause structures cannot

1. For the cases of Note18, I assume that the two items in question (amhlaidh and sé rud)

are lexically speciﬁed to take a complement clause headed by C0 [−pronominal, −anaphoric].

. is is also the position advocated in Saﬁr (1984).

Resumptive pronouns, Ā-binding, and levels of representation 

be semantically interpreted unless there is a binding operator in the COMP-position

closest to the head NP. Put another way, the presence of a WH-operator in the top-

most COMP-projection is an obligatory syntactic correlate of the semantic process of

λ-abstraction that forms the property predicated of the head NP in a relative clause

structure. As a consequence of this requirement, the topmost instance of the comple-

mentizer COMPpro is also required by Spec-Head agreement.

But it is also true that, given a structure such as (61), there is no reason why

an operator, although not required, should not be base generated in the specier

position of the lower C-projection, as well as in the specier position of the higher

C-projection.

(61) [NP [CP XP C0 [S … [CP XP C0 [S … pro …]]]]]

As long as the Ā-disjointness requirement in (44) is observed, nothing we have said

rules out a structure such as (62).

(62) [NP [CP Opj C0 [S … [CP Opj C0 [S … pro …]]]]]

is structure gives rise to the highly marked examples seen in (50)–(53), once Spec-

Head agreement has enforced appearance of the appropriate complementizer at each

C0-position associated with an operator.

is result will be a welcome one to the extent that we can now give some account

of the marked and peripheral status of these examples. One possible account would

be to appeal to the principle of economy of representation proposed and discussed by

Chomsky (1988). It is true, of course, that the intermediate operators and the inter-

mediate links in the Ā-chain that characterize (62) and (50)–(53) are unnecessary:

their appearance is not mandated by any principle of grammar, or by the demands of

semantic interpretation. is is just the kind of situation that the principle of economy

of representation disfavors. If that principle can be appropriately interpreted not as an

absolute prohibition on the structures that are in violation of it but rather as one that

assigns a high cost realized as markedness, then we will have an account of the mar-

ginal character of (50)–(53).

ere is one last issue in this domain that we must consider before proceeding.

Wehave just seen that a resumptive pronoun can be bound by an operator in the mini-

mal clause (CP) in which it is contained. We have also seen that this is a highly marked

option. It is important to observe at this point, however, that once this marked option

is chosen, certain other requirements must be met. In particular, ifa binding operator

is present in the C-projection that most immediately dominates the resumptive pro-

noun, then each successively higher C-projection must also contain a binding operator.

at is, the successive cyclic requirement reinstates itself, so to speak, once the option is

 James McCloskey

taken of placing a binding operator for the resumptive pronoun in the lowest projection

of C. us, (63) contrasts with (64) in being completely ungrammatical.

(63) *an talamh ar dhúirt tú gur mheas tú a raibh

the land pro said you  thought you pro was

an ceol ag teach as

the music come- out-of-it

“the land that you said you thought the music was coming from”

(64) an talamh ar dhúirt tú ar mheas tú a raibh

the land pro said you pro thought you pro was

an ceol ag teach as

the music come- out-of-it

“the land that you said you thought the music was coming from”

In (63), the chain of successively subjacent aN particles has been broken by the go

particle, which signals, on our assumptions, the absence of a binding operator in the

second C-projection of the series. In (64), the chain is not broken. I want to stress at

this point that, although I believe that the contrast represented by (63)–(64) is real, we

are working here at a level of delicacy and marginality that makes it extremely dicult

to be condent that the observations one is trying to account for are real. e struc-

tures exemplied in (50)–(53) are already extremely marginal. To test for the contrast

explored in (63) and (64), we must increase the level of embedding from that seen in

(50)–(53), thus increasing one of the features that contributes to uncertainty. ese

caveats aside, there does seem to be a strongly felt contrast between (63) and (64).

What does this indicate?

ese data in fact provide a clue toward the resolution of an issue that we have

not so far addressed, namely, the nature of the null operators that we are assuming as

binders for both WH-traces and resumptive pronouns. What these observations sug-

gest is that these operators themselves must be locally bound. What principle would

ensure this?

Lasnik and Saito (1984) have shown that a trace le in COMP by successive cyclic

WH-movement is dependent on its antecedent for purposes of satisfying the ECP,

just as is a subject or adjunct phrase. Now, if the binding operators that bind resump-

tive pronouns are themselves empty categories subject to the ECP, then the operator

whose presence is marked by aN in examples such as (63)–(64) will require anteced-

ent government, as does a trace in COMP. is condition is met in (64), since each

intermediate binder is also bound by the operator in the next highest COMP, but not

in (63). In (63), the appearance of go means that there is no binder in this intermedi-

ate position to function as antecedent governor for the lowest operator (which locally

binds the resumptive pronoun). is view of matters suggests that the operators that

Resumptive pronouns, Ā-binding, and levels of representation 

bind traces and resumptive pronouns in Irish are essentially indistinguishable from

traces in COMP – that is, they are empty categories subject to the ECP.23

We have at this point assembled an account for all the data considered so far.

We have argued that resumptive pronouns are syntactic variables (that is, they are

Ā-bound), that their immediate antecedents are not the head NP but rather null opera-

tors in the specier position of the COMP-projection (null operators that are them-

selves subject to the antecedent-government conditions of the ECP). We have also

argued that, despite being variables, resumptive pronouns preserve their pronominal

character in that they are subject to a disjoint reference (Condition B) constraint that

holds in the domain of Ā-binding – a condition that we have formulated (following

Aoun and Li, 1989) as the Ā-disjointness requirement in (44). We have also presented

a proposal about why and how the choice between use of a gap and use of a pronoun

is reected in the choice of complementizer.

e analysis as developed involves, in essence, the linking of a number of phe-

nomena. We take the appearance of the complementizer aN to signal the presence of

an operator of the appropriate type to bind a resumptive pronoun. When that operator

is present, the pronoun with which it is linked is a variable and exhibits Condition C

eects (strong crossover and an Ā-version of Condition B of the binding theory) in

the form of the highest subject restriction (analyzed in terms of the Ā-disjointness

requirement). For a given construction, then, we expect three variables to correlate:

(1) appearance of aN, (2) Condition C eects, and (3) HSR eects. I want, nally, to

consider some data that provide indirect support for the correctness of this correlation

by indicating that when one of the elements above is missing, they are all missing.

. Notice that this proposal requires (counter to Lasnik & Saito 1984) either that traces in

COMP cannot delete between S-structure and LF (on the assumption that they must meet

ECP requirements at LF), or that they must meet ECP requirements at S-structure. Stowell

(1986) argues against the proposal that traces in COMP can freely delete before LF on the

basis of certain properties of inﬁnitival WH-constructions.

In all of the examples we have seen so far, there is a complete match between the

complementizers that appear in intermediate C-projections. However, another marked

pattern that occurs is schematized in (i) and illustrated by the (real) example in (ii).

(i) [NP [aL [… [aN [… pro…]]]]]

(ii) an galar a meastar ar cailleadh muintir an

the disease t is-thought pro was-lost people the

oileáin leis

island- with-it

“the disease from which it is thought the people of the island died”

For discussion of the status (also highly marked) of this pattern, see McCloskey (1985).

 James McCloskey

Le dislocation in Irish is of a type familiar from other languages. A constituent

(typically “heavy”) appears in initial position, separated from its associated clause by

comma intonation. e place of the dislocated constituent in the associated clause is

lled by a pronoun:

(65) An t-airgead sin a ta tú a iarraidh orm, bhéarfaidh mé

that money t are you asking on-me, give- I

duit amárach é.

to-you tomorrow it

“at money you are asking me for, I shall give it to you tomorrow”

is is a construction which one might well interpret a priori as a construction involv-

ing resumptive pronouns. But notice that the complementizers that we have taken to

be morphological signals of the presence of a binding operator (aL and aN) are absent

here. A related observation is that the element that obligatorily demands an Ā-binder,

namely, WH-trace, is impossible in this construction:

(66) *An t-airgead sin a ta tú a iarraidh orm, bhéarfaidh mé

that money t are you asking on-me, give- I

duit amárach __.

to-you tomorrow

“at money you are asking me for, I shall give it to you tomorrow”

ese observations jointly suggest strongly that le dislocation in Irish is not an

operator-binding construction of the type with which we have so far been concerned.

Ifthere were an operator in (65), we would expect its presence to be signaled by aL

or aN in the manner we have seen. Furthermore, we would expect that it would be

possible to bind a WH-trace from the same position.

But if there is no operator-variable chain in an example such as (65), then the

analysis we have built so far leads us to expect that le dislocation contrasts with the

other constructions we have been dealing with in showing neither strong crossover

nor HSR eects. e pronoun should not be subject to Condition C, since, not being

Ā-bound, it is not a variable. Furthermore, since it is not Ā-bound, the pronoun should

trivially satisfy the Ā-disjointness requirement in (44), and there should be no HSR

eects. Both of these expectations are borne out:

(67) An donán beag táillúra sin thíos, deir an duine gránna go

the runt little tailor-  down says the person ugly 

bhfuil sé ag dul dhá phósadh.

is he going to-be-married

“at little runt of a tailor down there, the wretchj says hej is getting married”

Resumptive pronouns, Ā-binding, and levels of representation 

(68) a. An chéad ear eile a tháinig isteach d’aithin sé mé.

the rst other man t came in recognized he me

“e next man that came in, he recognized me”

b. Dream eile fós, mheasadar pro gur chóir

group other again think--3 + right

don Chumann comórtas a rith.

for-the Society competition run-

“Still another group, they thought that the Society should run

a competition”

Here we have strong, if indirect, conrmation that the set of connections that the anal-

ysis seeks to make is a real one.

Notice, though, that nothing we have considered so far provides any evidence

at all about the level of representation at which the crucial chains are formed. It is

presumably the case that the relevant Ā-chains must be formed by LF; otherwise, we

would have operators that bind no variables at that level. But this still leaves many

options open. Are the Ā-chains that bind resumptive pronouns visible as S-structure

or as D-structure phenomena, for instance? ese are the kinds of questions addressed

in the next section.24

. I leave open here the question of whether there is, in addition to the binding of a vari-

able by an operator in COMP, a syntactic binding between the relative head and the operator

in COMP (this is the relation dubbed “R-binding” by Saﬁr 1986). Saﬁr (1986) suggests that

languages diﬀer as to whether or not they require resumptive pronouns to be R-bound as

well as Ā-bound. He also suggests tentatively that Irish may be among the languages in which

resumptive pronouns need not be R-bound. e grounds for making this assessment lose

their force in the context of the analysis of questions in Irish later pursued here (an analysis

anticipated by Saﬁr 1986: 686). However, one indication that his conclusion may nevertheless

be correct is that there is no necessary agreement in person-number features between the

relative head and the resumptive pronoun. Disparities can arise in a number of circumstances,

such as the one illustrated in (i).

(i) Is sinne an bheirt ghasúr ar dhíol tú ár lóistín.

- we the two boy pro paid you our lodging

“We are the two boys that you paid our lodging”

Alternatively, the grammaticality of (i) might simply indicate that person-number features are

not transferred from the resumptive pronoun to the operator or operators that bind it in an

Ā-chain. is diﬀerence might in turn provide a basis for making the necessary distinction

between resumptive pronouns (which are not subject to the ECP) and the operators that bind

resumptive pronouns (which, according to our earlier discussion, are subject to the ECP).

 James McCloskey

. e level

ere are various ways in which one might think about trying to resolve the question

of the level at which resumptive pronouns are bound to their operators. An obvious

question to ask is whether resumptive pronouns can or cannot license the appearance

of parasitic gaps. If it is possible for resumptives to do this, then that would count as

strong evidence that they are recognizable as variables (i.e. are bound) at S-structure.

Unfortunately, the language is disobliging in this respect, since there seem to be no

parasitic gap constructions in Irish at all. e question, then, of whether resumptive

pronouns can license parasitic gaps in Irish cannot even be posed.

We have seen that resumptive pronouns must be recognizable as variables with

respect to Condition C of the binding theory since they give rise to strong crossover

eects. If it can be argued, then, that Condition C is a condition on S-structures,

these facts together will constitute evidence that resumptive pronouns are bound at

S-structure. Such an argument has been made (Chomsky 1981: 196–197, drawing on

earlier work by Brody 1979). If it can be maintained, then we have evidence for binding

of resumptive pronouns at S-structure.25

However, there is evidence of a fairly rm kind that bears on the question. is

evidence derives from a phenomenon that is peculiar to constituent questions.

Our discussion until now has focused on relative clauses (as convenient exem-

plars of WH-constructions in general). But, at this point, let us focus specically on

constituent questions, in particular, on a phenomenon characteristic of questions but

unknown in other WH-constructions. It is possible in questions to front a PP that

contains a resumptive pronoun to a position immediately to the right of the interroga-

tive phrase:

(69) a. Cé leis a raibh tú ag caint?

who with-him pro were you talk-

“Who were you talking to?”

. e evidence is delicate though, having to do with the status of coreference eﬀects under

pied piping:

(i) Which book that Johnj read did hej like?

(ii) *Hej liked every book that Johnj read.

S-structure provides the right class of representations to make the necessary distinction

between (i) and (ii): at LF they will he identical in relevant respects, and it is hard to see

how the necessary distinction could he drawn. Belletti and Rizzi (1988: 318, n. 16) discuss the

matter in terms of whether or not Condition C must be satisﬁed by D-structures as well as by

S-structures, but they draw no ﬁrm conclusions. e argument in the text would stand under

these assumptions.

Resumptive pronouns, Ā-binding, and levels of representation 

b. Céard leis a ndearna tú é?

what with-it pro did you it

“What did you do it with?”

is fronting is typically demanded in pedagogic grammars but in fact represents only

an option in informal speech. Doublets such as those in (70) are thus possible:26

(70) a. Cé a bhfuil fáth aige

who pro is reason at-him

“Who has a reason?”

b. Cé aige a bhfuil fuath ar a bheith ag fanacht

who at-him pro is hatred on be- wait-

“Who hates (lit. is hatred at him) to wait?”

In (70a), the PP containing the resumptive pronoun remains in the usual position in

which PPs are found; in (70b), however, the PP containing the resumptive is fronted

to a position between the interrogative phrase and the complementizer particle aN.27

e crucial initial observation about this construction is that only the PP that contains

the resumptive pronoun can be fronted. Consider an example with two unfronted PPs,

as in (71). is example is based on the collocation tá eagla ar X roimh Y (‘fear is on X

before Y′), which means “X is afraid of Y.”

(71) Cé a bhfuil eagla air roimhe?

who pro is fear on-him before-him

is sentence is ambiguous because there is no way to tell which of the two pronouns

is the resumptive pronoun. erefore, (71) can mean either ‘Who is he afraid of?’

(when the object of the second preposition is the resumptive) or ‘Who is afraid of

him?’ (when the object of the rst preposition is the resumptive). Once one of the PPs

is fronted, however, the ambiguity disappears, since fronting itself identies which PP

contains the resumptive:

(72) a. Cé air a bhfuil eagla roimhe?

who on-him pro is fear before-him

“Who is afraid of him?”

b. Cé roimhe a bhfuil eagla air?

who before-him pro is fear on-him

“Who is he afraid of?”

. e thorough optionality of the fronting is indicated by the fact that the two examples in

(70) occur on the same page of the same book.

. For discussion of this construction, see McCloskey (1979: 94–97), Ó Cadhlaigh

(1940: 422–424), Ó Searcaigh (1939: 103–104), O’Nolan (1920: 138), and de Bhaldraithe

(1953: 159–160).

1 James McCloskey

is PP-fronting is subject to a number of apparently idiosyncratic restrictions that

are ultimately, I believe, phonological in origin. In particular, the interrogative phrase

to which the PP containing a resumptive pronoun is attracted must be monosyllabic.

One consequence of this restriction is that fronting is impossible in the case of syntac-

tically complex (that is, nonpronominal) interrogative phrases:

(73) *Cén fear leis a raibh tú ag caint?

which man with-him pro were you talk-

“Which man were you talking to?”

Furthermore, northern dialects (unlike western and southern) have a disyllabic word

for ‘what’, goidé, with a strong stress on the second syllable. In these dialects, PP-fronting

is impossible with this interrogative phrase:

(74) a. *Goidé leis a ndearna tú é?

what with-it pro did you it

“What did you do it with?”

b. Goidé a ndearna tú leis é?

what pro did you with-it it

“What did you do it with?”

Fronting is, however, possible with the monosyllabic and weakly stressed variant form

ca “what”, which is used exclusively in this environment. at is, ca functions as a sort

of allomorph of goidé in the context dened by an application of PP-fronting.

(75) a. Ca leis a ndearna tú é?

what with-it pro did you it

“What did you do it with?”

b. *Ca a ndearna tú leis é?

what pro did you with-it it

“What did you do it with?”

ese considerations suggest that there exists a strong phonological dependency

between the fronted PP and the interrogative pronoun. e same conclusion is sug-

gested by the fact that the sequence of interrogative pronoun and fronted PP forms a

stress group. As is the case with all syntactically dened stress groups in Irish, the pri-

mary stress is on the rightmost element;28 in fact, in the fronting structures, there is a

very strong contrast between the interrogative pronoun and the following PP, the latter

being heavily stressed, the former being weakly stressed to the point where one might

regard it as a (phonological) clitic dependent on the fronted PP. So for phonological

. See Ó Cúiv (1944: 68–70), Breatnach (1947: 83–88), de Bhaldraithe (1966: 64–66), and de

Burca (1970: 75–77).

Resumptive pronouns, Ā-binding, and levels of representation 11

purposes at least, the bracketing in (76) seems to be what is called for in the case of

examples such as those in (69).

(76) [Cé leis] aN raibh tú ag caint?

Now, of course, it is dangerous and dicult to make assumptions about syntactic

phrase structure on the basis of evidence that suggests a certain bracketing is required

for phonological purposes. Nevertheless, there is also evidence of a more purely

syntactic nature which suggests that (76) is the right syntactic representation of the

fronting structures. is evidence has to do with the element eile, which translates to

English other or else but has a broader distribution.

(77) a. an fear eile

the man other

“the other man”

b. duine ar bith eile

person any other

“anybody else/nobody else”

c. Bhuail siad arís eile é

struck they again other him

“ey struck him yet again/one more time”

d. a oiread eile airgid agus a fuair sé roimhe

amount other money as t got he before

“as much money again as he had received before”

is element can modify interrogative pronouns:

(78) a. Goidé eile a chonaic tú?

what other t saw you

“What else did you see?”

b. Cé eile a raibh tú ag caint leis?

who other pro were you talk- to-him

“Who else were you talking to?”

When a PP is fronted in such an interrogative structure, however, it always appears

between the interrogative pronoun and eile:

(79) a. Cé aige eile a bhfuígfeadh siad é?

who at-him other pro would-leave they it

“Who else would they leave it with?”

b. Ca leis eile a raibh tú ag súil?

what with-it other pro were you expect-

“What else were you expecting?”

c. Cé h-air eile a ligfeadh sé amach an racht sin?

who on-him other pro would-let he out that outburst

“Before whom else would he let out that outburst?”

1 James McCloskey

is observation suggests strongly that when a PP is fronted in an interrogative struc-

ture, it is, as it were, folded into the interrogative pronoun itself. Expressed more mun-

danely, these facts suggest that the fronted PP is adjoined to the interrogative pronoun,

as in the (phonologically motivated) structure in (76). More precisely, the structure

will be as in (80).

(80) [CP [NP [NP Cé] leis] aN [S raibh tú ag caint]]

To explain the ordering facts with respect to eile, as illustrated in (79), that is, to account

for the fact that eile appears outside the fronted PP despite being a subconstituent of

the phrase to which the PP adjoins, we might make appeal to a rule that right-adjoins

eile to its containing NP. ere is a certain amount of motivation for such a rule, since

eile can appear either to the le or to the right of a demonstrative particle, as illustrated

in (81).29

(81) a. an fear eile seo

the man other 

“this other man”

b. an fear seo eile

the man  other

“this other man”

On this assumption, (81b) would derive from (81a) by way of right-adjunction of eile

to its containing NP. If there is anything to this line of thought, then there is reason to

believe that eile can be raised to adjoin to its containing NP, though why this process

should be obligatory in the context of PP-fronting, I do not know.

ere is much in this domain of facts that remains mysterious. But the cumulative

force of the observations suggests that what happens in PP-fronting in questions is

that the PP containing a resumptive adjoins to the interrogative phrase in clause-initial

position.

e question now is how this process interacts with the principles that govern

complementizer selection. In all the examples we have seen so far, the complementizer

that appears in C0-position is COMPpro (aN). It is as if the PP-fronting rule makes no

dierence to the way in which complementizer selection works; the clause contains

a resumptive pronoun and, therefore, is headed by aN. All of the examples we have

seen so far, however, are from the northern group of dialects. If we broaden our

dialectal perspective a little, then the facts quickly become more complicated and

. Both (81a) and (81b) are possible, but, at least in the dialects with which I am most fa-

miliar, there is a strong preference for (81b).

Resumptive pronouns, Ā-binding, and levels of representation 1

more interesting. In particular, if we examine the facts in southern dialects (those of

Munster), we see that many of those dialects show COMPt – the complementizer aL,

which is normally associated with the appearance in a clause of a WH-trace – rather

than the resumptive pronoun complementizer aN. e examples cited in (82) are all

from Munster dialects.

(82) a. Cuimhnigh cé leis a tánn tú ag caint.

remember who with-him t are you talk-

“Remember who you are talking to.”

b. Cad leis a chlúdaigh tú iad?

what with-it t covered you them

“What did you cover them with?”

c. Cad air a dhíolfá-sa do pháiste?

what on-it t sell--2 your child

“What would you sell your child for?”

I want now to address the question of how we can account for these facts and the

dialectal variation observed in this domain.30 Viewing the contrast between the two

groups of dialects in pretheoretical terms, one can say that for the northern dialects,

the principles that determine complementizer selection are blind to the fact of PP-

fronting. e southern dialects observed in (82), however, seem to take account of the

presence of the PP-gap created by PP-fronting and select the complementizer accord-

ingly. is is the intuition that we can try to make precise.

What is the structure of constituent questions in Irish? Notice that it already

follows from the analysis, as developed thus far, that interrogative phrases cannot be

in the specier position of CP. We have developed an analysis in which the appear-

ance of dierent complementizers is triggered by Spec-Head agreement between the

C0-position and the specier position within CP. If this general line of analysis is

right, then examples such as (83) show that the interrogative phrase cannot be in Spec

of COMP.

. Some Munster dialects allow both options, as seen in the following pair of examples

from the same book by the same author:

(i) Bhí ios ag cách cé dhó a bhí sé ag tagairt.

everybody knew who to-him t was he refer-

“Everybody knew who he was referring to”

(ii) Bhí ios agam cad dhó a rabh sí ag tagairt.

I-knew what to-it pro was she refer-

“I knew what she was referring to”

1 James McCloskey

(83) a. Cé a bhí __ ann?

who t was there

“Who was there?”

b. Cé a raibh tú ag caint leis?

who pro be- you talk- with-him

“Who were you talking to?”

Our analysis of complementizer selection is crucially local: it depends on a relation

holding between a head and a specier within the same projection. To make a dis-

tinction between (83a) and (83b) on these grounds is impossible, if the assumption is

made that the interrogative pronoun cé is in the specier position of CP in both cases.

For the specier position would, on this view, be occupied by apparently identical

elements in the two cases. e structure that is needed, then, must be (84). (I use “Q”

in[84] as a temporary abbreviation for the category of interrogative clauses.)

(84)

[+Q]

[αF]

Now, in fact, this is the structure for interrogative clauses in Irish that is also suggested

by many other considerations. at case is made in some detail (but on grounds inde-

pendent of our current concerns) in McCloskey (1979).31

Consider now the structure of an interrogative clause following application of

PP-fronting:

(85)

[+Q]

PPi

P NPj

[PP ti ]

OpjC1

proleis

cé

1. See Chapter3 in particular. In the way of things, not all of the arguments survive in the

context of current assumptions. e important ones do, however. See, in particular, McCloskey

(1979: 59–62, 70–71).

Resumptive pronouns, Ā-binding, and levels of representation 1

Note that in this structure the operator in Spec of CP does not c-command the

variable that it is, so to speak, “supposed” to bind, namely, the resumptive pronoun

within the preposed PP. is observation holds true if the relevant denition of

c-command is taken to be the classical denition deriving from the work of Reinhart

(1976, 1981, 1983):

(86) A node α c-commands a node β i the branching that most immediately

dominates α also dominates β, and neither α nor β dominates the other.

In (85) CP is a branching node that dominates the operator but not the preposed

resumptive pronoun. If the relevant notion is m-command (Aoun & Sportiche 1983;

Chomsky 1986a), as in (87),

(87) α m-commands β i every maximal projection that dominates α also

dominates β, and neither α nor β dominates the other.

then this relation still fails to hold between the operator in Spec of CP and the pre-

posed pronoun.32

If Opj fails to c-command the resumptive pronoun, then it also fails to bind it.

Structure (85) is thus, on the face of it, ill formed at LF, since it contains an operator

that fails to bind a variable. e operator fails to bind a variable because it fails to

c-command the preposed resumptive pronoun; but it is of the wrong type (as signaled

by the presence of the complementizer aN) to bind the PP-trace le by pronoun pre-

posing since this trace is a nonpronominal variable. erefore, (85) runs afoul of the

LF-requirement that every operator bind a variable. On this view of matters, exam-

ples such as (88) will be ungrammatical if they are inspected in their audible form by

LF principles.

(88) Cé leis a raibh tú ag caint?

who with-him pro were you talk-

“Who were you talking to?”

. e conclusion also holds if (as seems likely) Q is CP (or CP[+Q]), and one adopts the

segment analysis of adjunction structures (May 1985; Chomsky 1986b). e relevant sche-

matic structure is in (i):

(i)

βCP

αC′

Here, there is a maximal projection (i.e. CP) that includes α but does not include β (since β is

not dominated by every segment of CP). erefore, nothing in the lower CP can c-command

anything within β.

1 James McCloskey

But notice that (85) can be saved if an operator of the appropriate kind to bind a

PP-trace is base generated in Spec of CP. In that case, the trace-binding operator

(i.e. an operator that is [−anaphoric, −pronominal]) will c-command, and therefore

bind, the trace le by PP-fronting, and the structure will not be in violation of the

LF-requirement that every operator bind a variable. Now the presence of this kind

of binding operator will be signaled on the C0-position in the usual way, that is, by

the appearance of the aL complementizer. In this way, we derive an account of the

grammaticality of the Munster examples in which PP-fronting is accompanied by the

appearance (at rst blush, unexpected) of COMPt (see [82]). eir structure will now

be as in (89).

(89)

CPNP

[+Q]

[PP ti ]

Opi

[+Q]

PPi

cé leis pro

NPj

What emerges then is that, given the set of assumptions we have developed, it is only

when the complementizer aL, that is, COMPt, appears that PP-fronting can be legal at

all. us the existence of those Munster dialects that have (82) is rationalized.

Consider in a little more detail how this analysis will work. What is crucial for

the licensing of (89) is that the “wrong” operator, so to speak, be generated in CP. An

operator of the appropriate kind to bind a PP-trace must be generated there. is is,

presumably, simply an empty PP with the features [−anaphoric, −pronominal], the

kind of operator that appears, for instance, in a cle such as (90).

(90) Is [PP le Siobhán] [CP Opj a [S bhí mé ag caint [PP tj]]]

- with Siobhan t was I talk-

“It was with Siobhan that I was talking”

At D-structure there will be no variable for this operator to bind, but this will be a

harmless property of the structure since the requirement that every operator phrase

bind a variable is an LF-requirement. If an Ā-chain were to be formed between this

operator and the pronoun that later comes to be regarded as the resumptive pronoun,

there would be a featural mismatch, and the derivation would be aborted. But the

Resumptive pronouns, Ā-binding, and levels of representation 1

Ā-chain need not be formed at this level, and the operator and the pronoun will not

interact at all at this level. If PP-fronting now applies before S-structure, a PP-trace

will be le under S in (90). It is now crucial that this trace be bound by the operator in

Spec of CP; otherwise, this operator will be in violation of the LF-requirement that it

bind a variable. ere will now be the appropriate featural match between the operator

and its variable, and this featural match will be reected in the appearance of aL as the

complementizer in the usual way.

Notice too that in the nal structure, the fronted PPj c-commands Opj and is coin-

dexed with it. It thus identies the null operator in the familiar way and the PP-trace

under S is strongly bound in the sense of Chomsky (1986a).

is analysis is built on three assumptions: (1) chain formation is independent

of movement, (2) operators must bind variables at LF, and (3) Ā-chains can be

(but need not be) formed at any syntactic level. Assumption (2) is uncontroversial.

Assumption(3) seems like the minimal assumption one could make in this domain.

Assumption (1) has been argued for independently by Rizzi (1986) and by McCloskey

and Sells (1988).

But what of the northern dialects with which we began our discussion, those

in which PP-fronting is accompanied by the appearance of the complementizer aN

(COMPpro)? According to our earlier discussion, such structures are declared ungram-

matical by principles of LF. In particular, they violate the LF-requirement that every

operator bind a variable because the operator whose presence is signaled by aN can

bind neither the PP-trace of PP-fronting (it is of the wrong type) nor the fronted pro-

noun within that PP (it fails to c-command it). is in turn suggests that in those

dialects in which PP-fronting cooccurs with the appearance of the complementizer

aN, principles of LF never see or inspect those structures in which PP-fronting has

applied. at is, we can account for the dialectal dierences examined here by propos-

ing that in northern dialects PP-fronting applies between S-structure and PF. In Mun-

ster dialects, on the other hand, the rule applies between D-structure and S-structure,

and its eects are visible to principles whose domain of application is LF. For northern

dialects, in contrast, the eects of PP-fronting are invisible at LF; for LF purposes, the

two examples in (91) are indistinguishable.

(91) a. Cé leis a raibh tú ag caint?

who with-him pro were you talk-

“Who were you talking to?”

b. Cé a raibh tú ag caint leis?

who pro were you talk- with-him

“Who were you talking to?”

e structure that is seen is (91b). In this structure the operator whose presence is

signaled by aN c-commands and binds a variable (the resumptive pronoun), and the

1 James McCloskey

sentence is unproblematic.33 Notice that this interpretation of the facts depends cru-

cially on one of the salient features of the model of grammatical organization usually

assumed in government and binding, namely, the fact that there is no sharing of infor-

mation between the levels of LF and PF, except insofar as is mediated by S-structure.

is, I think, is a reasonably satisfying account of the dialectal dierences exam-

ined here. Both groups of dialects have essentially the same rule, but it applies in dier-

ent components. But notice that, if we accept this analysis, there is an implication for

our original question, the question of whether or not resumptive pronouns are bound

at S-structure and are identiable as variables at S-structure. It seems that they must

be. e version of PP-fronting that applies in northern dialects must be able to tell the

dierence between resumptive and nonresumptive pronouns. at is, it must be able

to recognize that certain pronouns are Ā-bound and that certain others are not, for

recall that it is a crucial property of PP-fronting that it applies only to PPs that imme-

diately dominate a resumptive pronoun. Its formulation must be akin to (92).

(92) Right-adjoin [PP P proj] to NPj

[+Q]

If such a rule is to apply correctly between S-structure and PF, it seems that the infor-

mation necessary to pick out resumptive pronouns from non- Ā-bound pronouns

must be available by S-structure.34

. Summary and conclusions

At this point, we have assembled answers to most of the questions with which we

began our discussion. We have seen that there is reason to believe that resumptive

pronouns are to be considered variables, and also that there is reason to believe that

the chain-formation rule that binds a resumptive pronoun to a null operator can apply

at any level. Certain pronouns will therefore be Ā-bound at S-structure. We have also

seen that the dual nature of resumptive pronouns – simultaneously pronouns and

. For those Munster dialects that allow either complementizer in the context of PP-fronting

(cf. Note30), we will, of course, allow the rule of PP-fronting to apply in either component.

. From the formulation in (92), we expect that PP-fronting will obey subjacency. is is

essentially correct: see McCloskey (1979: 95). e formulation in (92) assumes a richness of

descriptive apparatus not normally assumed in current work. I see no way at present to reduce

the statement of PP-fronting any further than (92). One might rationalize the need for this

apparatus by appealing to the peripheral and parochial character of the rule in question.

Resumptive pronouns, Ā-binding, and levels of representation 1

variables – is reected in their behavior with respect to binding theory. Resumptive

pronouns, being variables, are subject to Condition C. But they are also, being pro-

nouns, subject to a disjointness requirement that holds in the domain of Ā-binding.

We have also seen, though, that it is not the case that every construction that looks

initially like a resumptive pronoun construction actually involves Ā-binding of a pro-

noun by an operator. Le dislocation in Irish, in particular, seems at rst glance to

be just like the other resumptive pronoun constructions examined. When examined

more closely, however, it turns out not to involve an operator-variable binding at all.

In fact, the work of Engdahl (1985) on resumptive pronouns in Swedish suggests

that there is a third phenomenon lurking in the pretheoretical category of resump-

tive pronouns. Engdahl shows that the resumptive pronouns of Swedish behave

almost entirely like WH-traces, and she suggests that they be analyzed in terms of

a mechanism that “spells out” a trace created by WH-movement. e distribution of

resumptive pronouns in Swedish is extremely limited: they appear only in the subject

positions of tensed clauses next to lexical complementizers. Notice that this pattern

is, in essence, the complement of the distributional pattern found in Irish. at dis-

tributional pattern we have accounted for here by appealing to the binding theoretic

properties of an element that is simultaneously a pronoun and a variable. is view

and Engdahl’s evidence converge to suggest that the resumptive pronouns in Irish and

Swedish are quite dierent elements – pronominal variables in Irish, pure variables

in Swedish. eother class of resumptive pronouns, the class that appears in le

dislocation and exhibits no variable properties at all, is also instantiated in Swedish

(Engdahl 1985: 11).

Finally, I want to consider some of the broader theoretical questions for which

resolution of the resumptive pronoun issue is crucial.

One of the theoretical debates in which the question of the proper analysis of

resumptive pronoun structures has most prominently gured has been the debate on

weak crossover. Consider a standard weak crossover conguration such as (93).

(93) *Whoj does hisj mother dislike tj?

Since hisj in (93) is not c-commanded by the trace of WH-movement but rather is

c-commanded by and coindexed with the interrogative operator who, it is locally

Ā-bound by who. It is, therefore, an Ā-bound pronoun; in our terms, it is a resumptive

pronoun. Furthermore, both the pronoun his and the trace le by WH-movement

in(93) are variables. Given these observations, there are certain obvious ways in which

we might think of accounting for the ungrammaticality of (93).

First, we might say that pronouns (in English) cannot be Ā-bound. On this view,

(93) would be ungrammatical by virtue of containing an Ā-bound pronoun. Second, we

might maintain that what is wrong with (93) is that the operator who simultaneously

11 James McCloskey

binds two variables – the pronoun his and the trace of WH-movement. On this view,

(93) would be ungrammatical because it violates a principle that would hold that an

operator must bind one and only one variable. ird, we might maintain that what is

wrong with (93) is not that it contains an operator that binds two variables but rather

that the two variables bound by the same operator are of dierent types (pronominal

and nonpronominal). On this view, (93) would be ungrammatical by virtue of violat-

ing a principle that all variables bound by a single operator must he of the same type

(all pronominal or all nonpronominal). All three of these lines of thought have been

pursued in the literature on Weak Crossover. Versions of the rst have been pursued

in Chomsky (1976) and Jaeggli (1984). e second is the essential idea behind the

bijection principle of Koopman and Sportiche (1982). And the third is the parallelism

constraint on operator binding (PCOB) of Sar (1984).

What do the Irish data have to say about these issues? e basic observation is that

there are no weak crossover eects when a relative clause (for instance) contains two

resumptive pronouns, neither of which c-commands the other. e examples in (94)

are both attested.

(94) a. fear ar ág a bhean é

man pro le his wife him

“a man that his wife le”

b. an fear so ar mhairbh a bhean féin é

this-man pro killed his-own-wife him

“this man that his own wife killed”

But weak crossover eects are pronounced when the second pronoun in (94) is

replaced with a trace:35

(95) a. *fear a d’ág a bhean __

man t le his wife

“a man that his wife le”

b. *an fear so a mhairbh a bhean féin __

this-man pro killed his-own-wife

“this man that his own wife killed”

Now, the occurrence of any weak crossover eect at all in a language with a pro-

ductive resumptive pronoun strategy renders the rst approach, sketched above,

. In English, weak crossover eﬀects are barely noticeable in the case of restrictive relatives

(Chomsky 1982; Saﬁr 1984). is is not true of the Irish examples in (95), however, which

are robustly ungrammatical. It is as if the availability of a grammatical alternative (as seen in

[94]), renders the ungrammaticality of examples such as (95) more salient.

Resumptive pronouns, Ā-binding, and levels of representation 111

implausible (this observation is made in Koopman & Sportiche 1982; Engdahl 1985;

and Sells 1984a). erefore, the ungrammaticality of (95) gives us reason to reject the

rst approach. On the other hand, the grammaticality of (94) gives us reason to reject

the second approach above, based as it is on the bijection principle. It seems clear

that we have in such examples an instance of the binding of two variables by the same

operator, particularly in light of the evidence so far accumulated here that resumptive

pronouns in Irish are in fact Ā-bound.

e only analysis of the facts of weak crossover that I know of that carries over

in a natural way to the Irish data is Sar’s PCOB, the third approach sketched above.

Notice that this approach accounts for the contrast between (95) and (94) in a straight-

forward way, with one modication. Sar’s formulation of the PCOB demands that all

the variables bound by a single operator must agree in being all lexically specied or

all lexically unspecied. Now, given the Irish data, this cannot simply mean that all the

variables must be phonologically realized or phonologically unrealized. As usual, it

makes no dierence in Irish whether one is dealing with an audible pronoun or with

pro. Notice that the grammatical examples in (94) all involve pairings of pro and an

audible pronoun. Rather, the crucial factor must be whether or not the variables agree

in being pronominal or nonpronominal.

In fact, one might speculate that the PCOB is related to some of the eects we have

already had occasion to study here. If one examines the ungrammatical examples in

(95) in light of our earlier discussion, one might be tempted to explain their ungram-

maticality in terms of principles we have already developed. One might attribute their

ungrammaticality to the fact that, since the operator binds both a pronominal and

a nonpronominal variable, the Ā-chains formed thereby will impose incompatible

and irreconcilable featural demands on the binding operator. at is, the operator is

required to be simultaneously associated with COMPt and with COMPpro. But the

mechanisms that we have assumed to account for these distributional connections in

Irish are quite general (most notably, feature sharing among the links of an Ā-chain).

It is possible that the PCOB is ultimately a reection of this same conict in a language

in which the dierence between the two classes of binding operators is not otherwise

formally marked.36

. Lasnik and Stowell (1987) base their discussion of weak crossover on the observation

that, in a range of WH-constructions in English, weak crossover eﬀects disappear. One would

want to know now what the corresponding data are for resumptive pronouns and nonpro-

nominal variables in these constructions in Irish. is is an investigation that I must, however,

leave for future work.

11 James McCloskey

. Appendix: e distrubution of resumptive pronouns

.1 Construction types that permit resumptive pronouns

Restrictive relatives

(96) a. daoine nár shroich an tsibhialtacht fós iad

people pro+ reached the civilization yet them

“people whom civilization has not yet reached”

b. an fear a raibh mé ag caint leis pro

the man pro was I talk- with-3-

“the man that I was talking to (him)”

Nonrestrictive relatives

(97) a. áinig an saighdiúir eile, nach bhfaca mé

came the soldier other pro+ saw I

roimhe é, aníos chugainn.

before him up to-us

“e other soldier, whom I hadn’t seen before. came up to us”

b. Chonaic mé mo dheartháir, a mbíodh faitíos

saw I my brother pro be-- nervousness

i gconaí air pro ag éalú an doras amach

always on-3- slip- the door out

“I saw my brother, who was always nervous, slipping out the door”

Constituent questions

(98) a. d’inis siad cén turas a raibh said air pro

told they what joumey pro be- they on-3-

“they told what journey they were on (it)”

b. cén t-oigeach ar shíl tú go mbeadh sé i láthair?

which ocer pro thought you  would-be he present

“Which ocer did you think would be present?”

Cles

(99) a. Is tú a bhfuil an deallramh maith ort pro

- you pro is the appearance good on-2

“It is you that looks weIl”

. Traditional grammars oen claim (in eﬀect) that resumptive pronouns are impossible

in cles. In earlier work (McCloskey 1978: 181) I accepted this claim. More careful investiga-

tion, however, has shown clearly that the claim is wrong, as the examples cited in the text

demonstrate. Many more counterexamples could have been cited. e grammaticality of such

examples has also been documented and discussed recently by Mac Cana (1985).

Resumptive pronouns, Ā-binding, and levels of representation 11

b. Tig beag caol ar mhaireamar ann pro

house little narrow pro live--1 in-3-

“It was a narrow little house that we lived in (it)”

c. Siobhán a bhfuil buaite aici pro

Siobhan pro is won by-3-

“It is Siobhan that has won”

Lit. “It is Siobhan that has been won by her”

Comparative clauses

(100) a. áinig níos mó daoine ná a raibh

came more people than pro was

súil leo pro

expectation with-3

“More people came than were expected”

b. Do fuair sé leaba chó math agus ar lui sé

get- he bed as good as pro lie- he

riamh uirthi pro

ever on-3-

“He got a bed as good as he ever lay on (it)”

“Tough-movement” constructions

(101) a. Bhí Ristéard doiligh cur suas leis pro

was Richard dicult put- up with-3-

“Richard was hard to put up with (him)”

b. Beidh an gasúr sin deacair fail réitithe leis pro

be- that-boy hard get- rid with-3-

“at boy will be hard to get rid of (him)”

Ready-class of adjectival complements

(102) Bhí na putóga ullamh ansan chun iad a líonadh.

were the intestines ready then to them ll-

“e intestines were ready then to ll (them)”

Innitival relatives and purpose clauses

(103) a. Ní rabh a’n duine aige le labhairt leis pro

 was anybody at-him to talk- with-3-

“He had nobody to talk to (him)”

. It is notoriously diﬃcult to distinguish reliably between inﬁnitival relatives and purpose

clauses. e problems and issues seem to be much the same in Irish as they are in English. I

am content here with citing some relevant examples and trusting that their general similarity

to the corresponding English construction will be enough to establish the relevant point.

11 James McCloskey

b. Bhí mórán aige le smaointeamh air pro

was a-lot at-him to think- on-3-

“He had a lot to think about (it)”

. Clausal positions in which resumptive pronouns appear

Direct object of highest clause

(104) a. an ghirseach ar ghoid na síogaí í

the girl pro stole the fairies her

“the girl whom the fairies stole”

b. an ghirseach a ghoid na síogaí __

the girl t stole the fairies

“the girl whom the fairies stole”

Subject of embedded nite clause

(105) a. na caiple sin a n-abrann sealgairí go mbíonn siad

those horses pro say hunters  be- they

ag léimnigh agus ag damhsa

jump- and dance-

“those horses that hunters say are always jumping and dancing”

b. no caiple sin a deireann sealgairí a bhíonn __

those horses t say hunters  be-

ag leimnigh agus ag damhsa

jump- and dance-

“those horses that hunters say are always jumping and dancing”

(106) a. cúpla muirear a bhféadfaí a rá go

a-few families pro one-could say- 

rabhadar pro bocht

be-3 poor

“a few families that one could say (they) were poor”

b. cúpla muirear a d’hféadfaí a rá a bhí __ bocht

a-few families t one-could say- t be- poor

“a few families that one could say were poor”

Object of embedded nite clause

(107) a. an rud céanna a dtug sé orm mionnú go

the thing same pro took he on-me swear- 

gcoinneoinn ceilte é

keep--1 concealed it

“the same thing that he made me swear that I would keep (it) hidden”

Resumptive pronouns, Ā-binding, and levels of representation 11

b. an rud céanna a thug sé orm mionnú a

the thing same t took he on-me swear- t

choinneoinn ceilte

keep--1 concealed

“the same thing that he made me swear that I would keep hidden”

Subject of embedded nonnite clause

(108) a. duine ar bith ar mhaith leat é beannú sa

person any pro you-would-like him greet- in-the

tsráid duit

street to-you

“anybody that you would like him to greet you in the street”

b. duine ar bith a ba mhaith leat beannú sa tsráid duit

person any t you-would-like greet- in-the street to-you

“anybody that you would like to greet you in the street”

Object of embedded nonnite clause

(109) a. na Dílseoirí seo ar theastaigh ón Uasal Ó hEadhra

these Loyalists pro wanted from Mr. O’Hara

iad a thabhairt isteach sa ghluaiseacht

them bring- in into-the movement

“these Loyalists that Mr. O’Hara wanted to bring (them) into

the movement”

b. na Dílseoirí seo a theastaigh ón Uasal Ó hEadhra __

these Loyalists t wanted from Mr. O’Hara

a thabhairt isteach sa ghluaiseacht

bring- in into-the movement

“these Loyalists that Mr. O’Hara wanted to bring into the movement”

Object of progressive phrase

(110) a. an rud a raibh siad dhá chuartú pro

the thing pro were they 3- seek-

“the thing that they were seeking (it)”

b. an rud a bhí siad a chuartú __

the thing t were they seek-

“the thing that they were seeking”

. For detailed analysis of progressive phrases in Irish, see McCloskey (1983a). See also

Chung and McCloskey (1987) and Clements et al. (1983).

11 James McCloskey

Copular clauses

(111) a. rud arbh iú cuid mhór é

thing pro+ worth a-lot it

“a thing that it was worth a lot”

b. rud a b’iú cuid mhór __

thing t +worth a-lot

“a thing that was worth a lot”

is set of examples is given only for relative clauses, but an exactly comparable set of examples

could be provided (where applicable) for the full range of WH-constructions cited in the previ-

ous section.41

Acknowledgments

This article was completed under difficult circumstances. My thanks to Randy

Hendrick for his patience and persistence, and to Sandy Chung for her help. anks

are also due Siobhán Ní Laoire, Róise Ní Bhaoill, Liam Mac Con Iomaire, Caitlín Mhic

Niallais, Alan Harrison, and Nicholas Williams for their help in establishing the data.

e work reported on here was supported by NSF Grant BNS86-17274, awarded to the

University of California, Santa Cruz.

References

Andrews, Avery. 1990. Unication and morphological blocking. Natural Language and Linguistic

eory 8: 507–557.

Aoun, Joseph & Y.-H. Audrey Li. 1989. ree cases of logical relations: Relative scope, bound

pronouns, and anaphoric relations. University of Southern California, ms.

. e syntax of copular clauses in Irish is very poorly understood. But one point is clear:

their syntax is very diﬀerent from the syntax of verbal clauses. For a comprehensive account

of the facts, see Ó Siadhail (1989).

1. In almost every case where a free choice exists between use of the resumptive pronoun

structure and use of the WH-movement structure, the latter is more commonly used. is is

presumably a reﬂection of Chomsky’s (1981) avoid pronoun principle, interpreted not as an

absolute grammatical prohibition but rather as a measure of preferred usages.

ere is one exception to the generalization noted here: in copular clauses [cf. (111)], the

preferences are reversed, and the WH-movement option is felt to be formal, marginal, and

highly marked. is is just one of the puzzles that mark this domain of Irish syntax as special.

Resumptive pronouns, Ā-binding, and levels of representation 11

Aoun, Joseph & Dominique Sportiche. 1983. On the formal theory of government. Linguistic

Review 3: 211–235.

Belletti, Adriana & Luigi Rizzi. 1988. Psych-verbs and theta-theory. Natural Language and

Linguistic eory 6: 291–352.

Borer, Hagit. 1984. Restrictive relatives in Modern Hebrew. Natural Language and Linguistic

eory 2: 219–260.

Breatnach, Risteard. 1947. e Irish of Ring, Co. Waterford. Institute for Advanced Studies, Dublin.

Bresnan, Joan & Jane Grimshaw. 1978. e syntax of free relatives in English. Linguistic Inquiry

9: 331–391.

Brody, Micheal. 1979. Innitivals, relative clauses, and deletions. University of London, ms.

Browning, M. 1987. Null Operator Constructions. Doctoral dissertation, Massachusetts Institute

of Technology, Cambridge.

Chao, Wyn & Peter Sells. 1983. On the interpretation of resumptive pronouns. NELS 13:47–61.

Chomsky, Noam. 1976. Conditions on rules of grammar. Linguistic Analysis 2: 303–351.

Chomsky, Noam. 1977. On WH-movement. In Formal Syntax, P.W. Culicover, A. Akmajian &

T. Wasow (eds.). Academic Press, New York.

Chomsky, Noam. 1981. Lectures on Government and Binding. Foris, Dordrecht.

Chomsky, Noam. 1982. Some Concepts and Consequences of the eory of Government and

Binding. MIT Press, Cambridge, Mass.

Chomsky, Noam. 1986a. Barriers. MIT Press, Cambridge, Mass.

Chomsky, Noam. 1986b. Knowledge of Language: Its Nature, Origin, and Use. Praeger, New York.

Chomsky, Noam. 1988. Some notes on economy of derivation and representation. Massachusetts

Institute of Technology, ms.

Chung, Sandra. 1990. Sentential subjects and proper government in Chamorro. In Interdisci-

plinary Approach to Language. Essays in Honor of Yuki Kuroda, Carol Georgopoulos &

Roberta Ishihara (eds.). Dordrecht: Kluwer.

Chung, Sandra & James McCloskey. 1987. Government, barriers, and small clauses in Modern

Irish. Linguistic Inquiry 18: 173–237.

Clements, Nick, James McCloskey Joan Maling & Annie Zaenen. 1983. String vacuous rule

application. Linguistic Inquiry 14: 1–17.

de Bhaldraithe, Tomas. 1953. Gaeilge Chois Fhairrge – An Deilbhíocht. Institúid Ard-Léinn

Bhaile Átha Cliath, Dublin.

de Bhaldraithe, Tomas. 1956–1957. Nótaí comhréire. Éigse 8: 242–246.

de Bhaldraithe, Tomas. 1966. e Irish of Cois Fhairrge, Co. Galway. Institute for Advanced

Studies, Dublin.

de Burca, Sean. 1970. e Irish of Tourmakeady. Institute for Advanced Studies, Dublin.

Emonds, Joseph. 1978. e verbal complex V′-V in French. Linguistic Inquiry 9: 151–175.

Engdahl, Elisabet. 1985. Parasitic gaps, resumptive pronouns, and subject extractions. Linguistics

23: 3–44.

Guilfoyle, Eithne. 1988. Parameters and functional projection. Paper presented at the Eighteenth

Annual Meeting of the North-Eastern Linguistic Society.

Jaeggli, Osvaldo. 1984. Subject extraction and the null subject parameter. NELS 14: 132–153.

Koopman, Hilda. 1984. e Syntax of Verbs: From Verb-Movement Rules in the Kru Languages to

Universal Grammar. Foris, Dordrecht.

Koopman, Hilda & Dominique Sportiche. 1982. Variables and the bijection principle. Linguistic

Review 2: 139–160.

Kuroda, Sige Yuki. 1968. English relativization and certain related problems. Language 44:244–266.

11 James McCloskey

Lasnik, Howard. 1976. Remarks on coreference. Linguistic Analysis 2: 1–22.

Lasnik, Howard. 1989. On the necessity of binding conditions. In Principles and Parameters in

Comparative Grammar, R. Freidin (ed.). MIT Press, Cambridge. Mass.

Lasnik, Howard & Mamoru Saito. 1984. On the nature of proper government. Linguistic Inquiry

15: 235–289.

Lasnik, Howard & Timothy Stowell. 1991. Weakest crossover. Linguistic Inquiry 22: 687–720.

Mac Cana, Proinsias. 1985. Varia xii: A note on the prepositional relative. Éiru 36: 210–212.

May, Robert. 1985. Logical form. MIT Press, Cambridge, Mass.

May, Robert. 1989. Bound variable anaphora. In Mental Representations: e Interface Between

Language and Reality, R. Kempson (ed.), Cambridge University Press, London.

McCloskey, James. 1978. A Fragment of a Grammar of Modern Irish. Doctoral dissertation,

University of Texas. Published as Texas Linguistic Forum, 12. Department of Linguistics,

University of Texas at Austin.

McCloskey, James. 1979. Transformational syntax and model theoretic semantics. A Case Study

in Modern Irish. Reidel, Dordrecht.

McCloskey, James. 1980. Is there raising in Modern Irish? Ériu 31: 59–99.

McCloskey, James. 1983. On the binding of resumptive pronouns in Modern Irish. Paper pre-

sented to the annual meeting of GLOW. University of York.

McCloskey, James. 1985. e Modern Irish double relative and syntactic binding. Ériu 36: 46–84.

McCloskey, James. 1986. Inection and conjunction in Modern Irish. Natural Language and

Linguistic eory 4: 183–186.

McCloskey, James & Kenneth Hale. 1984. On the syntax of person-number inection in

Modern Irish. Natural Language and Linguistic eory 1: 487–533.

McCloskey, James & Peter Sells. 1988. Control and A-chains in Modern Irish. Natural Language

and Linguistic eory 6: 143–189.

Ó Cadhlaigh, Cormac. 1940. Gnás na Gaeilge. Oifg an tSoláthair, Dublin.

Ó Cuív, Brian. 1944. e Irish of West Muskerry, Co. Cork. Institute for Advanced Studies, Dublin.

Ó Nolan, Gerald. 1920. Studies in Modern Irish. Educational Company of Ireland, Ltd., Dublin

and Cork.

Ó Rahilly, omas Francis. 1946. Irish Dialects Past and Present. Institute for Advanced Studies,

Dublin.

Ó Searcaigh, Seamus. 1939. Coimhréir Ghailge an Tuaiscirt. Government Publications Oce,

Dublin.

Ó Siadhail, Micheal. 1980. Learning Irish. Institute for Advanced Studies, Dublin.

Ó Siadhail, Micheal. 1989. Modern Irish Grammatical Structure and Dialectical Variation.

Cambridge University Press, London. To appear.

Pollock, Jean-Yves. 1989. Verb movement, UG, and the structure of IP. Linguistic Inquiry

20:365–424.

Reinhart, Tanya. 1976. e Syntactic Domain of Anaphora. Doctoral dissertation. Massachusetts

Institute of Technology, Cambridge.

Reinhart, Tanya. 1981. Denite NP anaphora and C-command domains. Linguistic Inquiry

12:605–636.

Reinhart, Tanya. 1983. Anaphora and Semantic Interpretation. Croom Helm, London.

Rizzi, Luigi. 1986. On chain formation. In e Syntax of Pronominal Clitics: Syntax and Semantics

19, H. Borer (ed.). Academic Press, New York.

Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press.

Resumptive pronouns, Ā-binding, and levels of representation 11

Sar, Kenneth. 1984. Multiple variable binding. Linguistic Inquiry 15: 603–638.

Safir, Kenneth. 1986. Relative clauses in a theory of binding and levels. Linguistic Inquiry

17:663–689.

Sells, Peter. 1984a. Syntax and Semantics of Resumptive Pronouns. Doctoral dissertation, University

of Massachusetts at Amherst.

Sells, Peter. 1984b. Resumptive pronouns and weak crossover. WCCFL 3.

Sells, Peter. 1987. Binding resumptive pronouns. Linguistics and Philosophy 10: 261–298.

Sproat, Richard. 1985. Welsh syntax and VSO structure. Natural Language and Linguistic eory

3: 173–216.

Stenson, Nancy. 1981. Studies in Irish Syntax. Gunter Narr Verlag, Tübingen.

Stowell, Timothy. 1986. Null antecedents and proper government. NELS 16: 476–493.

Taraldsen, Knud Tarald. 1983. Parametric Variation in Phrase Structure: A Case Study. Doctoral

dissertation, University of Tromsø.

Zaenen, Annie, Elisabet Engdahl & Joan Maling. 1981. Resumptive pronouns can be syntactically

bound. Linguistic Inquiry 12: 679–682.

Pseudo Wh-Fronting: A Diagnosis of Wh-Constructions in Jordanian Arabic

Thesis

Full-text available

Dec 2016

Zeyad Al-Daher

This thesis provides an in-depth analysis of wh-question formation in Jordanian Arabic (JA) and presents a uniform approach that can accommodate all of its various wh-constructions. JA makes use of five different wh-constructions, four of which involve clause-initial wh-phrases and the fifth is a typical in-situ wh-construction. Although wh-phrases surface clause-initially in four different wh-constructions in JA, I propose that bona fide wh-movement to [Spec, CP] does not occur in any of these constructions, whether overtly in syntax or covertly at LF. I abandon the classification of JA as a wh-movement language (Abdel Razaq 2011) and focus instead on identifying the syntactic role that wh-phrases realize and the underlying structures that feed each wh-construction. I propose that the clause-initial position of the wh-phrase results either from the syntactic function that the wh-phrase serves or from other syntactic operations that are independently attested in JA. There are three clause-initial positions that the wh-phrase can occupy: it surfaces in [Spec, TP] when functioning as the subject of a verbal or verbless structure, in [Spec, TopP] when functioning as a clitic-left-dislocated element (as in CLLD questions and ʔilli-interrogatives involving PRON), or in [Spec, FocP] when undergoing focus fronting. Thus, all instances of clause-initial wh-phrases in JA constitute what I refer to as-pseudo wh-fronting‖, as the clause-initial position of the wh-phrase arises from mechanisms other than canonical wh-movement to [Spec, CP]. To account for the interpretation of wh-phrases in JA, I adopt a binding approach in which a null interrogative morpheme (Baker 1970; Pesetsky 1987; Chomsky 1995) unselectively binds the wh-phrase regardless of its surface position, whether clause-initial or clause-internal (in-situ). A major implication of this analysis is that JA is a concealed wh-in-situ language of the Chinese type although it looks at a cursory glance as though it were a wh-movement language of the English type. A broader typological implication of my analysis is the convergence of Cheng's (1991) Clausal Typing Hypothesis to which JA previously appeared to constitute a counterexample. The recognition of the null interrogative particle, or its optional overt realization as the Q-particle huwweh, as the locus of interrogative clause typing in all JA wh-questions entails that JA employs just one unique strategy to type a clause as a wh-question, as predicted by Cheng's Clausal Typing Hypothesis, regardless of whether the wh-phrase surfaces clause-initially or clause-internally. ii

Structural Change in Relative Clauses and the Autonomy of Heritage Grammars

Article

Full-text available

Apr 2022

Alberto Frasson

This paper addresses the question of structural change in relative clauses in heritage speakers of two varieties of Venetan, a northern Italo-Romance language. It will be shown that appositive and restrictive relative clauses are not structurally distinguished in Brazilian Venetan, while they display different structural properties in Italian Venetan. It will be proposed that the phenomenon described in the paper does not depend on transfer from another language and it is not exclusively a matter of processing. The approach presented here aims to account for structural change in syntactic terms, without resorting to extra-linguistic factors. Heritage grammars are autonomous systems and follow predictable paths of language variation, as such, variation may take place at an interface level and at a syntactic level alike. This does not exclude possible influences from the dominant language, which, however, do not need to be taken as the only triggers of change.

The Syntax of Relative Clauses in Heritage Languages

Preprint

Full-text available

Jan 2022

Alberto Frasson

Resumption in Igbo: Two types of resumptives, complex phi-mismatches, and dynamic deletion domains

Article

Full-text available

Nov 2022
NAT LANG LINGUIST TH

This paper investigates the morphosyntax of resumption in Igbo (Benue-Congo). The first part addresses the syntax and argues that Igbo has two types of resumptive pronouns (RPs): (i) RPs that terminate base-generation Ā-dependencies, and (ii) RPs at the bottom of Ā-movement dependencies. While similar splits have been claimed to exist in a few other languages, established with a limited data set, Igbo provides pervasive evidence for the co-existence of both types within the same language: type-(ii) RPs occur in all Ā-movement dependencies, and there is comprehensive evidence from a variety of movement tests, including also cyclicity effects, parasitic gap licensing, and language-specific diagnostics. We pursue a spell-out approach to type-(ii) RPs à la Pesetsky (1998) and Landau (2006), and discuss potential reasons behind their restricted distribution: type-(ii) RPs only surface in PPs, DPs, and &Ps, which are thus not (absolute) islands in Igbo. The second part of the paper deals with the morphological side of resumption, viz., with phi- and (alleged) case mismatches between the RP and its antecedent. The phi-mismatch provides further evidence for two types of RPs in Igbo. Moreover, it is more complex than the mismatches reported so far in the literature since the loss of phi-information depends on the type of antecedent (pro/noun, coordination). This pattern poses a challenge for previous accounts of phi-mismatches with movement-derived RPs that are based on static deletion domains. We propose that the cross-linguistic variation in phi-mismatches can be captured in a partial copy deletion approach along the lines of van Urk (2018) if the amount of structure that is deleted is defined dynamically. This further supports the relevance of dynamic domains in morphosyntax, in particular in postsyntactic operations, as previously identified in other areas, e.g., in Moskal’s (2015b) work on contextual allomorphy. Finally, we show that the “case” mismatch is a consequence of the (supra)segmental nature of the relevant exponent and the relative timing of the operations involved.

An Experimental Study on the Syntax of English and Egyptian Arabic: A Unified Account of Bilingual Grammatical Knowledge

Thesis

Jan 2022

Yourdanis Sedarous

This dissertation investigates the extent to which bilingual speakers’ cognitive representations of the syntactic structures of their two languages are interconnected. This is a substantially understudied domain despite the widespread existence of bilingual/multilingual communities in the world, including the US. To draw solid conclusions about bilingual individuals’ linguistic representations of syntactic structures, in this dissertation we test bilingual individuals’ sensitivity to island and non-island wh-questions in both unilingual and code-switched contexts. Code-switching is a particularly relevant domain of investigation for determining how structures with similar surface word orders, but either similar or different derivations across the two languages, are processed as by bilingual individuals. The logic here is: if two structures share a syntactic derivation, which we classify as a shared structure, then these structures will be sensitive to the same well-formedness conditions in both unilingual and code-switched contexts. If the two structures have different syntactic derivations, which we classify as separate structures, then we will see divergent sensitivities to well-formedness conditions across the two unilingual contexts, as well as the code-switched contexts. We focused our attention on wh-in situ and wh-resumptive structures in Egyptian Arabic and English. The wh-in situ structure was chosen because, in the right pragmatic context, it has been argued to have a similar syntactic derivation across the two languages, while the wh-resumptive structure was chosen because it has been argued to have a different syntactic derivation across the two languages. We conducted a four-block experiment administered within one experimental session, testing wh-in-situ, wh-resumptive and wh-gap structures in each block. The first block tested the wh-structures in unilingual Egyptian Arabic sentences. The second block tested them in unilingual English sentences. The third and fourth blocks tested the acceptability of the wh-structures in code-switched Egyptian Arabic/English sentences. To test both island and non-island wh-in situ and wh-resumptive structures, we used a factorial design to isolate island effects from extra grammatical processing effects. Based on the results of the experiment we concluded the following: First, in line with the predictions of previous literature, we argue that the wh-in situ structure is not only similar in the surface order across both languages, but the wh-in situ structure in Egyptian Arabic and English share the same derivational properties. Second, based on the reported island sensitivity in the code-switched conditions, we argue that the wh-resumptive structure of Egyptian Arabic is formed via movement in a manner like how clause-initial wh-constituents are formed in English, but that the island sensitivities are masked in the unilingual Egyptian Arabic contexts as the distribution of resumptive pronouns is subject to both phonological and syntactic well-formedness conditions. Based on this discussion, we conclude that the wh-resumptive structure across Egyptian Arabic and English is a partially overlapping structure for the population of speakers recruited in this study, since the clause initial wh-element is formed via movement in both languages, but the insertion of resumptive pronouns is generated as part of the derivation in Egyptian Arabic, but as the result of a production effect in English.

Resumption in Mandarin Chinese

Article

Full-text available

Oct 2022

Victor Junnan Pan

Resumptive pronouns (RPs) exist in different types of A'-dependencies in Mandarin Chinese, such as relativization, Left-Dislocation (LD) structures, cleft-sentences, Across-The-Board (ATB) constructions, etc. Diachronically, resumptive dislocation structures have already been documented in the literature around 502 B.C. In modern Chinese, the obligatory use, the systematical use, and the intrusive use of RPs are all observed. When a prepositional object is A'-extracted, the extraction site must be occupied by an RP given that Chinese does not permit preposition stranding. In certain island free contexts, an RP and a gap are free alternatives in relatives and in LD-structures. However, RP can only redeem the potential violation of island constraints in an LD-structure but not in a relative clause. Gap strategy is always subject to locality constraints. Resumptive strategy only gives rise to island effects in relatives but not in LD-structures. In addition, two empty categories should be distinguished one from the other: gap and pro. The extraction of the direct object of action verbs causing direct physical effects on the object-patient, such as ōudǎ 'beat', will leave a gap, which potentially gives rise to island effects. By contrast, the extraction of the object of stative and psycho verbs which do not cause any physical effects on the object, such as xīnshǎng 'appreciate', never gives rise to island effects. It is assumed that these verbs take pro as their complement and pro functions as an RP in these structures, which saves the sentence from the potential violation of island constraints. Based on the syntactic distribution of gap, RP and pro in relatives and in LD-structures, the minimalist derivation of A'-dependencies involving these elements will be demonstrated.

Subject marking in Awing

Article

Full-text available

Apr 2021

This paper investigates the principles that govern subject marking in Awing (Grassfields Bantu). We observe that the subject marker (SM) that doubles the subject is sometimes obligatory, sometimes optional and sometimes prohibited. We argue that it is the referentiality of the subject that controls the distribution of the SM in Awing, rather than factors such as its morpho-syntactic features or its information structural status, which have been identified to govern argument doubling in a number of other languages with a similar phenomenon. The empirical evidence leads us to conclude that the SM is a pronominal element in Awing rather than an agreement marker. When it occurs, it functions as the argument of the verb and the associated subject NP is base-generated in the left periphery of the clause; when it is absent, the NP is the verbal argument. Awing thus qualifies as a pronominal argument language in the sense of Jelinek (1984); Bresnan and Mchombo (1987); Baker (1996).

Grammatical resumption and dependency processing in Southern Jordanian Arabic

Article

Dec 2022

This study investigates the sensitivity of grammatical resumption to islands in wh -interrogative and relative clause dependencies in Southern Jordanian Arabic ( JA ). An offline acceptability judgment task and an eye-tracking reading experiment were conducted. The results reveal that resumption in southern JA exhibits sensitivity to strong islands, such as adjunct islands, in both dependencies. The findings also suggest that the southern JA parser posits a resumptive pronoun ( RP ) inside islands that allow resumption. However, the parser does not predict an RP inside islands that disallow resumption. Furthermore, quantitative data show that wh -interrogative and relative clause dependencies pattern similarly in their sensitivity to islands.

Parallel changes in pronominal clitic systems: A view from heritage Romance and Slavic (1st draft)

Preprint

Full-text available

Dec 2022

Alberto Frasson

This study discusses the structure and the properties of pronominal clitics from the perspective of heritage Romance and Slavic languages, focussing in particular on changes in clitic paradigms of heritage Venetan and Bulgarian. While the former displays a peculiar distribution of subject clitics, the latter displays a parallel behaviour for object clitics. The two types of clitics are clearly very different, both in terms of grammatical function as well as placement restrictions, but they allow for parallel analyses in heritage varieties of the two languages. The behaviour of these clitics challenges previous models of structural deficiency of functional words, allowing for an analysis that takes into account predictable change patterns in heritage languages. The final goal of the present contribution is to propose a theoretical framework for the analysis of clitics in heritage languages, with respect to their structural and distributional properties.

8 - Resumptive Pronouns in English

Article

Dec 2021

Grant Goodall

In many introspective and corpus studies, inserting a resumptive pronoun in place of a gap in island-violating wh-dependency structures in English is said to amnesty, ameliorate, or repair the island violation, improving the acceptability of otherwise unacceptable structures. Most experimental studies on the acceptability of such resumptive structures, however, report that native speakers of English do not judge island-violating dependency structures with resumptive pronouns to be more acceptable than the ones with gaps. But studies testing the comprehensibility and processing of resumption report that resumptive pronouns increase the comprehensibility of island-violating structures and facilitate processing of long dependencies. These results taken together suggest that although resumptive pronouns in islands do not have an ameliorating effect on grammaticality, they may confer a processing benefit. A question, however, remains as to whether the reported enhanced comprehensibility and ease in processing actually increase the accuracy in interpreting the resumptive pronouns.

Two Cases of Logical Relations: Bound Pronouns and Anaphoric Relations

Chapter

Jan 1996

The Extended Standard Theory elaborated in Chomsky (1977) and Chomsky and Lasnik (1977) postulates the existence of an autonomous interpretive level: the Logical Form (LF) component. Representations at this level are derived from S-Structure representations via transformations (see May (1977)). These representations represent the structural meaning of a sentence and abstract away from other aspects of meaning such as word meaning or pragmatics: I. D-Structure — → S-Structure— → LF-Structure

Multiple variable binding

Article

Jan 1984
LINGUIST INQ

Ken Safir

Sentential Subjects and Proper Government in Chamorro

Chapter

Jan 1991

Sandra Chung

The problems of representation posed by VSO languages arise in some form in every current syntactic theory. How is clause structure represented in these languages, what is its relationship to surface word order, and how does it differ (if at all) from the clause structure of languages such as English? One way of approaching these questions within Government-Binding Theory is to rephrase them as questions about the government properties of the subject. If the subject position in VSO languages could be shown to be properly governed, then there would be some reason for supposing that the clause has an S-structure different from that assumed for English: either a flat S-structure, as in (1a), or else an S-structure in which Infi supported by V (1b), or perhaps Infi alone (1c), properly governs the subject. (For discussion of the specifics of the structures shown below, see the works cited.)

On WH movement

Article

Jan 1977

N. Chomsky

On The Necessity of Binding Conditions 1986

Chapter

Jan 1989

Howard Lasnik

I will be concerned in this paper with the binding conditions — the structural requirements governing certain anaphoric relations. I will give particular attention to “Condition C” effects, and will argue that, contrary to a currently popular view, something like Condition C does indeed exist. That is, I will display a wide variety of facts motivating Condition C which cannot be handled by, for example, independently motivated pragmatic constraints (see Reinhart (1983) for extensive discussion of such constraints) or by core properties of the theory of “Linking” (Higginbotham (1983)). A number of the arguments will be seen to carry over to Condition B as well. [It is on these grounds, of course, rather than on logical or biological grounds, that I will attempt to motivate the “necessity” of binding conditions, as one can surely conceive of an organism, even an evolutionarily successful one, whose linguistic system allows, say, the binding of a pronoun within its governing category.] In the course of the discssion, it will become evident that a partial reformulation of Condition C is in order, but its basic nature as a structural constraint on binding will remain intact.

Bound variable anaphora

Article

Jan 1989

R. May

Is there raising in Modern Irish?

Article

Jan 1980