Pronouns and Non-coreference

HOWARD LASNIK

Abstract
In this 1990 conference paper, I examine the important arguments of Evans
(1980) that pronouns with quantificational antecedents represent exactly the same
phenomenon as pronouns with referential antecedents; and that there are no non-
coreference rules in the grammar, just constraints on antecedence. I claim that
while his arguments are partially correct, they are not entirely correct and hence
that they do not support as strong a conclusion as the one that he draws.

I would like to explore a number of the issues raised by Gareth Evans's interesting
and important paper "Pronouns" (1980). I will not be concerned with Evans's
arguments for the existence of "E-type pronouns" - pronouns with quantificational
antecedents that don't bind them. I take this as well established. Rather, I will
concentrate on the two other themes that Evans develops pertaining to the
analyses and proposals of Lasnik (1976): (a) that, apart from E-type pronouns,
pronouns with quantificational antecedents, as in (3), represent the same
phenomenon as pronouns with referential antecedents, as in (2); and (b) that there
are no non-coreference rules in the grammar.

(1) His mother is here.

(2) John loves his mother.

(3) Every man loves his mother.

Evans brings a substantial amount of data and argument to bear on these

positions. Indeed, many would agree with Soames (1989) in finding "Evans's
criticisms of [the views of Lasnik (1976)] powerful and persuasive." I, too, think
there is much of value in what Evans has to say. But I do not think he is entirely
correct. What I want to do is examine his arguments in some detail to see what we
are entitled to conclude from them.
One obvious, but none-the-less significant, observation of Evans's is that
to an overwhelming extent, the same lexical items are used in constructions like
(2) as are used in constructions like (3). Surely this is no coincidence, Evans
reasonably argues. There is, though, another correlation lurking in the
phenomena, as Evans concedes in passing. The same lexical items that we find in
(3) also tend to show up where there is no possibility of binding, as in (1). Evans
concedes that this is a "complication" for his theory, "but not a complication of

University of Maryland Working Papers in Linguistics, vol. 13, 214-227.

©2004 Howard Lasnik
 PRONOUNS AND NON-COREFERENCE 215

the same kind [as] in Lasnik's treatment." Whether this is so or not is unclear. At
least at one time, Evans evidently felt that a complication of the sort here
conceded for his own theory was, in fact, significant. Evans (1977) based one of
his arguments against the complex predicate formation analysis of pronouns of
Geach (1962) directly upon it:

If we collect together all the devices that can occur in a single term
position chained to some other referring expression, we discover
that they are all capable of being used, in other contexts, to make
independent references. This little generalization will surely come
as no surprise to one who holds a theory according to which
expressions occupying such positions do refer (albeit with the aid
of a co-referring device). But under Geach's theory, this
generalization must remain quite unexplained. [p.491]

It should also be noted that the (2)-(3) correlation of central interest to

Evans, while quite widespread, is not total. It is well known that in many
languages the pronouns used in a coreferential way are not uniformly available in
configurations involving quantificational antecedents. Montalbetti (1984)
discusses numerous cases of this type. (4)-(5) is a representative paradigm from
Spanish.

(4) Juan piensa que (él) es inteligente

Juan thinks that (he) is intelligent

(5) Quién piensa que (*él) es inteligente

who thinks that (*he) is intelligent

The generalization is roughly that both overt pronouns and phonetically null ones
in Spanish can take referential antecedents, but only null pronouns can take
quantificational antecedents. Similarly, in Japanese, the pronoun kare can take
only referential antecedents, while jibun and null pro-forms can take referential
and quantificational antecedents:

(6) John-ga {kare-ga/jibun-ga/0/} saiko-da to omotte-iru

'John thinks he is the best'

(7) Dare-ga {*kare-ga/jibun-ga/0/} saiko-da to omotte-iru

'Who thinks he is the best'
216 HOWARD LASNIK

It is less widely discussed that the implicational relation breaks down in

the reverse direction as well. That is, there are circumstances where a particular
pro-form can be used as a bound variable, but not as a coreferential pronoun.
Patterns of ellipsis with "sloppy" identity vs. "strict" identity are of relevance
here. For example, in Japanese, kare allows strict identity but does not support
sloppy identity (just as expected under a variable binding treatment of sloppy
identity like that suggested by Lasnik (1976), Sag (1976), Williams (1977),
Reinhart (1983), etc., since, as observed above, kare cannot function as a bound
variable). Jibun, on the other hand, allows sloppy identity, but does not support
strict identity, suggesting that it is only a bound variable. The following examples
from Kawasaki (1989) illustrate the pattern.

(8) John-wa kare-no hanashi-o shi-ta-shi, Bill-mo soo shi-ta

-TOP he-GEN story-ACC do-PAST-and also so do-PAST
[strict reading only: John told his story, and Bill also told John's story]

(9) John-wa jibun-no hanashi-o shi-ta-shi, Bill-mo soo shi-ta

-TOP 'self'-GEN story-ACC do-PAST-and also so do-PAST
[sloppy reading only: John told his own story, and Bill told his own story]

Thus, the correlation is not total. This may or may not be directly relevant
to Evans's argument, since his formulation is somewhat unclear. He observes that
"a quantifier can bind a pronoun only if it precedes and c-commands that
pronoun." He further observes that "it is sufficient (though not necessary) for a
pronoun to be interpretable as coreferential with a singular antecedent that it be
preceded and c-commanded by that antecedent". Evans indicates that this
correspondence argues powerfully for a particular sort of unification. But what is
the correspondence if there is not necessity?
Evans also presents what I take to be a language acquisition argument for
his position. As a cognitive scientist, I tend to put great store in such arguments.
According to Evans,

The connection [between pronouns used coreferentially and

pronouns used as bound variables] can be shown empirically by
demonstrating the speakers' capacity to understand new sentences -
- those which contain, as bound pronouns, expressions which had
not explicitly figured in that role before -- a capacity which
presumably relies upon their familiarity with sentences in which
the expression has a singular term as antecedent. [p.352]
 PRONOUNS AND NON-COREFERENCE 217

Evans does not actually provide any empirical basis for his claim, but it seems at
least plausible in general. However, there is some reason to believe that it is
incorrect in detail. Thornton and Crain (1989) report that a substantial percentage
of the subjects in a series of acquisition studies regularly used as bound pronouns
forms that were in some respects special. For example, productions of the sort
illustrated by (10) were a very frequent occurrence.

(10) Which girli said that the dog licked themi on the ear

Here, a morphologically plural pronoun is used with a singular quantificational

antecedent. On the other hand, plural pronouns do not seem to occur when the
antecedent is a singular referential expression, as in (11).

(11) (*)The girl said that the dog licked them on the ear

Thus, it appears that language learners can, indeed, use, hence presumably
"understand new sentences -- those which contain, as bound pronouns,
expressions which had not explicitly figured in that role before...", but even when
those expressions do not otherwise take a singular term as antecedent.
Finally, with respect to (1)-(3) above, I note that the two positions being
compared are not as different as they appear to be. Evans initially seems to argue
that (all instances of) categories (2) and (3) must be reduced to one type. Further,
he indicates that in Lasnik (1976), (all instances of) categories (2) and (1) are
reduced to one type. But appearances are somewhat deceiving here. Evans
eventually concludes that the grammar will

have to recognize as one among other interpretations of a sentence

like [(2)], that it is true iff John loves John's mother. Since it will
also have to recognize that pronouns can be used to make reference
to salient objects, the grammar must itself distinguish between
[these two uses of his], and will treat a sentence like [(2)] as
ambiguous - in the sense that interpretations can be provided for it
which exploit different rules of the grammar. [p.359]

Similarly, Lasnik (1976), in discussing the phenomenon of sloppy identity alluded

to earlier, suggests that a coreferential pronoun in the appropriate configuration
(such as the one in (2)) could be "an instance of a bound variable..."[p.20] Thus,
on both views, (2) is ambiguous, being analyzable in the manner of (1) or that of
(3).
The second major question I want to consider concerns the existence of
non-coreference rules. Evans gives a variety of arguments against non-
218 HOWARD LASNIK

coreference rules. Some of them seem to me questionable. Others seem strong,

but do not necessarily lead to conclusions as sweeping as Evans's. Discussing the
non-coreference (or disjoint reference) rule of Lasnik (1976), Evans reasons as
follows:

To deem sentences like John thinks that John is ill ungrammatical

is to claim the existence of a significant limitation upon the
expressive power of English. How are we to state John's
possession of a "non-self-conscious" belief that he may have about
someone who is in fact himself, but whom he does not know to be
himself?

I'm not certain what position is actually being criticized here. At first blush, it
seems as if Evans is taking issue with a prescriptivist point of view, one that
forbids speakers to use certain sorts of utterances. Needless to say, that is not the
position any generative grammarian would take, so I will not defend it or consider
it further. Alternatively, what Evans might be assuming is that the language
faculty is (necessarily) designed in such a way that anything that is thinkable is
expressible in any human language, and that, therefore, English could not have
the constraint at issue. But it is not a priori clear that the language faculty is
designed in that fashion. Further, even if it is, there is no reason to think that it
would disallow a version of English with a non-coreference constraint. As
Casteñeda (1966) discusses in some detail, such a sentence as (12) can be used to
express precisely the thought in question.

(12) John thinks that he is ill

Further, as Tanya Reinhart observes, even under the circumstances where Evans's
example is acceptable (circumstances which will be considered further below),
the "non-self-conscious" reading is not forced. Thus, contrary to what Evans
states, there is no particular communicative advantage or disadvantage at issue
here. But even apart from these considerations, work in syntax of the last few
decades strongly indicates that the potential communicative utility of a particular
construction does not guarantee its grammaticality. Thus, it might be convenient
to be able to extract an interrogative form out of any syntactic position whatever.
One might, for example, want to ask about the identity of the person that Mary
talked to at the same time that she was talking to John. (13) is the expected
construction, but, as discussed by Ross (1967), extraction in this case is
completely unacceptable.

(13) *Who did Mary talk to John and

 PRONOUNS AND NON-COREFERENCE 219

A further argument of Evans's is that such an example as (14) is not all

that bad, contrary to the predictions of the non-coreference rule.

(14) Oscar loves Oscar's mother

Evans observes that in an appropriate context "in which repetition of the name has
some point...the implication of disjointness of reference is canceled." Evans
further observes that, again contrary to the non-coreference rule, (14) is
significantly better than a corresponding example with a pronoun in place of the
first name:

(15) He loves Oscar's mother

Finally, he indicates that even such an example as (15) does not display "strict
ungrammaticality" if the pronoun is referentially dependent not upon the name
that it c-commands, but rather on some prior occurrence of the name.
I think that every one of these points has some merit. While I, like many
speakers, find both (14) and (15) substantially degraded, they do seem to differ
somewhat in status. The distinction between examples like (14) and those like
(15) was ignored in Lasnik (1976), mistakenly so, I have come to believe. Further,
the (perhaps small) distinction in acceptability between (14) and (15) in English
shows up in much more extreme form in numerous other languages, some of them
discussed in Lasnik (1989). For example, Vietnamese speakers agree with English
speakers in finding sentences with the relevant structural properties of (15), as in
(16), bad on the reading in question, but they find ones with the relevant structural
properties of (14), as in (17), fully acceptable, independent of any special
discourse circumstances.

(16) *No tin John se- thang

'He thinks John will win'

(17) John tin John se- thang

'John thinks John will win'

This strongly suggests that the two cases are to be distinguished. [It must be noted
at this point that Evans claims that it is crucial for the non-coreference rule theory
that these cases not be distinguished. I have to confess that I do not follow his
reasoning here.] But it also strongly suggests that there is something to explain
about (14), or the English translation of (17). Vietnamese and English appear to
be the same as each other in one respect and different in another respect. How
220 HOWARD LASNIK

should the property distinguishing them be characterized? English evidently has

some sort of constraint that Vietnamese lacks. As argued in Lasnik (1989) (and
Lasnik (1976), for that matter) the constraint crucially involves syntactic structure
('command' of some sort, in particular). Mere repetition of names causes some
decrease in acceptability (or, alternatively, some need for special discourse
reinforcement), but, according to the judgment of many speakers, not as much as
when c-command obtains. (14) is considerably more marked than (18) or (19):

(18) After Oscar arrived, Oscar's mother prepared dinner

(19) Oscar arrived. Then Oscar's mother prepared dinner.

Hence, here I disagree with Evans, as well as with Chomsky (1976), both of
whom imply that there is no difference, and that the relevant rule is therefore not
one of sentence grammar. But I concede that (14) is not gibberish. And it is
further true, as Evans claims, that if there is some good reason for repeating the
name, (14) gets better still. However, as Evans hints, this is all still consistent with
the possibility that the infelicity of (14) is grammatical in origin, just as long as
the treatment of (14) is distinct from that of (15). The observation that sentences
display different degrees, or even different types, of unacceptability is a familiar
one. The fact that (14) is not as bad as some other unacceptable example, or is bad
in a different way, does not mean that there is no grammatical fact to explain.
Rather, what we want to explain is why it has exactly the status it has, where part
of that status includes the fact that it might be felicitously used only in certain
special discourse situations (whereas (18) is quite good even in a completely
neutral discourse situation). There is no reason to suppose that all rule violations
yield the same acceptability judgments, that there is one monolithic notion
"ungrammatical". In fact, a vast amount of productive research in syntax in recent
years is based precisely on the differing statuses of different sorts of violations.
For example, the divergent statuses of the question versions of (20) and (21)
respectively have provided the impetus for a major research program.

(20) a. You wonder who built something

b. What do you wonder who built

(21) a. You wonder who built the house some way

b. How do you wonder who built the house

From the observation that (21b) is far worse than (20b), it does not follow that
(20b) is grammatical, or that there is nothing to explain about that example. On
the contrary, there are evidently two things that need to be explained: first, why
 PRONOUNS AND NON-COREFERENCE 221

(21b) is worse than (20b); and second, why they are both worse than (22b) and
(23b):

(22) a. You think John built something

b. What do you think John built

(23) a. You think John built the house some way

b. How do you think John built the house

There is no a priori reason to exclude the possibility of similar grammatical

patterning with respect to coreference.
I turn now to (15), where all parties agree that there is some sort of
grammatical constraint at work. Evans argues that the constraint could not be one
on coreference, but rather, is just on antecedence. He gives (24) as an approximate
formulation.

(24) A term can be referentially dependent upon an NP iff it does not precede
and c-command that NP.

The central evidence for this was mentioned earlier: an example like (15) is not
"strictly ungrammatical" when the pronoun is merely coreferential with the name
it precedes and c-commands, but is referentially dependent upon some other NP,
as in a discourse like (25).

(25) Oscar is not a very family oriented man. But he loves Oscar's mother.

Now if strict ungrammaticality is the status of, say, (15) read back to front, I
readily concede that the example in question does not display strict
ungrammaticality. But surely there is something wrong with it, and, as indicated
above, that is a fact in need of an explanation. Further, (25) seems substantially
worse than (26), again, a fact in need of an explanation.

(26) Oscar is not a very family oriented man. But Oscar loves his mother.

Examples similar to these were presented in Lasnik (1976), and formed the basis
for the argument there that the relevant constraint must be on coreference. A
constraint on antecedence will not suffice. As far as I can tell, that argument still
goes through in its essentials.
The above considerations argue for a non-coreference (or disjoint
reference) condition. This still leaves open the question of exactly what the
semantic import of these notions might be. And it is in this realm that we can find
222 HOWARD LASNIK

a solution to another puzzle noted by Evans. Alongside the examples already

discussed where special discourse circumstances are necessary to facilitate
coreference, we find a class of sentences that are perfect in even the most neutral
of discourse situations. These are the identity statements. Evans argues that any
sort of non-coreference rule would have the effect of making all true identity
statements ungrammatical, an absurd consequence. However, Heim (1982)
presents a simple way to avoid this absurdity. Following Postal (1970), she
proposes that non-coreference rules are prohibitions against presupposed
coreference. Examples like (27), or even (28) are then not problematic.

(27) He is John

(28) He must be John, because he put on John's coat

As Heim observes,

From the point of view of Postal's proposal, [(27) and (28)] are no
longer counterexamples. Whenever [(27) and (28)] are uttered, the
identity of the referent of "he" with the referent of "John" is ipso
facto under debate, and what is under debate is not presupposed.
So these examples are consistent with [non-coreference rules and
disjoint reference rules] when [such rules] are taken to rule out
presupposed coreference of "he" and "John". [p.316]

With the major apparent counterevidence to non-coreference rules

plausibly accommodated, I turn now to further arguments for such rules.
Higginbotham (1985) proposes that "obviation" captures the semantic intuitions
about the disjoint reference paradigm. In the relevant structural circumstances,
two expressions will be obviative, where obviativity is as in (29).

(29) If X and Y are obviative, then they cannot be determined by the

structure in which they occur to share a value.

Note that, as required, obviation is stronger than a mere prohibition of

antecedence. (Note further that along the lines of Heim's proposal, we might take
the determination in (29) to concern presupposition.) Higginbotham summarizes a
number of arguments that this stronger notion is needed. One of these arguments,
originally in Lasnik (1976) in slightly different form, involves so-called anaphoric
epithets (or, as Higginbotham more accurately labels them, "incomplete
descriptions"). (30)-(31) show that disjoint reference effects obtain with such NPs
 PRONOUNS AND NON-COREFERENCE 223

in the standard (precede and) c-command environment even though in other

configurations coreference (and seemingly even antecedence) is freely available.

(30) When John came in, the man looked tired

(31) *John thought the man looked tired

It is unclear what constraint on mere antecedence could account for the deviance
of (31); certainly Evans's constraint does not. Significantly, Vietnamese, which,
as noted above, allows repetition of names even in c-command configurations,
similarly allows anaphoric epithets in such environments, as shown in (32).

(32) John tin thang cho pe se- thang

'John thinks the son of a bitch will win'

Thus, the difference between Vietnamese and English seen in (17) above is not
merely a reflection of the fact that names are used more freely in Vietnamese than
in English.
The discussion thus far, like Evans's argument, has centered on what have
come to be called "Condition C" effects (based on the typology in Chomsky
(1981)): situations where an R-expression (a fully lexical NP) is c-commanded by
another NP. As is well known, there are also obviation effects when a pronoun is
c-commanded by another NP in a particular local domain ("Condition B" effects),
as in (33).

(33) John likes him

Evans briefly alludes to this phenomenon in a footnote:

...I make no allowances for the fact that a pronoun which is to be

referentially dependent upon a prior occurrence of a term which
stands in a certain grammatical relation to it must be a reflexive
rather than an ordinary pronoun... [p.358]

But, as observed by Lasnik (1981) and emphasized by Higginbotham (1985), this

will not suffice. We find obviation with pronouns even when there is no
alternative reflexive form available. Both (34) and (35) are quite odd even though
reflexive analogues to the c-commanded pronouns are impossible (and even
though in (34) antecedence could not be at issue, given that first person pronouns
do not take linguistic antecedents).
224 HOWARD LASNIK

(34) ?*We like me [ /**myself]

(35) ?*John and Mary saw him [ /**himself]

That (33), (34), and (35) are all instances of the same general phenomenon is
strongly suggested by the fact that the obviation effect disappears in all three
types of cases when the pronoun is structurally more distant from the previous
item with which it (partially) shares reference:

(36) John thinks I like him

(37) We think Susan likes me

(38) John and Mary think Susan likes him

Higginbotham also presents another highly significant type of Condition B

effect. Consider an interpretation of (39) where both he and him are referentially
dependent upon John, as indicated by the arrows, in Higginbotham's linking
notation.

(39) John said [S he saw him]

This is clearly impossible even though no constraint on antecedence is violated,

nor is Evans's implied condition on reflexives vis-à-vis pronouns relevant. Once
again, a stronger condition is demanded, seemingly one involving some notion of
overlap in reference.
There is one final argument that Higginbotham hints at that I would like to
develop more fully. Higginbotham indicates that there is no prospect of avoiding
the notion "obviation" in linguistic theory in general, since "switch-reference"
languages of the sort investigated by Finer (1984) mark obviation overtly. In such
languages, when two clauses are in a particular structural relation, a morpheme
appears on the verb of the subordinate clause to signal that the two subjects are
"different". (And, in many of these languages, another morpheme appears to
signal that the two subjects are the same.) Finer points out that "...when the same-
subject marker (SS) is present, the two subjects in question are coreferential, and
when the different-subject marker (DS) is present, the two subjects are
noncoreferential." Before Finer's description, and Higginbotham's conclusion, can
be accepted, we must consider the scope of the phenomenon to determine whether
the DS marker is just signaling lack of antecedence, a possibility that would be
consistent with Evans's general approach. Many of the core examples could be
 PRONOUNS AND NON-COREFERENCE 225

thus described, it seems, including the following Yavapi example that Finer
attributes to Kendall (1975).

(40) tokatoka-c savakyuva u-t-m cikwar-kiñ

Tokatoka-SUBJ Savakyuva see-TEMPORAL-DS laugh-COMPL
'When Tokatokai looked at Savakyuva, hej/*i laughed'

(41) tokatoka-c savakyuva u-t-k cikwar-kiñ

Tokatoka-SUBJ Savakyuva see-TEMPORAL-SS laugh-COMPL
'When Tokatokai looked at Savakyuva, hei/*j laughed'

The DS marker could be taken as indicating that the subject of the adverbial
clause cannot be the antecedent of the (null) pronominal subject of the main
clause, and the SS marker as indicating that there is such antecedence. But Finer,
in explicitly arguing that switch reference is not "simply an optional device to
reduce ambiguity", implicitly argues that it is also not simply a device for
marking lack of referential dependence. First, the whole switch reference system
is still exhibited even with 1st and 2nd person pronouns, as in (42) from Mojave
[Langdon and Munro (1979)]:

(42) ?inyec pap ?-^kxi:e-m Judy-c salyi:-k

I potato 1-peel-DS Judy-SUBJ fry-TNS
'After I peeled the potatoes, Judy fried them'

Presumably, antecedence is not at issue with other than 3rd person expressions.
Second, and perhaps even more significantly, switch reference occurs even with
fully lexical referential NPs, as in (43), again from Mojave:

(43) ?i:kwi:v^-c n^k^mic-m ?^-taly-c tu:pa

men-DEM-SUBJ return=PL-DS 1-mother-SUBJ crack=acorns
'When the men came back, my mother cracked acorns'

Here, even though there is semantically no possibility of antecedence or

referential dependence, the embedded verb still must carry the "different subject"
marker. Higginbotham's conclusion thus seems inescapable, that something
stronger than lack of antecedence is grammatically encoded in these languages.
And the last of these examples suggests that it might even be something as strong
as non-(presupposed)coreference.

Acknowledgements
This paper was presented at the 1990 Princeton Conference on Linguistic and
226 HOWARD LASNIK

Philosophical Approaches to the Study of Anaphora and never published. The

issues still seem to me relevant, so I decided to publish it here. I am indebted to
Jim Higginbotham for very helpful discussion those many years ago.

References
Casteñeda, H.-N. 1966. 'He': A Study in the Logic of Self-Consciousness. Ratio 8:
130-157.
Chomsky, N. 1976. Conditions on Rules of Grammar. Linguistic Analysis 2: 303-
351.
Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Evans, G. 1980. Pronouns. Linguistic Inquiry 11: 337-362.
Evans, G. 1977. Pronouns, Quantifiers, and Relative Clauses (I). Canadian
Journal of Philosophy VII: 467-536.
Finer, D. 1984. The Formal Grammar of Switch-Reference. Doctoral dissertation,
University of Massachusetts, Amherst.
Geach, P. 1962. Reference and Generality. Ithaca, New York: Cornell University
Press.
Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Doctoral
dissertation. University of Massachusetts, Amherst.
Higginbotham, J. 1985. On Semantics. Linguistic Inquiry 16: 547-593.
Kawasaki, N. 1989. Jibun-tachi and Non-coreferential Anaphora. Ms., University
of Massachusetts, Amherst.
Kendall, M. 1975. The /-k/, /-m/ problem in Yuman Syntax. International Journal
of American Linguistics 41.1.
Langdon, M. and P. Munro 1979. Subject and (Switch-)Reference in Yuman:
Folia Linguistica 13.
Lasnik, H. 1976. Remarks on Coreference. Linguistic Analysis 2: 1-22. [Reprinted
in Lasnik 1989.]
Lasnik, H. 1981. On Two Recent Treatments of Disjoint Reference. Journal of
Linguistic Research 1.4: 48-58. [Reprinted in Lasnik 1989.]
Lasnik, H. 1989. Essays on Anaphora. Dordrecht: Reidel.
Montalbetti, M. 1984. After Binding. Doctoral dissertation, MIT.
Postal, P. 1970. On Coreferential Complement Subject Deletion. Linguistic
Inquiry 1: 439-500.
Reinhart, T. 1983. Coreference and Bound Anaphora: A Restatement of the
Anaphora Questions. Linguistics and Philosophy 6: 47-88.
Ross, J. R. 1967. Constraints on Variables in Syntax. Doctoral dissertation, MIT.
Sag, I. 1976. Deletion and Logical Form. Doctoral dissertation, MIT.
Williams, E. 1976. Discourse and Logical Form. Linguistic Inquiry 8: 101-139.
Soames, S. 1989. Book Review: Gareth Evans Collected Papers. The Journal of
Philosophy 89: 141-156.
 PRONOUNS AND NON-COREFERENCE 227

Thornton, R. and S. Crain 1989. Acquisition of Bound Pronouns. Presented at the

Annual Meeting of the Linguistic Society of America.

Howard Lasnik
University of Maryland
Department of Linguistics
1106 Marie Mount Hall
University of Maryland
College Park, MD 20742
USA
lasnik@wam.umd.edu