Maria hafi sviki sig(i)
J believes that M hasSBJ
betrayed self
In accounting for the capacity of speakers to use and recognize anaphoric connection, linguistic theory has been much concerned with how reference is individuated. It turns out that coming up with a satisfying theory of the mapping between the grammatical system involved in whatever speakers use to make reference and their intent to refer is far from trivial. And so, some research, including the work I will be discussing here, tries to get at that correlation by looking at other acts of intension. In addition to individuating the objects and individuals in their conversation, speakers also individuate the utterances, beliefs and intensional states of other linguistic agents. And, though there is a rich literature surrounding the topic of belief ascription, discourse theory has had much less to say on this latter type of individuation.
Accounting for the distribution and interpretation of long-distance anaphora (LDA) provides a particularly focused site for intersecting individuation theory of these two types. LDA show up in a wide variety of languages, but I will mainly be going through the LDA data from Icelandic in order to make the basic points about this correlation. If we look at the interpretive constraints on LDA, the distribution facts fall out beautifully. As an account for the representation of the required interpretation is explicated, what emerges is a grammatically determined relationship between these two types of individuation.
First of all, as noted by Chierchia (1989) LDA are always given de se interpretation [1]. Take the following sentence from Icelandic as an illustration[2]:
(1) Jón(i) telur a
(1) is true only if Jón is aware that it is himself that he believes Maria to have betrayed; he would describe his belief as something like "Maria has betrayed me." (1) would be false under circumstances that do not include some such belief on Jón's part. These circumstances, under which a non de se interpretation would hold, might come about in the following way: Jón comes to the conclusion, through reading or hearing a report for example, that Maria has betrayed someone. What he fails to realize is that the betrayed person reported about is him. A sentence which would be true under these circumstances would be one where the speaker and Jón differ with respect to just this bit of knowledge.[3]
The use of the regular pronoun as in (2) allows the speaker to remain agnostic about Jón's perspectival relation to the proposition that Maria betrayed him.
(2) Jón(i) telur a
The contrast between the two interpretations is undoubtedly more subtle than this test, the difference being something more like public versus private relations to belief states, but however the distinction is demonstrated, the above pair exhibits a robust de se/non de se contrast. [4]
The challenge this data presents to a theorist is to find some way to distinguish representations that would yield de se interpretation from those which would yield non de se interpretation. Clearly no theory which merely marks the co-referential relation between the two arguments is going to be sufficiently fine-grained to do the job. Thus, neither co-indexation nor lambda-abstraction by themselves are candidates for expressing the distinction.
Without going into a formal description of this difference, we can observe that the distinction between the two cases turns on whether or not the speaker asserts that a particular kind of epistemic state holds of some entity referred to in that assertion.Proceeding with the hypothesis that LDA are a species of pro-form dedicated to de se interpretation, keeping in mind the informal characterization given above, it is predictable that there will be no relationship establishable between sig and a non-animate antecedent, an antecedent that a speaker would not be attributing any epistemic state whatsoever to. The following contrast bears this prediction out:
(Siggur
(3) Jón(i) krafess a
(4)
Where no epistemic relationship mediating the antecedent and the
LDA is involved, as in (5), it is also predictable that no de
se interpretation is forthcoming. This lack of any epistemic
role for the antecedent would thus preclude the given indexing:[5]
(5)
Before allowing into the picture whatever possibilities there
are for the formalization of the de se description I have
given so far, even this informal sketch leads us to expect yet
another constraint on the role of the antecedent. Exploring this
constraint narrows the gate to a smaller set of possible theories
of de se representation. This time, the constraint stems
from the requirement that the antecedent be aware that the belief
concerns himself. That is, the issue is the kind of relationship
that the antecedent bears to an epistemic state, rather than just
that there be an epistemic state. In particular, in order to make
sense of the claim about (1) that Jón would describe his
belief as "Maria has betrayed me", it is crucial that
the belief be Jón's belief. It will not be sufficient,
given this requirement, for the antecedent to be in the structural
relation of subject of an epistemic verb in order for de se
interpretation to go through. What is required is that the antecedent
be related to the epistemic state by agency of it. Thus the necessary
establishment of the identity of the intension, which a speaker
can make distinct by asserting an agent for it, predicts the contrast
between passive and active sentences like the following:*Peter(i) yri
glaur ef u
hjalpair sér(i)
P
would-be glad if you helpSBJ selfDAT
(Siggur
(6) Olaf(i) ba
(7)
The same constraint predicts the famous "subject-orientation" of LDA. Take (8) for example. Since the intension here cannot be identified as Jón's, no de se interpretation with respect to Jón can be established. This precludes Jón as a possible antecedent in (8).
(8)
Siggursson (1990b)
brings up another kind of case that bears on this question in
an interesting way. It is reasonable to ask why the matrix object
argument is not a possible antecedent in (9).
(9) Jón(i) sannfær
The question is pertinent at this juncture because the lexical properties of the matrix verb are insufficiently sharp to entail either that it is the matrix subject or the matrix object that is the agent of the intensional complement sentence. While it is perfectly conceivable that Maria would describe her belief as "You had forgotten me", and conceivable even that (10) implies that this is precisely what she believes, I think we have to look at the role that the subjunctive plays in such sentences before concluding that (10) actually commits the speaker to that assertion.
(10) Jón sannfær
J convinced M that you hadSBJ forgotten her
What formal properties the use of the subjunctive induce can be
uncovered by looking at evidence from verbs that may take either
subjunctive or indicative complements. Thráinsson (1990)
points out that upplysta (to reveal) is compatible with
Jón's inadvertent relationship to the revelation only where
the indicative rather than the subjunctive is used.
(11) Jón upplysti óviljandi hver haf
i
bari Harold
J revealed inadvertently
who hadIND hit H
(12) * Jón upplysti óviljandi hver hefi
bari Harold
J revealed inadvertently
who hadSBJ hit H
What this contrast suggests is that the subjunctive use of this
verb entails that Jón have an agentive relation to the
intension. Taking this as a general lexical property of subjunctive-taking
verbs, what I take this example to signify is that the subjunctive
is assigned with respect to the agent of that verb by the same
sort of lexical process by which verbs assign theta-roles to their
arguments. Assuming that something like this is the way to view
these examples, this forces subjunctive complements to be uniquely
associated with one and only one argument.[6]
In the case of (9) and (10), Jón has to be that
argument, since the complement is marked subjunctive, thus entailing
that there is a unique agent of the intension and nothing like
passive morphology, for example, indicates that this argument
is non-agentive.
The relationship between the subjunctive requirement and subject-orientation
is somewhat clarified by these data. They lead to the present
claim, based on evidence independent of the LDA data, that the
description of sig as subject-oriented is somewhat mistaken.
What I think should be concluded about subjunctives (of the type
that are subcategorized for by epistemic verbs, as opposed to
the conditional type seen in (5)) is that it is they rather than
the LDA themselves which are subject-oriented. If a reason for
the subjunctive requirement can be adduced, then the subject-orientation
of LDA follows. Otherwise, both conditions must be independently
motivated. The arguments to follow are intended to demonstrate
that it is precisely its capacity to signal the identity of the
intension that makes the subjunctive a particularly congenial
environment for the LDA.
To get some purchase on the representation, it will be necessary
to discover which of the properties that distinguish subjunctive
and indicative complement clauses might be the one upon which
the representation of LDA depends. First though, the place of
infinitives in this account can be quickly sketched in by moving
into view the postulation that they are tenseless subjunctives.
This would be to say that the properties relevant for LDA in subjunctives
hold also for LDA in infinitives. So saying would seem to be the
right move since the distributional facts differ not at all for
LDA. The distributional differences between the full pronoun and
LDA in these environments are discussed after the LDA account
is put together because the contrasts then become sharper.
The descriptive fact is that the LDA pro-form sig is not
grammatical on any indexing when it occurs in an indicative complement
clause. This is unexpected, especially since the full pronoun
on the indexing in (13) is interpretable as de se.
(13) Anna(i) veit a
A knows that M lovesIND her
That is, the antecedent is the agent of the epistemic state and
(13) is perfectly consistent with Anna's describing what she knows
with the statement "Maria loves me." Although this case
shows that the overlap is not complete between LDA and positions
that can host an argument interpreted de se, it is an important
pointer to the divide between two means of expressing de se
interpretation. This divide becomes clearer when the differences
between LDA and full pronouns is discussed but for the example
at hand, the difference can be approached by noting that the truth
of (13) requires that both the speaker and Anna be committed to
the truth of the embedded sentence; that is the nature of factives
.[7] Note as well, that the speaker
of (13) would not use the quoted statement to describe his knowledge.
These observations lend support to the above hypothesis that subjunctives
have the capacity to distinguish the agent of the intension so
marked, while the indicative does not.
The semantic differences between the two moods may well have the
significance rehearsed here. The question that arises is what
the actual representation responsible for these consequences has
to be in order to yield these differences and, at the same time,
be compatible with the suggestion that LDA are somehow related
to first-person interpretation. Suppose we take the strong position
that sig actually does appear as a first person pronoun.
If this is correct, then reason for the tight restrictions on
the agency of the intension imposed by the subjunctive is clearer.
Take the clause appearing in the subjunctive, for example in (14), to be the speaker's model of what the agent's intension is.
(14) Jón(i) taldi a
J believed that M hadSBJ hit H
The modelled intension being that given in (15), modulo the subjunctive marker, since (15) is meant to model Jón's unembedded intension.
(15) Maria hef
M hadSBJ hit H
The task is now to replace the morphological mood marker on the verb with representation that designates the intension's agent. Let us suppose that the phrase marker containing the intension can be annotated as in (16), where the annotation (i) represents this relation of agency. Linguistic Agent(i) is the "performer" of (16) in some extended sense which includes beliefs, expectations and the like as (represented) linguistic performances.
(16) [(i) Maria haf
M had hit H
However this modelling is achieved, something like it must be
in effect to explain both the differences between indicatives
and subjunctives as well as the particularized-to-agent effects
discussed above. Independent of this data, some system like this
has to be in effect to take care of indexicals even in ordinary
unembedded sentences. I am supposing that the notation above is
a special case of a more general form of performance constraints.
At whatever level performance and modelling constraints operate,
the relationship between the utterance index and NP(i)
in (17), they would need to entail that the form of NP(i)
be that shown below:
(17) [(i) Sally loves [NP(i) me] ]
This indexical relationship is meant to represent the assertion
a speaker makes when using a first person form. A speaker can
no more use an overt first person form to mean someone other than
herself than she can use the expression "that man"
to mean someone other than whoever she is pointing to at the time
of utterance.
One way to understand this idea is to see person features as involved
in a checking relation such that identity between the NP index
and the performance index (utterance or reported performance)
determines first person, and nonidentity determines third person.[8]
The idea here is to extend this briefly sketched performance theory
(See Reeves (1995)) for a more detailed
discussion) to accommodate the interaction of the role of subjunctives
being proposed and modelled first person interpretation of LDA.
Since there is a discourse need to individuate intensions in various
ways, prominent among these ways being the individuation by the
intension's agent, such notation should be considered part of the
syntactic apparatus employed in discourse representation.
Since embedded indicatives, as in (13), cannot be said to have
a unique agent, the representation would not involve the embedded
performance index. On this account, subjunctive morphology just
is the syntactic representation of the embedded performance index.
The first person understanding of sig is thus licensed by its
co-indexation with the performance index. The lack of featural
specification for this pro-form suggests that some sort of phi-feature
licensing is necessary. Keep in mind that this view entails that
the LDA is dependent on the performance index for its interpretation,
though not necessarily on the antecedent. Performance indexes
assigned to reported performances are themselves, however, agent-dependent,
as the ungrammaticality of independently occurring subjunctive
clauses attests.
So why should anyone believe that LDA are actually syntactically
represented as first person pronouns? There are two arguments
for this claim over and above the platitude that representation
determines meaning. The first argument starts from the observation
that Icelandic has no nominative version of sig. The claim
is simple. As a de se pronoun, the LDA would have to occur
as first person with respect to the indirect performance index,
however the agreement features act as an ordinary pronoun would.
That is, by checking the NP index (the nominative argument, in
the case of agreement, the argument itself in the case of pronouns)
against the utterance index. This claim, of course, assumes that
agreement would have no access to the indirect performance index
generated by the subjunctive. Consider (18), uttered by speaker(j),
where the indirect performance index appended to the embedded
bracket is (i):
(18) [(j) Jón(i) telur [(i) a
J believes that self isSBJ a-farmer
The agreement checking on the subjunctive verb checks its sister
NP, index (i) against the performance index (j), thus third person
features surfaces. The de se pronoun, on the other hand,
checks the NP index (i) against the indirect performance index
(i), this relation surfacing as first person. Agreement clash
will prevent a nominative LDA from ever surfacing. One might ask
why hann, the fully specified pronoun can occur in this
position and then be interpreted de se.
This question points up what I take the real difference between
the two pro-forms in this and other languages showing this LDA/pronoun
split. The fully specified pronoun is primarily a direct performance
pronoun while the under-specified LDA need licensing by indirect
performance representation. On this assumption the fully specified
pronoun behaves just as the agreement features do, checking only
against the performance index for its form.
The idea to be exploited here is that full pronouns, as direct
discourse entities, can support the changes in interpretation,
up to syntactic identity, that discourse normally practices; namely,
the perspectival variances that accrue as the discourse proceeds
from one participant to the next. These pronouns have the capacity,
unlike the underspecified LDA, to be interpreted first person,
hence de se, as a form of vehicle change (in the sense of Fiengo and May (1994)).
This sort of vehicle change is routinely practiced in everyday
discourse, as the following discourse progression from utterance
to inference demonstrates:
(19) I like you: said by John to Mary
(20) John told Mary that he liked her
So while the inference to a de se interpretation of a full pronoun
would indeed change it to a first person form, presumably it occurs
as a discourse phenomena and thus would have no interaction with
derivational phenomena like agreement checking. LDA, to reiterate
the position, receive first person interpretation as a matter
of licensed representation, this representation being present
throughout the derivation.
A similar argument can be used to explain what has been called
in the literature "obviation effects" in Romance languages.
The descriptive claim is that a matrix subject cannot be coreferenced
with a little pro subject of a subjunctive clause it directly
dominates. Thus a sentence like (21) would be ruled out:
(21) Tina(i) espera que [ pro(i) gane]
T hopes that (she) winsSBJ
The problem with this description is that it is false. What is
prohibited in (21) is for pro to be interpreted de se.
It is perfectly grammatical where it is interpreted non de
se. The same sort of story holds here as for the nominative
LDA case in Icelandic. On the assumption that pro, like
sig as an underspecified pro-form is theoretically able
access the indirect performance index in order to acquire person
features, the interpretation fails in the same way: by virtue
of feature clash with the verbal agreement features. Where the
cases is differ is that pro, unlike sig, also may
access the direct performance index. In this accessing, it behaves
exactly as the agreement does, and all is well under the non de
se interpretation because no differences in person features arise.
It is somewhat more mysterious why full pronouns are also ungrammatical
in this position. There are two facts that help explain why the
Romance case differs from the Icelandic case in this respect.
First of all, the subjunctive taking verbs in Romance are a somewhat
different set of verbs. The reason this matters is that most of
the verbs that take subjunctive also have infinitival counterparts.
Thus, for each little pro in a subjunctive clause there is a big
PRO in an infinitive clause. As noted by Chierchia (1989)
and others PRO (at least subject-control PRO) is always interpreted
de se. So, (22) the infinitival counterpart of (21) would
be false in non de se circumstances.
(22) Tina(i) espera [ PRO(i) ganar]
T hopes to-win
This is no surprise from the account discussed here. What has
traditionally been called "control" is the representation
that de se requires. It is the perfect candidate for receiving
such interpretation, in that it has no phi-features, it can never
clash with agreement features.
The Romance situation brings to light a novel question: what happens
where there are three forms and two interpretations? When it appears
postverbally, as in (23), an interpretation is available for the
full pronoun co-indexed to the matrix subject.
(23) Tina(i) espera que [ gane ella(i)]
T hopes that winsSBJ she
Though the non de se interpretation is covered by little pro and
the de se by big PRO, it ought to still be possible for the full
pronoun to be interpretable via vehicle change at the discourse
level.
What is interesting about intuitions with respect to this sentence
is that it seems to very strongly favor the de se reading, without
precluding the non de se reading. The de se interpretation for
(23) differs somewhat from the interpretation of (22). It is felt
to be a factual report about what it is that Tina hopes, whereas
(22), though compatible with that understanding, seems to be more
about the fact that there is something that she wants. I take
this split to be a third person reflection of the distinction
seen in nearly all first person attitudes. There is systematic
ambiguity in first person reports of hope, desire and the like.
One may report of one's relationship to a proposition: desire
that the circumstances that would make it true come about, or
one may use the same linguistic apparatus to express a desire.
In its capacity as direct discourse pronoun, it seems that the
de se interpretation of the pronoun in (23) is an inference to
the reportive interpretation that the agent (Maria) would make;
whereas (22) is presumably the third person report of the agent's
first person expression of desire.
The moral to be drawn from this array of distinctions is that
language is often multiply ambiguous, but where differing forms
are available, interpretations that appeared merged acquire discrete
properties as a function of the differing representations. These
three ways that the grammar has of representing other linguistic
agents' intensions are only distinguishable where there are separate
forms for expressing those ways. This means that while, at least
at the use theory level, the relationship between de se and non
de se interpretations are highly dependent on one another, their
representational statuses vary. There are two final pieces of
data from Icelandic infinitives that point to this interdependence
as well.
It is worth considering an ECM example since Binding Theory permits
only one form under co-indexation:
(24) Jón(i) telur sig(i)/*hann(i) vera skrytinn
J believes self /him to-be crazy
In every other case, a sentence with the LDA was false in in non
de se circumstances, but in this case sig is interpretable
as both de se and non de se. The puzzle is why it
should be that being bound in a local domain should give sig
access to the direct performance index. It may be that the access
is mediated through the antecedent because of this structural
relationship.
The final puzzle comes from a second version of misdiagnosed obviation
effects. It is reported that the full pronoun when it occurs in
an infinitive clause cannot be coindexed with a subject. Take (25)
as an example
(25) Hver einasti sjúklingur(i) telur einhver hafa sviki
Each and every patient believes someone to-have betrayed
him
As with the Romance case what is prohibited is a de se
interpretation, not co-indexation. On the non de se interpretation
the sentence is fine. Here the failure to get the de se
interpretation, like the ECM case has to do with the relationship
these two elements have to each other. It will be recalled that
there is no such split in the tensed subjunctive clauses. The
LDA there are all dedicated de se, and the full pronuns
which occur in the positions that also house LDA can be either
de se or non de se.
What prevents the vehicle change of the full pronoun hann
to the first person pronoun as part of some inference level process
is actually quite simple. Recall what sort of process this vehicle
change is postulated to be. The movement is from an utterance
to proposition, as seen in (19) and (20). So, the change to the
first person, which would license a de se interpretation on the
full pronoun has to occur where it is possible to derive a proposition,
but thes are infinitival clauses and hence will never constitute
complete propositions.
These data put together with the intuitions surrounding the full
pronoun in post-verbal position in Romance subjunctives provide
nice evidence that full pronouns acquire de se interpretation
by a distinctly post-derivational means. The closely related meanings
of a de se LDA form versus the de se meaning of the full pronoun
are quite difficult to tease apart, but focusing on the properties
of non de se anaphora in conjunction with these meanings has revealed
some of the several ways the interaction of grammatical representation
and discourse allow speakers to distinguish the indirect linguistic
performances of their fellow speakers.
* I wish to thank Eva Fernandez,
Francisco Ordenez, Carme Picallo and Jose Rojas for their help
with various speices of Romance data. For the Icelandic data,
I wish to thank Höskuldur Thráinsson, Egill Halldorsson,
Halbert Hallmundson, Vidar Hreinsson, and Sigri
[1] Chierchia notes this explicitly
as a de se fact, but many others have given descriptions
involving "orientation" or "perspective" that
are similar in spirit, if not in method. See Kuno, (1972); Thráinsson,
(1976); Sells, (1987); Siggur
[2] Note that the form sig
(as well as its other case-variants) does not vary in form according
to the phi-features of its antecedent. The non-variance holds
for most anaphora that have been classed as long-distance.
[3] Siggur
[4] The classic case demonstrating
the distinction for English was first discussed in Fodor (1975).
This pair may facilitate the intuition for readers unfamiliar
with the de se/non de se distinction:
(i) Only Churchill remembers giving the Blood Sweat and Tears
speech
(ii) Only Churchill remembers his giving the Blood Sweat and Tears
speech.
Since (i) and (ii) differ in truth value, there must be some semantic
difference between PRO in (i) and his in (ii).
[5] Embedding the if-then clause
under an epistemic verb provides the needed environment for de
se interpretation. This allows co-indexation between the LDA
and the subject of the non-epistemic predicate "yr
(A) Jón(i) telur a
(B) Jón(i) telur a
[6] A more generalized theory of
the subjunctive might be that it implies access to other possible
worlds. We would need something like this to account for the
use of the subjunctive in the "ef" clauses such as (5).
Also, the non-intensional status of the subject in (4) could be
explained in this larger theory. The verb in this case means "made
it necessary" rather than "demanded". The semantics
of necessity sentences are clearly consistent with this proposal.
[7]. The factive/non-factive cut
is not exactly the same as the divide between verbs taking subjunctive
complements and those taking indicatives, but it's pretty close.
While there are some factives, such as harma (to regret),
which take subjunctive complements, there are no indicative complement
clauses that are interprted non-factively. With harma, while
it may be true that the factivenesss requires that the speaker
and the agent of the matrix verb be committed to the truth of
the complement clause, it's plain that the agent may be unique
in the feeling of sorrow with respect to that truth.
[8] Second person can actually be represented as a function of
first person.ur
Siggurjónsdottir, who patiently gave their intuitions regarding
the truth of sentences with perversely difficult contexts.
sson
(1990); Hellan (1991), among others. The relation between these
terms of interpretation and logophoricity is not discussed here
for reasons of space. There appears to be no very congenial fit
between the latter and the notion surrounding de se interpretation
either.
sson's
(1990) notion of secondary ego point of view is very similar to
the description used here to get an intuitive notion of the properties
of de se interpretation. The formal representation I propose
for this interpretation yield predictions that his analysis does
not make.
i
glaur". However
as the contrast between (A) and (B) shows, it is the relationship
of the antecedent to the epistemic predicate that is the necessary
one.
hann(i) yriglaur
ef u hjalpair
ser(i)
J thinks that he would-be glad
if you helpedSBJ selfDAT
Petur(j) yriglaur
ef u hjalpair
ser(i/*j)
J thinks that P would-be glad
if you helpedSBJ selfDAT
Bibliography
Anderson, Stephan (1983) "Types of
Dependencies in Anaphors: Icelandic (and Other) Reflexives"
Journal of Linguistic Research 2, 1-23
Burzio, Luigi (1992) "Weak Anaphora"
paper presented at CUNY Graduate Center Syntax Lunch
Casteñeda, Hector-Neruda (1966),
"'He': A Study in the Logic of Self- Consciousness"
Ratio 7, 130-157
Chierchia, Gennaro (1989) "Anaphora
and Attitudes De Se" in Semantics and Contextual
Expression, R. Bartsch, J. van Benthem and P. van Emde Boas,
eds., Foris Publications, Dordrecht
Cole, Peter and Sung, Li-may (1991) "Head
Movement and Long Distance Reflexives" paper presented at
the University of Southern California
Einarsson, Stefán (1945) Icelandic:
Grammar, Texts, Glossary, The Johns Hopkins Press, Baltimore
Evaraert, Martin (1986) The Syntax of
Reflexivization, Foris, Dordrecht
Fiengo, Robert and May, Robert (1994) Indices
and Identity, MIT Press, Cambridge, MA
Fodor, Jerry, A. (1975) The Language of
Thought, Hassocks, Harvester Press
Hellan, Lars (1991) "Containment and
Connectedness Anaphors" in Long-Distance Anaphora,
J. Koster and E. Rueland, eds., Cambridge University Press, Cambridge
Hestvik, Arild (1992) "LF Movement
of Pronouns and Antisubject Orientation" Linguistic Inquiry
23, 557-594
Huang, James C.-T. and Tang, Jane, C.-C. (1991)
"The Local Nature of the Long- Distance Reflexive in Chinese"
in Long-Distance Anaphora, J. Koster and E. Rueland, eds.,
Cambridge University Press, Cambridge
Katada, Fusa (1991) "The LF Representation
of Anaphors",Linguistic Inquiry 22, 287-313
Kuno, Susumu (1972) "Pronominilization,
Reflexivization, and Direct Discourse", Linguistic Inquiry
III, 161-196
Maling, Joan (1984) "Non-Clause-Bounded
Reflexives in Modern Icelandic" Linguistics and Philosophy
7, 211-241
Maling, Joan (1990) "Clause-Bounded
Reflexives in Modern Icelandic" in Syntax and Semantics:
Modern Icelandic Syntax, volume 24, J. Maling and A. Zaenen,
eds., Academic Press, Inc., New York
Picallo, Carme (1985), Opaque Domains,
City University of New York Graduate Center PhD Thesis
Pica, Pierre (1991) "On the Interaction
Between Antecedent-Government and Binding: the Case of Long-Distance
Reflexivization" in Long-Distance Anaphora, J. Koster
and E. Rueland, eds., Cambridge University Press, Cambridge
Platzack, Christopher (1987) "The
Scandinavian Languages and the Null-Subject Parameter" Natural
Languages and Linguistic Theory 5, 377-401
Platzack, Christopher (1986) "COMP,
INFL, and Germanic Word Order" in Topics in Scandinavian
Syntax, L. Hellan and K. Christensen, eds., Reidel Publishing
Company, Dordrecht
Progovac, Ljiljana (1992) "Relativized
SUBJECT: Long-Distance Reflexives without movement" Linguistic
Inquiry 23, Squibs and Discussion, 671-680
Reeves, Ruth (1995) "Reference and Representation in Belief Attribution," CUNY manuscript
Reinhart, Tanya and Reuland, Eric (1991)
"Anaphors and Logophors: an Argument Structure Perspective"
in Long-Distance Anaphora, J. Koster and E. Rueland, eds.,
Cambridge University Press, Cambridge
Rizzi, Luigi (1989) "On the Anaphor -
Agreement Effect" paper presented at the annual meeting of
the Revista di Grammatica Generativa, 1986
Rögnvaldsson, Eiríkur and
Thráinsson, Höskuldur (1990) "On Icelandic Word
Order Once More" in Syntax and Semantics: Modern Icelandic
Syntax, volume 24, J. Maling and A. Zaenen, eds., Academic
Press, New York
Sells, Peter (1987), "Aspects of Logophoricity"
Linguistic Inquiry 18, 445-479
Siggurjónsdottir, Sigrídur
and Hyams, Nina (1992) "Reflexivization and Logophoricity:
Evidence from the Acquisition of Icelandic" paper presented
to the Proceedings of the 7th Workshop on Comparative
Germanic Syntax, Kluwer, Dordrecht
Siggur
Siggursson,
Halldór (1990) "V1 Declaratives and Verb Raising in
Icelandic" in Syntax and Semantics: Modern Icelandic Syntax,
volume 24, J. Maling and A. Zaenen, eds., Academic Press, New
York
sson, Halldór
(1990b) "Long Distance Reflexives and Moods in Icelandic"
in Syntax and Semantics: Modern Icelandic Syntax, volume
24, J. Maling and A. Zaenen, eds., Academic Press, New York
Soames, Scott (1992) "Attitudes and
Anaphora" paper presented at the CUNY Graduate Center Philosophy
Colloquium
Thráinsson, Höskuldur (1976)
"Reflexives and Subjunctives in Icelandic", NELS 6,
225-239
Thráinsson, Höskuldur (1990)
"A Semantic Reflexive in Icelandic" in Syntax and
Semantics: Modern Icelandic Syntax, volume 24, J. Maling and
A. Zaenen, eds., Academic Press, New York
Thráinsson, Höskuldur (1991)
"Long-Distance Reflexives and the Typology of NPs" in
Long-Distance Anaphora, J. Koster and E. Rueland, eds.,
Cambridge University Press, Cambridge
If you wish to make a public comment about this paper; please send
your message to:
linconf@tamvm1.tamu.edu
Discussion of this paper is available at the Discussion Web Site at: