CC BY
29
ш
ЭПИСТЕМОЛОГИЯ & ФИЛОСОФИЯ НАУКИ • 2014 • Т. XLI • № 3
I
DIRECT REFERENCE FOR INDEXICALS
AND AMBIGUOUS SELF-IDENTIFICATION
Alexei Chernyak -
Candidate of philosophical sciences, docent at the Department of Social Philosophy of the Faculty of Social Sciences and the Humanities of the People's Friendship University of Russsia. E-mail:
The article is devoted to a philosophical discussion of semantics of indexical expressions inspired mainly by theories of D. Kaplan and P. Schlenker. The author considers Schlenker's arguments contra Kaplan that referents of indexicals sometimes may not be considered as directly provided by contexts of their use. He argues that the idea that indexicals can have shifted reference can and should be developed further. The author discusses cases which seem to call for a broader understanding of context dependency of indexicals. In particular the cases that introduce contexts where the speaker does not referto herself as a unique individual specifically located in space and time. It is argued that references in such cases can hardly be explained as eithershifted or not shifted in the standard way suggested by Schlenker. These examples are followed by more examples of conditional shifting and examples of split shifted parameters. The author argues that such real speech situations along with communicative intentions of speaker suggest that a broader understanding of context dependency of indexicals should be taken into account.
Key words: indexicals, semantics, context shift, conditional shifting, split shift parameters, D. Kaplan, P. Schlenker.
1. Introduction
Sentences like (1) "Mary said that I am stupid", (2) "I am here now", (3) "We command you to obey!", (4) "John hopes that he will be elected", (5) "Mary left two days ago", (6) "The present King of France is bald" etc. contain indexical expressions such as pronouns "I", "he", "you", "we" or indexicals of time, place and other coordinates ("two days ago", "present", "here", "now" etc.). References of such expressions usually change together with the circumstances in which they occur. This is generally understood as context dependency of reference.
In (1), (3), and (4) sentences contain embedded clauses, which might be evaluated relative to a different context or set of circumstances, than that of the utterance of the sentence. Thus in (4) "he" may refer to John or to someone else, dependent on whom John had meant and what was the speaker's idea of the report (what content he/she wanted to ascribe to John's reported hope attitude). Such expressions are called attitude reports. But "John" may refer to different persons if, for example, there are two men named "John" in the room where (4) was uttered and each John believes that he will be elected. If so, then context dependency of the meaning of an expression (indexical as well as nonindexical) can be context dependent no matter whether the expression occurs within an intentional context or not. "I" in (2) evidently refers to
1 The study has been supported by Russian Foundation for Humanities, project 14-33-01043.
1
82 Language and Mind
different individuals depending on who utters the sentence, and "here" and "now" refer to different places and times depending on the place and time of the one who utters (2), but also perhaps dependent on what the speaker thinks about the place and the time of the utterance. In (6) "The present king of France" would refer to different individuals depending on the time of utterance (hence to no one in real world).
If there are rules representing semantic competence of a normal user of a natural language2 some of them should cover the ways indexical expressions of the language have references. Thus it is supposed to be normal in English to assign to "the present", and to "I" in direct speech references, correspondingly, to the time of the utterance and to the one who makes it (uses the words meaningfully).
Terms with essentially context-dependent meanings are usually interpreted as variables; their contribution to the meaning of a sentence makes the latter context-dependent too. In fact there might be found few (if any) instances of completely context-independent assignments of meanings to sentences.
The general question for semantic theories is: what principles of context-dependency might fit the behavior of indexicals, as demonstrated in attitude reports and beyond them? The popular answer to be analyzed below is that contexts determine references, i.e. that an expression, if meaningfully used, 'chooses' its denotation strictly relative to a certain context. I doubt though that this answer fits equally well all relevant cases; in what follows I will investigate some such cases.
2. The norm of direct reference
A very popular theory of reference for indexicals which was almost
commonly accepted up to recent time was introduced by D. Kaplan
[Kaplan, 1989]3: according to him denotations of indexicals within a
sentence are fully determined by parameters of the context of the utterance
of the sentence. This kind of reference Kaplan calls direct: intuitively
directness of reference implies that semantic rules directly prescribe to
expressions of certain type (singular terms, which indexicals belong to) s
that their referents in any possible circumstances would be their actual ®
referents, i.e. individuals assigned to expressions in contexts of their ^
utterances. This feature is provided by two levels of sense which Kaplan IB
associates with linguistic expressions: the content and the character. ®
S)
-;- ®
2 What the one should know to use an expression correctly or effectively. Many theorists,
including M. Dummett (e.g., [Dummett 1993]) and R. Montague ([Montague 1974]), suppose a there are such rules; but cr. the well-known criticism by L. Wittgenstein: [Wittgenstein, 1953,
§§ 185-289]. ¿1
3 See also [Perry, 1977]. ™
Content he defines as a function from possible worlds, a commonly accepted notion for a way that things (in the actual world) might have been, to an extension of the expression with this content in this world. And characters are semantic functions from possible contexts of the utterance to its intensions (contents) in these contexts [Kaplan, 1989: 506]. Characters of natural language terms function as common semantic rules for their users. Thus an appearance of «I» in the utterance must refer to the one who is actually speaking by semantic rule governing the use of this term. In the same way "you" must always refer to the one whom the speaker addresses, "he" - to the one whom speaker is talking about, "now" and "here" - to the time and the place of the utterance etc.
Of course this wouldn't work if the term was used in an abnormal way or according to some other rule (e.g., in citation "I" wouldn't ordinarily refer to the speaker); but at least it is supposed to be the case that the rules of natural languages support direct references of indexicals in their most ordinary uses. An expression may refer to different things in different contexts or relative to different contexts. Thus (2) usually means that the speaker is in the place and at the time of the utterance, hence "I" in it should refer to the one who occupies certain place at certain time. In other words the context of the utterance is the one which the indexical has to 'choose' its denotation from; since different individuals may be speakers of (2) the expression would refer to a particular person only in the context in which this person is the speaker of (2).
But the rule described above wouldn't work if "I" was used in an unusual manner (i.e. to mean someone who is not strictly the speaker): therefore "I" might refer to different persons (and perhaps even things) relative to different contexts, each corresponding to some way of its use (or corresponding semantic rule). Ordinarily model-theoretic accounts based on Kaplanian notion of context-dependency rule out contexts alternative in this sense, for characters are associated with rules by one-to-one relation (one semantic rule corresponds to one character).
As a common rule the character of a term should have a descriptive expression to be learned. But Kaplan notes that rules do not include enough details to determine referents of singular terms; therefore no known ^J description could determine a context-dependent reference. The character ■S does it by introducing contexts to supply details not mentioned in S descriptions [Kaplan, 1989: 498]. Thus "I" is not synonymous to "The one ^ who utters this token of "I"" or whatever even in the context where its
reference is governed by this rule. 0 Many cases seem to support the idea of direct reference of indexicals:
jj we actually rarely hesitate to give priority to the speaker, time of the utterance etc. in assigning denotations to "I", "now"and alike in most ® ordinary communicative situations. And perhaps there are certain rules of natural languages. But there are also counterarguments to the idea that direct —' reference is the only way of indexicals to have meaning in communication.
3. 'Monstrous' indirect reference
Some counterevidence against direct reference for indexicals comes from ambiguities of the notion of the context of utterance. Thus any utterance may be made at one time and heard at another. Kaplan claims that "insofar as the agent and audience of a given context can differ in location, the location of a context is the location of the agent" [Kaplan, 1989: 526 ft.]. But different times and places may correspond to the very act of uttering: if the sentence is long enough or discontinued by some long pause, the speaker might change his/her location while uttering the sentence. An agent may be in one place and make her utterance in another through some mechanism. Some pretty ordinary cases (e.g. SMS or phone talks) generate ambiguities of the utterance's location4. When "I am exhausted" uttered in the actual world is evaluated relative to a world where all physical processes in living beings go much faster than we actually think, the speaker's state might have been changed from exhausted to well-being in a period of time shorter than the period she needs to finish the utterance. Since this world could be the actual one the utterance in its actual context might be simultaneously true (of the time of the beginning of utterance) and false (of the time of its ending).
On the other hand not only the one who actually produces an expression may be identified with the agent of its utterance if there is someone else whose thought or words were represented and if the speaker has nothing to add to its content but signifying its source5. Meanwhile the time and the place of the utterance might be still identified with the time and the place of the source.
There are theories in contemporary linguistics which explain behavior of indexicals, in particular in attitude reports, as indirectly referential. Much weight in them has the claim that this is due to the existence in natural languages of 'monstrous' modal operators quantifying over contexts. Such is the position of [Schlenker, 2003]. Schlenker argues that though English in general supports direct references of indexicals some cases of indirect speech in some other languages are interpreted otherwise. Thus if (1) was said in Amharic it could report Marie's talking about herself.
Schlenker explains such cases as containing 'shifted' indexicals, i.e.
such that the term is evaluated with respect to a context that is different s
from the context of the actual utterance. So there should be natural ®
language terms functioning as special modal operators which command ^
assignments of references to indexicals in attitude reports in a way that IB
allows shifting contexts of speakers to ones of agents of attitudes. ®
8)
—:--n
4 Also consider (3) uttered by a speaker on behalf of a group of people which of course are not in the same place as the speaker, or someone pointing at some place on the map while a
saying "We are here now".
5 For the notion of a source introduced as a type of agency distinguished from those of the self and the pivot, cf. [Sells, 1987: 455-456].
IB
ALEXEI CHERNYAK
Much linguistic evidence on the existence of «monstrous» shifters was accumulated from different languages including Zazaki, Amharic and Russian ([Anand, Nevins, 2004], [Anand, 2006], [Schlenker, 2003]). But for the sake of brevity I will reproduce here only some examples from English: "Each year, it was clear to John that, some day, all of the students now studying with him would be on the Editorial Board of Linguistic Inquiry". The «now" operator won't help in this case, because it displays the behavior of a variable bound by the time quantifier "each year". I slightly modified the example). The English sentence "John says I like cheese" is ambiguous without punctuation or special intonation: one reading is quotational ("John says: 'I like cheese'") while the other is non-quotational: "John says <that> I like cheese"; and the latter may be read as a monstrous construction if assigned a de se reading: "John says <about himself> <that> I like cheese"6.
Interpreting the data [Schlenker, 2003, 2010] explains "monsters" as modal operators which manipulate context-parameters directly. In general his account can be treated as belonging to the framework in which contexts determine references for indexicals in a one-to-one relation, when a denotation should be «chosen» either from one or from another context.
In what follows I will call Schlenker's account The Account. Not trying to argue against 'monsters' as they are represented in The Account, I will explore some cases where indexicals seem to be used a manner which challenges it, i.e. so that they cannot be explained away as providing indexical references shifted (or not) between contexts.
4. Ambiguous self-identification
According to the Account contexts are understood as tuples of parameters. But they seem to be not as fine-grained entities as the circumstances which determine the reference of indexicals. If agents are invited to be parts of contexts with all their relevant attitudes defined as dispositions, then individuals and their personal histories should be parts of contexts because in order to ascribe a disposition one needs to explore a subject's past and relate it with her present state. And this could mean that no personal state corresponding to "here" and "now" of the utterance, hence identifiable in principle with the corresponding parameter, might be a real agent of an attitude of saying the uttered sentence; for to be such an agent one would have to be properly historically connected with at least one of her past thoughts or belief states.
If, on the other hand, the time of utterance is strictly identified with the period of time during which an expression was pronounced (left aside
6 Ibid., 22.
IB 0 a
IB
s
a
c
IB
a
problems with synchronization of audio effects) the agent of utterance without her personal history would be at best equal to some slice of the full-fledged agent of the attitude-set. How such slice may be the one who thinks (since thinking requires having some epistemic attitudes), and expresses thoughts (since that requires an attitude of knowing at least one public language), not to mention accomplishing historically more grounded intentions, projects and objectives, is a question in need of answer in terms of some more refined objects, than those of an agent and a time of an utterance7.
However I would like to focus on cases where the problem of agency has a different nature, following rather from pragmatic ambiguities than from those of model-theoretical concepts. I claim that some communicative situations are such that whatever is chosen as the context of an utterance it is unable to provide a unique set of parameters relative to which denotations of indexicals used in that context should be chosen.
Suppose someone says
(7) I tell you that I am not the one you think I am.
Each of the three occurrences of "I" in (7) may have a different denotation. In a minimal sense consistent even with the rule English use of "I" in both direct and indirect speech, the first occurence denotes the speaker as she is at the moment of this occurrence (t1), the second denotes the speaker as she is at the moment of this occurrence (t2), and the third denotes the speaker as she is at the moment of this occurrence (t3). All three moments are parts of the time of the utterance, and nevertheless the world of the utterance might be such that the speaker did not pass unchanged through the sequence of moments thus distinguished.
But in a more obvious way there are three different senses or roles which may be and sometimes are associated with these three occurrences of the same indexical: that of the speaker, that of the speaker's self, and that of the addressee's image of the speaker. In yet more obvious way we may distinguish between our bodies and souls so that the distinction is not represented by our self-identifying uses of words due to some shared linguistic habits, limited thesaurus, considerations of simplicity and convenience or whatever. Let the body be always where and when the agent is; but the soul need not: it may be supposed located by the agent in any virtual circumstances.
Thus due to the simple force of my imagination "I" in (2) may mean something ambiguous: that my body is in the place and at the time of the utterance and my soul is in some other place and at some other time. There is an inherited way of treating such examples, according to which for each
7 Relevant refinements were suggested, in particular, in: [Sells, 1987], [Higginbotham,
2003].
(8
dl
occurrence of «I» it chooses either the body or the soul of the agent to be gj located in some certain place (at some certain time). But I think this is not W always the best explanation. gi
IB
The simple way to insist that in some cases like (7) self-reference should be treateed as ambiguous is to claim that since an agent may actually mean to refer to herself in an ambiguous way the reference to be assigned to certain occurrence of «I» should reflect this. But certainly what is said must not depend on what was thus meant in that crude manner.
Nevertheless, to identify, e.g., (4) as a de se report we need (at least) two kinds of evidence: that "he" in it is not bound anaphorically by an occurrence of a term with reference to someone different from John, and that John's reported attitude was the hope that John be elected8. This is just one of the standard ways of dependency of semantic assignments on pragmatic and psychological features which could hardly be denied even by the one who do not share sympathy toward pragmatic accounts of semantics in general.
Now consider a situation in which the speaker truly reports:
(8) The world of politics is highly cynic, John believes that he should change himself essentially to succeed in this world, and that after doing this he will be elected.
Should the last occurence of «he» be treated as a substitute (made by convention) for «the one he then becomes», not contributing to the meaning of the sentence? I doubt this. An «essentially» operator implies that the meaning of this token of "he" is not identical to the agent But if asked «Do you mean that someone who shares with you only the body and no part of the soul will be elected?» John might deny this on the grounds that that he (= agent) should retain the ability to claim «I am elected», where the meaning of "I" shares some essential features (some personality) with the agent of utterance. In that case it cannot be truly said that either John thus believes de se to be elected or believes someone else to be elected. In that case the last occurrence of "he" in (8) would be ambiguous from the point of view of the agent of a reported attitude, and if the truth of the report depends essentially (as it is supposed to be) on the reported attitude's being as it was reported, the semantics of this token of "he", rather than just its pragmatics, has to be sensitive to the agent's self-identifying idiosyncrasies, when they influence her reported attitude ™ (as in the scenario described above).
2 So, I suppose, there is some evidence, independent enough from the
^ most fundamental theoretical presumptions, to accept that in some (S situations ambiguous self-identifying dispositions of participants in ® communication may contribute to assignments of references to indexicals (g in the manner which could challenge their normal explanation in terms of
a c
IB
a
switches between concurring readings. Thus if we choose a de se reading of all embedded pronouns in (8) (which would correspond to the scenario) we
8 When the reference of «John» is presumed invariant for all occurrences.
would therefore predict that the reported attitude was about John relative to the set of John's beliefs reflected in the attitude, or in other words, that it was about John in all worlds consistent with John's current beliefs. But that is not the case, since there is some sense (world) of meaning himself by John according to which not all embedded content of (8) was about John.
5. Partial shifting
Ambiguous self-reference may lead to indexical shifting which is essentially different from the one proposed in The Account. Consider a scenario where playing a computer game the gamer exclaims
(9) Here I am!
when her virtual character enters the virtual enemies' site. "I" may count denoting the speaker or the computer character, and "here" would correspond either to speaker's location or to the one of the character. In the first case the agent might mean that she (= the agent of the context) appeared suddenly before the eyes of her virtual enemies. This introduces the reading (9'), according to which the agent is in loc(x, y, z) of the game. But the real speaker did not change position in the actual world, so it would be just false for her to state that she suddenly appeared in the location she have already been occupying (if only invisible up to the actual moment, which is not the case).
An alternative meaning of (9) might be either that the agent (^ the agent of the context) appeared suddenly before the eyes of her (^ the agent of the context) (virtual) enemies, or that the same agent appeared before the eyes of her (= the agent of the context) enemies. But normal gamers, like actors in dramas, rarely fully identify themselves with their computer characters and more often recognize the difference between real and virtual surroundings of the «center of deixis». Thus neither reading seems to be plausible enough if it requires choosing denotations for indexicals of (9) from the one or the other context exclusively.
The problem here does not consist in ambiguous self-reference, which could be overcome by the use of some more fine-grained concept of agency. Rather it consists in that the agent of the required type looks unavailable in any context or set of. In other words, it does not seem standing in one-to-one relation to any of them.
The solution for such cases in which more than one context or set of parameters is involved may consist in allowing some the reference of an W indexical to be partially dependent on any one of them, shifted or not. Q)
Consider the case of Sherlock Holmes who says jg
(10) I live on Baker Street.
(8 0 a
a
Who is the agent of the context here? We have two candidates: the author and the character, hence two possible contexts of utterance. We cannot say categorically that in such case the author is always the speaker. The author is the narrator in relation to the whole of the text; but inside the text the author's relations to different phrases may and usually do vary: from a mere observer to an active participant of interactions described. To choose a set of relevant readings for (10) the evaluator must decide which discourse relations it might be accomplishing: could it be relative to its position in the text from the author's point of view, for instance?
Since the question concerns counterfactual matters there may hardly be enough evidence available from the real world evaluating position to rule out some competing hypotheses. Let's take (10) as reduced to just two: (10') cc [live on Baker Street]Wa and (10'') cc [live on Baker Street]Wc, where «cc» is the character of the context and «wa», «wc» are worlds of the author and the book correspondingly.
Let's suppose further that the author (from the point of view of evaluator) does not assign existence to any other thing such that it is Baker Street than that of her actual world (wa); in that case to interpret (10) as (10') would be the same as to assign it the content Sherlock Holmes does not live on Baker Street. But (10') and (10'') give cc [live on Baker Street]wa or [live on Baker Street]Wc, which cannot be true in the scenario just introduced. Then the content of (10) would rather be represented as cc live on [Baker Street](Wa^Wcy But «Wa^.Wc» here should not be read as ordinary shifting, to be described by the rule: substitute denotation for "Baker Street" in the author's world by the corresponding entity of the world of novel, - for, according to the scenario, there is no corresponding entity in wc. Rather the relation between worlds should represent them as a contained and a containing respectively9. Thus three roles attached to three occurrences of the indexical in (7) make their denotations result from some such relations. For the self of the speaker is how the speaker in her context views herself, and her image is how the speaker from that very context sees
9 Possible worlds are often (in formal semantics) thought to be mere parameters convenient for logical analysis, not objects; but since they are bound by operators (at least by ^-function, even if not by existential quantifier) they look proper parts of models providing denotations for certain expression types. Now the world w may consist of sentences ® representing facts of this world only or allow some sentences which represent facts or states of
Q affairs of some other world w' possible from the point of view of w. Thus if Sherlock Holms
B dreams in the novel that he is Conan-Dole, the world of his dream may be considered as the
IB possible world relative to the world of the novel, and afortiori as the possible world relative to
Bl the actual one. That it becomes possible from the actual point of view through being possible
8) from the point of view possible from the actual one (and perhaps sometimes only this way) may be, and I think would better be interpreted, as the reason to count such world not independent part of the set of worlds possible relative to the actual one, but such that belongs to the set due to the relation (of being contained in) to some other world in the set. I consider Jg this feature to be characteristic of virtual worlds which would then be distinguished from
merely possible ones. Of course, metaphysics of possible worlds is a very subtle matter which ^^ I wouldn't like to sink in, but I think that what I've saidhere looks at least intuitively plausible.
the addressee seeing the self of the speaker. Then each occurrence of "I" in the sequence will choose denotation partially from the utterance's context (the world of the speaker's point of view, to be more precise), and partially from the worlds representing one of the rest points of view.
The corresponding rule then might be: the denotation should be taken from the containing world as if existing in the contained one; but it wouldn't fit cases when (10) is supposed to express something literally true. A better way to define semantic condition like «Wa^.Wc» seems to describe it as partial shifting, where (as in the game-case) real object is taken as bearing some counterfactual properties, hence the denotation is partially determined by each of the relevant sets of parameters (worlds or contexts, if wp is considered the world of c' where cp is the one who speaks (10)), and completely determined by neither of them.
This feature characterizes a considerable number of cases which include beside games, some patterns of dreaming, text-reading or movie-seeing; i.e. such when an agent of the context of action is drawn deeply enough into the plot, though still recognizes the difference between reality and fantasy. Consider a description of replicas of dramatic characters within a screenplay made in common notation:
Peter: I came to make you pay!
Sam: But who are you to claim this?
In these roles the sentences described are considered to be representing direct speeches of the characters whose names should be read as no parts of the sentences themselves. The names or even indexes of the characters to whom the sentences are to be attributed could be eliminated from the text, but the competent reader knows anyway how to read properly the sentences thus used: i.e. as rather replicas of personages than the author's ones. Now, when Peter says in ordinary direct discourse
(11) I came to make you pay!
I should be obviously read as denoting Peter; but if we knew that Peter identifies himself as a person (unlike the one who now here speaks) with someone like the literature character we would be as well licensed to assign to I in the utterance the denotation of not just Peter as the speaking one in the context of the utterance, i.e. to use some other context as a source of relevant data.
6. Conditional shifting and split shifted parameters
(8 0 a a
a
Some partial shiftings may be further explained as conditional and some as originating from split parameters. Since a speaker in situations like 9,10 or 11 may intend some priority in self-identification swaying between W contained and containing worlds, such that the relevant denotation should ~
Q
be taken primarily or in its most essential parts from one context against all others, the result could be described as conditional shifting. Consider
(12) John said I be a hero.
Since a grammatically inappropriate linguistic form is used in (12) the evaluator (whether she is a participant of the current communication or an external theoretically motivated observer) has to choose whether to count it ill-formed and hence unavailable for interpretation, or well-formed, though in a manner corresponding to parroting English or bad use of it. If the decision was made in favor of well-formedness of (12) it might be read de se in spite of the fact that according to the rules of English this should not be done.
Let there be two pieces of evidence - A and B - one supporting the first decision against the second, while the other - the second against the first. Let A witness that John's original utterance was "I be a hero", and B shows that in producing (12) the speaker believed that John's use of «I» was unusual, i.e. tending to refer not to himself but to his hearer (which happened to be the speaker of (12))10. In particular, evidence A supports assigning John as the reference of I, while evidence B supports assigning the speaker as the reference of I: let's accept that these inferences are strictly contradicting for the evaluator in the scenario.
Since it is up to the agent how to use a piece of evidence she may assign to A and B unequal weights relative to semantic decisions to be made about (12). Then, if equal for the evaluator under comparison, A and B together might justify the choice of two separate readings for (12), one predicting shifted and another unshifted reference of "I". But what if A and B are weighted unequally by the agent? In this case it could be rational to choose a reading which reflects both concurring hypotheses concerning the reference of «I». The reading then should represent these hypotheses in accordance with the weight assigned to the evidence for each of them. If A has more weight than B for the evaluator of (12), though neither is ignored, an adequate reading might then be such that assigns to "I" the shifted denotation conditionally, i.e. so that if an expression under such an assignment fails to refer to anything of the presupposed kind (right individual in our case) it should be assigned denotation as if not shifted. The procedure thus would roughly correspond to the rule: «choose the agent of thus reported attitude instead of the speaker (which is to be chosen in normal attitude reports), but if the sentence fails to have truth-value choose the one who speaks».
Suppose the speaker mistakenly ascribes the reported attitude of (12)
d)
m to John: since John's uttering of the embedded clause of (12) may be not identical with its uttering by the agent of the reported attitude, there could gi be a reading of (12) such that it satisfies the subject-matter of the report,
10 This hypothesis would not completely irrationalize the speaker's point of view if, for example, John is not a fluent speaker of English.
and would say (under an assignment of shifted reference to I) (12') that the one who is (in fact) the agent of the attitude said that the one who is (in fact) John be a hero. Evaluated in the world when there is no individual for the role of John being supposed to be a hero by the agent of the attitude, (12') would lose truth-value due to its embedded clause.11 In these circumstances the choice of shifted denotation for I would look irrational, hence the shift might be constrained so as to return unshifted denotation if the shifted one produces failure of the truth-value.
This way of shifting might be called conditional; it looks more like a recursive than a one-way procedure, i.e. implies at least two steps where the first substitutes one context (or set of parameters) for another, and the second verifies the result of the first step. Conditional shifting makes denotation assignments to indexicals in its scope disjunctive, but in a peculiar way which does not presuppose the use of two or more competing readings.
Of course the result might be discharged if circumstances of evaluation were supposed to have no independent contribution to the meaning of an utterance; but I think they still make such a contribution even when the speaker may not be attributed non ambiguous identification of at least one of parameters contributing to the meaning of the utterance. In cases like that of the game observed above this cannot be done relative to the agent parameter, in other cases (like in the one of Sherlock Holmes above) - with the location parameter, etc. Semantic decisions are licensed to be made by the evaluator if the speaker is unable to provide content completely representable by a set of readings.
In principle any consciously made utterance may be associated with a pair of contexts such that one is the context of thought and the other is the context of expression of this thought. In most cases these contexts might be said coinciding. But if we consider
(13) I am tired
we may say that it expresses the thought which came to speaker's mind a minute ago, exactly in the time when she was not yet tired, though she is tired at the very moment of the utterance. We presume that «am» relates the sentence with the time and the place of utterance, and from this we can infer that the sentence is true iff the speaker is tired at the moment of the utterance. And (13) may be true in this sense, though its tense morpheme still refers to the moment which is ten minutes before the utterance, if the speaker foresaw that she would become tired in a few minutes and, having calculated roughly the time the production of the report would take, chose 81 present tense of the expression.
__a
11 According to another treatment it would become just false. But then the rule might be jg
corrected so as to recommend choose alternative denotation if the sentence turns out false, ^^ when it is supposed to be true.
(8 0
We may not read (13) in such circumstances as The speaker was tired at the moment which is 10 minutes before the moment of utterance; neither should we read it as The speaker is tired at the moment of utterance. These readings would rather correspond to the thought I will be tired in a few minutes, when denotation of few embraces 10 minutes after the thought. But the speakers communicative intentions may be (as it is in the case) contradicting: to say about herself that she was tired at certain moment before the utterance, and to point to herself as the speaker of the utterance. This is exactly the case in which conditional shifting rather than the strict one becomes applicable.
But sometimes there are no priorities of the sort just introduced to be assigned to the agent of the utterance; in this case it might be analyzed as introducing split parameters under the shifting reading, when neither world (or context) should be taken as containing or contained: the denotation assigned to an agent-variable in a relevant reading might be thus composed from metaphysically equal parts constituting strictly competing sets of parameters. Unlike conditional-shifting-case, where we can choose denotations for all other variables semantically dependent on the one of an agent from some world or another, split-parameter case seems to allow only strictly ambiguous denotations for these variables: such that needs both (if just two) worlds to contribute to them at any time of evaluation. Thus if time and location of 10 are assigned in the reading as time and location of an agent's self (or personality), this component of ca parameter, if split, will make corresponding times and locations in both worlds equally relevant in providing denotations in question, hence the latter not to be supplied by any relevant context alone.
Conclusion
If what was said is in general right then, I think, that indexicals demonstrate not a uniform kind of indirectness. At least cases when an utterance was made for the sake of or in accord with some idea of self-identification do not completely support the model of indirectness ^ with 'monstrous' operators changing corresponding references by shifting jg indexicals' contexts. There are some shifts of indexical meanings which do qj not look as of context shifting type. In particular if semantic theory should W tolerate communicative endeavors of ordinary people, however strange, it has to allow at least an agent of the context to be of the kind not assimilated
a
M
to any context parameter. Partial, conditional and split-parameters ways of
fl shifting seem to show that denotations of indexicals may be still
underdetermined by their characters and contexts.
References
Anand, 2006 - AnandP. De De Se. Ph.D. Dissertation. MIT, 2006. Anand, 2004 - Anand P., Nevins A. Shifty Operators in Changing Contexts // Proceedings of Semantics and Linguistic Theory (= SALT) XIV; K. Watanabe, R.B. Young (eds.) Ithaka, N.Y. : CLC Publications, 2004.
Dummett, 1993 - Dummett M.A.E. What is the Theory of Meaning? (II) // The Seas of Language. Oxford : Clarendon Press, 1993.
Higginbotham, 2003 - Higginbotham J. Remembering, imagining, and the First Person // Epistemology of Language; A. Barber (ed.). Oxford: Oxford University Press, 2003.
Kaplan, 1989 - Kaplan D. Demonstratives // Themes from Kaplan; J. Almog, J. Perry, H. Wettstein (eds.). Oxford University Press, 1989.
Montague, 1974 - Montague R. English as a Formal Language // Formal Philosophy. Selected Papers of Richard Montague ; R.H. Thomason (ed.). New Heaven, 1974.
Perry, 1977 - Perry J. (1977) Frege on Demonstratives // Demonstratives. P. Yourgrau (ed.). Oxford : Oxford University Press, 1990.
Schlenker, 2003 - Schlenker P. Plea for monsters // Linguistics and Philosophy. 2003. Vol. 26.
Schlenker, 2010 - Schlenker P. Indexicality and De Se Reports // von Heusinger, Maienborn and Portner (ed.) Handbook of Semantics. Mouton de Gruyter, 2010.
Sells, 1987 - Sells P. Aspects of Logophoricity // Linguistic Inquiry. 1987. № 18(3).
Wittgenstein, 1953 - Wittgenstein L. Philosophical Investigations. Oxford: Blackwell, 1953.
(8
0
a
a
s
a
c
IB
