SORITES, ISSN 1135-1349

Issue #02. Juy 1995. Pp. 46-56.

Indexicals and Descriptions

Copyright © by SORITES and Fernando Garcia-Murga

Indexicals and Descriptions

Fernando Garcia-Murga


Reference is a highly intricate question at the very core of philosophy of language, linguistics, and cognition. In this paper, we maintain that reference is a common feature to indexicals, definite descriptions and, at least to some uses of indefinite descriptions.

The main aspect of a referential expression, from the addressee's point of view, is that it triggers search for a referent, the search ranging over the linguistic context, physical environment or encyclopedic knowledge.

As regards the antecedent's location, traditionally only the last two ranges gave rise to existential presuppositions. That's why the referent's source had great theoretical importance and, since no referential expression automatically triggered presuppositions, presuppositional theories had to be complemented with projection theories for presuppositions.

Following research on the relation between language and other cognitive abilities, we maintain that it is salience, and not the referent's source, that conforms speaker's choice of referential expressions. Accordingly, indexicals and definite descriptions indicate that the speaker acts as if the object he intends to refer to were salient, whereas indefinite descriptions guide the search for a non salient object. Therefore, salience and presupposition divide reference expressions equally.

Moreover, we find a pervasive referential/attributive distinction in indexicals and descriptions. Our claim is that the dubbed attributive reading is a consequence of a search failure that provides a «weak understanding» by «accommodating» referential expressions to their utterance context. This paper is an initial step toward a unified and abstract theory of referential expressions.

1. Definite descriptions.

In this section, we are concerned with the analysis of expressions headed by a definite article.Foot note 72 Obviously, such expressions appear in sentences playing different grammatical roles: subject, direct object, embedded under different clauses, and, according to the «pragmatically defined» topic/comment structure, definite descriptions also occur in both sentential divisions.

Traditionally, a main linguistic feature of definite descriptions has been their role triggers for existential presuppositions.Foot note 73 The different roles definite descriptions play have given rise to puzzling questions such as the projection problem for presuppositions and topic assignment. We hope our proposal will help to dissolve such intricate questions. As a methodological claim, we will analyze the definite article and the descriptive content of the description independently.

The role of definite articles, stated in procedural terms,Foot note 74 lies in the search for an antecedent. Definite articles give the addressee the search order automatically. The search range includes linguistic, encyclopedic, and the «physical» environments.Foot note 75 So, assuming that «the President» is not present in situation of utterance of (2) and «a man playing chess» is present in (3), we have the following respective examples:

(1) If there is a car in Hyde Park, the car runs slowly.Foot note 76

(2) The President has resigned.

(3) The man playing chess is Karpov.

The speaker, by using the definite article, acts as if the antecedent were a salient objectFoot note 77 for the addressee. We maintain that while salience implies familiarity, the converse does not hold.Foot note 78 Therefore, familiarity is a necessary but not a sufficient condition for the speaker to refer to an object by using a definite description.

Now, let us assume that definite articles provide a semantic marker to semantic representation, whatever type of semantic representation we are disposed to assume, and let us use «R» as such a semantic marker.Foot note 79 We assume that the semantic representation of expressions of the form «the X» looks like this: [X]R. On the basis of such a semantic representation, we adopt the following inference pattern:

[X]R -> EXIST([X])

Before discussing the rationale for the inferential rule just established, we must face the referential/attributive distinctionFoot note 80 in the light of our frame.

Suppose the following context: Jones is on trial charged with Smith's murder. Jones begins to cry, and the speaker utters:

(4) The murderer of Smith is a coward.

It seems obvious to us that, in this situation, the speaker can act as if Jones were a salient object. Automatically, the addressee looks for an object that the speaker takes to be salient and that satisfiesFoot note 81 the descriptive content. The addressee easily finds Jones and interprets the speaker's utterance as referring to Jones. In other words, the utterance receives referential interpretation.

Suppose, by contrast, that a person finds Smith's body with a knife hanging on his shoulder. Nobody knows who the murderer is, but the speaker utters (4). In this case, we want to maintain, nothing alters the article's linguistic behaviour. So, as usual, the addressee searches for an object, but now he cannot find any salient object that satisfies the descriptive content. Nevertheless the crucial point here is that the inference pattern applies because the definite article is present in semantic representation. Therefore, although in this context there is no other reference specification but linguistic description, the addressee takes it that there is an object that satisfies the descriptive content. This is the so-called attributive reading.

The specificity of this reading lies in the fact that the addressee creates the reference using the available linguistic description. In a sense, the attributive reading is the result of an accommodation process.Foot note 82 Nothing prevents us from saying that the attributive reading triggers existential presupposition.Foot note 83

Moreover, the referential/attributive distinction is parallel to the distinction between a strong and a weak understanding.Foot note 84 In our view, the attributive reading arises from a failed search; the interpretation being, therefore, the pure procedure, the concept, the expression's character.Foot note 85

If the general frame we are developing is close to the mark, we can state a precondition for attributive readings: the addressee is not able to find the salient object signalled by the descriptive content. So, in some sense, we can say that all definite descriptions -- even under attributive reading -- are referential.

We must discuss, at least briefly, the descriptive content we have represented semantically as X. We want to maintain that X is a representation of an object-as-perceived.Foot note 86 As such, we think X represents a concept.Foot note 87

Now, clearly the existence stated by the inference rule we have introduced represents an existence in a world-as-perceived. Therefore, no ontological claims follow from such an existential rule.

Up to now, we have argued that definite descriptions are referential expressions that refer to salient objects-as-perceived. As salient, the object referred to by definite descriptions are familiar to the speaker and, therefore, he presupposes the existence of the object referred to. Our next task lies in the extension of these assumptions to indexicals.

2. Indexicals.

We will characterise indexicality on two features. On the one hand, indexicals are highly context-dependent expressions, and, on the other, they have low descriptive content.Foot note 88 We assume that both pronouns and demonstratives are indexical expressions:

(5) There is a man at the door. He is crying.

(6) He is Karpov.

The thesis we want to maintain can be stated as follows: the role of indexicals is the very same we established for definite descriptions, that is, indexicals indicate to the addressee that he must start a search for a salient antecedent, be it linguistic or physical.Foot note 89 So, we will use the same semantic marker for the semantic representation of indexicals, and, of course, we will maintain the inferential pattern adopted for such marker.

As we have just mentioned, the alleged antecedent must be a salient object, as in the case of definite descriptions, but now the addressee has less descriptive content available for the search than in the case of definite descriptions.Foot note 90 This lack may be surmounted thanks to a very prominent salience accompanied, in some cases, by a demonstration.

Let us analyze, as an example, the first person personal pronoun «I». We assume that «I» indicates the addressee is looking for a singular person who could be responsible for the assertion:

(7) I am hungry.

We can find two possible readings again. On the one hand, if the addressee finds a salient object that satisfies the descriptive content, the addressee understands that object as the one the speaker is trying to refer to. This reading corresponds to the known referential reading or strong understanding.

On the other hand, the addressee will not always be able to find the proper salient object. In that case, as we have already seen, the addressee has to create an object. In other words, indexicals have attributive readings. We maintained in the previous section that a precondition for an attributive reading is that the addressee not be able to find the salient object. Obviously, this means that we should think of a quite strange context for (7) to read it attributively.Foot note 91 However, there are other examples where such a reading is mandatory:Foot note 92

(8) Condemned prisoner: I am traditionally allowed to order whatever I like for my last meal.

An obvious problem arises when we ask why the speaker of (8) is not the salient object the addressee is looking for. Intuitively, the salient object (8) refers to is not the speaker as a particular person, but the role the speaker is playing. In fact, were the speaker the object referred to, a contradiction arises between a last dinner as a necessary unique event and the multiplicity introduced by the expression «traditionally».

We assume that in a first attempt at interpretation, the addressee takes the speaker as the reference of the personal pronoun «I» until the presence of the expression «last dinner» forces a reinterpretation. This is not, to our mind, a strange or an ad hoc resource. On the contrary it is a widely extended phenomenon (think of «garden-path sentences»). Anyway, the salient object referred to is the speaker not as a particular person but as a player of a role.

If the picture sketched on indexicals is basically correct, then we maintain that indexicals provide the semantic marker «A» to the semantic representation of such expressions and the inferential rule we defined applies: [X]R -> EXIST([X]). Nothing prevents us from saying that indexicals are -- even under attributive readings -- referential terms.

3. Indefinite descriptions.

Indefinite descriptions have the form «a(n) X». Our thesis tries to reflect a close similarity between indefinite descriptions and the referential expressions we have already analyzed. So, we maintain that indefinite descriptions are also semantically represented by the semantic marker «R» that triggers the inferential rule we introduced in section 1. Following the method we used in the first section, we analyze the indefinite article separately from the descriptive content.

We keep the procedurality we are assuming throughout this work by maintaining that indefinite articles indicate to the addressee to look for an antecedent:

(9) There is a woman in the bank.

It is crucial to note that the antecedent the addressee is looking for is not a salient one -- or, more precisely, the speaker acts as if the antecedent were not a salient object. So, the object lacks salience in the context up to the moment of its utterance. In fact, utterance of an expression is a usual mode of providing salience to the object referred to by an expression.Foot note 93

Accordingly, since a linguistic utterance makes the uttered elements salient, it is not possible to take linguistic antecedents for indefinite descriptions:

(10) John has a cat and feeds a cat.

The cat John has and the cat John feeds is not the same cat -- at least under normal intonation.

Now, the two readings -- referential and attributive -- we have encountered for definite descriptions and indexicals apply also for indefinite descriptions. Following a similar example by Wilson (1978), suppose Mary is trying to seduce Peter and since the speaker knows Mary is a chess player, utters the following sentence:

(11) A chess player is trying to seduce Peter.

According to our assumptions, the addressee begins to look for a non salient object that satisfies the description «be a chess player». If the addressee finds such an objectFoot note 94 he will construct what is traditionally known as a referential reading for (11). In sum, in this situation, the addressee understands the indefinite description as referring to Mary.

Alternatively, if the addressee is not able to find an appropriate object, since the semantic marker has triggered the inferential rule, the addressee creates a mental representation for a new object on the basis of the descriptive content available to him. In other words, the addressee ends with an attributive reading for (10).

Since the referential/attributive distinction depends exclusively on the success of the search the addressee performs on mandatory grounds, we can maintain that indefinite descriptions are referential expressions.

4. Conclusions.

We have argued for a unified theory of reference for definite descriptions, indexicals and indefinite descriptions based on a common procedural task these expressions share. However, indexicals and definite descriptions refer to salient objects while indefinite descriptions refer to non salient objects. The descriptive content attached to each expression (varying from the low content of pronouns to the higher content of descriptions) provides information that makes it possible for the addressee to find an object the speaker has referred to.Foot note 95 Ostension and other non linguistic knowledge helps the addressee's search.

Now, the traditional referential/attributive distinction relies on the success of the search process common to indexicals and descriptions. This means that the ambiguity hypothesisFoot note 96 should be ruled out. The referential/attributive distinction is a kind of strong/weak understanding distinction and we find inferential process as its theoretical ground.

From the frame we have presented, we conclude that salience, and not mutual knowledge or givenness, is the crucial aspect the speaker considers when he performs a referential act.

We think it is not fortuitous that the difference in salience between definite descriptions and indexicals on the one hand, and indefinite descriptions on the other coincides with the difference presupposition imposes on such expressions, since while indexicals and definite descriptions presuppose the reference existence, indefinite descriptions do not presuppose such existence.

From our thesis on referentiality, and, since we claimed at the beginning that our analysis would not differentiate the roles these expressions play in sentences, it follows that all indexical expressions and all definite descriptions trigger existential presuppositions. However, all theories on presupposition projectionFoot note 97 maintain inheritance mechanisms for presuppositions such that, for instance, it is claimed that (the speaker of) our example (1) does not presuppose the existence of the car. So, non cumulative strategies seem to be inconsistent with our theory. However, since all presuppositional theories I am aware of require projection mechanisms, we have to define a new presuppositional theory. But these questions require further investigation.Foot note 98


Ariel, M. (1988), «Referring and accessibility», in Journal of Linguistics 24.

Ariel, M. (1990), Accessing noun-phrase antecedents. Routledge.

Bar-Hillel, Y. (1954), «Indexical expressions», in Mind 63.

Chomsky, N. and Lasnik, H. (1991), Principles and Parameters Theory. Ms. MIT.

Clark, H.H. and Marshall, C.R. (1981), «Definite Reference and Mutual Knowledge.» In A. Joshi, B. Webber and I. Sag, Elements of Discourse Understanding. Cambridge University Press.

Donnellan, K. (1966), «Reference and definite descriptions», in Philosophical Review, 65. Reprinted in Steinberg, D. & Jakobovits, L. (1971)

Frege, G. (1892), «Über Sinn und Bedeutung», in Zeitschrift fuer Philosophie und philosophische Kritik.

Garcia-Murga, F. (in prep.), Los fundamentos conceptuales del fenomeno presuposicional. Diss. University of the Basque Country.

Gazdar, G. (1979), Pragmatics: Implicature, Presupposition, and Logical Form. Academic Press.

Grice, H.P. (1969), «Vacuous names.» In Davidson, D. and Hintikka, J. (eds.), Words and objections. Reidel.

Heim, I. (1982), The Semantics of Definite and Indefinite Noun Phrases. Ph.D. thesis. University of Massachusetts.

Heim, I. (1983), «On the projection problem for presuppositions», WCCFL 2. MIT.

Jackendoff, R. (1983), Semantics and cognition. MIT Press.

Jackendoff, R. (1987), Consciousness and the Computational Mind. MIT Press.

Kaplan, D. (1989), «Demonstratives. An essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals.» In Almong, J., Wettstein, H. and Perry, J. (eds.), Themes from Kaplan. Oxford University Press.

Karttunen, L. (1973), «Presuppositions of compound sentences», in Linguistic Inquiry, IV, 2.

Karttunen, L. (1974), «Presupposition and linguistic context», in A. Rogers, B. Wall and J. Murphy (eds.), Proceedings of the Texas Conference on Performatives, Presuppositions and Implicatures. Center for Applied Linguistics.

Kempson, R. (1986), «Definite NPs and Context-Dependence: A Unified Theory of Anaphora.» In Myers, Brown and McGonigle (eds.), Reasoning and Discourse Processes. Academic Press.

Kempson, R. (1988), «Grammar and conversational principles.» In F. Newmeyer (ed.), Linguistics, The Cambridge Survey. vol. II, Linguistic Theory Extensions and Implications. Cambridge University Press.

Langendoen, T. and Savin, H.B. (1971), «The projection problem for presuppositions.» In Ch. Fillmore and Langendoen, Studies in linguistic semantics. Holt.

Lewis, D. (1979), «Score-keeping in a language game.» In R. Bauerle, U. Egli and A. von Stechow (eds.), Semantics from different points of view. Springer-Verlag.

Miller, G.A. and Johnson-Laird, P.N. (1976), Language and perception. Cambridge University Press.

Nunberg, G. (1978), The pragmatics of reference. Indiana University Linguistics Club.

Nunberg, G. (1993), «Indexicality and deixis», in Linguistics and Philosophy 16.

Recanati, F. (1989), «Referential/Attributive: a Contextualist Proposal», in Philosophical Studies, 56.

Recanati, F. (1993), Direct reference: from language to thought. Blackwell. Oxford.

Russell, B. (1905), «On denoting», in Mind 14.

Soames, S. (1982), «How presuppositions are inherited: a solution to the projection problem», in Linguistic Inquiry 13.

Sperber, D. and Wilson, D. (1986), Relevance: communication and cognition. Blackwell.

Sperber, D. and Wilson, D. (1987), «Precis of relevance: communication and cognition», in Behavioral and Brain Sciences 10.

Steinberg, D. & Jakobovits, L. (1971), Semantics. An interdisciplinary reader in philosophy, linguistics and psychology. Cambridge University Press.

Strawson, P. (1950), «On referring», in Mind 59.

Strawson, P. (1964), «Identifying reference and truth-values», in Theoria 30. Reprinted in D. Steinberg and L. Jakobovits (1971).

van der Sandt, R. (1992), «Presupposition projection as anaphora resolution», in Journal of Semantics 9.

Wilson, G. (1978), «On definite and indefinite descriptions», in Philosophical Review, 87.

Fernando Garcia Murga

Institute for Logic, Cognition, Language and Information

University of the Basque Country

Apdo. 1249. 20080 San Sebastian. Spain