SciELO - Scientific Electronic Library Online

 
vol.29 número2Nota introductoriaDefinite descriptions, misdescriptions and semantic content: different ways to solve a tricky puzzle índice de autoresíndice de materiabúsqueda de artículos
Home Pagelista alfabética de revistas  

Servicios Personalizados

Articulo

Indicadores

  • No hay articulos citadosCitado por SciELO

Links relacionados

  • En proceso de indezaciónCitado por Google
  • No hay articulos similaresSimilares en SciELO
  • En proceso de indezaciónSimilares en Google

Bookmark


Análisis filosófico

versión On-line ISSN 1851-9636

Anal. filos. vol.29 no.2 Ciudad Autónoma de Buenos Aires nov. 2009

 

SIMPOSIUM

First Symposium. Definite descriptions: The referential/attributive distinction
Introduction.
Referential descriptions: for and against*

Eleonora Orlando
Universidad de Buenos Aires – CONICET - GAF
e_orlando@fibertel.com.ar

Abstract

In this introduction I start by presenting and examining the main positions on the current debate concerning the semantic analysis of sentences containing definite descriptions. As is known, the debate in question has started off with Russell's proposal (Russell 1905), which has been initially criticized by both Strawson (1950) and Donnellan (1966). Nowadays, waters are divided on this issue: some philosophers, representing the so-called univocality approach (Kripke 1977, Neale 1990), defend Russell's original analysis, according to which all definite descriptions are quantificational expressions, whereas there are others who, following Strawson´s and Donnellan´s objections, consider that at least some descriptions are genuine singular terms and hence have referential meanings. Among the defendants of the last approach, known as the ambiguity theory, we can find Michael Devitt (1981, 2004, 2007a, 2007b). In the second part of the paper, I then survey Devitt's stance on some traditional arguments for the ambiguity theory –mainly, the arguments from misdescription, incompleteness, opacity, and anaphora–, and finally I consider the new ones that he has put forward in favor of the existence of referential descriptions: the argument from regularity, the one based on the comparison with demonstratives, and the argument based on weak rigidity.

KEY WORDS: Definite description; Quantificational or attributive vs. Referential meanings; The semantics-pragmatics divide.

Resumen

En esta introducción comienzo por presentar y analizar las distintas posiciones que caracterizan el debate actual en torno al análisis semántico de las oraciones que contienen descripciones definidas. Como es sabido, tal debate se origina a partir de la propuesta de Russell (1905), la cual fue inicialmente objetada por Strawson (1950) y por Donnellan (1966). En la actualidad las aguas están divididas: algunos filósofos, representantes del aquí llamado enfoque de la univocidad, defienden el análisis russelliano, según el cual todas las descripciones son susceptibles de ser analizadas como expresiones cuantificacionales (Kripke 1977, Neale 1990); otros, por el contrario, haciéndose eco de las objeciones anteriores, consideran que algunas descripciones son términos singulares genuinos y, en tanto tales, expresiones referenciales. Entre los defensores más conspicuos de este enfoque, conocido como teoría de la ambigüedad, se encuentra Michael Devitt (1981, 2004, 2007a, 2007b). En la segunda parte de esta introducción, resumo la posición de Devitt respecto de los argumentos tradicionales a favor del enfoque de la ambigüedad –fundamentalmente, los argumentos basados en las descripciones fallidas, las descripciones incompletas, los casos de opacidad y los de anáfora–, y luego presento y analizo los nuevos argumentos por él ofrecidos, a saber, el argumento basado en el uso regular, aquél basado en la comparación con los demostrativos y el que se vale de la noción de rigidez débil.

PALABRAS CLAVE: Descripción definida; Significado cuantificacional o atributivo vs. significado referencial; La distinción entre semántica y pragmática.

1. The semantic analysis of definite descriptions has been the focus of a current debate. On the one hand, there are those who think that any statement containing a definite description is to be analyzed along the lines of Russell's theory of descriptions, namely, as semantically equivalent to an existentially quantified statement including a uniqueness condition, expressing thereby a general proposition. The paradigmatic representatives of this stance, which may be referred to as the univocality approach, are Kripke (1977) and Neale (1990). On the other hand, other philosophers, following both Strawson's and Donnellan´s suggestions (1950 and 1966, respectively), think that any statement containing a definite description can be thought to express either a general proposition or a singular one, depending on whether the definite description in question is interpreted to have either a quantificational –in Donnellan's terms, attributive– meaning, or a referential meaning respectively. From this perspective, the referential-attributive distinction is thus the distinction between the two possible meanings that can be assigned to a description used in an utterance; in other words, any description-type is considered to be semantically ambiguous: it can be construed either referentially or attributively.1 It is mainly Devitt (1981, 2004) and Wettstein (1981, 1983) who defend the alternative perspective, known as the ambiguity theory.
Getting more specific, according to the univocality approach, a statement such as

(1) The last book written by Borges contains a poem about war

is to be analyzed in terms of

(2) There is at least and at most an object such that it is a book written by Borges after all others and contains a poem about war2

According to this, (1) expresses a general or object-independent proposition –a proposition partly constituted by the property of being a book written by Borges after all others, which in this world is uniquely instantiated by Los conjurados but in other, counterfactual worlds, is instantiated either by other works of fiction or by none. But, and this is the core idea, the particular book Los conjurados is not itself part of the proposition expressed by (1). The main question is thus the following one: how can this position account for the fact that, on most occasions, in uttering (1), one is expressing a thought about a particular object that one has in mind, namely, Los conjurados? In other words, the Russellian has to explain why Los conjurados itself is not part of what is expressed by means of (1) when it is uttered with that particular book in mind.3 This is the intuition that gives grounds to Donnellan's proposal of a semantic ambiguity: there are some uses of definite descriptions where they seem to work as devices for singular reference, in much the same way as names and demonstratives –namely, the paradigmatic instruments for singular reference– do. Likewise, they seem to serve to express singular or object-dependent propositions, namely, propositions that are partly constituted by objects, such as Los conjurados.4
At this point, the defendants of the univocality approach make use of the Gricean distinction between what is literally said and what is meant or, in other terms, the proposition semantically expressed and the proposition pragmatically conveyed (Neale 1990). According to Grice, in uttering a sentence, what is thereby communicated may differ from what is literally said by means of it –namely, the proposition that constitutes the semantic content of the uttered sentence. The main process that underwrites this possibility is the one of conversational implicature: as is known, it allows for the derivation of a pragmatic content on the basis of the grasp of the literal, semantic one, some background knowledge and the assumption that the so-called 'Cooperative Principle' and its maxims is being respected. Back to the example, in uttering (1), while literally expressing a general thought along the lines of the proposition expressed by (2), the speaker conversationally implicates –hence, manages to pragmatically convey– a singular one, concerning a specific, particular book, the same she would have literally expressed had she uttered (3) instead:

(3) Los conjurados contains a poem about war

More specifically, the interpreter's thoughts may be rationalized as follows:

(i) the speaker has literally expressed the general proposition <There is at least and at most an object such that it is a book written by Borges after all others and that object contains a poem about war>5
(ii) there are no reasons to suppose that the speaker is not respecting the Cooperative Principle
(iii) on the assumption that the speaker is observing the Maxim of Relation, she must be trying to communicate something beyond the above-mentioned general proposition; on the assumption that the speaker is observing the Maxim of Quality, she must have enough evidence for the claim that the last book written by Borges contains a poem about war; the speaker knows that Los conjurados is the last book written by Borges; therefore, the speaker must think that Los conjurados contains a poem about war
iv) the speaker knows that I know that Los conjurados is the last book written by Borges, that I know that she knows that Los conjurados is the last book written by Borges and that I realize that the supposition that she thinks that Los conjurados contains a poem about war is required
v) the speaker has not done anything to stop me thinking that Los conjurados contains a poem about war
vi) the speaker has conversationally implicated that Los conjurados contains a poem about war.

Moreover, it has been suggested that the conversational implicature at stake is a generalized one, namely, one that does not depend on the particular features of a certain conversational context but can be derived in different ones; in other words, what is thought to constitute a pragmatic regularity.6 This does not imply that it is not cancelable, though: a context in which it cannot be inferred is perfectly imaginable, namely, a possible world in which the last book written by Borges is different from Los conjurados. The upshot is that the semantic content of (1) is exhausted by the general proposition –the pragmatically conveyed singular one is by no means part of it.
The contention is that since it is both possible and viable to subscribe to a pragmatic explanation like the above summarized one, there is no point in positing a semantic ambiguity with its corresponding multiplication of entities, namely, the existence of a referential meaning along with the quantificational one: the pragmatic explanation is ontologically simpler than the other one. Moreover, it is argued that the phenomenon of referential use is not peculiar to definite descriptions: far from that, it involves all quantificational phrases, which nobody wants to regard as semantically ambiguous expressions.7 To vary an example introduced by Neale, someone may utter

(4) Certain politicians do not like the press to publish critical views on their political decisions

while having ex-president of Argentina, Néstor Kirchner, in mind, but this does not make Kirchner the semantic referent of 'certain politicians', which is not a referential expression at all. To summarize, from the Russellian + Gricean perspective, the referential/attributive distinction turns out to be not a distinction between two kinds of meanings but a distinction between two kinds of uses. Description-types are not (semantically) ambiguous; they have just one kind of conventional meaning, namely, the so-called 'quantificational' one. However, they can be used either attributively, namely, with the intention of communicating just their conventional meanings, or referentially, namely, with the further communicative intention of pragmatically conveying a singular thought by means of them. In other terms, their referential use is not a semantic phenomenon, but a purely pragmatic one.8

Now, four main arguments have been put forward in favor of the ambiguity theory, and hence, of the existence of, along with the quantificational one, a referential meaning for definite descriptions: (i) the argument from misdescription, (ii) the argument from incompleteness, (iii) the argument from opacity and (iv) the argument from anaphora (Neale 1990).
To start with the first one, the argument from misdescription is based on the alleged fact that certain uses of definite descriptions that fail to select a particular object give rise to nonetheless true statements. Typical in this respect are Donnellan´s examples:

(5) The man drinking a martini is a friend of mine

An utterance of (5), said of a man, visible for both speaker and hearer, who is drinking not a martini but just water in a martini glass, can be taken to be true of the man at stake –if he is in fact a friend of the speaker's.9
Moving onto the second argument, the argument from incompleteness, it purports to show that the existence of referential descriptions –descriptions with referential meanings– is based on the fact that statements containing an incomplete description cannot be taken to express a unique condition of application, as demanded by Russell's analysis. The typical example, taken from Strawson, is

(6) The table is covered with books

uttered in the presence of a table, visible for both speaker and hearer, that is in fact covered with books: in most contexts of utterance, someone uttering (6) can be taken to be saying not the obviously false general proposition that there is a unique object that is a table and is covered with books but a true one, which does not include a uniqueness condition. Now, three main kinds of responses have been put forward by the defendants of the univocality approach: the syntactic ellipsis approach –also known as the explicit approach (Neale 1990)–, the semantic explanation –also known as the implicit approach (Stanley and Szabó 2000)– and the pragmatic one (Bach 2004).
According to the first, any instance of (6) is a syntactic ellipsis for a more complex statement, which has some grammatical constituents that are not phonologically articulated –they are unarticulated constituents–, nonetheless present in the logical form of the sentence and recoverable from the context of utterance. Context is then supposed to provide the hearer with certain grammatical items that have been elided –as much as it is supposed to provide the hearer with that kind of items in usual cases of syntactic ellipsis such as

(7) Peter hopes Mary will win the first prize and I hope so too

thought to be both grammatically and semantically equivalent to

(8) Peter hopes Mary will win the first prize and I hope Mary will win the first prize too

The semantic explanation or implicit approach has it that the logical form of (6) has a variable that is contextually given as value not a grammatical item but a set, which provides a restriction for the domain of the quantifier. Accordingly, if (6) is uttered in the above-described context, the domain of the variables of the quantifier must be thought to be restricted to the set of tables that are in the room where both speaker and hearer are located. Finally, from a pragmatic point of view, the proposal is that sentences of the likes of (6) must be taken to be literally false, since they express a false proposition, but they can be taken to pragmatically convey a true one, derived from the proposition expressed by means of both a conversational implicature and a process of pragmatic enrichment. According to this, context is thought to provide the hearer with a conceptual completion of the corresponding incomplete description –in the above mentioned example in which (6) is uttered in a context where both speaker and hearer are sitting at a big round table, it can be taken to pragmatically implicate the true proposition that the table they are both sitting at is covered with books.10 11
Turning to the argument from opacity, it attempts to establish that referential meanings are needed in order to account for the de dicto-de re ambiguity of propositional attitude ascriptions. As is known, the ambiguity in question can be exemplified by a sentence, such as

(9) Tom thinks that the author of Los conjurados died in Buenos Aires

The claim is that the best explanation for the de re reading of (9) involves ascribing a referential meaning to the description at stake, namely, 'the author of Los conjurados'. To put it in more Quinean terms, it is precisely the referential or designative position of the description that allows for its exportation outside the scope of the attitude verb, as required by the de re reading. The de dicto reading can be in turn cashed out by ascribing a quantificational meaning to the description. Russellians, in contrast, intend to account for the ambiguity in question in terms of the different scopes that can be assigned to the attitude verb and the quantifier: the de re reading is determined when the quantifier has wider scope than the verb, whereas the de dicto one involves that the scope of the verb is wider than the one of the quantifier.
The argument from anaphora is based on the analysis of sentences containing anaphoric pronouns whose antecedents are definite descriptions. Let's consider, for instance, an utterance of

(10) By the end of the summer, he finished his first script. It gave rise to a magnificent movie, which was immediately filmed in Colombia.

where 'it' is an anaphoric pronoun for 'his first script', namely, it exemplifies the phenomenon of cross-sentential anaphora. The point is that 'it' cannot be interpreted as a variable bound by the quantifier –allegedly involved by the presence of the definite description– because quantifiers cannot bind variables across different sentences. As is known, a pronoun that is anaphoric on a quantifier can be considered to be a bound variable only if the quantifier c-commands it –where the concept of c-command is to be construed in the framework of Generative Grammar.12 Consequently, the argument runs, 'it' can only be interpreted as a referential pronoun that inherits its referent from its equally referential antecedent; thus, 'his first script' must have a referential meaning. As is known, there are three main alternatives to be considered, both of which are compatible with the endorsement of the univocality approach concerning definite descriptions. The first one has been defended by Lewis (1975) and Kripke (1977), according to whom the anaphoric pronoun is referential but this is not the case with its descriptive antecedent, which is thought to have just a quantificational meaning that brings, by means of the usual Gricean mechanism, a certain individual to saliency in the context of utterance –thereby providing a referent for the pronoun. Secondly, there is Evans'position (1977), for whom the anaphoric pronoun is an E-type expression, namely, an expression that is descriptive in as far as it gets its referent fixed by description but is, at the same time, rigid. Finally, according to Neale (1990), the anaphoric pronoun is a D-type expression, namely, an expression that goes proxy for the previously occurred definite description, comparable to a pronoun of laziness; (10) comes thus out as

(10*) By the end of the summer, he finished his first script. His first script gave rise to a magnificent movie, which was immediately filmed in Colombia.

From this perspective, cross-sentential anaphora does not provide us with a motivation to ascribe referential meanings to descriptions.

2. As is known, in different papers, Michael Devitt has set out to defend the thesis that allows for the existence, along with attributive descriptions, of semantically referential ones, to which he refers to as RD (1981, 2004, 2007a, 2007b). To this aim, he has made significant contributions to both the argument from incompleteness and the argument from opacity, but has not put much weight on the argument from misdescription. Besides, he has offered three new arguments in support of the existence of referential descriptions: the argument from regularity, the argument based on the comparison with demonstratives, and the argument based on weak rigidity. First of all, I will survey Devitt's stance on the more common arguments, mentioned in the previous section, to then consider the novel ones.
To start off with the argument from misdescription, it must be said that its force depends on whether one has the semantic intuition that sentences such as (5) above, representing a typical case of misdescription, are in fact true: while some people do have the corresponding intuitions, some others do not, and think that there is a tension involved in such cases –regarding an instance of (5), uttered in the above-mentioned circumstances, they would not say that it is true but that it lacks truthvalue or that nobody is drinking a martini. More importantly, there are some ambiguity theorists who think that those intuitions are not clear enough, so that they do not want the referential-attributive distinction, semantically understood, to depend on them. This is, for instance, the case with Devitt, who does not consider this argument to be one on which the case for referential descriptions should be taken to rest.
As for the incompleteness argument, Devitt has maintained that the syntactic, the semantic and the pragmatic approaches previously summarized have problems in accounting for the referential use of incomplete definite descriptions. More specifically, both the syntactic and the pragmatic approaches have to face the problem of lacking a principled basis for choosing the right completion, and a related problem of subsequent undesired ambiguity: if there are many different possible completions of any given incomplete description, each one will turn out to be many-way ambiguous –which does not seem to be intuitively the case. More importantly, he claims that all the views are open to an ignorance and error objection: since the speaker can be ignorant or mistaken about either the accurate description of the object referred to or the sets of things providing the proper restriction for the quantifier domain, neither the explicit nor the implicit approach seems to be offering an acceptable explanation of the successful referential use of descriptions. One of the advantages of RD is that incomplete referential descriptions can be handled without having to resort to any of the above views: referential success is explained in terms of the existence of a referential semantic mechanism underlying all the referential uses of descriptions –namely, a causal-perceptual mechanism.13 In my opinion, both the ignorance and error objection to the above-mentioned approaches and the proposal of an alternative, nondescriptive semantic mechanism for referential descriptions are significant contributions of Devitt´s position to the debate that have not been sufficiently appreciated.
As far as the argument from opacity is concerned, a Gricean-Russellian such as Neale has replied that, if the context is not an extensional one –as is the case with some contexts introduced by attitude verbs– the ambiguity theorist will have the same problem as the supporter of the univocality approach: neither of them can treat the position of the description as a characteristically referential or designative position, be it by means of identifying the description with a singular expression or by means of taking it as replaceable by a bound variable.14 But this does not prevent the Gricean-Russellian from applying her own quantificational strategy wherever the substitution of the description by a bound variable is possible –namely, whenever the occurrence of the variable is purely referential. Accordingly, from Neale's perspective, there does not seem to be any clear reason to deny that the de re reading can be cashed out in terms of the wider scope of the description relatively to the attitude verb. However, I think that Devitt has a point here, against Neale. The RD supporter has an explanation of why the description can be exported out of the attitude verb context: it is a referential expression, and as such it can be replaced either by a variable that can be bound by an external quantifier or by another singular term that is co-referential with it –as is known, those are the two Quinean criteria to tell apart a transparent context and a subsequent de re reading of the ascription sentence from an opaque context and a subsequent de dicto reading. But, insofar as the Gricean-Russellian does not take the description to be referential, she seems to have no grounds for justifying its exportation and consequent replacement by a bound variable.15 That seems to be for her an ultimate fact.
As far as the new arguments offered by Devitt are concerned, according to the first above-mentioned one, the argument from regularity, the univocality approach, for which we can rest content with referential uses without postulating referential meanings, does not take sufficiently into account the fact that referential uses of descriptions are not rare and occasional but regular. As pointed out by Neale, it is certainly the case that there are referential uses of other quantificational phrases, such as the ones exemplified by the above (4) and the following (11)

(11) I know of someone who would like to have some chocolate

uttered with a particular eleven-year-old boy in mind, Lucio, who loves chocolate. But nobody would dare to suggest that such phrases are thereby semantically ambiguous, so that 'someone' may on occasion refer to Lucio. However, as emphasized by Devitt, referential uses of such quantificational phrases are occasional, and require a lot of stage setting; in contrast, referential uses of definite descriptions are ubiquitous: definite descriptions are regularly used to refer to particular individuals, which gives grounds to the thesis that there is an underlying semantic convention.
Bach has answered that there are also pragmatic regularities, such as the above-mentioned generalized conversational implicatures –namely, the ones that, not depending for their derivation on the hearer's grasp of features that are peculiar to a certain context of utterance, can be derived across different contexts. Bach takes such inferential processes not to require full-blown consciousness, involving what he calls 'standardization'. From his standpoint, there are then certain regularities that should not be taken to be a sign of the existence of a semantic convention. Devitt, in turn, interestingly replies that Bach's answer can be taken to involve a fundamentalist Gricean maneuver: on pain of becoming a Gricean fundamentalist, there does not seem to be any reason to deny that the regular use of definite descriptions as referential devices has given rise to a semantic convention, as is the case with most of the originally merely pragmatic regularities in natural language use. In a nutshell, he claims it is not clear enough why standardization, in the case of referential uses of definite descriptions, should be considered something different from semantic convention. He also argues that even if it is possible to give a pragmatic explanation of referential uses of descriptions, it is not certain that the explanation in question is a good one –let alone the best one. From his perspective, the same point can be made with regard to dead metaphors: the fact that they can be accounted for in terms of a contextual derivation of a pragmatic content from a literal semantic content does not imply that this is the best account of their semantic functioning we can come up with.
As far the argument based on the comparison with complex demonstratives is concerned, Devitt claims that a definite description 'the F' and a complex demonstrative 'that F' are similar in two respects. First of all, their conventional meaning is very similar, to the point that they can be interchanged in almost any context without any communicative loss. Secondly, both descriptions and complex demonstratives partially depend, for their reference, on the causal-perceptual mechanism that determines the object that the speaker has in mind. As a consequence, the statement containing a referential description, as much as the statement containing a complex demonstrative, semantically expresses, at least in part, a singular thought. A singular thought is one that is grounded on an object by means of a causal-perceptual mechanism. That mechanism is central to explaining why the expression refers to a particular object –instead of any other. As I mentioned before, an advantage of Devitt's position is that he provides us with a clear explanation of the referential mechanism at stake.
It is important to notice that, according to Devitt, in both cases the predicate 'F' also contributes to the meaning of the complex expression –be it 'the F' or 'that F': that is why the resulting meaning is, in both cases, taken to be partly referential and partly descriptive.16 One of the reasons offered by him in support of the inclusion of the predicative material in the semantic content of the referential description is the contradictory character of certain statements, whose structure is 'The F is not F', such as the rather dramatic

(12) Your father is not your father

With regard to the argument based on weak rigidity, Devitt takes it to depend on the others. According to him, once the existence of referential descriptions has been established, it must be observed that, due to the presence of the descriptive component, they turn out to be weakly rigid expressions. More specifically, a weakly rigid expression is one that designates the same object in every possible world in which (i) that object exists and (ii) any descriptive element of the expression applies to that object. This makes for another aspect –their modal behaviour– in which they are similar to complex demonstratives.
It is worth noticing that the interesting notion of weak rigidity has enabled Devitt to rebut an argument by Salmon, which can be reconstructed in the following terms (Salmon 1982):

(i) If there were referential descriptions, they would have to be rigid.
(ii) Referential descriptions cannot be rigid.
(iii) There are no referential descriptions.

According to Devitt, Salmon is right in subscribing to (ii); his point is that this should not be taken to imply the thesis that referential descriptions are flexible, and hence too different from typical referential devices, such as names and pure indexicals: they can be just weakly rigid. It is (i) that is wrong.

3. The discussion that follows contains three short articles in which different positions concerning definite descriptions are defended, each one involving a point of divergence with respect to Devitt's hereby summarized stance on the matter. The first one, written by Justina Díaz Legaspe, presents a version of the ambiguity theory, different from Devitt's, in which the properties involved by the predicative material of definite descriptions are epistemically relativized. In the second paper, due to Laura Skerk, a straightforward Donnellian position is defended, according to which any statement containing a definite description in referential use is taken to express, not a mixed proposition as suggested by Devitt, but a singular one. The third paper, written by Ramiro Caso, proposes a different, non-Gricean-Russellian pragmatic account, based on Relevance Theory, on which there is no derivation of a singular proposition from a general one by means of the mechanism of conversational implicature, that is alleged to be beyond Devitt's criticisms of pragmatic accounts. Finally, Michael Devitt will further defend RD by responding to the objections involved in the afore-mentioned articles.

Notes

*I want to thank Michael Devitt, Ramiro Caso, Laura Skerk, Justina Díaz Legaspe and Juan Manuel Berrós for giving me the opportunity of analysing and discussing the present topic. Funding for the research done has been provided by both the PIP 204, subsidized by the CONICET -the Argentine National Council for Research in Science and Technology-, and the UBACyT F 123, subsidized by the University of Buenos Aires.

1 This seems to suggest that the ambiguity at stake is a lexical one, namely, similar to the ambiguity affecting lexical items, such as 'bank'.

2 This is a natural language version of the corresponding Russellian analysis given in the quantificational language of Principia Mathematica. As emphasized by Neale, the notation of Principia Mathematica is not essential: restricted quantifiers, more akin to natural languages, could be used instead.

3 This problem is of course worse in the case of statements containing either incomplete or mistaken descriptions, as will be clear later on.

4 I leave it as an open question what kind of object is at stake –as must be clear, a work of fiction, such as Los conjurados, cannot be identified with any of its concrete exemplars.

5 I use angle brackets as a convention to designate propositions.

6 It is interesting to notice that this kind of generalized conversational implicature arises out of the observance of the maxims –rather than arising from their violation, as is the case with the particularized ones. It may be worth exploring whether this is the case with all kinds of generalized conversational implicatures. 7 See the following fragment by Neale (1990, p. 88; the emphasis is mine): "Thus the Gricean-Russellian views the referential use of definite descriptions as an instance of a more general phenomenon associated with the use of quantified noun phrases. Of course, definite descriptions are particularly susceptible to referential usage because of their own particular semantics. As Klein (1980) points out, if S is observing the Maxim of Quality, S will typically believe that one and only one objects satisfies the description used. [...] And quite often S will believe this because S knows of some particular object b that b is uniquely F. The beginnings of an explanation of the quite general phenomenon of communicating object-dependent propositions using quantified sentences surely liein the fact that the grounds for a quantificational assertion are very often objectdependent beliefs of one form or another. [...]".

8 Afurther point made by Kripke in support of the Russellian account is the following one: speakers of a purely Russellian language –by hypothesis- may end up using definite descriptions referentially. I agree with Devitt (2004) that this is not very persuasive: it seems to be begging the question against the referentialist.

9 As is also known, Donnellan himself is not completely clear on this point, since he is not clearly committed to descriptions actually having referential meanings.

10 Notice that the proposition conversationally implicated, though general, includes an indexical element, namely, the meaning of the indexical 'they'.

11 Summarizing the last paragraphs, on each approach, context is taken to play a different kind of role: a grammatical, a semantical and a pragmatic one. For more on this, see Stanley and Szabó (2000).

12 A phrase á c-commands a phrase â if and only if the first branching node dominating á also dominates â and neither á nor â dominates the other.

13 As must be clear, this does not preclude incomplete descriptions from being used attributively –there is no argument from incompleteness to referentiality. As far as attributive uses are concerned, Devitt thinks that we can settle for the kind of completion recommended by the implicit approach –where a semantic value, constituted by a set restricting the quantifier domain, is given by context to a domain variable that is present in the logical form of the utterance. See, for instance, Stanley and Szabó (2000). In such cases, the ignorance and error objection does not hold.

14 See Neale (1990, chapter 4).

15 Notice that she [the Gricean-Russellian] cannot even consider applying the second Quinean criterion: since no description is a referential expression, none of them can have a co-referential term.

16 This aspect of Devitt's position will be criticized in two of the essays below.

References

1. Bach, K. (2001), "You Don't Say", Synthese, 128, pp. 15-44.        [ Links ]

2. Bach, K. (2004), "Descriptions: Points of Reference", in Reimer, M. And Bezuindenhout, A. (eds.), Descriptions and Beyond, Oxford, Clarendon Press, pp. 189-229.        [ Links ]

3. Bach, K. (2007), "Referentially Used Descriptions: a Reply to Devitt", European Journal of Analytic Philosophy, 3, pp. 33-48.        [ Links ]

4. Bezuindenhout, A. (1997), "Pragmatics, Semantic Underdetermination and the Referential/Attributive Distinction", Mind, 106 (423), pp. 375-409.        [ Links ]

5. Devitt, M. (1981), "Donnellan's Distinction", in French, P. A., Uehling, T. E. and Wettstein, H. K. (eds.), Midwest Studies in Philosophy VI, Minneapolis, University of Minnesota Press, pp. 511-524.        [ Links ]

6. Devitt, M. (2004), "The Case for Referential Descriptions", in Reimer, M. And Bezuindenhout, A. (eds.), op. cit., pp. 280-305.        [ Links ]

7. Devitt, M. (2007a), "Referential Descriptions and Conversational Implicatures", European Journal of Analytic Philosophy, 3, pp. 7-32.        [ Links ]

8. Devitt, M. (2007b), "Referential Descriptions: a Note on Bach", European Journal of Analytic Philosophy, 3, pp. 49-53.        [ Links ]

9. Donnellan, K. (1978), "Speaker Reference, Descriptions and Anaphora", in Cole, P. (ed.), Syntax and Semantics, vol. 9, Pragmatics, New York, Academic Press, pp. 129-142.        [ Links ]

10. Donnellan, K. (1966), "Reference and Definite Descriptions", Philosophical  Review, 77, pp. 281-304.        [ Links ]

11. Evans, G. (1977a), "Pronouns, Quantifiers, and Relative Clauses (I), Canadian Journal of Philosophy, 7, pp. 187-208.        [ Links ]

12. Evans, G. (1977b), "Pronouns, Quantifiers and Relative Clauses (II), Canadian Journal of Philosophy, 7, pp. 777-797.        [ Links ]

13. Grice, H. P. (1957), "Meaning", Philosophical Review, 66, pp. 377-388.        [ Links ]

14. Grice, H. P. (1967), "Logic and Conversation", William James Lectures, in Studies in the Way of Words, Cambridge, Mass., Harvard University Press, (1989), pp. 22-40.        [ Links ]

15. Grice, H. P. (1969), "Utterer's Meaning and Intentions", Philosophical Review, 78, pp. 147-177.        [ Links ]

16. Kripke, S. (1977), "Speaker Reference and Semantic Reference", in French, P. A., Uehling, T. E. and Wettstein, H. K. (eds.), Contemporary Perspectives in the Philosophy of Language, Minneapolis, University of Minnesota Press, pp. 6-27.        [ Links ]

17. Lepore, E. (2004), "An Abuse of Context in Semantics: the Case of Incomplete Definite Descriptions", in Reimer, M. and Bezuindenhout, A. (eds.), op. cit., pp. 41-67.        [ Links ]

18. Neale, S. (1990), Descriptions, Cambridge, Mass., The MIT Press.        [ Links ]

19. Neale, S. (2004), "This, That and the Other", in Reimer, M. and Bezuindenhout, A. (eds.), op. cit., pp. 68-182.        [ Links ]

20. Russell, B. (1905), "On Denoting", Mind, 14, pp. 479-493.        [ Links ]

21. Salmon, N. (1982), "Assertion and Incomplete Definite Descriptions", Philosophical Studies, 42, pp. 37-45.        [ Links ]

22. Stanley, J. and Szabó, Z. (2000), "On Quantifier Domain Restriction", Mind and Language, 15 (2-3), pp. 219-261.        [ Links ]

23. Strawson, P. (1950), "On Referring", Mind, 59, pp. 320-344.        [ Links ]

24.Strawson, P. (1971), Logico-Linguistic Papers, London, Methuen.        [ Links ]

25. Wettstein, H. (1981), "Demonstrative Reference and Definite Descriptions", Philosophical Studies, 40, pp. 241-257.        [ Links ]

26. Wettstein, H. (1983), "The Semantic Significance of the Referential-Attributive Distinction", Philosophical Studies, 44, pp. 187-194.         [ Links ]