The Semantic Challenge to Russell's Principle 1

Mark Textor

1. Introduction

[1] Russell's paper "Knowledge by Acquaintance and Knowledge by Description" (KAKD) stands at the beginning of a philosophical tradition that sees epistemology and the theory of reference as closely connected. In KAKD Russell states his Principle of Acquaintance:

(PA) Every proposition which we can understand must be composed wholly of constituents with which we are acquainted. (Russell 1910-11: 209)
Russell holds that only if I know an object m, or, as he says, I am acquainted with m, this object can figure in the content of my propositional attitudes. PA has also consequences for the philosophy of language. Russell identifies the meaning of a real or logically proper name with the object named. The proposition expressed by an utterance containing a logically proper name can according to PA only be grasped by someone who is acquainted with the object named. This epistemic feature distinguishes genuine naming from other forms of referring to things. Thus the epistemological notion of acquaintance is of central importance to the philosophy of mind and language. Russell's followers, most prominently Gareth Evans, have not accepted every element of the Russellian approach, but they have tried to preserve the basic thought embodied in PA while, among other things, developing a more plausible version of acquaintance.

[2] Now the Principle of Acquaintance has been challenged, and some philosophers would even say, refuted, by the so-called "New Theories of Reference". These theories are primarily semantic proposals for indexicals, proper names and natural kind terms. Their proponents are sometimes labelled "Neo-Russellians" because, like Russell, they hold that utterances containing proper names and indexicals express singular propositions. But "Neo-Russellian" is misleading: some so-called "Neo-Russellians" have drawn from their semantic theories conclusions for the philosophy of mind which seem to refute Russell's Principle of Acquaintance. In the following I will use "Neo-Russellian" therefore only for Philosophers who accept PA. Scott Soames provides an example of the misleading character of the label "Neo-Russellian" when he lists among the main theses of the "contemporary Russellians" the thesis that:

to believe a proposition, it is sufficient for one to understand and accept any sentence or representation that expresses that proposition. (Soames 1995: 518)
"Understanding" means here "knowledge of linguistic meaning of the words and of their mode of combination", "proposition" means "singular proposition".2 According to this Anti-Russellian Thesis one need not stand in the relation of acquaintance to an object to entertain a singular proposition containing it or to name it. All that is needed to entertain such a proposition is knowledge of the linguistic meaning of a sentence that expresses a singular proposition!

[3] In this paper I will discuss whether arguments based on the New Theories of Reference really refute the basic thought of the Russellian view. Although the issues discussed here bear immediately on issues in the philosophy of language, for instance, the question whether definite descriptions and proper names form a "natural semantic kind" (G. Evans), I will mainly discuss the more fundamental question whether we can grasp singular propositions without possessing special knowledge about their constituents. This question is one of the central questions of Gareth Evans's book Varieties of Reference and I will develop some of Gareth Evans's arguments in my critical discussion. But I will apply these arguments against a critique of PA not discussed by Evans himself.

[4] In Varieties of Reference Evans defends a version of PA against the so-called "Photograph Model" of singular reference and thought. Many New Theorists of Reference accept some version of the Causal Theory of Reference for proper names. This theory says roughly that the object a use of a proper name refers to is determined by a causal chain. Some philosophers have integrated the basic thought of this theory of reference into an account of singular thought.3 Evans calls the resulting theory the "Photograph Model". "The name 'Photograph Model'", Evans says, "is apt because we do speak of a photograph's being a photograph of one object rather than another solely on the basis of which object was related in the appropriate way to its production" (Evans 1982: 78). The proponent of the Photograph Model models the object-directedness of a singular propositional attitude on the of-ness of a photograph: which object a singular propositional attitude is about is determined solely by the causal history of the attitude. The Photograph Model implies that a thinker can have a singular propositional attitude towards m, even if he does not know which thing m is. Evans criticizes the Photograph Model because it conflicts thus with the basic tenet of his Russellianism.

[5] While the Photograph Model is encouraged by a proposal for the semantics of proper names, another model for singular thought that has a similar Anti-Russellian implication builds on recent proposals for the semantics of indexicals and demonstratives. This, as one might say, "Function Model" uses the notion of a semantical rule. The semantical rule is pictured as a function which returns for contexts of use objects. According to the Function Model my thought concerns an object m, because I "entertain" a function whose value is m. The Function Model is most forcefully put forth by David Kaplan. According to him my knowledge of the semantical rules connected with indexicals and demonstratives (a form of descriptive knowledge) is sufficient to grasp singular propositions. Thus linguistic knowledge replaces the more substantial knowledge Neo-Russellians require for this feat. I will discuss the Function Model and not the Photograph Model, since it is now widely accepted, yet seldom scrutinized in detail. Therefore, indexicals and demonstratives will be at the center of the discussion in this paper, and I will devote special attention to the arguments sketched in Evans's seminal paper "Understanding Demonstratives".

[6] The plan of this paper is as follows: In section 2 I will prepare the ground for the further discussion by clarifying notions crucial for understanding PA. In section 3 and 4 I will present and criticize a counter-example against PA given by Kaplan which prima facie rests on a thesis about the semantics of demonstratives. The example has puzzled many philosophers, in section 4 I will try to explain why. Sections 5, 6 and 7 criticize Kaplan's argument against PA which is based on his theory of indexicals. This critique will use notions from the theory of thought, especially Evans's so-called "Generality Constraint".

2. Russell's Principle

[7] Let us take a closer look at the Principle of Acquaintance. Crucial are the notions of proposition and acquaintance. I will start with Russellian propositions. Sometimes philosophers use "proposition" to refer to a complete sentence. This is not the use Russell makes of "proposition" in the formulation of PA. For Russell propositions are not linguistic items at all. Russellian propositions are objects of thought. Now as Prior remarked the phrase "object of thought" may be used in two different ways (Prior 1971: 1). An object of thought may be what we think or what we think about. If I think that Mont Blanc is a high mountain, what I think is the content of my thinking, what I think about is Mont Blanc. Russellian propositions are objects of thought in the first sense.

[8] An object can be only what we think if it can play a part in the explanations of a thinker's doings. We explain that S does x by saying that he believes that p and wishes that q. If propositions shall be the accusatives of believing and wishing they minimally must fulfill the condition that a rational subject cannot have conflicting attitudes towards them.

[9] According to PA only objects we are acquainted with can be constituents of the Russellian propositions we grasp. Acquaintance is a "direct cognitive relation" (ibid.: 200), the relation which constitutes "presentation" and not judgement (ibid.: 201). Russell holds that we are acquainted with universals and some particulars. In KAKD he lists as the only particulars to which we can stand in the relation of acquaintance sense-data, memories and ourselves. In his later writings Russell will exclude ourselves, whatever we are, from his list. Together with PA this entails that only the objects listed can be constituents of propositions we grasp.

[10] Russell's restriction that only objects of acquaintance can be constituents of propositions we grasp can be seen as an attempt to ensure that his propositions can be what we think. If the ordinary particulars named by "Hesperus" and "Phosphorus" could be constituents of the propositions expressed by "Hesperus shines brightly" and "Phosphorus shines brightly", these sentences would express the same proposition. Yet a rational person may honestly and with comprehension affirm that Hesperus shines brightly, but that Phosphorus does not. If Russell placed no restriction on propositional-constituency, his theory of propositions would lead to the result that such a person believed at the same time that p and that it is not the case that p. Russell concludes that the propositions we grasp cannot contain the referents of ordinary proper names. If fleeting sense-data are the only particulars that can enter propositions, the danger that the same singular proposition is the content of conflicting attitudes is minimized. Yet this maneuver also minimizes the usefulness of Russell's theory for the theory of reference. In the end only objects we usually never refer to, sense-data, are for Russell the only possible objects of genuine reference.

[11] Due to this problem Neo-Russellians try to save the intuition behind Russell's idea that a proposition contains an object, without making essential use of Russell's containment-metaphor. The metaphor is cashed out by appealing to reports of utterances. Here is a straightforward example. I make a demonstrative utterance with the words "That tree is blossoming" and I manage to refer demonstratively to the tree before me. In a correct report in indirect speech the reporter re-expresses what the utterer said and thereby entertains the proposition expressed. If my use of the demonstrative pronoun does not latch on to an object, the informed reporter cannot give a report of my utterance in indirect speech which captures the proposition I intended to express. The informed reporter cannot say, for instance, "M. T. said that that tree is blossoming". In this situation the informed reporter could report in other ways, but these reports cannot be faithful to the proposition the speaker intended to express. If these considerations are convincing, they show that my utterance cannot be adequately reported if I use an empty referring expression. This does establish that the expressed proposition is object-dependent if we add the assumption that an utterance that cannot be adequately reported in indirect speech expresses no proposition at all.

[12] John McDowell exploits this object-dependency intuition to clarify the talk of a proposition that contains an object (such propositions are usually called "singular propositions"):

The intuition recognizes a relation between objects and thoughts so intimate that it is natural to say that the objects figure in the thoughts.... (McDowell 1982: 308)
Now your utterance of "He is a fool" (said about me) and my utterance of "I am a fool" each express a proposition that depends for its existence on the same object, yet they express different (object-dependent) propositions. If we characterize singular propositions as object-dependent propositions, singular propositions can be what we think and not only what we think about: the object-dependency conception allows us to individuate propositions fine enough for them to fulfill the role of propositional attitude contents. Thus the interesting relation between the singular proposition I express with "I am a fool" and me is not that of containment, but that of existential dependence. This move makes room for singular propositions which are individuated by Fregean criteria of thought difference. The idea of singular Fregean thought is not incoherent.

[13] We can say that a singular proposition is what someone thinks iff the content of his thinking is individuated in part by the objects he thinks about.4 Now we individuate the content of a propositional attitude in different ways for different purposes. But for our ordinary explanations of behavior we need to individuate propositional attitudes in part by the objects they are about. The actions we are interested in are typically described with the use of singular terms. Harry married Joan. Why did Harry marry Joan? Because he thought (among other things) that Joan is beautiful. In this explanation we cannot substitute "Joan" for a non-indexical definite description without the explanation loosing its force. The fact that Harry believes (among other things) that the nicest blonde girl in Nottingham is beautiful does not explain why Harry married Joan. Given this belief Harry could marry another woman as long as she fulfills the condition of being the nicest blonde girl in Nottingham. Why Harry married Joan is left unexplained. This way of explaining Harry's action by his belief that Joan is beautiful is only open to us if Joan exists. Therefore we can say that Harry's belief that Joan is beautiful is individuated in part by the object it is about, Joan.

[14] The above reflections make the following thesis plausible: A singular proposition about m is the content of a propositional attitude if the attitude can explain behavior directed upon m. My actions directed upon you can only be explained by my beliefs about you. This can of course just be a sketch of the conditions under which a person has a propositional attitude whose content is a singular proposition. But I think it has great intuitive appeal. And this will be sufficient for the purposes of our discussion. The account given can serve as a guide to answer the question whether the content of a propositional attitude is a singular proposition. The guiding idea is that a propositional attitude has a singular proposition as its content if (i) this attitude can be used to explain (actual or potential) behavior that is relationally characterized by the use of genuine singular terms and (ii) if the acquisition of the attitude can be explained by causal interaction with a particular object.

[15] We have now seen that a Neo-Russellian need not and should not endorse Russell's restriction on the possible constituents of propositions. Consequently, Neo-Russellians need not and should not use a notion of acquaintance according to which we are only acquainted with sense-data. Gareth Evans finds the key-concept for a less restrictive account of acquaintance in Russell's reason for holding PA,

that it seems scarcely possible to believe that we can make a judgment or entertain a supposition without knowing what it is that we are judging or supposing about. (Russell 1910-11: 209; my emphasis)
Evans calls Russell's reason for PA "Russell's Principle" (RP) (Evans 1982: 89). Other authors prefer the term "Discrimination Requirement" (Q. Cassam). In the formulation of RP the semi-technical expression "acquaintance" gives way to the colloquial expression "knowing who (what) someone (something) is". This is not a mere shift of terminology. One is acquainted with an object m if one knows m. Acquaintance is knowledge of objects. In contrast our ordinary concept of knowing who someone is is a concept of knowledge that something is thus-and-so. We say that S knows who m is if and only if S can provide a (contextually appropriate) answer to the question "Who is m?". This answer will express a complete proposition, something that can be true and known by someone. Grammatically speaking, what follows "know" in "S knows who (what) x is" is the nominalization of an interrogative clause. In his argument for PA Russell seems to confuse this nominalization with the nominalization of a relative-clause (McDowell 1969/70: 188).

[16] Russell's use of the concept of knowing who (what) (I will use "discriminatory knowledge" as an alternative and less cumbersome term for this kind of knowledge) enables his readers to get a better grip on PA, but on a second look it seems to falsify it. In a natural sense of the phrase "knowing who someone is" the amnesiac does not know who he is if he cannot answer the question "Who are you?" at the police-station by supplying information about his name and address. Yet the contents of his "I"-judgements are dependent on him in the same way as the contents of my "I"-judgements are dependent on me. So the question arises which concept of discriminatory knowledge makes PA a substantial and true principle? Evans tries to solve this problem by replacing our common sense notion of "knowing who" with a theoretical one motivated by the demands of a theory of propositional thought. He does not analyze our usage of phrases like "S knows who m is". Even if there is a fixed usage of these phrases, it is too context-sensitive to ground fundamental distinctions in the philosophy of language (Evans 1982: 66). Evans asks what discriminatory knowledge of m must be if its possession shall enable me to grasp an indefinite variety of thoughts about m in which different properties are ascribed to m.

[17] Evans's theoretically motivated answer to this question is, roughly speaking, that one possesses discriminatory knowledge about m iff one is able to discriminate m from all other things. Quassim Cassam offers a metaphor which helps to focus on the intended notion of discrimination: "[t]o delineate an object in thought is to draw its boundaries..." (Cassam 1997: 123). Sufficient conditions for the possession of discriminatory knowledge of m are according to Evans: (i) the current perception of m, (ii) the ability to recognize m when one encounters it and (iii) the knowledge of distinguishing facts about it (Evans 1982: 89). Evans's reason for the conception of discriminatory knowledge that is partially characterized by these conditions will be developed in section 7. I will use arguments drawn from Evans's work to show that the entertainment of reference rules for indexicals does not give one discriminatory knowledge.

[18] It is important to note that the concept of acquaintance and the notion of discriminatory knowledge do not coincide. Only the first two varieties of discriminatory knowledge, current perception and the ability to recognize something, enable a thinker to have propositional attitudes in which the object of the attitude figures, only they can be seen as varieties of acquaintance. For example, someone who can perceive an object m can (under normal conditions) act upon it. The fact that he perceives m will figure crucially in explanations of his actions towards m. The third variety of discriminatory knowledge is according to this standard a variety of descriptive knowledge.

[19] In this section I have outlined a Neo-Russellian version of PA. In the next two sections I will present and criticize David Kaplan's attempt to overthrow PA in any version by using arguments from the theory of demonstratives. Kaplan's counter-example to PA has puzzled many philosophers and I will try to locate the source of the puzzlement in what follows.

3. "Dthat" And Grasping Singular Propositions

[20] In his paper "Quantifying In" Kaplan subscribes to the view that S can only have a propositional attitude reportable de re about m iff s is en rapport with m. S is only en rapport with m if (i) she knows which thing m is and (ii) if she has mental and linguistic representations of m (Kaplan 1969: 229). "Of" is a technical term. Kaplan uses the model of a portrait of a particular person to explain its meaning. Consider the case of two visually indistinguishable twins Fred and Ted. If we take a picture of Ted, the picture not only resembles him but also Fred. But it is a picture of Ted because he has played the right causal role in the production of the picture. Even if the picture resembles another person it is still a picture of Ted.

[21] So far Kaplan seems to work with a conception of a propositional attitude having a singular proposition as its content very similar to the one presented in the last section. But after the publication of "Quantifying In" Kaplan starts working on the logic and semantics of demonstratives. This leads to a radical change in Kaplan's views on propositional attitudes. He reports the startling result of these investigations in his paper "Dthat" thus:

All this familiarity with demonstratives has led me to believe that I was mistaken in "Quantifying In" in thinking that the most fundamental cases of what I might now describe as a person having a propositional attitude (believing, asserting, etc.) toward a singular proposition required that the person be en rapport with the subject of the proposition. It is now clear that I can assert of the first child to be born in the twenty-first century that he will be bald, simply by assertively uttering,
Dthat ["the first child to be born in the twenty-first century"] will be bald.
I do not now see exactly how the requirement of being en rapport with the subject of a singular proposition fits in. (Kaplan 1978: 229-230)
According to Kaplan I may acquire a belief about a person (a belief reportable in the de re style) simply by accepting the sentence:
(S1) Dthat ["the first child to be born in the twenty-first century"] will be bald.
without being en rapport with the object of the attitude. 5 In "Demonstratives" Kaplan sees the semantics of "dthat" as a "mechanism" which allows us to apprehend singular propositions containing individuals formerly known to us only by description:
Normally one would not introduce a proper name or a dthat-term to correspond to each definite description one uses. But we have the means to do so if we wish. Should we do so, we are enabled to apprehend singular propositions concerning remote individuals (those formerly known to us only by description).... The introduction of a proper name by means of a dubbing in terms of description and the active contemplation of characters involving dthat-terms -- two mechanisms for providing direct reference to the denotation of an arbitrary definite description -- constitute a form of cognitive restructuring; they broaden our range of thought. (Kaplan 1989a: 560, fn 76)
Thus "dthat" broadens our range of thought by enabling us to apprehend singular propositions which were otherwise beyond our cognitive reach.

[22] Kaplan has already given us a hint why he thinks that it is not necessary to be en rapport with the object a propositional attitude is about: his familiarity with demonstratives has led him to reject his former epistemic views on singular thought and reference. This is at first sight a puzzling claim. How can "familiarity with demonstratives" lead to this result? For "dthat" in Kaplan's example can hardly be seen as a variant of the demonstratively used demonstrative pronoun in English.6 This is already clear from syntactic considerations. The demonstrative pronoun combines with predicates to yield complex singular terms, but it does not combine with nominal phrases like "the first child to be born in the twenty-first century" to a well-formed singular term: "that the first child to be born in the twenty-first century" is not a well-formed term in English.

[23] Kaplan's own theory of the demonstrative use of the demonstrative pronoun and his explanation of the meaning of "dthat" confirm the suspicion that Kaplan's example does not contain a demonstratively used term at all. What is special about the demonstrative use of a singular denoting phrase is that "the speaker intends that the object for which the phrase stands be designated by an associated demonstration" (Kaplan 1978: 220). A demonstrative utterance like "That man is married" needs completion by demonstration, the demonstration determining the object referred to in the context of utterance.7

[24] Demonstration is one of Kaplan's theoretical notions. He takes the demonstration "typically, [to be] a (visual) presentation of a local object discriminated by a pointing" (Kaplan 1989a: 490). The same demonstration may present different objects to different demonstrators. Essential for a demonstration is that it presents an object from a certain perspective. Kaplan specifies the way a demonstration presents an object by a definite description. For example, in (S2) "That ['the brightest heavenly body now visible from here'] is Hesperus" the definite description in square-brackets specifies the way in which the demonstration associated with a demonstrative utterance of this sentence presents its object.

[25] Since Kaplan is mostly interested in the demonstrative use of the demonstrative pronoun, he introduces by stipulation the expression "dthat", the "d" simply serves to signal the demonstrative use of "that" (Kaplan 1978: 222, 1989a: 521ff).

[26] Kaplan is intrigued by the observation that a demonstration is a "way in which objects are given" which must be grasped by the competent auditor of a demonstrative utterance, without the demonstration being part of the content of a demonstrative utterance. "Content" is here best understood as corresponding to an intuitive notion of what is said by an assertoric utterance (in a context). The demonstration which a competent auditor must grasp to understand my demonstrative utterance does not contribute to the content of the utterance, although it determines the reference of "that" in the particular utterance-context and directs the attention of the audience to the object determined. Thus, Kaplan argues, a demonstrative utterance containing "that" or Kaplan's "dthat" (which is specially designed for demonstrative use) expresses a singular proposition: two demonstrative utterances in which the demonstrative pronoun or Kaplan's demonstrative surrogate is coupled with a predicate F express the same utterance content iff the same object is demonstrated. The identity of utterance content does not depend on identity of demonstration but only on identity of the object demonstrated (Kaplan 1978: 221). Although this way of talking is misleading, Kaplan frequently says that the utterance content of a "dthat"-utterance contains the object demonstrated. Later he will coin the expression "directly referential term" for a term that only serves to introduce an object into a proposition. The demonstratively used "that" is the natural language paradigm of a directly referential term. Neo-Russellians will agree with this claim about the content of a demonstrative utterance. The complete understanding of a demonstrative utterance requires that one perceives the object demonstrated oneself. The content one grasps depends for its existence on the existence of the object demonstratively referred to. So we can say that a demonstrative utterance expresses a singular proposition.

[27] Let us now look back at Kaplan's counter-example to the Russellian theory. It is important to notice that up to now "dthat" was just a stylistic variant of the natural language demonstrative pronoun especially designed for demonstrative use. So "dthat" cannot achieve more than "that" in its ordinary use. It is hard to see how one can express with (S1) "Dthat ['the first child to be born in the twenty-first century'] will be bald" a singular proposition or grasp such a proposition by understanding an utterance of it. Kaplan's reason why a demonstrative utterance expresses a singular proposition was that although the demonstration determined an individual, the demonstration was not part of the content of the demonstrative utterance. But in Kaplan's example there is no object one can visually discriminate and point to! Perhaps we can refer to future individuals, but we cannot point to these individuals now. "Dthat" cannot function in an utterance of (S1) as a demonstratively used demonstrative pronoun: there is no demonstration to determine its reference. (S1) cannot be used by us now to make demonstrative utterances, thus we have no reason for holding that an utterance of (S1) expresses a singular proposition.

[28] Kaplan's hint seems to be misleading. Familiarity with the semantics of natural language demonstratives cannot make plausible that the en rapport requirement for entertaining singular propositions is false. How can we understand then Kaplan's alleged counter-example? In the next section I will answer this question.

4. Kaplan's Fallacy of Equivocation

We do not get ourselves into new belief states by "the stroke of the pen" (in Grice's phrase) simply by introducing a name into the language. (Evans, "Reference and Contingency")

[29] Kaplan misrepresents his example when he says that familiarity with demonstratives has led to the theoretical changes he outlines. These changes are rather due to a far-reaching extension of the notion of a pointing. Kaplan assimilates in "Dthat" and "Demonstratives" reference-determining demonstrations to descriptions. Kaplan asks in "Dthat":

If pointing can be taken as a form of describing, why not take describing as a form of pointing? (Kaplan 1978: 223)
and he answers himself in "Demonstratives":
Now why not regard descriptions as a kind of demonstration, and introduce a special demonstrative which requires completion by a description and which is treated as a directly referential term whose referent is the denotation of the associated description? Why not? Why not indeed! I have done so, and I write it thus:
where a is any description, or more generally, any singular term. "Dthat" is simply the demonstrative "that" with the following singular term functioning as its demonstration. (Kaplan 1989a: 521-522)
After Kaplan has extended the notion of a pointing, the ordinary "that" has become a "special demonstrative" indeed! Kaplan's "dthat" is no longer a syntactically complete singular term, now it requires completion by a definite description. Therefore it is best classified syntactically as an operator which returns for a description a complete singular term.

[30] Kaplan notes this result in his "Afterthoughts" to "Demonstratives". There he points out that the expression "dthat" has two non-equivalent interpretations in his work: sometimes it is a demonstrative surrogate, sometimes it is an operator (Kaplan 1989b: 579). My criticism of Kaplan's argument against PA simply develops the consequences of the ambiguity he himself notes.

[31] If "dthat" is a demonstrative surrogate, it is syntactically a complete singular term. The demonstration which is needed to fix the reference of a particular token of "dthat" in a context of utterance is not itself an expression on which "dthat" could be seen to operate. A pointing may fix the reference of a demonstrative, but it is not the kind of thing that can be operated on by a singular term forming operator of any kind. Like "dthat" "I" needs completion by contextual factors, but that does not make "I" an operator. Why should this be different with "dthat"? Kaplan's method of specifying the way a demonstration presents an object by a definite description and his symbolism in which these descriptions are added to the demonstrative tend to obliterate this. The relevant description must be seen as describing the part of the utterance context which determines the reference of a "dthat"-token, but the "dthat"-token is already a complete singular term.

[32] If "dthat" is an operator on definite descriptions, it is syntactically not a complete singular term. Kaplan characterizes the dthat-operator semantically as a rigidifier (Kaplan 1989b: 580): "dthat['the f']" rigidly designates whatever object "the f" designates (flexibly). This characterization falsifies Kaplan's original claim in "Demonstratives" that dthat-terms completed by definite descriptions are (i) complex directly referential terms in which (ii) "dthat" functions as a demonstratively used demonstrative pronoun. In contrast to the demonstration that determines the reference of the demonstrative surrogate "dthat" the definite description which completes the dthat-operator cannot be seen, like a gesture, as "pre-semantic" and thus irrelevant for utterance content. A reflection on the above characterization of the semantics of the dthat-operator provides an intuitive argument for this claim.

[33] Kaplan sometimes conceives of the content of the definite description "the " as an instruction that determines exactly one object. In order to determine the object designated by "dthat ['the f']" one must apply the "instruction" expressed by "the ". The application of "dthat" to the definite description adds, so to say, an additional instruction: hold the object determined by the original instruction constant when evaluating the sentence containing the definite description with respect to different circumstances.

[34] This illustration shows that although dthat-rigidified descriptions are rigid, the content of the descriptions they contain is part of the utterance content. Kaplan's original characterization of the dthat-rigidified descriptions as directly referential confused obstinate rigidity and direct reference. Soames makes the same confusion in writing:

When the dthat-operator is applied to a description D, it completely obliterates the descriptive content of D, and leaves the rigidified description dthat D with no descriptive content at all. As a result, coreferential dthat-rigidified descriptions have the same content.... (Soames 1998: 17)
"Content" means different things in different theoretical contexts: Sometimes it means intensions, understood as functions from possible worlds to extensions, sometimes it means constituents of Fregean thoughts that are relevant for notions like cognitive significance.

[35] If "content" means something like cognitive content in the passage just quoted, Soames's claim about co-referential dthat-rigidified descriptions is falsified by a counter-example. Consider Kaplan's original example (S1) and the similar sentence (S3):

(S1) Dthat ["the first child to be born in the twenty-first century"] will be bald.

(S3) Dthat ["the first child to be born in the first century after the death of Albert Einstein"] will be bald.
Neither in (S1) nor in (S3) can "dthat" function as a demonstrative surrogate. I take it to be uncontroversial that only the operator-reading is intelligible. If there will be a first child born in the twenty-first century, it will be denoted by both definite descriptions with respect to all possible worlds. The content, in the sense of intension of the rigidified descriptions contained, in (S1) and (S3) is thus the same: both descriptions are associated with the same constant function from possible worlds to extensions.

[36] But someone may understand the sentences (S1) and (S3) in the readings under discussion and accept what is said with the first, while reject what is said with the second. This is sufficient to grant that the rigidified descriptions used in these sentences express different contents in the sense of Fregean modes of presentation. The same holds for content in the sense of structured intensions. This point can again be unfolded by using Kaplan's instruction-metaphor: someone who understands (S1) and (S3) knows thereby different instructions to determine an object. The fact that these instructions determine the same object is neither evident nor knowable a priori. What the competent speaker additionally knows in virtue of his knowledge of the semantics of the dthat-operator is that each instruction determines the same object with respect to every possible world.

[37] If two rigidified descriptions produced by the application of the dthat-operator express different Fregean modes of presentation or structured intensions, then there are (in two important meanings of "content") different contents they express. Co-referential dthat-rigidified descriptions need not have the same content.

[38] This conclusion about dthat-rigidified descriptions undermines arguments which, if convincing, would show Kaplan's claim that one can grasp a singular proposition about a future individual by knowing the linguistic meaning of (S1) to be true. If I honestly assert (S1), I believe that dthat ["the first child to be born in the twenty-first century"] will be bald. Is the ascribed belief reportable in the de re style? A positive answer could rely on the following principle used by Donnellan in a critique of Kaplan's counter-example (Donnellan 1979: 54):

(DR) If an ascription of a propositional attitude contains a term without any descriptive content in its "that"-clause, the ascribed attitude is de re.
The dthat-rigidified description in Kaplan's example (S1) has descriptive content. The pecularity of this content is that it determines the same object with respect to every possible world. This shows that we cannot establish Kaplan's thesis that we can acquire de re propositional attitudes by the comprehending use of dthat-rigidified descriptions via DR.

[39] Some authors endorse a principle related to DR which, if true, would give us a reason for Kaplan's thesis. For instance Mark Sainsbury applies the notion of rigid designation to thought in order to give a principle for thought-constituency independent of epistemic notions:

(R) "[Object] o is a constituent of [thought] t if and only if the truth of t with respect to any possible situation turns on how things are with o. (Sainsbury 1986: 239)
Relying on R one could say: if an ascription of a thought contains a (obstinately) rigid term "a" in its "that"-clause, the truth of the ascribed thought turns with respect to any possible situation on how things are with the referent of "a", namely a. Therefore the content of the thought contains a and must consequently be de re.

[40] But I think we have a good reason to reject R. Consider a parallel example. The definite description "the current Prime Minister of Great Britain" is temporally rigid: a use of this description in a context of utterance denotes with respect to every time of evaluation the same object: the object which is the prime minister of Great Britain at the time of utterance. But from the fact that I believe that the current Prime Minister of Great Britain has grey hair, it does not follow that I believe of the current Prime Minister of Great Britain that he has grey hair (see also Donnellan 1979: 54). The truth of the ascription is compatible with me believing that the property of being a unique current Prime Minister of Great Britain and the property of having grey hair are co-instantiated.

[41] Up to now I have argued that we have no convincing reason to take Kaplan's (S1) as expressing a singular proposition, it expresses a general proposition which can be pictured as containing a rigid individual concept. Therefore his alleged counter-example to the en rapport condition for entertaining a singular proposition is unconvincing. We now have the material needed to develop a reason against Kaplan's thesis that the competent use of "dthat" enables us to grasp a singular proposition without being en rapport with the individual the proposition is about. Take the sentences (S4) and (S1):

(S4) The first child to be born in the twenty-first century will be bald.

(S1) Dthat ["the first child to be born in the twenty-first century"] will be bald.
Considered from the perspective of New Theories of Reference (S4) is a paradigmatic example of a sentence which expresses a general content: the expressed content is exclusively made up from conditions on things and not of things themselves. Now can someone who understands (S4) and (S1) at the same time believe what is said with (S4) and reject what is said with (S1) (and vice versa)? I take it that this is not possible. (S4) and (S1) should be grouped with sentence-pairs like "The president of the USA came to dinner" and "The actual president of the USA came to dinner" (Stanley 1997: 574ff) which, on the face of it, impart the same information about the world. Frege would therefore say that utterances of (S4) and (S1) express the same Fregean thought. This Fregean thought can then be seen as the common content of the beliefs one can express by uttering (S4) and (S1). Now if (S4) does not express an object-dependent (singular) Fregean thought, how could (S1) express such a thought, if both sentences (and the beliefs one voices with those sentences) have he same Fregean thought as their content?

[42] The thesis that there must be a notion of content according to which the beliefs one can express with utterances of (S4) and (S1) have the same content can be made plausible by an independent argument. Consider the following example. In 1997 Richard understands (S4) and accepts what is said by it. In 1998 he acquires the use of "dthat" and accepts what is said by (S1). Does this bring about that Richard acquires in 1998 a belief whose content is a singular rather than a general proposition? According to our guide to the ascription of propositional attitudes whose contents are singular propositions, the answer must be No. The belief that Richard acquired in 1998 cannot dispose him to act upon a particular person in the future or to treat information he receives from a particular person as germane to the evaluation of his belief. Only the additional information he can acquire in the twenty-first century that this or that child is the first child born in the twenty-first century will enable him to act upon a particular child. So there is no good reason to see Richard as holding true a singular proposition rather than a general proposition that, if true, will be verified by a particular child. The mere acquisition and application of the "dthat"-operator does not change this. Although Richard's acceptance of (S1) will cause some changes in his overall belief-state, it does not change his dispositions to act upon things nor to justify his beliefs. If we take Richard to acquire a new belief in 1998, we must therefore construe his new belief as sharing something with his old belief connected to (S4). Let us call what they share "rational content" because this kind of content is closely bound up with the rationalization of actions. "Rational contents" are, to say the least, close relatives of Fregean thoughts. They fulfill the role we previously marked by the terms "what we think" and "accusatives of thinking".

[43] Now while the rational content of Richard's beliefs stay the same, there is also a difference to be noticed: The beliefs Richard would express with (S4) and (S1) are true with respect to different possible worlds. To avoid the terminology of possible worlds by saying that a sentence S is true with respect to such a world, we should instead say that complex modal sentences containing S are true absolutely. Complex modal sentences containing (S4) and (S1) differ in truth value. Take for instance (S6) and (S7):

(S6) It is possible that the first child to be born in the twenty-first century will be bald.

(S7) It is possible that dthat ["the first child to be born in the twenty-first century"] will be bald.
If (S7) is true, its truth depends on the baldness of the particular that fulfills in the actual world the condition of being first child to be born in the twenty-first century. In contrast the truth of (S6) does only depend on the joint fulfillment of the condition of being first child to be born in the twenty-first century and being bald. In different possible worlds different things can jointly fulfill these conditions. So we must assign to the beliefs expressed by (S4) and (S1) different objects (classes of possible worlds, etc.) which become relevant when these beliefs are modally evaluated. Let us call these objects "modal contents". The arguments adduced so far suggest that two beliefs can have the same rational content, while differing in their modal contents.

[44] In "Reference and Contingency" Evans pointed out that, to use the terminology of this section, two sentences may have the same rational, but different modal contents, the difference in modal content being due to the fact that the relevant sentences embed differently under modal operators. In this section I have argued that this thesis holds for sentences containing dthat-rigidified descriptions and the corresponding sentences which contain only the definite description. A sentence containing the dthat-rigidified description "dthat D" may have the same rational content as the corresponding sentence containing the description "D". We can adopt and extend Evans's insight: if the rational content of a belief is a general proposition, we do not change this content into a singular proposition by applying a rigidifying operator to a definite description or the analogue of such an operator in the realm of concepts to a descriptive concept.8 What we change is the modal content.

[45] Let us briefly take stock. (S1) and (S4) have the same general proposition as their rational content, at the same time they represent different ways the world might have been. This interesting finding is compatible with the Russellian view that grasping a singular proposition requires being en rapport with the object a propositions depends on. The contrary impression is due to an equivocation between "dthat" as a demonstrative surrogate and a rigidifying operator. "Dthat" as a demonstrative surrogate enables us to express and entertain singular propositions, but in Kaplan's central example (S1) "dthat" cannot function as a demonstrative surrogate. Only the operator interpretation is available and in this interpretation utterances of (S1) express the same rational content as utterances of (S4). So understood utterances of (S1) do not express singular propositions as their rational contents. Consequently one cannot entertain a singular proposition, make such a proposition the rational content of one's thinking by understanding (S1). The basic assumption of Kaplan's counter-example to PA turns out to be false.

5. Direct Reference And Grasping Singular Propositions

[46] We have now rejected Kaplan's initial counter-example to the Russellian acquaintance condition. Kaplan's counter-example has not convinced everyone, for example Donnellan and Soames have been puzzled by Kaplan's thesis that the use of "dthat" should enable one to entertain singular thoughts by understanding sentences containing it. The previous section tried to explain why Kaplan's example is so puzzling: according to the diagnosis given, it failed because "dthat" turned out not to be a directly referential expression. In "Demonstratives" Kaplan challenges what he calls the "Direct Acquaintance Theories of direct reference" by arguments based on the semantics of indexicals like "here", "now" and "I". The notion of a pointing or a demonstration does not play a crucial role in their semantics. Therefore the problems connected to "dthat" should not arise. Let us now see whether the semantics of indexicals yields counter-arguments against PA.

[47] In order to develop Kaplan's argument I will sketch his semantics for indexicals. Kaplan associates in "Demonstratives" two sorts of meaning with indexicals: character and content. First to character. The entry under "I" in my English dictionary reads:

I: the person speaking
This entry gives (in one sense of the word "meaning") the meaning of "I" by providing a rule which systematically fixes its reference with respect to contexts of use. Kaplan calls more refined versions of such rules "characters". They are, so to say, the theoretical successors of dictionary entries. Formally characters are represented as functions from contexts of use into intensions. Since the formal representation blurrs the intuitive point Kaplan wants to make, I will not use it from now on.

[48] A first critical remark is in order here. It is questionable whether the character of "I" incorporates the information that the producer of the "I"-token is a person. I take it that only a sortally neutral reference rule for "I" can plausibly be held to specify what a speaker of English knows when he knows the meaning of "I". This point will become important in section 7. I will then give an argument for my critical remark.

[49] Central for our purposes is Kaplan's thesis that indexicals are directly referential. Normally one expects a dictionary entry for a word to provide a synonym or at least something very like it. In the case of indexicals this is different: the rules English associates with indexicals do not supply synonyms, they have the different function to fix the reference of individual tokens of the indexicals in a systematic way. For instance, "I" and the reference-fixing definite description "the one speaking" known by the competent user of "I" cannot be exchanged salva sensu, as the following sentence-pair shows:

I am a fool!

The one speaking is a fool!
The rules governing "I" provide a referent for each context of utterance, but not an intension or descriptive content for the indexical. This brings us to content. The indexical, says Kaplan, only introduces the referent into the content of an assertoric utterance made with its help.9 The content is a singular proposition. Indexicals are therefore directly referential. Since the rules for indexicals do only determine "implicitly" that their referent in a particular context of use does not change with respect to circumstances of evaluation, the sense of an indexical cannot be seen as given by a rigidified description.

[50] The second crucial point in Kaplan's semantics for indexical expressions is that the semantic rules for indexicals do not require that their user is acquainted with the object referred to (in a context of use). A sensory deprived amnesiac could use "I" to express a singular proposition "containing" himself if his use of the first person pronoun confirms to the rules which govern this expression in English. He is surely not acquainted with himself, neither in Russell's demanding nor in a more liberal sense of "acquaintance" preferred by Neo-Russellians. This lack of knowledge does not deprive him from using the first person pronoun.

[51] Kaplan exploits the two features of indexicals just mentioned in his argument against the Direct Acquaintance Theory of direct reference. Someone who utters a sentence containing a directly referential term expresses a singular proposition containing the referent of this term. What enables us to express singular propositions is our knowledge of the semantic rules connected with indexicals and not our acquaintance with the referent of the indexical on an occasion of use. If this argument is sound, one can express singular propositions without being acquainted with their constituents, on any construal of acquaintance. But there is a step to be taken from (i) expressing a singular proposition by utttering a sentence to (ii) entertaining the so expressed singular proposition. For a speaker may express a singular proposition which he cannot grasp himself.10 For example I may assert that nikotin is dangerous by uttering the sentence "Nikotin is dangerous", but, not an expert in chemistry, fail to grasp the proposition expressed by my utterance. Kaplan takes the step from (i) to (ii) when he says that

[f]rom [the Theory of Direct Reference] follows that a special form of knowledge of an object is neither required nor presupposed in order that a person may entertain as object of thought a singular proposition.

There is nothing inaccessible to the mind about the semantics of direct reference, even when the reference is to that which we know only by description. What allows us to take various propositional attitudes towards singular propositions is not the form of our acquaintance with the objects but is rather our ability to manipulate the conceptual apparatus of direct reference. (Kaplan 1989a: 536)

[52] Kaplan presents this consideration as an argument against the view that one can only grasp the proposition expressed by an indexical utterance if one is acquainted with the indexically referred objects. He aims to refute the Direct Acquaintance Theories of direct reference. But the conclusion of Kaplan's argument does not conflict with these theories at all. According to Kaplan's theory a propositional attitude is a triadic relation: a person S f-s that p under a character c. In the case of indexicals the character expressed by an indexical sentence determines in a context of use a singular proposition. In Kaplan's triadic conception singular propositions are not what we think. What we think are the characters of the expressions used, they are relevant for individuating propositional attitudes and giving psychological explanations.11 Someone who "entertains the character" of "I" will think of its referent in a context of use under the description "the producer of this "I"-token". What we think when we assert an indexical sentence and entertain its character is a general, not a singular proposition. This is confirmed by our intuitive conception of what has to be the case for a propositional attitude to have a singular proposition as content. The belief that the producer of a particular "I"-token speaks too loud cannot explain why I lower my voice. I might believe that the producer of the particular "I"-token speaks to loud without recognizing that I am the producer of the relevant token. What must be added for the explanatory purpose at hand is the additional belief that I am the producer of the particular "I"-token. The belief I have when I accept the character of an indexical sentence cannot explain relationally characterized behavior. It cannot therefore be seen as a belief with a singular proposition as its content.

[53] The conclusion that the entertainment of the character of an indexical is a form of descriptive knowledge already points to a difficulty for the view that characters can serve as modes of presentation in indexical thought. For instance, if I think that I am wounded I am disposed to act upon one particular object, namely myself, but if I entertain the character of "I", i.e. think of myself as the producer of a particular "I"-token, I may lack these dispositions. In the next section I will elaborate this critical remark.

[54] Seen this way Direct Reference Theory does not amount to a refutation of the Direct Acquaintance Theory of direct reference because both parties use the phrase "entertaining a proposition as an object of thought" differently. The Direct Acquaintance Theory holds for singular propositions in the role of what we think. The Neo-Russellian who argues for epistemic constraints on grasping singular propositions can happily agree that these constraints do not hold for singular propositions in a different role. Of course a theorist who chooses to employ the term "object of thought" for something which is not intended to play an explanatory role in action explanations etc. can hold that a singular proposition can be an object of thought, although the thinker has only knowledge by description of the propositional constituents.

[55] Kaplan's argument against the Direct Acquaintance Theory of direct reference misses its target. This is so because the theory of indexicals on which the argument is based assigns a new role to singular propositions. They are not, as Russell and the Neo-Russellians thought, the contents of propositional attitudes. So there is no real conflict between Kaplan and Russell and his followers. Both parties agree that if a speaker entertains the character of an indexical, he knows the referent only by description. The content of his thinking, in a psychologically relevant sense, is a general proposition. Kaplan shows that a propositional attitude with a general content may be verified by the states and doings of a particular thing or person. Every Russellian can accept this thesis without abandoning the acquaintance restriction on the contents of propositional attitudes.

[56] As already pointed out, what we think are according to Kaplan the characters of the expressions used. This idea is independently interesting for it poses a real challenge to the Russellian tradition as well as to Fregean theories of sense. Therefore I will discuss it in more detail in the next two sections.

6. "Entertaining the Meaning Of An Indexical"

[57] Kaplan takes the meanings of the indexicals, his characters, to play the role of Fregean modes of presentation (Kaplan 1989a: 530). The basic notion of Kaplan's account is that of entertaining the character of an indexical. The natural understanding of this notion is that one entertains the character or meaning of an indexical iff one grasps the sense of the definite description which English associates with the indexical , say, in the case of "I" the definite description (DI) "the producer of x and x is a token of 'I'". If one thinks of oneself under this description, Kaplan and Perry will say that one entertains an ego-thought, a thought in which I think of myself self-consciously. Evans gives a condensed version of this view in "Understanding Demonstratives":

By "entertaining" the meaning or role of the demonstrative "I", Hume thinks of himself. Similarly, by entertaining the meaning or role of the demonstrative "here", Hume thinks of a particular place. (Evans 1981: 320)
Evans deals with "self-conscious" thought and not just with thought about oneself (with "here"-thought and not just with thoughts about one's location). Now ego-thoughts are paradigm cases of thoughts whose contents are singular propositions. So we can grasp some singular propositions by entertaining the meaning of indexicals.

[58] Kaplan additionally holds that the particularity and primitiveness of the first-person way of thinking Frege was interested in, "can be fully accounted for using only our semantical theory" (Kaplan 1989a: 534). Moreover one can entertain the character of an indexical without fulfilling any substantial epistemic requirements, like being en rapport with the object referred to. Kaplan's account seems nonetheless to fulfill the Discrimination Requirement. If I entertain the character of "I" in a context of utterance, it is prima facie justified to say that I know who the referent of the indexical is, namely the producer of this "I"-token. So Kaplan seems to be able to do everything Frege wanted to do with modes of presentation on the basis of his semantic theory. If Kaplan's account is correct, semantic theory can resolve the issues that are puzzling about indexical thinking without invoking such slippery notions as that of acquaintance. For instance, a philosophical explanation of our ability to refer to ourselves in the first-person way and to think ego-thoughts about ourselves could be very easily given. All one has to do is to look up the meaning of "I" in the dictionary. By contrast a Neo-Russellian philosopher like Evans holds "that such an account must presuppose some of the profoundest philosophy. In the case of 'I' for example, one might think that an account of the relation R1 which explicates the 'self-identification' must incorporate the insights, as well as illuminate the struggles, of Descartes, Kant, and Wittgenstein and many others" (Evans 1981: 318). The crucial question is now whether entertaining the meaning of an indexical really amounts to thinking an ego-thought. Let us try to answer it.

[59] The first problem with Kaplan's account is that it faces the question what it takes to grasp a singular proposition that contains a token of an expression. This is, as Evans readily pointed out, a variant of the general question that should be answered by the account. But Evans's main objection is:

[W]hat has the idea in [(DI)] got to do with one's capacity to think of oneself self-consciously? The problem is this. No one can give an account of the constant meaning (=role) of a demonstrative without mentioning some relational property (relating an object to a context of utterance) which an object must satisfy if it is to be the referent of the demonstrative in that context of utterance, but the idea of that property plays no part in an explanation of what makes a subject's thought about himself, or the place he occupies, or the current time. (Evans 1981: 320)
At first sight Evans's argument seems to be this: My entertaining the role of an indexical may constitute my understanding of this expression. But the thought I might express by using the indexical is not about a certain object because I entertain the role of the indexical. Take as an example my thought that I have overweight. To have this thought it is not necessary to identify myself by using the definite description associated with the first person pronoun. The property of being the producer of an "I"-token plays no part in an explanation of what makes my thought a thought about myself.

[60] Even if the argument so construed is convincing, it will miss its target. The argument refutes the thesis that the capacity to think of oneself self-consciously is nothing over and above the capacity to understand "I". In order for me to think an ego-thought it is not necessary to entertain the role of "I", i.e. to think of me as the utterer of an "I"-token. What has yet to be shown is that entertaining the role of "I" is not sufficient to think an ego-thought. The thesis that it is sufficient would be sufficient to give Kaplan the interesting point that entertaining the character of "I" (and ascribing properties to the object so conceived) is a variety of self-conscious thought. Evans needs to show that in general the relational properties which figure in the linguistic meanings of indexicals are not the right properties to distinguish the objects of indexical reference in thought. What is lacking up to now in Evans's argument is a positive reason to say that entertaining the meaning of an indexical does not amount to the form of knowing which thing something is that required for indexical thought.

[61] One can also give a positive reason for the view that entertaining the meaning of "I" is sufficient to have an ego-thought. One may argue that the sensory deprived amnesiac can have ego-thoughts in virtue of using "I" in foro interno without being in the position to have any knowledge about himself.

[62] Now such examples are rather extreme and our intuitions about "what to say" in these cases are less than crystal clear (Cassam 1997: 150). Theoretical consideration can overrule pre-theoretical intuition. So what we need is a theoretically motivated reason which (i) makes plausible that entertaining the meaning of an indexical is not sufficient to think an ego-thought and (ii) allows us to correct the intuitions appealed to in the counter-examples. In the next section I will try to show that Evans's Constraints of Generality give us such a reason.

7. Roles And Adequate Ideas

Strawson has explored the consequences of one kind of generality, and I am exploring the consequences of the other. (Evans 1982)

[63] Evans sees propositional thoughts (thoughts ascribing a property to an individual) as subject to two Constraints of Generality (1982: 209):

[W]e cannot avoid thinking of a thought about an individual object x , to the effect that it is F, as the exercise of two separable capacities; one being the capacity to think of x, which could be equally exercised in thoughts about x to the effect that it is G or H; and the other being a conception of what it is to be F, which could be equally exercised in thoughts about other individuals, to the effect that they are F. (Evans 1982: 75)
The Constraints of Generality are the Principle of Predicate Variability:
One can think that a is F only if one can think that a is G, for every categorically adequate property of being G of which one has a conception. 12
and the Principle of Subject Variability:
One can think that a is F only if one can think that b is F, for every object b which is in the range of application of the property F.
In this section I will show that Kaplan's characters do not enable a thinker to meet the requirements imposed on ego-thought by the Generality Constraints. Before exploring the Generality Constraints further, a brief comment on Evans's terminology. Evans prefers to say that propositional thoughts are necessarily the exercises of different capacities because he thinks the talk of thought-constituents misleadingly suggests that there is a language of thought. He introduces the notion of an Idea which is a particular person's capacity to think of something. Ideas of different people are eo ipso different Ideas. But people who exercise different Ideas may think of the same thing in the same way. It will also be useful to distinguish between singular Ideas and ascriptive Ideas. The last ones are Ideas of properties which we express with predicates.

[64] The Principle of Predicate Variability serves (among other things) to introduce a theoretically motivated notion of discriminatory knowledge. A thinker possesses discriminatory knowledge about m if he has an Idea of m which "makes it possible to think of [m] in a series of indefinitely many thoughts, in each of which he will be thinking of [m] in the same way" (Evans 1982: 104). This is what discriminatory knowledge must be if it shall play a role in our description of propositional thought.

[65] The legitimacy of this notion of discriminatory knowledge depends on the plausibility of the Principle of Predicate Variability. Both Constraints of Generality have a certain Kantian flavor. "We cannot avoid", says Evans to see the Constraints of Generality as applying to propositional thought.13 Evans does not attempt to give a complete justification of this view, yet that his basic idea is correct can be seen from considerations about objectivity. It is part of our conception of an enduring and objective particular that it can bear different properties at different times, some of which we cannot apply on the basis of perception. For my propositional thinking to represent m as an enduring and objective particular, I must be able to apply a whole variety of properties to it, I must be able to think that m is F, m is G,.... Considerations like these provide a "local" or piecemeal explanation why our propositional thinking must conform to the Generality Constraints if it concerns objective spatio-temporal particulars at all (Campbell 1984-85: section IV). In the following I will not rely on the thesis that the Constraints of Generality are conceptually necessary because I take it to be independently plausible that the variety of thinking I will be concerned with fulfills the Principle of Predicate Variability.

[66] Let us now return to our initial question whether my entertaining the character of "I" (together with the exercise of property concepts) is sufficient to think ego-thoughts. We have seen that according to a notion of discriminatory knowledge shaped by the theory of thought my discriminatory knowledge of an object must enable me to ascribe all the properties to it that I can conceive of. Let us add to this the plausible thesis that the Principle of Predicate Variability holds for first person thought: I am able to entertain indefinitely many ego-thoughts which ascribe to me mental (thinking about Vienna, being in severe pain) and physical properties (looking pale, having overweight). If my entertaining the character of "I" does not enable me to think of me in a series of indefinitely many thoughts, in each of which I will be thinking of me self-consciously, the answer to the above question must be No. The mental state of entertaining the role of "I" and ascribing a property to the so conceived object would lack a feature crucial for the first person thoughts we express with "I". Similar things apply to other indexicals. I will now argue for the negative answer.

[67] In its most plausible version the character of "I" is the rule (CI):

(CI) Any token of the English word "I" refers to whoever uses it. (Woods 1968: 569; Campbell 1994: 102)
A thinker S who entertains the character of "I" can grasp thoughts about S which ascribe properties to S like makes strange noises or has a deep voice. S discrimination of S by the property of uttering an "I"-token puts S in a position to ascribe these auditory properties to S. But if this is the only way in which S can discriminate S and he is not able to think further that he himself is the producer of the "I"-token or that the object is a person, S cannot do what everyone capable of ego-thinking can do: entertain ego-thoughts in which such diverse properties as weighing too much or proving a theorem of arithmetic are ascribed to one and the same object. "Active contemplation" of the character of "I" does not enable one to entertain all the thoughts I may think about myself! Three arguments make this negative thesis plausible.

[68] 1. In order to think that a is F I must think of a as being the kind of thing to which it makes sense to ascribe being F (Woods 1968: 576). For most properties the question whether an object has it or not is only defined for certain kinds of objects. For instance, the question whether the property of being prime applies to a horse or not has no definite answer. A thinker who ascribes a property to an object (and not only plays at doing so) must conceive of this object as being in the range of application of the property. If a thinker can conceive of an object only via the relational property of producing an "I"-token, he will not be in a position to apply most properties we take to be applicable in ego-thought. I can apply character-traits in ego-thought. The range of application of the character-traits is the set of persons, or an analogical extension of this set. The property of producing an "I"-token leaves it open whether the object is in the range of application of the character-traits. So it does not enable a thinker to ascribe such traits to himself if this is the sole mode of presentation of himself that is available to him.

[69] 2. The ascription of certain properties is sensitive to the spatio-temporal boundaries of an object.14 Take for instance the mental property of proving a theorem of arithmetic. If I can only discriminate the referent of "I" in a context of use from other things by its property of uttering a specific token of "I", I may be completely in the dark what it means, for instance, to ascribe to it the mental property of proving a theorem of arithmetic. In order to ascribe this property to something I must know its persistence-conditions. Without sensitivity to these conditions it would be indeterminate whether I ascribe the property to an object or to a temporal part of it. And I can discriminate something by the property of being the current producer of an "I"-token without knowing anything about its persistence-conditions. The producer may be an automatic device, a human being etc.

[70] Similar things hold for the ascription of the physical property of weighing too much to an object. In order to ascribe this property to an object I must know what it means to discriminate it from other physical objects. I must be able to draw its physical boundaries in thought. The Idea constituted by the entertainment of the character of "I" does enable me to discriminate an utterer from other noise producing things in a context of utterance, but it does not give me the information about its identity-conditions needed to discriminate it from other physical objects. Thus, I cannot ascribe properties to the object which require sensitivity to the physical boundaries of an object for their ascription.

[71] 3. The Principle of Predicate Variability can serve as the starting-point of a more general argument against the thesis that the characters of indexicals are modes of presentation. According to this principle I can only think that a is F if I can also think that a is G, for every categorically adequate property of being G of which I have a conception. If I can think that a is F, I must be able to think that a is G, that a is H, ... . But then I must also be able think that a is F and G. This is the thought that a single thing (the same thing) is F as well as G. To grasp such thoughts one needs a conception of singularity that applies to a, i.e. one needs to know something about the identity- and persistence-criteria for a. One specifies such criteria by subsuming a under a sortal concept, for example the concept of a horse. Now the concept expressed by "producer of this "I"-token" specifies identity-criteria, but not identity-criteria that would enable us to grasp every thought we can grasp. I can grasp the thought that I was born in Australia and that I have blonde hair. The concept of sameness which allows me to grasp that the same object that was once born in Australia has the property of being blonde, must be one which applies to me from my birth to my current state. For instance, I am now the same human being I was back then in Australia. But the baby born in Australia is not the same producer of "I"-tokens than the man now. It did not produce "I"-tokens at all. A sortal term like "language user" does not enable us to specify what it would be to say of an arbitrary object w that it is identical with me. But this is required if my ego-thinking should enable me to grasp an indefinite variety of ego-thoughts in which different properties are ascribed to one and the same thing. Our ego-thinking can only confirm to the Principle of Predicate Variability if the Ideas we employ in ego-thoughts specify identity- and persistence-conditions that determined what it would be to say of an arbitrary object w that it is identical with me. Without fulfillment of this demanding condition our ego thinking cannot have the productive quality it has.15 Evans has a positive answer what one's "Ego"-Idea must be to answer these requirements: it must be the ability to distinguish oneself from other things by the property of being a person that occupies such-and-such a spatio-temporal location (Evans 1982: 209, 211). We need not decide whether Evans's view about this is correct. Sufficient for our purposes is the conclusion that the entertainment of the role of "I" does not enable us to grasp indefinitely many ego thoughts about ourselves.

[72] But let us not go too fast. In the situations outlined I can of course explain verbally that the thought I may express with my utterance of "I am born in Australia and have blonde hair" ascribes the properties of being born in Australia and having blonde hair to the one object referred to in the context of utterance by "I". Does this show that I grasp the thought that I was born in Autralia and have blonde hair by entertaining the character of "I" without any further substantial identification of the speaker as a so-and-so, most prominently as the person I am? If we accept with Evans that there should be some empirical restrictions on the ascription of the ability to grasp a proposition, the answer must be No. The verbal explanation is, as Evans puts it, "purely formal and insubstantial". I may give it and, for example, lack the ability to distinguish between situations in which the object referred to by "I" (in the relevant context of use) lacks the property of being born in Australia and those in which it has this property. In order to do this I would need to have some idea of what it would be to say of an object born in Australia that it is identical with the producer of the "I"-token (in the context c). But without further knowledge about the referent of "I" in c I do not know enough about this object to subsume it under more fundamental sortal concepts. But only the application of such concepts allows me to specify under which conditions the persisting particular a is identical with the persisting particular b. Compare: what does the claim that the object I thought of yesterday is the same object I thought of today amount to? Without more sortal information we cannot say. But if the object I thought of yesterday and the one I thought of today is the same material object, I can give an answer: there is a continous space-time path which connects the object I thought of yesterday and the one I thought of today and both fall under the same sortal concepts. That I may very well lack the information to give such an answer when I entertain the meaning of "I" shows that in this situation the ascription of the ability to grasp the thought that I am born in Australia and have blonde hair to is not justified.

[73] I take it that these are the reasons why Evans says about the relational property specified in the role of the indexical that it "plays no part in an explanation of what makes a subject's thought about himself, or the place he occupies, or the current time". Evans calls an Idea of an object m that enables a thinker only to ascribe a subclass of all properties he can conceive of to an object "inadequate" (Evans 1982: 148, 208). The characters of indexicals all rely on properties relating an object to an aspect of a context of utterance, to take another example, the time of the production of a token of "now". So we can generalize our last consideration: the Ideas constituted by the entertainment of the characters of indexicals cannot fulfill the Principle of Predicate Variability, entertaining them cannot amount to having adequate Ideas of the things referred to.

[74] At first sight there is an easy way out for the adherent of the view that characters can serve as modes of presentations in propositional thoughts. The producer of a certain "I"-token will fall under further and more fundamental sortal concepts than the "episodic" sortal concept of being a language user. Can we protect the account under discussion against the objections made to it by incorporating more sortal information into the character of an indexical? The first problem is to decide which sortal concept should be incorportated in the character of "I". Should the language user be discriminated by the sortal property of being a human being or the sortal property of being a person? But even if this problem can be resolved, another objection arises against the proposal to incorporate more sortal information into the character of the indexical. For instance, we cannot specify the character rule for "I" as follows:

(CI) Any token of the English word "I" refers to the person who uses it
without robbing Kaplan's claim that "it is natural to think of [character] as meaning in the sense of what is known by the competent language user" (Kaplan 1989a: 505) of its plausibility. According to (CI) it would be a priori true that the object referred to by a (literal and non-elliptic) use of the first person pronoun in a true utterance is a person. But that the object referred to by such a use of "I" in a true utterance is a person can sometimes be known only a posteriori. We understand what someone says with his words "I am here" fully, but it will take much more than a proper understanding of his words to determine that the utterer is a person. Moreover if (CI) would be the character of "I", every competent user would know that "I" refers in all its correct (literal and non-elliptic) uses to persons. According to this thesis utterances of "I don't know what I am, but at least I know I think" would show that the speaker either does not know the meaning of "I" or has a Cartesian concept of personhood. These are consequences hard to swallow. Therefore we must see "I" as governed by a sortally neutral reference rule. Such a rule will not determine the reference of the first person pronoun "in so far as it does not settle what kind of thing is using the pronoun, except in so far as the concept of a language user imposes some restriction" (Woods 1968: 570). And the sortal restrictions imposed by the concept of a language-user are rather liberal, for example, things that are not persons may use language.

[75] Let us now take stock. I have developed a reading of Evans's argument in which it establishes via the Constraints of Generality that entertaining the character of an indexical cannot amount to having an adequate Idea of the object indexically referred to. Now since our ego-thinking does conform to the Generality Constraint and consequently must incorporate adequate Ideas of ourselves, entertaining the character of "I" is not sufficient to think such ego-thoughts. Characters qualify as the linguistic meanings of indexicals, but they cannot be the ways in which we think of ourselves when we are thinking self-consciously.

[76] An inadequate Idea of an object is still an Idea of an object. So Evans's argument does not show that I cannot grasp a singular proposition by entertaining the role of a an indexical (demonstrative). But even if we can grasp singular propositions "containing" things by entertaining the roles of indexicals or demonstratives these ways of grasping them are uninteresting because they are not connected to the ways we in fact think indexically.

8. Conclusion

[77] Kaplan's Direct Reference theory may be seen as a significant contribution to the semantics of intensional languages, but this semantics cannot replace a theory of singular thought as Kaplan explicitly and others implicitly have assumed. The linguistic rules English connects with indexicals are not suited to play the role of modes of presentation if this notion is developed in a theory of propositional thought. Kaplan can take singular propositions to be utterance contents, but, as McDowell correctly points out, Kaplan's theory "locate[s] them with a purview of a discipline of 'semantics' further removed from than Russell's semantics was meant to be from aiming to delineate the contours of thought" (McDowell 1986: 138). Rather than replacing or refuting the main tenets of Neo-Russellian semantics Kaplan's claims show that he and the Neo-Russellians are involved in different projects. Acquaintance or its modern successor, the Discrimination Requirement, plays no role in the explanation of the truth-conditions of modal statements, yet a theory of reference which is mainly concerned with our understanding of singular terms cannot do without it.

Mark Textor
Philosophisches Seminar
UniversitŠt Hamburg


Almog, J., J. Perry, and H. Wettstein (eds.)(1989) Themes from Kaplan. Oxford: Oxford University Press.

Burge, T. (1986) "Individualism and Psychology". Philosophical Review 95: 3-45.

Campbell, J. (1984-85) "Possession of Concepts". Proceedings of the Aristotelian Society 85: 149-170.

----- (1994) Past, Space and Self. Cambridge, MA: MIT Press.

Cassam, Q. (1997) Self and World. Oxford: Oxford University Press.

Deutsch, M. (1989) "On Direct Reference". In Almog et al (1989).

Donnellan, K. (1979) "The Contingent A Priori and Rigid Designators". In Contemporary Perspectives in the Philosophy of Language, P.A. French, T. E. Uehling, H. K. Wettstein (eds.): 45-60. Minneapolis: University of Minnesota Press.

----- (1990) "Genuine Names and Knowledge by Acquaintance". Dialectica 44: 99-113.

Evans, G. (1982) The Varieties of Reference. John McDowell (ed.). Oxford: Oxford University Press.

----- (1985) Collected Papers. A. Phillips (ed.). Oxford: Oxford University Press.

Kaplan, D. (1969) "Quantifying In". In Words and Objections, D. Davidson, J. Hintikka (eds.). Dordrecht Reidel.

----- (1978) "Dthat". Reprinted in Meaning and Truth, J.L. Garfield, M. Kiteley (eds.): 212-234. New York: Paragon.

----- (1989a) "Demonstratives". In Almog et al (1989): 481-563.

----- (1989b) "Afterthoughts". In Almog et al (1989): 565-614.

McDowell, J. (1969/70) "Identity Mistakes: Plato And The Logical Atomists". In Proceedings of the Aristotelian Society 181-195.

----- (1982) "Truth-Value Gaps". In Logic, Methodology and Philosophy of Science VI, C. J. Cohen, J. Los, H. Pfeiffer, K.P. Podewski (eds.): 299-315. Amsterdam: North Holland Publishing Company.

----- (1986) "Singular Thought and the Extent of Inner Space". In J. McDowell, P. Pettit (eds.), Subject, Thought, and Context, 137-168. Oxford: Oxford University Press.

Peacocke, C. (1981) "Demonstrative Thought and Psychological Explanation". Synthese, 47: 187-217.

----- (1992) "Externalist Explanation", in: Proceedings of the Aristotelian Society 93, p. 203- 230.

----- (1997) "Metaphysical Necessity: Understanding, Truth and Epistemology", in: Mind 106, p. 521-575.

Perry, J. (1977) "Frege on Demonstratives", in: Philosophical Review 86, p. 474-497.

Prior, A.N. (1971) Objects of Thought P. T. Geach and A. J. P. Knny (eds.). Oxford: Oxford University Press.

Russell, B. (1910-11) "Knowledge by Acquaintance and Knowledge by Description". Reprinted in Mysticism and Logic, 1986, 200-221 London: Unwin.

Sainsbury, M. (1986) "Russell on Acquaintance". In Philosophers -- Ancient and Modern, G. Vesey (ed.): 219-224. Cambridge: Cambridge University Press.

Soames, S. (1995) "Beyond Singular Propositions". Canadian Journal of Philosophy 25: 515-550.

----- (1998) "The Modal Argument: Wide Scope and Rigidified Description". Nous 32: 1-22.

Stanley, J. (1997) "Names and Rigid Designation". In Companion to the Philosophy of Language, R. Hale, C. Wright (eds.): 555-585. Oxford: Blackwell.

Taschek, W. (1987) "Content, Character and Cognitive Significance". Philosophical Studies 52: 161-189.

Textor, M. (1998) ""Unbewu§te †berzeugungen, subdoxastische ZustŠnde, und die Allgemeinheitsprinzipien". In Bewu§tsein und ReprŠsentation. F. Esken, H.D. Heckmann (eds): 459-89. Paderborn: Schšningh.

Travis, Ch. (1994) "On Constraints of Generality". Proceedings of the Aristotelian Society 93: 165-186.

Wiggins, D. (1995) "Substance". In Philosophy -- A Guide through the Subject, A. C. Grayling (ed.): 214-249. Oxford: Oxford University Press.

Woods, M. (1968) "Reference and Self-Identification". Journal of Philosophy 65: 568-578.


1 My thanks go to the "Hamburg Circle"-Group for discussion. Special thanks go to Bill Brewer, Ben Hoefer, Wolfgang Kuenne and Mark Siebel for commenting on various drafts of this paper. (back)

2 Soames doubts whether knowledge of linguistic meaning is sufficient to entertain a singular proposition in the case of a sentence containing "dthat". Soames' doubts will be confirmed in sections 3 and 4. (back)

3 Evans lists as prominent examples Michael Devitt and Gilbert Harman. (back)

4 This thesis is elaborated and defended in Burge (1986) and Peacocke (1981) and (1992). (back)

5 In his "Afterthoughts" Kaplan favours a more careful version of this thesis in which "propositional attitudes reportable de re" is replaced by "propositional attitudes reportable de dicto". But it is difficult to see that this is a modification of Kaplan's position, since, for example, in propositional attitude sentences with proper names like "Peter believes that Hesperus shines brightly" the proper name will ensure that they have a de re reading. (back)

6 The cumbersome phrasing "demonstratively used demonstrative pronoun" is necessary because demonstrative pronouns can also be used anaphorically. I will use "demonstrative utterance" for an utterance which contains a so used demonstrative pronoun. (back)

7 Kaplan in Kaplan (1989b: 582ff) has revised this thesis. He now holds that the intention of the speaker determines the object referred to. In this paper I will stick to Kaplan's anti-intentionalist view. As I see it, the choice between the two views does not bear on the issues discussed here. (back)

8 For this generalization see Peacocke (1997: 534ff). (back)

9 The attentive reader will have noticed a tension in my representation of Kaplan's theory. Kaplan asserts that

1. the character of a directly referential term is a function whose value for a contexts of use is an intension, and

2. the semantical rule conventionally connected with a directly referential term is its character, and

3. the semantical rule of a directly referential term determines only a referent in a context of use.
Max Deutsch critizices this tension (Deutsch 1989: 173). He argues that one should drop 3.: Directly referential terms introduce rigid intensions into the proposition expressed. Instead I propose to see the tension due to the fact that the intuitive conception cannot be expressed in Kaplan's formal framework. But that the intuitive conception does not fit one theoretical framework does not show that is does not exist nor that it cannot expressed at all. Kaplan uses therefore the "metaphysical picture" of structured propositions to convey the basic idea of direct reference. (back)

10 K. Donnellan uses Kaplan's semantics for indexicals to argue against the thesis that genuine naming requires acquaintance (cp. Donnellan 1990). Because of the different requirements for naming (saying) and thinking Donnellan's and Kaplan's arguments are independent. (back)

11 There are good reasons to doubt that characters can fulfill this function; see (Tascheck 1987). (back)

12 Evans calls this principle the "Generality Constraint". I think it is more apt to speak of two Constraints of Generality. (back)

13 Travis (1994) tries to show that the Generality Constraints do not hold for propositional thinking. Textor (1998) hopefully shows that this critique is unsuccessful. (back)

14 Evans argues for this point in his paper "Identity and Predication" reprinted in Evans (1985). (back)

15 This is frequently emphasized by Evans; see (1985: 319). (back)

©1998 Mark Textor

EJAP is a non-profit entity. EJAP may not be printed, forwarded, or otherwise distributed for any reasons other than personal use. Do not post EJAP on public bulletin boards of any kind; use pointers and links instead. EJAP retains the right to continuously print all the articles it accepts, but all other rights return to the author upon publication.

EJAP is hosted at The University of Louisiana at Lafayette. Please send comments and questions to