Exports and Imports: Anaphora in Attitudinal Ascriptions

Exports and Imports: Anaphora in Attitudinal Ascriptions Tomis Kapitan Northern Illinois University Philosophical Perspectives Volume 8, edited by James E. Tomberlin (Ridgeview Publishers), pp. 273-292 I. Introduction Propositional attitudes are at the interface of several philosophical concerns with thinking, language, and reality. Some philosophers approach the attitudes with an interest in semantics, others with a concern about contents and the proper objects of cognitive states. For yet others, the attitudes are critical elements of a correct account of agency inasmuch as agents are responsible only for actions which are caused by the appropriate mix of intentions, beliefs, desires, etc.. Here, discriminations among contents are vital in specifying relevant causal mechanisms. Finally, the attitudes are studied from the standpoint of mind/body metaphysics, with a concern for what must exist in order for attitude ascriptions to be true, assuming, of course, that such talk is scientifically respectable. A good deal of fuss about the attitudes centers on reference and attributions of reference, stimulated by familiar problems of substituting apparently coreferential singular terms in attitude ascriptions. To illustrate; I recall my friend Robert's astonishment upon learning that the local priest, who lived in the parish house in our town, was also the venerated fire chief we had all seen in action. Prior to this discovery, one might have correctly reported one of Robert's beliefs with, (1) Robert believes that the priest lives in the parish house. but not with, (2) Robert believes that the fire chief lives in the parish house. The latter was false if taken as ascribing to Robert a disposition to assent to the content sentence 'the fire chief lives in the parish house' or acceptance of the proposition that the fire chief lives in the parish house. To be sure, Robert referred to the one who happened to be the fire chief, but not in those terms. Still, it might seem that (2) should follow from (1) and, (3) The priest is the fire chief. If not, one is faced with the problem of explaining the apparent breakdown of substitutivity in a manner consistent with one's perspectives on cognitive states, reference, and identity. Some seek a solution in a syntactical comparison of singular terms with quantifiers. For instance, by taking 'someone on the team' to fall within attitudinal scope in, (4) Henry believes that someone on the team is a thief. we report that Henry believes the generalization that someone on the team is a thief. Here, the quantifier phrase occurs internally in that the speaker uses it, and the entire content sentence within which it occurs, to reveal what he, the speaker, takes to be Henry's exact content. Quite a different interpretation is in order for, (5) Henry believes of someone on the team that he is a thief. 1

where the occurrence of 'someone on the team' is meant to be outside the scope of 'believes,' suggesting that Henry has some definite individual "in mind" albeit unspecified. The quantifier phrase occurs externally in (5), expressing only the speaker's manner of describing content without being used to reveal the exact content of the agent's. 1 Can a similar internal/external contrast be drawn for singular terms? Those who say yes take the de dicto ascription (2) as false if understood as imputing to Robert a manner of referring to the priest that he did not employ, i.e., qua the fire chief. But this is avoided in the de re ascription, (6) Robert believes of the fire chief that he lives in the parish house. in which we export 'the fire chief' to a position outside attitudinal scope. We thereby account for the sense in which (2) is true while distinguishing what is only the speaker's from what is properly attributed to Robert. 2 But this raises only more questions. What does such exportation achieve? Does it make sense to treat singular terms as having scope like quantifiers? How does the 'of' locution manage to secure a relevant difference between what (2) and (6) express so far as what is attributed? In particular, just how does the anaphoric 'he' in (6) differ in expressive content from its antecedent? These questions must be answered if the contrast between external and internal occurrences in accounting for attitude ascriptions is to be philosophically illuminating. This is the task of what follows. In it, attitudes will be taken as cognitive states whose contents are propositions, the latter being complex entities exhibiting various structures. Language is regarded as a tool for causing the apprehension of contents, as such, a tool for our own references as well as for communicating with others. Following Hector-Neri Castaneda (1977, 1989a), reference is first and foremost "thinking reference," with symbols deriving their expressive powers from the intentions of symbol-users, be these community-wide dispositions that underlie shared symbol systems or idiosyncratic intentions of an individual manifested in particular occasions of usage. Upon this observation is mounted an important methodological maxim: (M) To understand what a given sentence or discourse expresses, conceive it as embedded within attitudinal scope, viz., within an "I think," a "we think" or, better, an "it is thought that." 3 The advice is often neglected; we speak as though there were direct language-to-world connections, taking a sentence like 'the carpet is grey' to directly express a fact about objective mind-independent realities while overlooking the role of the interpreter whose expression this is. Yet, the sentence is obviously ours, yielding an interpretation or description of reality as we see it, that is, as filtered through our conceptual apparatus. If so, the study of propositional attitudes is a topic at the very core of the broader domain of philosophical semantics. 4 The partial account of attitude ascriptions offered herein is inspired by Castaneda's efforts to steer a middle course between the popular "Fregean" and "Russellian" alternatives on attitude ascriptions. Fregeans (McGinn 1983, Forbes 1987a, 1990) invoke a semantic duality by taking embedded singular terms to express their senses (representations, concepts) only, and not the referents which they customarily express in direct speech. In contrast, the Russellians (Richard 1983, 1987, Salmon 1986a, Soames 1988, 1990, Kaplan 1989) treat genuine singular terms as uniformly referring to individuals regardless of their embedments, regarding these individuals as 2

subjects in propositions whereas Fregeans take propositions to be constituted by senses. Castaneda blended these perspectives by taking individuals as characterized -- objects qua their identifying attributes -- to be propositional components. Embracing a cognitive approach to content, he held that one thinkingly refers to an item O by thinking of it, that is, by having accessible to one's consciousness a thought content with O itself in subject position. Singular reference, on his account, is always to individuals qua individuals, and is achieved only in the entire episode of thought wherein a unique individual is "picked out." 4 2. Exportation By 'exportation' is meant not the rule which is sometimes thought to sanction inferences from 'a believes that b is F' to '( x)(x=b and a believes that x is F),' but a procedure for revealing speaker's intentions concerning which terms are to be construed externally and which internally in attitude ascriptions. The claims for its virtues are familiar. It is thought to preserve the truth of a wide array of puzzling belief reports which paradigmatically block substitution, e.g., those involving the speaker's use of indexicals as in, (7) Poor Richard believes that I am wealthy where the externally occurring 'I' is not used with the intention to ascribe a first-person belief to Richard. Moreover, it anchors inferences that utilize principles of substitutivity and identity, e.g., the move to (2) from (1) and (3) provided that both 'the priest' and 'the fire chief' are read externally. To answer the aforementioned queries facing this strategy, additional moves are needed. One supposition that seems vital to explaining the internal/external contrast and, thus, to the exportation strategy, is that agents refer to items only via some mode of presentation. Typically, exporters endorse a principle of the following sort: (E1) Reference to an item O by an agent X is accomplished only via referential modes, each being a way of characterizing O through which X cognizes (thinks of, conceives, perceives, etc.) O, and at least one of which is an identifying mode whereby X picks out and distinguishes O from all else. 5 While some modes of presentation are properties of a referent, others are best classified as, just that, modes of presentation, that is, manners whereby referents are cognitively related the referring agent. Being this orange book, for example, may be a mode through which I refer to a certain object on the table, but 'this' and, perhaps, 'orange' apply to the item only in virtue of its causal relation to an experient subject, in this case, myself. A plurality of modes is evident here; in referring to this orange book on the table as my only copy of Tahafut at-tahafut, I might also think of it as an item on this table, as Averroes' greatest work, as a book, as my possession, or as a physical object. There might be a large number of modes that I associate with any given mode, even though I might not access them equally in a given act of reference. Further, an act can embody more than one identifying mode, e.g., when I report: "this orange book on my desk, my only copy of Tahafut at-tahafut, was purchased for twenty pounds." Grasping a mode that uniquely characterizes an item is a necessary, though not a sufficient condition of reference to that item. I certainly understand the words 'the first proposition Einstein wondered about on March 12, 1943' and might believe what is expressed by 'the first proposition Einstein wondered about on March 12, 1943 concerned Niels Bohr,' but I do not thinkingly refer to the proposition that satisfies this description. By the same token, I do not 3

refer to the first person Einstein spoke to on March 12, 1943 even though I might understand many descriptions which that person satisfies. The referential "picking out" of an item involves consciousness of it as it enters into the content of an experiential act (see note 4). Principle (E1) isn't enough to motivate exportation. Exporters who distinguish between external and internal occurrences are disinclined to accept free substitution of apparently coreferential terms, e.g, 'the fire chief' for 'the priest' in (1) to yield (2). In this, their position contrasts with Russellian approaches which permit substitution yet accept (E1). Terms occur externally to the extent that they express modes falling outside the agent's content, internally otherwise, in which case, exportation seems wedded to a further claim.: (E2) The referential modes through which an agent X refers to an item O constitute part of content of the propositional attitude within which X refers to O. This principle represents a real theoretical divide between Fregeans and Russellians. One initial point in its favor is its ability to accommodate some intuitive implications. For example, one who accepts the proposition that Cicero was a clever philosopher, and who associates with 'Cicero' the referential mode of being a Roman Senator, can legitimately infer from this proposition that some Roman Senator was a clever philosopher. Again, it seems obvious that that this orange book cost twenty pounds by itself entails that some book cost twenty pounds, or that 7 > 3 entails that some prime number is greater than 3, without additional premises predicating the properties of being a book or being a prime number, respectively, to the said referents. To be sure, the claim that these are entailments calls for additional theorizing about individuals and their essential properties. But to dismiss the alleged entailments as enthymemes, or as pragmatic features of utterances, thereby depriving singular terms of much of their inferential punch, is reminiscent of the questionable thesis that all entailments are logically valid or that all necessity is logical necessity (see Kapitan 1982). Refusing this restriction, the validity of these inferences can be explained by allowing their premises to contain the requisite content, e.g., Cicero qua Roman Senator, content that supports inference from that Cicero was a clever philosopher to that some Roman Senator was a clever philosopher just as much as it does to the more abstract proposition that some individual was a clever philosopher. With (E2) the significance of exportation emerges. In treating any attitude ascription, there is a duality of content to contend with, that expressed by the speaker and that which the speaker ascribes to the agent. If the speaker is to succeed in revealing the exact content of the agent's, then the speaker must cognize not only the items to which the agent refers, but, by (E2), the agent's referential modes as well. That is, ascription is governed by a principle of Shared Accessibility: (E3) Whatever content a person X ascribes to an agent Y must be content that is cognitively accessed by X, or, at least, accessible to X. In particular, if X ascribes to Y reference to an item O, then X must himself refer to O (Castaneda 1989a). However, the ascriber might wish to acknowledge a partial content sharing only. If Norman ascribes a propositional attitude to Sal by means of a content sentence S containing a singular term T, but does not take the identifying mode through which he himself refers with T as identical to Sal's, then Norman does not use T to specify an exact component within Sal's content. For this reason, T is said to occur externally in Norman's ascription. By 4

contrast, Norman's use of T is to be construed internally just in case Norman claims to specify not only Sal's referent, but also a mode through which Sal identifies this referent, and in such a case we may speak of a cumulation of Norman's reference with Sal's (Castaneda 1989a, 1990b). Only with an internal construal of a term can an ascriber claim to use it in specifying an ascribed propositional component, and to the extent that the content sentence contains any external occurrences then the exact proposition escapes. As a consequence, ascriptions with exported elements reveal only part of the agent's total content. Their content sentences do not express attributed propositions at all, but something significantly less. 3. Problems for Exportation A number of hurdles face attempts to consistently apply the exportation strategy and the external/internal distinction. Since there are not always syntactic or semantic indicators determining when external construal is permitted, context becomes the decisive factor, be this the surrounding text or other pragmatic features such as the speaker's intentions. This saddles ascriptive language with an annoying ambiguity (see note 1) and requires that disambiguation proceed with considerable care. Subtle distinctions are called for when one psychological verb is embedded within another, e.g., (8) Castaneda thought that Columbus believed that Castro's island was China. for here we face the task of carefully sorting out what the speaker is attributing to whom. Finally, since there remains the breakdown of substitution within attitudinal scope, the sense in which exporters can claim to save substitutivity or Leibniz's law remains to be explained. Two types of problem, one dealing with substitutivity and the other with anaphora, pose particularly stiff challenges to the claim that exportation sheds any light on attitude ascriptions. Let us first examine two cases that turn on substitution. The Distinguished Janitor Problem (Tienson 1987). Suppose a man, Alfred, who comes across a difficult problem which he thinks the Distinguished Professor of Metaphysical Logic can solve. Letting 'B' abbreviate 'believes that,' 'S' the predicate 'can solve the problem', 'a' the singular term 'Alfred,' and 'p' the description 'the Distinguished Professor of Metaphysical Logic,' we have B a Sp. Since the professor exists, we can generalize to ( x)b a Sx. Now Alfred happens to be on friendly terms with the janitor of the building but does not believe that he can solve the problem, in which case we have both B a Sj and ( x) B a Sx. Unbeknownst to Alfred, the Distinguished Professor, with a sufficient change of clothing, moonlights as a janitor. So, p=j. Substitution within attitudinal scope being blocked, we cannot infer B a Sp to produce a contradiction. But given that the janitor exists, we can apply the rule of exportation to B a Sj to yield, (9) ( x)(j=x & B a Sx). Allowing that B a Sp is okay with 'p' given internal construal, we may still export and infer a de re counterpart, namely, (10) ( x)(p=x & B a Sx) given that p exists. Since p=j, we can certainly substitute in (9) to get, (11) ( x)(p=x & B a Sx) which no exporter should protest against. However, (11) is inconsistent with (10). The upshot? 5

Problems with substitution continue to plague attitudinal ascriptions even after we have exported in deference to the contrast of external and internal occurrences. The Phone Booth Problem (Richard 1990). Richard objects to the "happy face" exportation strategies which allow ascriptions to fall short of specifying total content. He presents an independent "context-hopping" argument favoring a Russellian point of view whereby substituting coreferential names doesn't change the proposition which that a sentence determines. Suppose person A is speaking to person B on the phone while also watching B, across the street, talking in a phone booth. A doesn't realize that the person to whom he is speaking is the very one he is watching. Seeing that the person in the phone booth is in danger, A says to himself, (12) I believe that she [pointing across the street] is in danger. Not taking his addressee, B, to be in danger, A would not accept, (13) I believe that you [speaking into the phone] are in danger. even though A's 'she' and 'you' are coreferential. Moreover, for those who adopt the exportation strategy and its tenets (E1)-(E3), (13) seems false if 'you' is construed internally. But suppose A is within B's field of vision and, seeing A point at her with alarm, B says into the phone, (14) The person watching me believes that I am in danger. Taking B seriously, A can now report what B has just told him with, (15) The person watching you believes that you are in danger. Obviously, if (14) is true so is (15). Since A is watching B then, (16) I am the person watching you. would also be true when spoken (or entertained) by A. Yet, (13) follows from (16) together with (15) by means of the following pattern of inference: (S) a believes that S a = b b believes that S Granting (S) which even exporters should admit short of repudiating Leibniz's law and substitutivity into it follows that substitution into attitudinal scope must be permitted. Since A's 'she' and 'you' are coreferential, then if (12) is true we have evidence that substitution does preserve truth since we have found that (13) is also true. Assuming that a similar argument can be run for other problematic cases, there are grounds for a Russellian approach to attitude ascriptions and against the need for exportation. So, exporters can export till they have depleted their stocks; they have still not blocked inference to the paradoxical ascriptions which their program was designed to avoid. The second type of problem deals with an issue that was already raised in section 1 above, namely, the interpretation of embedded anaphors. How do the external de re readings of ascriptions differ from their internal de dicto counterparts so far as what is ascribed? If the anaphoric 'he' in (6) is coreferential with its antecedent, how does (6) really differ from (2)? On the simple coreferential reading, 'he' expresses what its antecedent 'the fire chief' does, in which case the difference between (2) and (6) is stylistic at best. If it is something more, then an explanation is in order. This worry does not disappear when (6) gives way to a quantified variant with 'he' interpreted as a bound variable, as in, (17) ( x)(x = the fire chief & Robert believes that x lives in the 6

parish house). But if the very thing Robert believes to live in the parish house is identical to what the speaker specifies with 'the fire chief,' how can (17) fail to inform us of anything less than Robert's total propositional content, viz., that the fire chief lives in the parish house? How does it manage to ascribe anything less than what (2) does, and how is it anything more than a "baroque version of Russellianism" (Richard 1990, 104)? Taking the anaphor as a bound variable engenders yet a further problem. Suppose I hear someone whom I correctly identify as the richest philosopher say, 'I am wise.' How do I report what I take him to assert? Neither the de dicto nor de re readings of, (18) The richest philosopher believes that I am wise. seem correct, since not only does 'I' expresses first-person reference by the speaker, me, but the belief is not about me. Nor can I offer, (19) The richest philosopher believes that the richest philosopher is wise. with the description taken internally, since the richest philosopher might not believe that he is the richest. An internal construal seems required: (20) The richest philosopher believes of the richest philosopher that he is wise. The question of what 'he' expresses for me arises anew. If it expresses no more than what the antecedent already gives, we are back at the very problem we began with. On the other hand, by quantifying to yield, (21) ( x)(x = the richest philosopher and the richest philosopher believes that x is wise). we correctly abstract from the agent's cognitive perspective (Soames 1989, 197), but are in danger of exporting too much. From his words, I have judged that the richest philosopher is making a first-person reference, but this is not captured in (21). The latter falls short of the expressive power of ordinary English within which we may report the belief as follows: (22) The richest philosopher believes that he himself is wise. where 'he himself' is a quasi-indicator (Castaneda 1967, 1980, 1989a), viz., a device for attributing indexical reference. Here's the problem. Does 'he himself' occur externally or internally in (22)? It is not merely external if I, or anyone else, uses it to report how the richest philosopher thinks of himself, namely, in a first-person way. But nor is it purely internal; I cannot refer to a distinct individual in a first-person way and, therefore, I do not specify the exact proposition the richest philosopher asserted with the content clause 'he himself is wise.' Apparently, 'he himself' is both; internal in revealing referential modes, yet external in its semantic dependence upon its antecedent. The question is whether the simple external/internal distinction supplies a rich enough mechanism for accommodating such mixed attributions and, consequently, whether exportation is of any help at all. To make matters worse, this problem is not confined to quasiindicators, for the same can be said of 'he' in (6); though reliant upon an externally occurring antecedent, the pronoun can be used to attribute a referential mode to the speaker, specifically, being a male human, in which case it too has an internal aspect which is lost with exportation. 7

4. Restoring the Happy Face Let us now attempt to resolve these problems in favor of the exportation strategy. One reason the Phone Booth Problem is so interesting is that there is no reason for thinking that the exported counterparts of (13) like, (23) I believe of you that you are in danger. or the more exact, (24) You = she and I believe that she is in danger. or the quantified variant, (25) ( x)(x = you and I believe that x is in danger). should differ in truth-value from the de dicto reading of (13). Part of the difficulty is due to the presence of indexicals which reflect only the speaker's mode of reference from a unique perspective. 6 As such, giving 'you' an internal construal, seems in order. Any speaker who selfascribes present beliefs presumably knows how he or she is representing the relevant items; "who is better placed to know the content of the ascribee's beliefs?" (Richard 1990, 131). But as we have seen, allowing (13) to be true on a de dicto reading is to give up the motivation behind the exportation strategy. Can a de re reading of (13) be defended? Why not? The person that A refers to as you is such that A believes that person -- to whom he also refers to with 'she' -- to be in danger. It may be true that a minimally rational person who believed any of (23)-(25), that is, who is willing to self-ascribe the content ascribed by (23)-(25), should automatically believe (13), and so the de dicto reading of (13) seems correct. This is a pragmatic feature of indexical usage; it would be a psychological aberration for one to accept any of (23)-(25) without also accepting (13), but it does not follow that the truth of any of (23)-(25) requires the truth of the de dicto reading of (13). Have we escaped the force of Richard's argument given the truth of (15) and (16) and the validity of (S)? Plainly, (15) itself is true only on the external reading; A's report of what B told him over the phone embodies A's reference to B, a reference which A has no reason to attribute to whatever he refers to with 'the person watching you.' Consequently, an internal reading of 'you' in (15) is inappropriate, and we must take (15) as, (26) ( x)(x = you and the person watching you believes x is in danger). where A, the speaker, makes no assumption about how B is referred to by the person whose belief B reports over the phone. Now the problem with substitution evaporates, for given (16), substituting 'I' for 'the person watching you' in (26) yields the true (25). Form (S) is not at stake here; exporters need not and should not question its validity. 7 Tentatively, we may reinstate exportation, and do so with a smile. 5. Sameness and Identity Still, we wonder about what is accomplished; both the Phone Booth Problem and Distinguished Janitor Problem overlap with the problem of the embedded anaphors; how do we interpret (23)-(26)? Just what is attributed with external construals, and how do embedded anaphors differ in expressive content from their antecedents? If a is identical to b, and if an agent X believes that a has some property F, then why shouldn't we say that X believes that b has F? Isn't the Russellian affirmative answer to this question the only one which makes sense? Further theoretical steps must be taken. One might reject Leibniz's Law, but besides 8

being a desperate solution, it would undermine the motivation behind exportation. There are, of course, the Fregean alternatives which exploit the sense/reference distinction (e.g., Forbes 1989, 1990), but like all representational accounts of understanding, they deprive us of direct cognitive contact with the concrete realm of actual entities. This is hard to stomach when we think about cognitive contact with our loved ones, a hard shot of strong liquor, or running our fingers through our favorite copy of Tahafut at-tahafut. We are in the world with concrete entities, and we are so as experient subjects into which they intrude, like it or not. The cumbersome sense/reference dichotomy is strained in any case if we adopt maxim (M) with a Fregean understanding (E1)-(E3); if all language falls within the scope of an implicit "I think" or "it is thought that," we never move beyond senses in interpreting language. 8 The usual Russellian theories are also deficient as cognitive theories of content. While we might agree that some of the propositions we think contain individuals, it is another matter to explain what these individuals are. If we reject referential modes as content, as direct reference theories are wont to do, we are left with the individual itself as the subject of a proposition. But what is this, say, Mont Blanc, when I say 'Frege and Russell disagreed about Mont Blanc'? The whole massive individual "with its snowfields" (Frege 1904, 56)? You might come to this conclusion if you live up high enough in the French Alps, though ultimately, the view is untenable. My act of thinking seems to be here in space-time, but the whole massive individual, that infinitely complex mountain weighing millions of kilograms that I believe to exist in the external world, is not here in my state of consciousness or qualifying my act of thinking, despite my reference to it. So, if it is not entire massive individuals -- the values of the bound variables in gravitational laws -- that we thinkingly refer to, what other options do we have? Uniformed matter? Bare particulars? Search for an alternative is in order. A third way is available to exporters. It agrees with the Fregean that substitution failures inside attitudinal scope refute a simple identification of content with an external structure (fact, situation) consisting of massive individuals exhibiting properties and relations. Contents must be more fine-grained, more manageable in order to be objects of finite acts of thinking. While accepting (E1)-(E3), it is Russellian in allowing bonafide individuals as referents, thus, as propositional components, though only when properly clothed. Inasmuch as we refer to Mont Blanc we do so qua some identifying mediator, e.g., qua the highest mountain in Europe, or qua that steep rise in ground over there. Just as we do not refer to Mont Blanc nakedly, without identifying modes, so we do not refer to it in toto, though we may believe that an entity of this sort exists (see note 8). There are different varieties of this third way. Meinongian theories and Castaneda's Guise Theory constitute one family, Whitehead's metaphysics another. Whatever the final account, let us adopt the terminology of Castaneda 1989a in calling the immediate object of thinking reference -- the object qua mode(s) -- a strict referent. It is the item "in mind" (Donnellan 1966 and Kaplan 1968), the thing "objectified" in us (Whitehead 1929), with distinctions among strict referents being equivalent to distinctions among associated identifying modes. This seems to be the most straightforward non-fregean way of accommodating the exporter's (E2). Once this step is taken, here's the immediate corollary. If strict referents are objects qua modes, then what Robert thought of through his referring use of 'the priest' was distinct from what he thinkingly refers to by means of 'the fire chief.' By (E3) this is also true for 9

the speaker of (1)-(3). Consequently, (3) is false if taken as an identity claim. Still, we can acknowledge a "congruence" in meaning of the two terms (Suppes 1991), that is, a sense in which the priest is the same as the fire chief. Assertions to the effect that a is the same as b, or that a is b, are informative precisely because they are not identities, but, rather, are expressive of a different sort of equivalence relation among distinct entities. Which entities? The strict referents, the objects qua referential modes, for only these are subjects of graspable propositions. How are strict referents to be differentiated? How is the "sameness" of strict referents to be analyzed? Answers will inevitably require developed species of our suggested third way, though, presumably, the first will be met by citing distinctions among referential modes or, given the plurality of modes usually present in reference, determinable-determinate hierarchies of such modes. Answers to the second might also resort to modes, specifically, to the idea that the identifying modes of same strict referents characterize a single concrete entity, viz., what we have referred to as an complex massive object. Barring a more extensive discussion of that latter, however, this is likely to be but a necessary condition. We can, however, make a distinction that tightens the parallel between the exportation of externally occurring singular terms and quantification into. Could a speaker use (5) to attribute a reference to Henry? Yes and no. Unlike (4), the speaker of (5) purports that Henry is referring to a specific individual even though the speaker does not identify Henry's strict referent. In such a case, let us say that the speaker makes an indefinite attribution of reference to Henry, revealing, at best, the type of strict referent Henry referred to. Similarly, the speaker does not attribute belief in a specific proposition to Henry, but, at best, belief in a type of proposition. A speaker who specifies Henry's strict referent, on the other hand, makes a definite attribution of reference, requiring him to employ a singular term T to refer to item O via an identifying mode M while claming that Henry also refers to O via M. An attribution by means of singular term T is indefinite if the attributer asserts that there is a strict referent of the attributee's which is the same as what he himself refers to with T, though the latter need not be identical to the attributee's. Therefore, unlike identity, sameness is not governed by Leibniz's law; terms referring to what are distinct, though same, strict referents are not universally intersubstitutable. 6. Tentative Solutions to the Remaining Problems Now we can approach anew the remaining problems for exportation. First, attributions of reference with exported singular terms are indefinite, and, therefore, the embedded anaphors are variables bound by external quantifiers. Second, the ranges of these variables are constituted only by strict referents which are the same as the strict referent which the speaker refers to with the exported singular term. Hence, it is wrong to analyze (6) as (17), but instead as, (27) ( x)(x the fire chief & Robert believes that x lives in the parish house). where ' ' represents sameness and the variable ranges over the speaker's strict referents. The solution to one aspect of the anaphora problems is immediate; the embedded anaphors are bound variables ranging over strict referents, and hence, they do not express exactly what their antecedents express. The other case concerning substitutivity, The Distinguished Janitor Problem, is resolved by noting that 'p=j' is false in the specified context where 'p' and 'j' express different modes, even if p j. As such, (11) cannot be derived from (9) and there is no 10

inconsistency. The problem concerning quasi-indicators, or, more generally, anaphors which embody both internal and external elements, requires an additional explanation. As bound variables, their range is externally specified by the exported term expressing the speaker's strict referent, yet they also carry an internal element determining their values as items referred to qua some specified non-identifying or generic mode. This places, in effect, a further restriction on the ranges of these variables. Formally, the restriction is expressible by an operator on variables falling within attitudinal scope, specifically, by a superscript to indicate an ascribed referential mode (construed, thus, internally). For example, taking 'x' to range over strict referents, then 'x M ' ranges over strict referents which are referred to qua male human. (27) can be further refined as, (28) ( x)(x the fire chief & Robert believes that x M lives in the parish house). Let us refer to such a superscript as an intensional monitor which discriminates among admissible substituends for bound variables, though no assumption is made that the speaker might actually be able to provide those substituends. Similarly, where the substituends of 'x I ' express strict referents which are thought of qua I, that is, qua self, then (22) is, (29) ( x)(x the richest philosopher and the richest philosopher believes that x I is wise). On the other hand, if the richest philosopher uses the sentence 'you are wise' in addressing the speaker, then the latter could report this with, (30) ( x)(x myself and the richest philosopher believes that x YOU is wise). 9 Castaneda's use of indexical types to function as common nouns expressing generic indexical modes is adopted here (Castaneda 1989a, 76), allowing us to speak of the referential modes of being an I, being a this, and the like. It is the irreducibility of such modes that grounds the irreducibility of indexical reference, yet it is their generic character that underlies the ascribability of indexical reference. It is the need to import some element of what had previously been exported that leads to the introduction of intensional monitors. The occurrences of 'he himself' in (22) and 'he' in (6) are partly internal in conveying modes which the speaker attributes to the agent. The superscripts are imports -- or, had we never been tempted to export in the first place, they are domestic productions integral to the attributed content. What's more, the contrast between internal and external occurrences must now be relativized not only to speaker's intentions, but to referential modes. Def. 1. A speaker's use of singular term T in an attitude ascription A is internal relative to M iff the speaker intends to ascribe to the agent of A reference via M; it is external relative to M iff the speaker does not intend to ascribe to the agent reference via M. Every term used to attribute content is partly internal in revealing some referential mode of the agent's, even if a very abstract mode like being an individual. On the other hand, some terms might be wholly internal: Def. 2. If a speaker X attributes to an agent Y reference to an item O by means of an occurrence of singular term T within an attitude ascription A, then that occurrence of T is 11

wholly internal for X iff every referential mode by which X refers to O through that occurrence of T is also a mode that X attributes to Y through A. Then we have a further principle: (E4) X attributes to Y an E-attitude towards a proposition P by means of an attitude ascription of the form 'Y E's that S' only if X expresses P by means of 'S' and every occurrence of a singular term in 'S' is wholly internal for X in that use of 'Y E's that S'. If it turns out that all reference requires indexical modes of presentation, then every singular attributive term is partly external in reflecting only a mode which is used by the speaker. The way is now clear for disambiguating ascriptions with multiple psychological operators. Suppose Mary says of Bill, (31) The big slob thinks that I love him. after hearing Bill talk about her with the sentence 'She loves me.' John, hearing what Mary said about Bill, can report what she says as follows: (32) Mary believes of Bill that he thinks that she herself loves him. Here, 'she herself' reflects Mary's way of characterizing Mary, not Bill's, that is, John is ascribing a first-person reference to Mary. Suppose, that with (32) John also wishes to characterize Mary as attributing a first-person reference to Bill while attributing to Mary a reference to Bill via the mode being a BIG SLOB. I know of no elegant English prose for expressing this, but our superscripting device permits the following rendition: (33) ( x)( y)(x Bill and y Mary and Mary believes that ( z)( w)(z y I and w x BS and x BS thinks that z loves w I )). On the other hand, if John also refers to Bill as the Big Slob, (31) can be reported with, (34) ( x)(x Mary and Mary believes that ( y)( z)(y x I and z the Big Slob and the Big Slob thinks that y loves z I )). Which modes are attributed to whom and by whom with these sorts of construction? The rule governing the use of superscripts is this: (R) A superscript on a bound variable embedded within n psychological verbs in an ascription S expresses a referential mode which is (1) attributed to the agent of the nth psychological verb in S, and (2) either attributed by the speaker if n=1, or claimed by the speaker to be attributed by the agent of the n-1th psychological verb in S if n>1. For example, the sole occurrence of 'y I ' in (33) expresses a mode attributed to Mary by the speaker, while the superscript in 'w I ' expresses a mode, the very same generic mode, attributed to Bill by Mary. If John does not claim that Mary attributed to Bill self-reference, then (32) would be, (35) ( x)( y)(x Bill and y Mary and Mary believes that ( z)(z y I and ( w)(w x BS and x BS thinks that z loves w))). Here then, is an partial account of attitude ascriptions in which the de dicto mode is restored to its rightful position as the more accurate mechanism for revealing propositional content. The de re mode, far from revealing the res with which an agent is en rapport, reflects the speaker's caution in making an attribution and, at best, is a device for showing the type of proposition attributed. The contrast does not support a weighty ontological distinction between 12

two different families of attitudes, yet is essential to an accurate display of attitudinal ascriptions. 13

REFERENCES Almog, J., J. Perry and H. Wettstein, eds., 1989. Themes From Kaplan (Oxford). Anderson, C. A. and J. Owens, eds., 1990. Propositional Attitudes (Center for the Study of Langauge and Information). Bach, K., 1982. "De Re Belief and Methodological Solipsism," in A. Woodfield, ed., Thought and Object (Oxford). Bealer, G., 1992. "A Solution to Frege's Puzzle," manuscript. Boer, S. and W. Lycan, 1986. Knowing Who, (MIT). Brown, M., 1992. "Doing it With Mirrors," manuscript. Burge, T., 1977. "Belief De Re," The Journal of Philosophy 74, 338-362. Castaneda, H-N., 1967. "Indicators and Quasi-Indicators," American Philosophical Quarterly 4, 85-100. Castaneda, H-N., 1975. "Identity and Sameness," Philosophia 5. 121-150. Castaneda, H-N., 1977. "On the Philosophical Foundations of Communication: Reference," Midwest Studies in Philosophy, II: Studies in the Philosophy of Language (Minnesota), 165-186. Castaneda, H-N., 1980. "Reference, Reality, and Perceptual Fields," Proceedings and Addresses of the American Philosophical Association 53, 763-822. Castaneda, H-N., 1986. "Self-Profile," in Hector-Neri Castaneda, in J. Tomberlin 1986, 3-137. Castaneda, H-N., 1989a. Thinking, Language & Experience (University of Minnesota Press). Castañeda, H-N., 1989b. "Direct Reference, Realism, and Guise Theory," in J. Almog et al, eds., 105-144. Castañeda, H-N., 1990a. "Indicators: The Semiotics of Experience," in Jacobi and Pape 1990, 57-93. Castañeda, H-N. 1990b. "Indexicality: The Transparent Subjective Mechanism for Encountering a World," Nous 24, 735-749. Chastain, C., 1975, "Reference and Context," Minnesota Studies in the Philosophy of Science, VII, 194-269. Crimmins, M., and J. Perry, 1989. "The Prince and the Phone Booth: Reporting Puzzling Beliefs," Journal of Philosophy 86, 685-711. Devitt, M., "On Removing Puzzles About Belief Ascriptions," Pacific Philosophical Quarterly 71, 165-181. Donnellan, K., 1966. "Reference and Definite Descriptions," The Philosophical Review 75, 281-304. Donnellan, K., 1989. "Belief and the Identity of Reference," in P. French et al, 275-288. Evans, G., 1985, Collected Papers (Oxford). Forbes, G., 1987a. "Indexicals and Intensionality: A Fregean Perspective," The Philosophical Review 96, 3-31. Forbes, G., 1987b. "A Dichotomy Sustained," Philosophical Studies 51, 187-211. Forbes, G., 1989. "Cognitive Architecture and the Semantics of Belief," in 14

P. French et al, 84-100. Forbes, G., 1990. "The Indispensability of Sinn," The Philosophical Review XCIX, 535-563. Frege, G., 1904. "Selections from the Frege-Russell Correspondence," in Salmon and Soames, 1988. French, P., T. Uehling and H. Wettstein, eds., 1989. Midwest Studies in Philosophy Contemporary Perspectives in the Philosophy of Language (Notre Dame). Higginbotham, James, 1991. "Belief and Logical Form," Mind & Language 6, 345-369. Jacobi, K. and H. Pape, eds., 1990. Das Denken und die Strucktur der Welt (Berlin: de Gruyter). Kapitan, T., 1982. "On the Concept of Material Consequence," History and Philosophy of Logic 3, 193-211. Kapitan, T., 1991. "Review Essay: Thinking, Language and Experience," Philosophy and Phenomenological Research LII, 203-214. Kaplan, D., 1969. "Quantifying In," in Words and Objections, D. Davidson and J. Hintikka, eds. (Reidel). Kaplan, D., 1989. "Demonstratives, An Essay on the Semantics, Logic, Metaphysics, and Epistemology of Demonstratives and Other Indexicals," in J. Almog et al, eds., 481-564. King, J., 1987, "Pronouns, Descriptions, and the Semantics of Discourse," Philosophical Studies 51, 341-363. Kripke, S., 1979. "A Puzzle About Belief," in A. Margalit, ed., Meaning and Use (The Magnes Press), 238-283. McGinn, C., 1983. The Subjective View: Secondary Qualities and Indexical Thoughts (Oxford). McKay, T., 1988. "De Re and De Se Belief," in D. F. Austin, ed., Philosophical Analysis (Kluwer, 1988), 207-217. McKay, T., 1991. "Representing De Re Beliefs," Linguistics and Philosophy 14, 711-739. Millikan, R. G., 1991. "Perceptual Content and Fregean Myth," Mind C, 439-59. Neale, S., 1990, "Descriptive Pronouns and Donkey Anaphora," The Journal of Philosophy 87, 113-150. Owens, J., 1990. "Cognitive Access and Semantic Puzzle," in Anderson and Owens, eds., 147-174. Perry, J., 1979. "The Problem of the Essential Indexical," Nous 13, 3-22. Perry, J., 1983. "Castaneda on He and I," in Tomberlin, ed., 15-42. Quine, W. V., 1955. "Quantifiers and Propositional Attitudes," reprinted in The Ways of Paradox (Harvard, 1976). Recanati, F., 1990. "Direct Reference, Meaning, and Thought," Nous 24, 697-722. Richard, M., 1983. "Direct Reference and Ascriptions of Belief," Journal of Philosophical Logic 12, 425-452. Richard, M., 1987. "Quantification and Leibniz's Law," The Philosophical Review 96, 555-578. Richard, M., 1989. "How I Say What You Think," in P. French et al, 317-37. Richard, M., 1990. 15

Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them (Cambridge). Saarinen, E., 1986. "Castaneda's Philosophy of Language," in J. Tomberlin 1986, 187-214. Salmon, N., 1986a. Frege's Puzzle (MIT). Salmon, N., 1986b. "Relexivity," Notre Dame Journal of Formal Logic 27, 401-429. Reprinted in Salmon and Soames 1988, 240-276. Salmon, N., 1989, "How to Become a Millian Heir," Nous 23, 211-220. Salmon, N. and S. Soames, eds., 1988. Propositions and Attitudes (Oxford). Schiffer, S., 1990. "The Mode-of-Presentation Problem," in Anderson and Owens, eds., 249-268. Searle, J., 1983. Intentionality (Cambridge). Soames, S., 1988. "Direct Reference, Propositional Attitudes, and Semantic Content," in Salmon and Soames, eds., 197-239. Soames, S., 1990. "Pronouns and Propositional Attitudes," Proceedings of the Aristotelian Society (1989/1990), 191-212. Stich, S., 1983. From Folk Psychology to Cognitive Science (MIT). Suppes, P., 1991. Language for Humans and Robots (Blackwell). Tienson, J., 1987. "An Argument Concerning Quantification and Propositional Attitudes," Philosophical Studies 51, 145-168. Tomberlin, J., ed., 1983. Agent, Language, and the Structure of the World: Essays Presented to Hector-neri Castaneda with His Replies (Hackett). Tomberlin, J., ed., 1986. Hector-Neri Castaneda (Reidel). Wettstein, H., 1989. "Cognitive Significance Without Cognitive Content," in J. Almog et al eds., 421-454. Whitehead, A. N., 1929. Process and Reality (Macmillan). Zalta, E., 1988. Intensional Logic and the Metaphysics of Intentionality (MIT). 16

NOTES 1. Some are uncomfortable with the ambiguity that affects attitude ascriptions if the contrast between internal and external occurrences of terms is taken seriously (e.g., Stich 1983, chp. 6; Richard 1990, chp. 3, section 1). At the level of quantifiers, at least, the distinction seems impregnable and fully representable, assuming that quantification into attitudinal contexts makes sense at all. Why do we think that (4), with the quantifier taken internally, might be true even though (5) is not? Consider. Upon returning from the showers Henry may have discovered that his wallet is missing. Believing that only those on the team were in the locker room at the time he comes to believes that some team member or other is a thief. As a result, he becomes angry and begins to curse his team members even though he has not singled out any individual and believed of him that he stole the wallet. Had he done so, his behavior would have been different. Quantifier scope is precisely how we record these crucial differences in attributing content, and we need not endorse Quine's distinction between two types of believing to justify a scope distinction (Quine 1955). A uniform treatment of belief is desirable, and misgivings about quantification into attitudinal contexts are not shared here. 2. This use of 'internal' and 'external' in treating the reference of singular terms is taken from Castaneda 1980, 1989a, 1990b. Others have emphasized that the de re/de dicto distinction is best applied not to belief reports wholesale, but to the occurrences of terms vis-a-vis the scope of various operators, in our case, psychological verbs, e.g., Zalta 1988, 171, and Brown 1992. I follow Searle 1983 and Richard 1990 in taking the de re/de dicto contrast as a distinction between different kinds of ascription, not between differents sorts of belief. The labels 'propositionally transparent' and 'propositionally opaque' are also used to make the internal/external contrast in Castaneda 1980 and 1989a. This usage is clarified in Kapitan 1991, 208. 3. This maxim echoes Kant's principle that an "I think" must be capable of accompanying all my representations (Critique of Pure Reason B132). A similar version of (M) appears in Castaneda 1975, 128. I prefer to employ the prefix 'it is thought that' rather than 'I think' since indexical terms like 'I' and 'we' express a degree of self-consciousness which neither need nor can accompany every meaningful use of language. The unnamed speaker or thinker need not be an "I," for an I or "self" exists only when a thinking being is itself the object of reflexive consciousness (Castaneda 1990c). 4. Castaneda 1989a, 33-35, 53, 66 for more on thinking reference. For his full approach to attitudes and attitude ascriptions see Castaneda 1977, 1980, 1989b and 1990a. 5. It is allowed that a given act of reference may involve a number of referential modes. For example, when I refer to someone with the term 'that man drinking a martini in that corner,' I might also be referring to him through the modes of being that man in that corner, being a man, being a living thing, etc. Qua mode, each "mediates" the presentation of O to the agent, even though some of them might fail to actually characterize the object, e.g., the man might not be drinking a martini at all. However, some modes must be "successful" in that they permit the referential "picking out" to occur, especially the indexical ones, viz., being that man in that corner. Schiffer 1990, 255, expresses skepticism that recourse to modes of presentation helps us to understand any sort of reference, but his regress argument against such modes as "individual 17