I. Introduction
The debate between Simulation Theory (ST) and Theory Theory (TT) concerns three questions, not always clearly separated in the literature.
- 1) What process do we use in attributing mental states to others?
- 2) How do we conceive of mental states? What is the nature of our mental concepts?
- 2a) How do mental concepts develop in childhood?
- 3) What is the nature of mental states?
I will present a conceptual argument for a simulationist answer to (2). Given that our conception of mental states is employed in attributing mental states to others, a simulationist answer to (2) supports a simulationist answer to (1). I will not address question (3). Answers to (1) and (2) do not yield an answer to (3), since (1) and (2) concern only our actual practices and concepts. For instance, an error theory about (1) and (2) would say that our practices and concepts manifest a mistaken view about the real nature of the mental. Finally, I will not address question (2a), which is an empirical question and so is not immediately relevant to the conceptual argument that is of concern here.
Simulationists maintain that, in order to explain or predict another's behavior, we imaginatively project ourelves into her situation. For instance, when I observe Sally reaching for a glass of water, I imagine performing a similar action and find that my action would be motivated by thirst. I then hypothesize that Sally is thirsty. Similarly, when I observe Sally eating salty pretzels I predict that she will accept a proferred glass of water, because I imagine the effects eating those pretzels would have on me. This is, in rough outline, the ST answer to question (1). (As we shall see, simulationists agree on little more than this rough outline, for they disagree about how simulation allows one to attribute states to others.) And simulationists believe that this process reflects our conception of mental states (question (2)): at a minimum, mental states are conceived as the sort of states which can be identified through this process.
Theory theorists deny that simulation is typically, or fundamentally, involved in attributing mental states to others. On the TT model, other-attributions are subserved by a theory, a largely implicit set of generalizations relating behavior to mental states (question (1)). And this theory may well exhaust our mental concepts. Even if we do occasionally engage in simulation, simulation has no bearing on our fundamental conception of mental states (question (2)).
The debate between ST and TT is now waged primarily on empirical turf. Theory theorist critics of ST claim that ST conflicts with empirical evidence, including developmental data and the prevalence of certain inferential fallacies in everyday reasoning. ST's proponents have, in response, tried to show that the empirical evidence is in line with ST or even favors ST. And they have offered empirical reasons to doubt that each of us knows, even implicitly, a theory of mind suitable to ground our impressive ability to understand others.(1) This debate has been productive. It has forced disputants on both sides to clarify and refine their positions, and it has led them to develop increasingly sophisticated strategies for interpreting empirical evidence. However, I believe that the focus on empirical evidence has obscured an important feature of ST, namely, that the original and best grounds for accepting ST are conceptual. This paper will reveal the importance of conceptual matters in this debate, by providing a new conceptual argument for ST.
The paper proceeds as follows. In Section II, I articulate and develop a conceptual thesis--the Indexicality of the Mental, or IM--which strongly supports ST. Unsurprisingly, this thesis is controversial. While conclusively establishing IM would require a paper of its own, Section III provides powerful reasons to accept it. Section IV demonstrates that IM is at odds with TT. In Section V, I argue that the three leading simulation theorists-- Alvin Goldman, Robert Gordon, and Jane Heal--are each committed, by their larger views, to IM. This section also explains how commitment to IM at least partially motivates their endorsement of ST. Finally, Section VI shows that the key differences among simulation theorists stem from the fact that they provide divergent accounts of the mental's indexicality.
If successful, my discussion shows how important indexicality is to this debate. IM provides strong support for ST, and sheds light on what unifies these three ST views while also illuminating the deep differences between them. More broadly, my discussion shows that it is a conceptual issue which forms the substantial core of simulation theorists' agreement, and which differentiates them from theory theorists. It therefore implies that the conceptual dimensions of the "ST vs. TT" debate deserve renewed attention.
II. The Indexicality of Mental Concepts
Here is the thesis which I will use to ground a conceptual argument for ST.
IM: Concepts of mental states are indexical, in that possessing a mental concept requires the capacity to make direct indexical reference to a state which satisfies that concept.
According to IM, concepts such as [belief] and [desire] are shaped by the capacity for direct indexical reference to beliefs and desires. (Throughout the paper, I use square brackets to indicate concepts.) Hence, while we may at times refer to beliefs descriptively, such as "that which led the Federal Reserve Chairman to raise interest rates", thinking of Greenspan's belief that inflation was rising as a belief requires the capacity to refer to a belief with a direct indexical. I now turn to explicating IM; the next section will provide reasons for accepting this thesis.
I must first specify the scope of IM. Since I am using IM to support ST, I limit the scope of IM to those mental state-types which are relevant to the "ST vs. TT" debate. Claims of indexicality are relatively familiar regarding non-propositional phenomenal concepts, such as [pain] and [tickle], and I will address these in passing. But as the "ST vs. TT" debate centers on propositional attitudes such as beliefs and desires, my discussion will center on showing that IM applies to these attitudes.
I claim, then, that possessing the concepts [belief] and [desire] requires the capacity to make direct indexical reference to beliefs and desires. Note that this statement concerns the propositional attitude modes belief and desire rather than particular propositional attitude types, such as the belief that snow is white or the desire for lemonade. I do not claim that possessing the concept [belief that snow is white] requires the capacity to make direct indexical reference to a belief that snow is white. Obviously, one grasps [belief that snow is white] only if one has the concept [belief]. As long as IM is true of the attitude mode belief--i.e., as long as possessing the concept [belief] requires the capacity to directly indexically refer to some belief or other--grasping a particular belief will require the capacity to directly indexically refer to some belief or other.
The next order of business is to explain the kind of reference at work in IM. According to IM, it is not just any sort of indexical reference that is implicated in possessing the concept [belief], but specifically direct indexical reference. I can refer indexically to Greenspan's belief, as "the state now causing Greenspan to raise interest rates"; yet my capacity to refer indexically in that way seems irrelevant to my concept [belief]. "The state now causing Greenspan to raise interest rates" secures its referent by use of a description, whereas the sort of indexical reference which does contribute to my concept of belief is, instead, direct indexical reference. Direct indexical reference is unmediated by descriptions; direct indexicals secure their referents independently of any description that the referents satisfy. (I develop this notion of direct indexical reference in the following section.)
IM's use of direct indexical reference has a notable consequence. Since the starting-point for reference is the referrer's own thought, and direct reference requires that one's access to the referent is unmediated, one can directly refer only to one's own states, or to states which share a "mental location" or "perspective" with the referring thought.(2) IM thus implies that an adequate conception of a mental state requires a first-person grasp of the state-type. For instance, IM implies that to understand "Sally desires lemonade", one must also grasp what it would mean for "I desire something" (or perhaps, using direct indexicals, "there is a desire here", or "this is a desire") to be true. This directly indexical thought is available only to one who can conceptualize the desire from the first-person perspective.
Basically, then, IM implies that an adequate conception of a mental state must draw on the capacity to refer to a state of that type with a direct indexical; it thereby implies that an adequate conception of a mental state requires conceptualizing it from the first-person perspective. I have offered a relatively superficial description of IM; this will be deepened by examining the grounds for accepting IM. Let us turn, then, to arguments for IM.
III. Establishing IM
The role of qualitative states
One possible argument for IM should be addressed straight away.
Every mental state, including every propositional attitude, has essential qualitative features; so properly conceptualizing a mental state requires knowing what it's like to instantiate the state. Knowing what it's like to instantiate a state requires conceptualizing it from the first-person perspective, which in turn secures the capacity to make direct indexical reference to a token of that state-type. Hence, IM is true.This line of reasoning may establish IM for mental states that have essential qualitative features, such as pains and tickles. But there are grounds for accepting IM, as regards propositional attitudes, that are wholly independent of any qualitative features which propositional attitudes have. If IM rested on a qualitative view of propositional attitudes, it would be unpalatable to most simulation theorists, since most deny that every propositional attitude has qualitative features.(3) (Hence "imagination", with its qualitative connotations, is a somewhat misleading term for "simulation": simulation is not assumed to be qualia driven, or to involve mental imagery.)
While I believe that there are at least some mental states which have essential qualia properties, this paper will remain neutral on that question. But if there are such states, IM is likely to be true of them. For one can adequately conceptualize a state which is defined by qualia properties only if one conceptualizes those qualia properties, and to conceptualize those qualia properties just is to conceptualize the state from the first-person perspective--to grasp, at least roughly, "what it's like" to instantiate that state. So IM squares with the existence of essentially qualitative mental states, without presupposing that there are such states.
The discussion of qualitative states raises another claim which may seem to follow from IM but which, in fact, does not. On one obvious construal, the capacity to make direct indexical reference to an essentially qualitative state constitutes a special access to evidence regarding that state-type. IM is neutral not only as to whether there are any essentially qualitative states; it is also neutral as to whether subjects enjoy special access to any kind of evidence, qualitative or not, regarding state-types which they instantiate. Perhaps the subject's capacity to refer to a belief with semantic directness doesn't provide her with any evidence about the nature of beliefs, but instead simply constitutes, in part, her concept [belief]. In any case, IM is neutral about how direct indexical reference contributes to one's grasp of mental concepts.
B. The argument for IM: Beliefs
There are obvious reasons for taking IM to apply to mental states with essential qualia properties. But how would the capacity for direct indexical reference, or for conceptualizing a state from the first-person perspective, contribute to a grasp of propositional attitudes, which are (on most views) non-qualitative states? The remainder of this section is devoted to answering this question.
Many philosophers maintain that, for rational subjects, there is a basic asymmetry between first-person and third-person awareness of propositional attitudes involved in practical reasoning, including both beliefs and desires. According to this view, part of possessing the concepts [belief] and [desire] is being disposed to self-attribute a belief that p, or a desire that q, whenever one has them--or, at least, whenever they play a promiscuous role in practical reasoning. And crucially, this disposition to self-attribute a belief underwrites one's competence in attributing beliefs to others. Christopher Peacocke endorses this view, which he expresses as follows.
[T]here is a sense in which the concept of belief is a first-person concept. The sense is not that there is something it is like to have a belief--there is not. . . . [It is rather that possessing the concept of belief] requires the thinker to have the capacity to ascribe beliefs to himself and relates mastery of third-person predications to that capacity to ascribe beliefs to himself. (Peacocke 1992, 164)Peacocke thus holds that possessing the concept [belief] requires a first-person grasp of the state-type belief, because he maintains that our concept [belief] is shaped by the process a rational person uses to self-attribute beliefs. At the same time, he denies that beliefs have essential qualitative properties.
On this view, rational subjects must be capable of recognizing their own beliefs and desires. For--according to Peacocke, Sydney Shoemaker (1988), André Gallois (1996), and others--a subject incapable of such recognition would run the risk of pragmatic inconsistency. The most famous instances of pragmatic inconsistency are utterances described by G.E. Moore, such as "It is not raining but I believe that it is raining." Such statements are logically consistent, of course, given the possibility of false belief. But they are pragmatically inconsistent. In a sincere assertion one aims to express one's beliefs, and willingness to sincerely assert this statement shows a failure to grasp the relation between an expressed belief and a self-attributed belief content. This failure in turn manifests the subject's failure to grasp [belief], according to the view at hand. Any rational subject who sincerely uttered this statement would, then, be pragmatically inconsistent; her willingness to utter a Moore-paradoxical statement would show that she lacked the concept [belief].
The truth of IM does not require that we generally enjoy self-knowledge; it requires only that one who grasps [belief] is capable of the sort of reference which self-knowledge typically involves. The threat of Moorean pragmatic inconsistency stems from one's capacity for direct indexical reference to a belief. "It is not raining but the tallest person in the room believes that it is", which (if "the tallest person in the room" is used as a genuine description) refers to the alleged believer descriptively, is non-paradoxical even if the speaker is the tallest person in the room. For she may intend to refer to someone else, and may be unaware that she is herself taller than her intended referent. In that case, the speaker does not intend to self-attribute the belief that it's raining. Similarly, "It is not raining but the person now reflected in the mirror believes that it is", which uses a non-direct indexical ("the person now reflected in the mirror"), is non-paradoxical even if the speaker is the person now reflected in the mirror. For she could fail to recognize herself as the person reflected in the mirror, a person opening an umbrella. In these cases, the statements are false but non-paradoxical. Only attributions of belief which exploit directly indexical access to the belief run the risk of pragmatic inconsistency.
Now for grasping [belief] it is not enough that one avoid Moore-paradoxical statements; after all, one could avoid these simply by resolving never to indexically refer to beliefs. Grasping [belief] requires that one avoid Moore-paradoxical statements because one understands, if only implicitly, the incoherence they involve. This requirement gives us the first premise in an argument for IM, regarding [belief].
This argument is clearly valid; opponents of IM will likely reject Premise 2. Defending Premise 2 requires closely examining the work done by direct indexical reference.
Direct indexical reference is necessary to generate Moore paradoxes, since it is only such reference which guarantees that the speaker cannot (rationally) fail to recognize that she is the subject of the attribution ("I believe that it is raining"). The reason that direct indexicals guarantee this, on my view, is that only direct indexical reference is unmediated by a mode of presentation. Francois Recanati describes a case which illustrates the mediating role ordinarily played by modes of presentation.
Just as a ship cannot both possess and lack a single property, being a steamer, a belief cannot both possess and lack a single property, being currently instantiated by me. (Alternatively, we could say: I cannot simultaneously possess and lack the property believes that p.)Consider the utterance "This ship <pointing to a ship through one window> is a steamer but this ship <pointing to a ship through another window> is not a steamer." The speaker is not irrational even though she says of the same ship both that it is and that it is not a steamer. She is not irrational because she does not realize that there is only one ship; and she does not realize this because she thinks of the ship under two different (psychological) modes of presentation. In other words, we have to posit two different (psychological) modes of presentation of the reference, one corresponding to each token of the expression "this ship" in order to make sense of the utterance. (Recanati 1990, 711)
Now if a speaker can rationally both deny p and self-attribute the belief that p, then she must think of the believer (or the belief) under two different "psychological modes of presentation". We saw above some ways in which this could happen: e.g., she could think of the believer as the tallest person in the room, or she could think of the belief as that which is causing her (the person reflected in the mirror) to carry an umbrella. As long as the speaker's relation to herself or her belief is mediated by a mode of presentation in even one of the Moorean conjuncts, the paradox is defused.
In the case of directly indexical reference, there are two possible positions about modes of presentation. Either direct indexical reference occurs without any mode of presentation, or the mode of presentation is identical to the referent and hence plays no truly mediating role.(4) I will not decide between these. The crucial point is that, unlike the ship case, there cannot be a mediating mode of presentation at work in direct indexical reference. So the reason why someone cannot comprehendingly, rationally and sincerely utter "it is not raining but I believe that it is raining" is this: the belief in question, first contradicted and then self-attributed, is not mediated by any psychological mode of presentation.
The main premise of the above argument for IM is Premise 2: Understanding that it is incoherent to say "Not p but I believe that p" essentially requires the capacity to make direct indexical reference to some belief. I have defended this premise as follows. The pragmatic inconsistency of Moore statements rests, on my diagnosis, on the fact that the (expressed and self-attributed) belief is grasped independently of any mediating psychological mode of presentation. The lack of any mediating psychological mode of presentation ensures the capacity for direct indexical reference to the belief--indirect or merely descriptive reference would render the statement non-paradoxical. So even implicitly understanding the paradoxical nature of these statements requires the capacity for direct indexical reference to some belief or other.
C. The argument for IM expanded: desires and intentions
While this discussion has focused exclusively on beliefs, the capacity for direct indexical reference also plays a key role in grasping desires, and for broadly similar reasons. Consider this statement: "Bring me some lemonade, but I don't want you to." (I am indebted here to Charles Siewert.) Like the Moore paradoxical utterances, this statement trades on a tension between what is expressed by the directive "bring me some lemonade", and the self-attribution--in this case, the speaker self-attributes a lack of desire. Assuming that by saying "I don't want you to" the speaker disavows any desire that lemonade be brought, someone who sincerely uttered this would fail to adequately grasp the concept [desire]. (The paradox requires this assumption, to exclude non-paradoxical cases of conflicting desires like the following. I know that lemonade is dangerous for my weak-willed diabetic friend, and I believe that if I don't drink it, he will. So I want it drunk by me, and hence want it brought over. Yet I hate the taste of lemonade, and so in a sense I strongly desire that it not be brought over. This case allows a non-paradoxical gloss of the statement, but only by limiting the range of desires disavowed with "I don't want you to". In one sense, I do want you to bring it to me.)
As with [belief], the danger of pragmatic inconsistency is present only when no psychological mode of presentation mediates the subject's grasp of the desire. This unmediated grasp of the desire is available only to the desiring subject, and hence only to one capable of direct indexical reference to the desire. (There is no threat of pragmatic inconsistency in "Bring me some lemonade, but the person now reflected in the mirror doesn't want you to".) Hence, since grasping [desire] requires understanding the pragmatic inconsistency of certain utterances, and that understanding in turn requires the capacity for direct indexical reference to some desire or other, grasping [desire] requires a capacity for direct indexical reference to some desire or other. So standing in a direct indexical relation to a desire, and being guided by the pragmatic constraints which that relation imposes, is crucial to grasping [desire].
Richard Moran provides a different sort of pragmatic consideration which supports IM. Drawing on Sartre, Moran argues that there is a fundamental difference between a first-person and a third-person perspective on one's own states, including intentions. While "I resolve to stop gambling but I doubt that I will" is not a strict paradox on a par with a Moore statement, for Sartre and Moran there is a pragmatic tension between resolving to carry through a course of action, on the one hand, and recognizing that one is unlikely to act in accordance with one's resolution, on the other. "What I am aware of concerning myself empirically, as a facticity [as that exhaustively grasped from a third-person perspective], cannot substitute for what I am committed to categorically [as that which requires a first-person perspective]; commitments I have simply in virtue of having any beliefs about the world or any decisions about my action." (1997, 150) On this view, adequately conceptualizing intentions requires understanding the role of agency in intending; that understanding requires grasping intentions from a first-person perspective; and that grasp secures the capacity to refer to intentions with a direct indexical.(5)
On the views I have sketched, adequately conceptualizing beliefs, desires, and intentions requires conceptualizing these attitude types from the first-person perspective, the perspective of one who is committed to some beliefs, one who has some desires, one who owns some intentions. To take a merely third-person (descriptive, non-indexical, or indirectly indexical) stance towards these attitude types is to miss something crucial about the nature of the attitudes.
I have not conclusively established IM. But I have strongly motivated it, by showing that it fits a familiar picture of qualitative states; that it is entailed by a plausible diagnosis of the pragmatic inconsistency involved in Moore-type paradoxes; and that it captures the deep intuitive contrast between self and others described by Moran. My next task is to show that IM provides reason to favor ST over TT.
IV. How IM favors ST over TT.
In this section I examine ostensible points of tension between TT and IM and show that, while some apparent conflicts between them are resolvable, TT ultimately cannot do justice to IM.
According to TT, possession of mental concepts uses the same sorts of cognitive resources which are required to possess non-mental concepts. In particular, TT holds that "an individual's mastery of the concepts of the propositional attitudes is . . . constituted by his or her tacit knowledge of . . . a theory embedding these notions." (Davies and Stone 1995c, 2) So if TT's answer to our question (2) is correct, mental states are not conceptualized by any special means, relative to non-mental states. We conceive of mental states as theoretical entities, on a par with (other) scientific postulates.
Here is one reason TT may appear at odds with IM. Since theories are usually construed as sets of propositions, TT seems to imply that our propositional beliefs about mental states exhaust our mental concepts. (Heal has suggested, in conversation, that she takes TT to be committed to the exhaustiveness of propositional knowledge.) This is in tension with IM if the indexical nature of mental concepts implies that mental concepts are non-propositional in nature--i.e., if it implies that possessing mental concepts is possessing some irrreducible knowledge how, rather than possessing expertise which is reducible to knowledge that propositional attitudes are thus-and-so. The claim that a true grasp of mental concepts requires an ability, or knowledge how, resonates with some versions of ST.(6) If TT construes the grasp of mental concepts as purely propositional, and if indexical knowledge is irreducibly knowledge how, then TT cannot accommodate the indexicality of mental concepts.
This apparent source of tension between TT and IM is, I think, merely apparent. First, there seems no reason why TT must be committed to the claim that propositional beliefs about the mental exhaust our mental concepts. TT does claim that mental concepts are on a par with non-mental concepts, but there are no obvious grounds for denying that non-mental scientific concepts can involve irreducible knowledge how. More importantly, even if TT must construe mental concepts as propositional, it does not immediately follow that such knowledge cannot be indexical. A simple example of indexical knowledge which appears to be propositional is this: "It is now six p.m.".(7)
There is, however, a more promising argument to show that TT is incompatible with IM, one which also exploits TT's assimilation of mental concepts to non-mental, indisputably "third-person" concepts. This alternative allows that propositional knowledge may be indexical, but denies that an adequate grasp of non-mental properties requires conceptualizing them from the first-person perspective, and so it denies that having purely propositional knowledge of non-mental properties requires the capacity to refer to their bearers with a direct indexical. Here is the argument.
Grasping ordinary non-mental concepts--the concept [star] or [liquidity], say--doesn't require conceptualizing stars or liquids from the first-person perspective, or effecting direct indexical reference to them. One can understand [star] and [liquidity] without being able to conceive of being a star or being liquid. More to the point, one cannot indexically refer to a star, as "this", without such reference being mediated by some psychological mode of presentation: perhaps an implicit description such as "the source of the light now striking my retina". So the continuity TT envisions between concepts of the mental and concepts of the non-mental implies that, since direct indexical or first-person reference to instances isn't required for non-mental concepts, it isn't required for mental concepts. This is at odds with IM, and so TT cannot accommodate IM.In evaluating this argument, we should first note that TT is consistent with the claim that a capacity for some direct indexical reference is necessary for having ordinary (non-mental) scientific knowledge. Perhaps possessing ordinary scientific knowledge requires that one can distinguish here (this world or time or spatial region) from there. David Lewis implicates indexicals in ordinary scientific knowledge when he claims that all de re reference to objects is grounded in indexical self-reference. (Lewis 1983; compare Castañeda 1966, Perry 1979.) For instance, according to Lewis, when one thinks "the North Star is bright" one refers to the North Star only because, by thinking that thought, one implicitly self-attributes the property of occupying a world where there is a bright North Star. Since on Lewis' view all de re reference is mediated by directly indexical self-reference, the need for directly indexical self-reference does not distinguish a grasp of the mental from a grasp of non-mental items. TT can, then, accommodate the claim that properly conceputalizing a belief requires being able to engage in some direct indexical reference. (Lewis himself endorses TT.)
But while Lewis's claim entails that one who refers (de re) to a belief is capable of some directly indexical self-reference, IM goes further than this. IM says that one who fully grasps [belief] must be capable of direct indexical reference to some belief. It is this further implication of IM that TT cannot accommodate. This is because TT assimilates concepts of the mental to concepts of the non-mental, and for some--probably, for every--non-mental property P, one can possess a concept of P without grasping what it would mean to exemplify P oneself. Hence, one can grasp a non-mental property P without being in a position to effect direct indexical reference to any instance of P.
Some may object to my argument about the domain to which we can directly refer, claiming that we can directly refer even to ordinary physical objects such as rocks and trees. I have two responses to this objection. First, as explained above, the notion of direct reference I'm using is that which is tied to psychological modes of presentation. So the claim that one can refer directly even to rocks and trees is either very implausible (since our reference to rocks and trees is psychologically mediated by their perceptible properties), or it draws on a linguistic notion of direct reference which allows psychologically mediating descriptions to be a part of linguistic directness.
Second, even allowing direct indexical reference a wider domain, TT cannot accommodate IM. For TT holds that the capacities used to grasp the mental are generally the same as those used to grasp non-mental items. And any view about reference will deny that we can directly refer to some non-mental items, such as electrons. Electrons are theoretical posits, never directly observed. Hence, assuming that some people grasp [electron], a grasp of [electron] does not rest on the capacity to effect direct indexical reference to electrons. According to TT, propositional attitudes are no different from electrons on this score--they are theoretical posits, accepted for their power to make sense of observations. TT entails, then, that your understanding of "Sally desires lemonade" is entirely independent of whether you can make direct indexical reference to some desire, just as your understanding of "the White House is partially composed of electrons" is independent of whether you can make direct indexical reference to some electron. Hence, TT is at odds with IM.
It is a laudable aim which moves theory theorists to claim that mental concepts are analogous to non-mental concepts: they want to improve the prospects for a simple, elegant ontology. Their answer to our question (2), regarding the nature of mental concepts, is meant to pave the way for an answer to question (3), regarding the nature of the mental itself. The overarching hope is to show that the mental does not form a special, sui generis kind, fundamentally different from the non-mental. While theoretical simplicity is a worthy goal, it does not outweigh explanatory adequacy. If the arguments of the last two sections are correct, TT's construal of mental concepts cannot capture a basic feature of our mental concepts. There are thus clear conceptual grounds for favoring ST's answer to question (2).
V. How IM motivates Gordon, Heal, and Goldman to accept ST.
The prospects for finding any substantive unity among the three leading advocates of ST as to the relation between simulation and concepts appear dim. For Gordon, Heal, and Goldman differ on seemingly fundamental points. First, they differ in how they see the relation between ST and mental concepts: while Gordon takes ST to bear directly on our concepts of mental states, and Heal believes that ST is (in her special sense described below) an a priori matter, Goldman claims that ST is a purely empirical theory which doesn't essentially concern mental concepts. Second, of the three only Gordon holds that engaging in simulation does not require the prior possession of mental concepts. Third, only Goldman maintains that introspection is essentially involved in possessing mental concepts. I now show that, despite these disagreements over the relation between simulation and concepts, an implicit commitment to IM partially motivates Gordon, Goldman, and Heal to accept ST.
(a) Gordon
Gordon is, of the three, most explicitly committed to IM. This commmitment is embodied in his statement that "an understanding of mental states as such [requires] the ability to reconceptualise such 'objects' as the pain in the foot ... as having a mental location." (Gordon 1996, 19) For the notion of locatedness is an indexical notion, and grasping this notion requires the capacity for direct reference to a mental location. This position fits well with Gordon's "radical" simulationist claim that simulation is required for attributing genuinely mental states to other beings, that is, for representing others as thinkers. In his words, "even our ability to grasp the concepts of mind and the various mental states depends on our having the capacity to simulate others." (Gordon 1996, 11) What renders Gordon's simulationism "radical" is the non-accidental nature of this dependence. So taking the first-person point of view is not merely a useful heuristic in understanding others, for Gordon; it is absolutely essential to seeing another as a mental being, a seat of mental states. Gordon is thus drawn to the Kantian view that "it is only by performing an egocentric shift and taking the point of view of x ... that one represents x as a thinking being, and thus as having mental states." (Gordon 1995, 56) And this "egocentric shift" enables one to make direct indexical reference to the other and her states. On Gordon's view, then, one must be capable of direct indexical reference in order to represent others as a thinkers, that is, as instantiating genuinely mental states. Clearly, Gordon accepts IM.
(b) Heal
According to Heal, we have no choice but to assume that those we interpret are rational. Otherwise, we would be totally unable to determine how mental states are connected to each other and to behavior, and hence we could not predict or explain behavior. But the rationality we unavoidably attribute is not an ideal rationality, nor is it derived from mind-independent facts about what sorts of inferences qualify as rational. Rather, each of us understands rationality through our own inferential practices. Because one's purchase on rationality depends on one's own reasoning practices, in attributing mental states to others we must "think with" them, that is, we must simulate their relevant states so as to determine the outcome which instantiating those states would have for us.
Heal's rationality assumption thus supports her version of ST, which she labels "co-cognition". She expresses the co-cognition thesis as follows. "It is an a priori truth that thinking about others' thoughts requires us, in usual and central cases, to think about the states of affairs which are the subject matter of those thoughts, i.e. to co-cognize with the person whose thoughts we seek to grasp." (Heal 1998b, 484) The use of "a priori" here will suggest, to many, that this is a conceptual truth about the mental. But Heal is suspicious about conceptual claims. This suspicion, which is based on her doubt about a strict analytic/synthetic distinction, leads her to qualify her use of "a priori". For Heal, the a priori status of a statement lies in the fact that "we are thoroughly at a loss to describe realistically or in any detail how we would carry on intellectually if we could not rely on it." (ibid., 480) Her further remarks show that abandoning such a statement would, at the very least, result in a sharp "mutation" of the concepts originally at issue. And if abandoning a statement about the mental would cause a "mutation" in the original concepts, then the truth of the statement would seem to be entailed by the nature of the original concept. Heal's construal of the a priori thus allows her co-cognition thesis to be conceptual or very nearly so. Her deeper commitments thus imply that Heal takes co-cognition to at least approximate a conceptual truth about the mental.
As I read her, then, the line of reasoning which leads Heal to accept co-cognition as a quasi-conceptual truth is as follows. It is central to our concept of thought that grasping others' thoughts requires seeing them as rational creatures; our grasp of rationality is fundamentally a first-personal grasp; hence, it is central to our concept of thought that fully understanding others' thoughts requires grasping those thoughts from the first-person point of view. Moreover, Heal cannot abandon the quasi-conceptual status of the co-cognition thesis without sacrificing a consequence of her view which she takes to be one of its central virtues, namely, that it allows for other-knowledge without analogical inference. If the notion of rationality is only contingently (or, as she would say, empirically) tied to the idea of mentality, then to ascribe a particular mental state based on simulation one must infer, by analogy, that the other is rational--that is, that the other's reasoning patterns tend to follow one's own. The need for analogical inference can be sidestepped only if it is a priori that other mental beings are rational.
As with Gordon, it is the indexicality of the mental that explains Heal's belief that only a first-person understanding can adequately capture the mental. She claims that, ultimately, to represent a thought as a thought is to directly and indexically represent oneself. "Lack of conscious indexical referring thoughts would be lack of sense of oneself as a perceiving, acting and spatio-temporally located being." (Heal 1997, 636) This view, that indexical reference is necessary to secure a sense of self as a locus of thought and action, echoes Moran's view cited above. Accepting ST is, for Heal, a strategy for remaining loyal to the mental's indexicality (or, as she sometimes puts it, the fact that our access to it is irreducibly demonstrative).
(c) Goldman
Goldman denies that ST answers conceptual questions, asserting instead that the truth of ST is an empirical matter. That is, he sees ST merely as a promising answer to our question (1), What process do we use in attributing mental states to others? Still, Goldman suggests a conceptual basis for his acceptance of ST when he observes that the truth of ST would nicely suit his larger project of defending "the first-person approach to the understanding of mental concepts." (Goldman 1989, 94) He is more explicit about mental concepts in a recent unpublished manuscript. "My hypothesis is that mental-state concepts such as belief, desire, and so forth are understood at least in part in terms of non-dispositional characteristics of conscious experience, characteristics that can be introspected by the subject of the experience." (Goldman unpublished, 6) Crucially, introspecting these characteristics is not only possible, but also necessary for fully grasping the mental-state concepts. "What introspectivism says is that acquiring the concepts of mental representational states involves the task of latching these complex conceptual structures [i.e., conceptual structures of the sort TT postulates] onto the right sorts of conscious characteristics". (ibid.) While mental states may be as TT describes them, TT cannot secure the connection, required for possessing truly mental concepts, between these descriptions and conscious characteristics.
Goldman's introspectivism entails that possessing mental concepts necessarily depends on exercising introspection. Since what introspection affords is direct indexical access to the bearer of mental properties, a commitment to IM is, at the very least, the best way to explain Goldman's introspectivism. And while he denies that ST is itself a conceptual claim, his acceptance of ST appears motivated by his commitment to IM, a conceptual claim, combined with his introspectivism.
I have argued that Gordon, Heal, and Goldman all believe that mental concepts have an indexical component, and that this belief motivates, in one way or another, their acceptance of ST. I now turn to consider the step from IM to ST. How, according to ST's proponents, does ST do the explanatory work made imperative by the truth of IM? That is, how does simulation secure indexical concepts?
How simulation secures indexical concepts: introspectivism, semantic ascent, and co-cognition.
(a) Gordon and Goldman
Goldman's introspectivism bears a close affinity with the traditional "argument from analogy" regarding the existence and nature of other minds; it is therefore the most historically familiar account of how simulation secures mental concepts. Gordon and Heal reject introspectivism precisely because they deny that our attributions of mental states fundamentally rely on analogical inference. But important differences exist between Gordon's and Heal's anti-introspectivist versions of simulationism. In this section I show how each of these three simulationists accounts for simulation's role in securing indexical mental concepts.
Goldman sees the first-person perspective as essentially bound up with the self. This means that, for him, any understanding that relies on the first-person perspective relies on introspection (i.e., on self-understanding), even if what is being introspected is merely a simulation or pretense. The identification of the first-person (or direct indexical) perspective with a perspective one has on oneself, combined with IM, leads Goldman to accept introspectivism about mental concepts.
Gordon, by contrast, maintains that IM is compatible with a non-introspectivist view. This is because he rejects the identification of the first-person perspective with a perspective one has on oneself, in favor of the claim that, by "recentering my egocentric map" onto another person's position, I can achieve a first-personal grasp of the other's perspective. (Gordon 1995, 60)
[The other] becomes in my imagination the referent of the first person pronoun "I" .... And I, RMG, cease to be the referent of the first person pronoun.... (ibid., 55)
This is a striking claim. It explains how Gordon's non-introspectivist view can share, with introspectivism, the idea that a first-person grasp of a mental state is necessary for a mental state concept. The special mode of access which simulation provides is a mode one can use to grasp others' states as well as one's own. It thereby fails to be introspective, since "[a]s the term 'introspection' is commonly understood, one can introspect only one's own mental states." (ibid., 57) So Gordon believes that, while access to mental states as mental is conceptually distinct from access to non-mental states, this special mode of access isn't introspective.
The thesis that one can grasp another's perspective in a first-personal way makes room for a non-introspectivist view. But this thesis is compatible with the claim that analogical inference from introspection is a legitimate way, or even the most legitimate way, to attribute others' states. Gordon uses an additional premise to defend his non-analogical view, namely, the claim that the process of ascribing mental states to others is conceptually prior to any introspective process. On his account, I understand my (actual) self as a subject of mental states only by contrasting it with an already conceptualized other (or with counterfactual possibilities).
If ... we are speaking of genuine, comprehending ascriptions of belief, ascriptions made with the understanding that the beliefs ascribed may be false, then present-tense self-ascriptions ought to come last of all. For these require ... taking a vantage point outside one's present perspective, whether by simulating another or by memory demotion, and then reflecting back from that vantage point on what one now counts as fact pure and simple and relegating it to a perspectival status. (ibid., 62-3)
In other words, it is only by leaving one's home perspective that one is in a position to recognize that it is a perspective, and thereby to grasp the notion of mental locatedness. Hence, simulating others necessarily precedes grasping one's own mental states, as such.(8)
Though Gordon's view is anti-introspectivist, it shares some of introspectivism's deep motivations. Gordon's claim that one can grasp another's state in a first-personal way, combined with his denial that simulation depends on analogical inference, suggests that he holds a fairly radical externalism about mental contents. For consider: introspectivism allows that simulation enables one to entertain a mental state that is type-identical with the target's state. The need for analogical inference arises precisely because, when simulating, one is in a state which is a distinct token from the target's state. It is to avoid the need for analogical inference that Gordon maintains that I can "recenter my egocentric map" onto another's position. This strategy succeeds only if there is no logical gap between the other's state and my simulation of it. Hence, Gordon's claim that the other "becomes in my imagination the referent of the first person pronoun 'I'" must be understood quite literally in order to accomplish the work he requires of it, namely, to avoid any logical space for inference between the simulator and the target. And the absence of this logical space for inference allows the indexical reference to the other to be direct.
If this is correct, then the dispute between the introspectivist and Gordon over whether one can grasp another's state in a first-personal way turns on a dispute about the scope of content. This reading of the dispute need not, however, saddle the introspectivist with a strict content internalism. Goldman is no strict internalist, but he does maintain that our concept of the mental includes characteristics that are accessible only through introspective self-awareness. Gordon's claim that one can fully grasp these indexical features of mentality through direct indexical awareness of others shows that, as he conceives of the mental, there are no characteristics which are accessible only to the subject herself: I can grasp another person's state by standing in a first-person relation to that state, in a sense "becoming" that person. Gordon and Goldman are united in requiring, for a comprehensive conception of a mental state, that one have direct indexical access to the state; it is a disagreement about the scope of mental content which leads them to disagree about whether all direct indexical access is introspective access.(9)
Finally, it is worth mentioning that my diagnosis of the disagreement about introspectivism also applies to another, closely related dispute. Gordon maintains that one can acquire mental concepts through simulation, while Goldman charges that "simulation assumes a prior understanding of what state it is that the interpreter assigns to S". (Goldman 1989, 94; compare Fuller 1995; Heal 1986.) Goldman thinks that this prior understanding comes from a mixture of theoretical and introspective knowledge. I have argued that Gordon rejects the assumption that first-person access requires introspection, and thereby allows that contents are wholly external to a subject in a way which makes first-person other-knowledge possible. If this is correct, then both can agree that acquiring mental concepts requires first-person access to a state; the dispute here concerns whether such first-person access requires introspection. My diagnosis of the disagreement about introspectivism applies here as well: the dispute about whether simulation requires a previous grasp of mental concepts is also a symptom of competing views about content scope.
In sum, then, I interpret the relationship between Gordon's view and Goldman's view as follows. Gordon's rejection of introspectivism stems from his endorsement of a fairly radical content externalism, according to which there is no aspect of a mental state which is ineliminably "inner" or "private" to the subject. Goldman rejects this view of content in favor of the claim that, according to our mental concepts, mental states have some features which are knowable only through introspection. Goldman therefore believes that our conceptual purchase on the mental must be introspective. (He is, however, wary of drawing conclusions about the actual nature of the mental--answers to our question (3)--from such conceptual facts.) Gordon agrees that it must be first-personal, but because he believes that one can have first-person access to another's states, he denies that first-person access is necessarily introspective. The debate about content scope is not one I can address here, nor need we address it. The key point is that Gordon and Goldman agree about the essentially indexical nature of mental concepts, in the sense of IM, while their differing construals of content result in different views about what this entails. Specifically, they result in a disagreement about whether other-knowledge must rest on analogical inference.
(b) Heal
Where does Heal figure in this dispute? As I read her, Heal disagrees with Goldman on two central points. First, she resists the claim that other-knowledge requires empirical inference. On her view, there is no need for an empirical presupposition that I am rational, i.e., cognitively competent; and there is no need for an empirical presupposition that the target resembles me in being cognitively competent. These presuppositions are a priori in Heal's special sense--roughly, accepting them is a precondition of conducting our lives as we do. (Perhaps Heal would say that it is central to our form of life that we see ourselves, and others, as rational.)
The second point on which Heal and Goldman disagree concerns the level of processing at which simulation occurs. Heal is committed to the claim that simulation occurs at the personal level. This view derives from her estimation of ST as an a priori thesis, and thus as capturing necessary preconditions for treating others as mental beings. For the assumption that I am rational, and that other humans are similar to me in that respect, concerns our mental lives at the personal level. Goldman allows that simulation may occur only on a subpersonal level. This fits with his strict empiricism about ST, for presumably the discovery that subpersonal processes are simulationist could only be empirical. Goldman also allows, of course, that simulation could occur at a personal level; but whether it does is also an empirical matter, for him.
Like Goldman, Heal believes that simulation requires, and so cannot provide, mental concepts. As she puts it, "to pretend I must, at some level, know what I am up to and have some concept of that which I pretend." (Heal 1995, 44) It is important to distinguish the necessity of indexical (or demonstrative) access, for possessing a mental concept, from the sufficiency of such access. Heal holds that mental concepts, such as [belief that p], are irreducibly indexical (or demonstrative) because she thinks that one can't think about a subject matter, such as that relevant to the content p, without having a suitable set of abilities and dispositions. These include inferential abilities and behavioral dispositions. And a purely descriptive, non-indexical theory simply couldn't capture all of this. However, she denies that a purely indexical (or demonstrative) grasp of a concept suffices for possessing the concept. My "concept of that which I pretend" may be exhausted by features which I cannot specify but can only demonstrate, in which case I certainly do not possess the concept in any informationally rich way. "I may have no explicit way of specifying what I pretend other than by pointing to the pretence itself". (Heal 1995, 45)
On Heal's view, then, understanding another's mental state requires thinking about the subject matter which the target is thinking about, because one can have indexical/demonstrative access to the state in question only by co-cognizing with the target. This co-cognizing goes beyond what IM requires: it not only puts one in a position to effect direct indexical reference to some belief, it also puts one in a position to effect direct indexical reference to a belief that p, for the relevant p. Add to this Heal's claim that thinking about a subject matter requires suitable abilities and dispositions, and it follows that co-cognizing is essential to understanding another because it is required for bringing oneself to exemplify the set of dispositions which the other exemplifies. Again, co-cognizing may not suffice for a robust descriptive understanding of how the target conceptualizes the subject matter. Co-cognition is crucial because, without it, one cannot have indexical/demonstrative access to the indefinitely large set of dispositions which grasping a subject matter requires. And she believes we simply cannot exhaustively represent these dispositions descriptively.
We are now in a position to compare the answers which these three versions of ST give to our question, How does simulation capture the indexicality of mental concepts? Goldman's introspectionism gives the following explanation. The indexicality of mental concepts entails that they can be fully grasped only in a first-personal way; first-person access is introspective access; therefore, one can understand a mental state only by introspecting an instantiation (or a simulation) of it in oneself. In particular, one can have adequate knowledge of another person's mental state only via analogical inference from one's own introspected state. On Gordon's semantic ascent account, the indexicality of mental concepts entails that they can be fully grasped only through an understanding of mental perspective; we attain the understanding of mental perspective through grasping a perspective different from our own, and thereby recognizing the indexicality of our own perspective; therefore, one can grasp mental states, as such, only by a first-personal grasp of another's perspective. Finally, Heal's rational co-cognition picture says, with Goldman and Gordon, that the indexical nature of mental states entails that adequately conceptualizing them requires a first-person grasp of them. For Heal, however, this requires no empirical analogical inference; nor can one have a first-personal grasp of others' states. Rather, one's grasp of others' states is mediated by a first-person grasp of one's own, together with an a priori analogical assumption, namely that others share my cognitive competence.
While the three simulationist views vary in important respects, they are also importantly united. The previous section showed that the indexical nature of our mental concepts is pivotal in motivating these three simulationists to accept ST, for they agree that only simulation can effect a grasp of mental states which is true to the mental's indexicality. And the current section has showed that the points on which they diverge stem from disagreements about how best to accommodate IM.
Conclusion
My purpose here has been to reveal a deep conceptual source of the dispute between ST and TT. The endorsement of IM is, then, a principled faultline which divides simulation theorists from theory theorists. Many have doubted whether anything of substance lies behind the "ST vs TT" debate. My analysis uncovers a substantial, conceptual difference between these two camps, even while recognizing that there are vast differences within each camp. And it provides a strong conceptual basis for favoring ST.
1. The growing emphasis on consequences of empirical findings is clear from three excellent anthologies on this debate: Davies and Stone 1995a, Davies and Stone 1995b, and--especially--Carruthers and Smith 1996.
2. As we will see below, Gordon believes that two persons can share a single mental location, so that "I" can refer, in thought, to another person. In fact, he claims that one must indexically refer to another before one can indexically refer to oneself. By contrast, Goldman believes that indexical reference to one's own mind is fundamental, and doubts that mental locations can be shared in Gordon's sense. Heal appears uncommitted as to whether self-reference and other-reference must exhibit any logical priority.
3. But Goldman (1993) suggests that intentional states may have a distinctive phenomenology. The view that intentional states have qualitative or phenomenological features appears to be regaining popularity. See especially G. Strawson (1994), Siewert (1998).
4. Compare Chisholm's claim that some mental states are "self-presenting", in the sense that the states' modes of presentation just are the states themselves. (Chisholm 1966)
5. Roth (2000) also argues that intentions have a self-referential indexical component. On Roth's view, when one intends to X, one thereby intends that X occurs as a result of this very (reflexively referred to) intention.
6. As we will see, Gordon and Heal maintain that grasping (indexical) mental concepts requires knowing how to simulate (or "co-cognize with") another; and this know-how requires other abilities, such as the ability to imaginatively identify with the target and, at least for Heal, to think about the subject matter the target is thinking about.
7. The question whether indexical knowledge can take a propositional form is familiar from responses to the "Knowledge Argument" for dualism. (Jackson 1982) This argument claims that exhaustive physical knowledge would leave out "what it's like" to experience a certain sensation, and so the phenomenal property of the sensation is not a physical property. Some try to block this argument by claiming that what's left out is an ability rather than a piece of propositional knowledge; and since we don't generally expect exhaustive propositional knowledge to yield abilities, there is nothing special about sensations in this regard. Other opponents of the Knowledge Argument allow that propositional knowledge can be indexical, but argue that we shouldn't expect even exhaustive non-indexical knowledge to yield indexical knowledge.
8. The deep sources of this picture are Sellarsian. In a forthcoming article, Gordon says "I agree with Sellars that even the first person use of mental predicates is essentially an intersubjective achievement"; self-reports "piggyback on the public language". (Gordon, forthcoming) The Sellarsian picture excludes the possibility of introspection, of knowing one's own states, prior to a concept of mental locatedness or perspective, a concept which can be gotten only through simulation (of another person or oneself in a counterfactual situation). Gordon's rejection of introspectivism thus derives from a picture of mental concepts according to which one can make direct indexical reference to another person, with "I", and this capacity for direct indexical reference to another is conceptually prior to any indexical grasp of oneself.
9. Another point of difference between Gordon and the introspectivist concerns how we acquire our basic idea of mental perspective, that is, it concerns the type of process which creates "a logical space ... that enables us to think of pains and facts as located in (or at) individual minds". (Gordon 1996, 19) Gordon holds that a first-personal grasp of another's mental location is conceptually required for any grasp of mental perspective, and so this grasp is conceptually prior to grasping one's own mental location as such. Introspectivism entails that this conceptual order is reversed, i.e., that one can conceive of another's states, as mentally located, only by first noting the locatedness of one's own (introspected) states. But beyond the issue about content scope discussed above, nothing further appears to be at stake here. This disagreement is merely a byproduct of the disagreement about whether I can grasp another's perspective with a direct indexical such as "I". This is not to say that the disagreement is insignificant; whether I can adopt a first-person perspective vis-à-vis someone else's mental state is a critically important issue. But this point of convergence between Gordon and Goldman is equally important: they agree that a capacity for first-person or direct indexical reference ("I") is a necessary precondition for grasping the concept of mental perspective more generally. The idea that direct indexical reference is essential to mental concepts motivates both of these versions of simulationism.
Carruthers, P. and Smith, P., eds., (1996) Theories of Theories of Mind. Cambridge: Cambridge University Press,
Casteñeda, H.-N. (1966) "'He': a Study in the Logic of Self-Consciousness." Ratio 8:130-57.
Chisholm, R. M. (1966) Theory of Knowledge. Englewood Cliffs, N.J.: Prentice-Hall.
Davies, M. and Stone, T. (1995a) Mental Simulation: Evaluations and Applications. Oxford: Blackwell Publishers.
Davies, M. and Stone, T. (1995b) Folk Psychology: The Theory of Mind Debate. Oxford: Blackwell Publishers.
Davies, M. and Stone, T. (1995c) "Introduction". In Davies and Stone (1995a), 1-18.
Evans, G. (1982) The Varieties of Reference, ed. J. McDowell. Oxford: Oxford University Press.
Fuller, G. (1995) "Simulation and Psychological Concepts". In Davies and Stone (1995a), 19-32.
Gallois, A. (1996) The World Without, the Mind Within: An Essay on First-Person Authority. Cambridge: Cambridge University Press.
Goldman, A. (manuscript) "The Mentalizing Folk".
Goldman, A. (1993) "The Psychology of Folk Psychology". The Behavioral and Brain Sciences, 16:15-28.
Goldman, A. (1989) "Interpretation Psychologized". Mind and Language 4, 161-185; reprinted in Davies and Stone (1995b), 74-99.
Gordon, R. (forthcoming) "Sellars's Ryleans Revisited". Protosociology: An International Journal of Interdisciplinary Research, vol. 14.
Gordon, R. (1996) "Radical Simulationism". In Carruthers and Smith (1996), 11-21.
Gordon, R. (1995) "Simulation Without Introspection or Inference From Me to You". In Davies and Stone (1995a), 53-67.
Gordon, R. (1986) "Folk Psychology as Simulation." Mind and Language 1:158-71; reprinted in Davies and Stone (1995b), 60-73.
Heal, J. (1998a) "Understanding Other Minds from the Inside". Contemporary Issues in the Philosophy of Mind, ed. A. O'Hear. Cambridge: Cambridge University Press.
Heal, J. (1998b) "Co-cognition and Off-line Simulation: Two Ways of Understanding the Simulation Approach" Mind and Language 13:477-98.
Heal, J. (1997) "Indexical Predicates and Their Uses". Mind 106:619-40.
Heal, J. (1996) "Simulation and Cognitive Penetrability". Mind and Language 11:44-67.
Heal, J. (1995) "How to Think About Thinking". In Davies and Stone (1995a), 33-52.
Heal, J. (1986) "Replication and Functionalism". In Language, Mind, and Logic, ed. J. Butterfield. Cambridge: Cambridge University Press.
Jackson, F. (1982) "Epiphenomenal Qualia". The Philosophical Quarterly 32: 127-36.
Lewis, D. (1979) "Attitudes De Dicto and De Se". The Philosophical Review 88:513-43.
Moran, R. (1997) "Self-Knowledge: Discovery, Resolution, and Undoing". The European Journal of Philosophy 5:141-61.
Peacocke, C. (1992) A Study of Concepts. Cambridge, MA: MIT (Bradford).
Perry, J. (1979) "The Problem of the Essential Indexical". Nous 13:3-21.
Recanati, F. (1990) Direct Reference: from Language to Thought. Oxford: Blackwell.
Roth, A. (2000) "The Self-Referentiality of Intentions". Philosophical Studies 97: 11-51.
Shoemaker, S. (1988) "On Knowing One's Own Mind". Philosophical Perspectives 2:183-209. Atascadero, CA: Ridgeview Press.
Siewert, C. (1998) The Significance of Consciousness. Princeton: Princeton University Press.
Strawson, G. (1994) Mental Reality. Cambridge, MA: MIT Press.