2. Searle’s theory of intentionality

On Searle’s theory, intentional states may be represented as an intentional or psychological mode with an intentional or representative content.Footnote ³ The intentional mode is the way the state is directed—a believing, say, as opposed to a desiring. Borrowing an idea from speech act theory, these modes are said to have direction-of-fit. He maintains that beliefs like statements can be true or false and have a mind-to-world direction of fit. The belief is to accord with the world. In contrast desires and intentions have a world-to-mind direction of fit in that satisfaction of the state requires bringing the world into accordance with the desire or intention. States such as sorrow or pleasure have the null direction of fit since though such cases presuppose the world is a certain way—an event or state of affairs that one is sorrowful or pleased about—they themselves are neither directed at the world nor is any change in the world required for their satisfaction.

On Searle’s theory, it is the intentional content that both specifies what the intentional state is about and accounts for the aboutness of every intentional state. The intentional content is a representation of what it is about under some aspect(s); it provides conditions of satisfaction for the intentional state; these are internal to the intentional state (Searle Reference Searle1983, 11). These conditions may or may not be satisfied. When they are satisfied, whatever satisfies them is the intentional object for that state. When no object satisfies them, the intentional state lacks an intentional object.Footnote ⁴ He says:

For example, the intentional content of Ponce de Leon’s intentional state of desiring to find the Fountain of Youth is not satisfied, thus, his intentional state has no intentional object and his desire cannot be satisfied. Even so, what the state is about is specified by its (unfulfilled) conditions of satisfaction. In such cases, there is a purported intentional object via conditions of satisfaction. In sum, on Searle’s theory, every intentional state must have an intentional content, which provides the answer to (Q3), and (Q1) is answered negatively. Searle’s answer to (Q2) is simply that intentional objects are whatever satisfies the intentional content if anything does. Thus, for Searle, intentional objects are nothing unusual; they do not have any special ontological status.Footnote ⁵ Before examining (Q4), I turn to some objections to these answers.

3. The idea of intentional object versus intentional objects

There are those that would argue, contra Searle, that every intentional state must have an intentional object: even thoughts about nonexistents require intentional objects. They thus give an affirmative answer to (Q1). Tim Crane is among those just described. He holds that if an intentional state could lack an intentional object, we would be “at a loss to say what makes this latter class of states intentional … [and later] … I don’t see how we should do without the idea of intentional object” (Reference Crane2001, 22–23; last italics added. Similar remarks in his Reference Crane2013, 5).

Crane’s affirmative answer to (Q1) serves another purpose for him: it is the basis for his answer to (Q3), for he holds that it is the having of an intentional object—a nonexistent one if need be—which accounts for an intentional state being about something—what it is directed at or to.Footnote ⁶ Crane’s conclusion that thoughts about nonexistents require intentional objects is ultimately based on two claims: we can and do think about the nonexistent, and such thoughts would not be about anything if they lacked an intentional object. The first is uncontroversial, the second is not. The claim that the idea (concept, or notion) of intentional object plays an essential role in a theory of intentionality is one thing; whether every intentional state must have an intentional object is another. One may assent to the former and deny the latter. The idea of intentional object and the existence of an intentional object must not be conflated. Let me explain.

On Searle’s theory, what makes an intentional state that lacks an intentional object intentional is exactly what makes any intentional state intentional—intentional content. Intentional content specifies what the intentional state is about via conditions of satisfaction.Footnote ⁷ As we have seen, if there is an object that satisfies the conditions, it is the state’s intentional object; if no object satisfies the conditions, then there is no intentional object for that state. Nevertheless, in either case, the satisfaction conditions specify what would be the intentional object if the state had one. Hence, satisfaction conditions implicate the idea of intentional object, but they do not, should not, need not, and cannot guarantee that there is an object which satisfies them. Thus, the idea of intentional object plays a role in every intentional state for Searle, whether it is about an existent or not. In no way does Searle’s theory contravene the certainly true claim that the idea of an intentional object is required for any theory of intentionality. Crane apparently is under the misimpression that by holding that some intentional states lack an intentional object one is doing without the idea of intentional object, but this, I have just argued, is simply false.

A related worry that one might have once one denies that all intentional states have an intentional object is this: if, on Searle’s theory, thoughts about nonexistents do not have intentional objects, how does one distinguish, say, a thought about Zeus from a thought about Pegasus?Footnote ⁸ Given what has been recently said the answer is clear: intentional contents for thoughts concerning different nonexistents have different satisfaction conditions, thus, though neither Pegasus nor Zeus exist (since the respective satisfaction conditions of either are not fulfilled), they have distinguishable purported intentional objects via their different satisfaction conditions.

Thus, Searle’s theory of intentionality clearly accounts for the aboutness of intentionality via intentional content with its conditions of satisfaction. It also uniformly explains thoughts about both existents and nonexistents, explains why some thoughts lack intentional objects, and explains how it is that different thoughts about nonexistents are different—all this without assigning any special or unique type of ontological status to intentional objects. (“Do not multiply entities beyond necessity.”)

4. Is intentionality relational?

Intentionality superficially appears to be a relation between an intentional state and what it is about or directed to. Perhaps this partly explains why some think that every intentional state must have an intentional object. The apparent relational nature of intentionality worried Brentano. In the appendix of Brentano’s [1874] 1911, he states that mental activity seems to be the relation between a person and what the person’s mental activity is about. The problem he saw is that typically two-place relations require both relata for the relation to obtain, e.g., iff ‘A kicks B’ is true, both A and B must exist. Both relata, a person and what their state is about, may exist for many intentional states, but for others the second relatum does not, e.g., Ponce de Leon did exist, but The Fountain of Youth, which he desired, did not. True negative existentials pose the same but more trenchant problem, for if one thinks that The Fountain of Youth does not exist, what the thought is about—the apparent second relatum—must not be, if one’s thought is correct. Given these considerations, Brentano concludes, “… one could doubt whether we are really dealing with something relational here, and not, rather, with something somewhat similar to something relational in a certain respect, which might, therefore, better be called ‘quasi-relational’” (Reference Brentano1911, 212).

The view that intentionality is relational persists in modern times; indeed, it is arguably the orthodox view.Footnote ⁹ For example, on this view, if Amanda believes that Brentano discussed intentionality, this is explained as Amanda being in the believing relation to a proposition expressed by the sentence “Brentano discussed intentionality”; when the intentional state is directed toward an object—e.g., if Amanda desires the glass of scotch on the table—she is in the desiring relation to the object designated by the definite description. The pervasiveness of the orthodox view that intentionality is relational is what generates the perceived requirement that all intentional states require an intentional object, and, in consequence, some intentional states require nonexistent intentional objects. This is not a productive path. Brentano was right when he expressed reservations whether intentionality is relational and suggested that it is “quasi-relational” since intentionality is “only somewhat similar to something relational in a certain respect…” ([1874] 1911, 212).

The previous two sections explained aboutness or directedness of intentional states in terms of intentional contents and their conditions of satisfaction. Though Searle does not explicitly reject the relational view of intentionality, he avoids presupposing it. On his theory, while every intentional state must have intentional content, the latter’s satisfaction conditions are not always satisfied by an object. In such cases, the intentional state simply has no intentional object; even so, the aboutness of the state is secured by the intentional content. No need to posit mysterious nonexistent intentional objects to accommodate the (false) view of the relational nature of intentionality. There is a different problem, however, common to both Searle’s theory and the orthodox view since both hold that intentional states are univocal in what they are about. I turn to this, (Q4), next.

5. How the new theory identifies intentional contents

To my mind, Searle’s theory of intentionality is of major significance and groundbreaking. It elegantly dispels many longstanding confusions regarding intentionality and intentional objects. Still, I also think there is a fundamental and essential ambiguity between a subjective and objective reading of the question as to what an intentional state is about, an ambiguity that neither Searle’s theory nor anyone else’s accommodates, though appeal to it is sometimes obscurely and implicitly made; this promotes confusion and untoward consequences.Footnote ¹⁰ I argue for the ambiguity and present the central ideas that accommodate it. I show how the resulting new theory of intentionality, one which extends Searle’s, is successfully deployed to address the problem of substitution in propositional attitudes. I indicate here and develop in detail elsewhere (Georgalis Reference Georgalis2006, Reference Georgalis2015) how it resolves others. In section 12, I demonstrate a problem in Searle’s theory that results from his failure to explicitly recognize and accommodate this fundamental ambiguity.

Is what an agent’s thought about univocal? If having a thought is held to be standing in some relation to a proposition, as Frege and many philosophers do, then what a thought is about is univocal.Footnote ¹¹ I purport to show that this is a mistake. Let us see why.

I have a thought that I can appropriately express with the sentence “Up quarks have color.” An expert elementary particle physicist could utter the same sentence to express her thought. Do we have the same thought? The answer seems to be yes and no. The answer is yes since, in accordance with ordinary linguistic practices of thought attribution, we both can appropriately use the same sentence to express our thought, even though I know next to nothing about quarks or color and the physicist knows much. But the answer is also no since what I am thinking—what I have in mind—when I utter that sentence bears almost no resemblance to what the expert physicist is thinking when she utters the sentence.Footnote ¹²

To reconcile the conflicting answers, I exploit the difference between a thought-token and a thought. The concept ‘thought-token’ is used to indicate the actual content of what one thinks on a given occasion. Thought-tokens are episodic. The content of a thought-token is something ‘going on in an agent’ at a particular time. While this locution is vague, it is an undeniable fact that an agent is able from her first-person perspective to somehow identify what is going on in her with sufficient clarity to utter an appropriate sentence, as I do when on some occasion I utter ‘Up quarks have color.’ The fact that an agent has this ability to express what is going on in her—what she is thinking at a given moment—with a sentence which would be deemed appropriate in the context of utterance and would be in accordance with our ordinary linguistic practices of thought ascription suffices for my purposes.

To be clear, when I say that an agent is able to identify “something going on in her,” I am neither saying nor presupposing that she examines and identifies some inner mental content, which she then proceeds to express linguistically. No examination of an inner-theater is appealed to. What I am saying and claiming is that something unspecified occurs in her head/mind which prompted her utterance. Because this something pertains to what one is thinking at a given moment, I label it ‘thought-token,’ as distinct from a ‘thought,’ where the latter is not indexed to a moment. (More on ‘thought’ below.)

As indefinite as “something going on in her” is, it still appears to be a simple fact of the matter that we do express what we are thinking at a given moment in this direct way. An agent is directly aware of her thought-token to the extent that if asked, for example, “What were you thinking about?” she would unhesitatingly and unequivocally say she was thinking about up quarks, that they have color, or some such. Without hesitation or inner inspection she gives linguistic expression to her thought-token—what she is thinking at the moment.

I make no pretense of knowing just what constitutes a thought-token—neural events, neuronal representations, or something else. Still, it seems unproblematic to suppose, as in the quark example, that what the two agents are actually thinking when they utter the same sentence is different and, therefore, that the actual contents of their respective thought-tokens are different.

A reader might have noted a similarity between my example and some Stich discussed in his Reference Stich1983.Footnote ¹³ Two of Stich’s examples are that of Mrs. T, who suffers from dementia—an actual case. Mrs. T. would assent to or assert ‘McKinley was assassinated,’ while not remembering what dying is or whether an assassinated person dies. Another, one very similar to the quark case, is of a child who believes E = mc ², but knows little of physics.

Stich’s purpose was to lay some groundwork to advance a holistic theory against a narrow causal theory of mental sentences. While I am sympathetic to holistic theories, as may be evident in my appeal to the subject’s understanding of the terms she utters in expressing her thought-token, I am not arguing for that here, nor does what I have argued presuppose a holistic theory of semantic meaning. Rather, I use these types of cases (more later) to bring out a crucial fact for my extended theory: there is no one–one mapping between thought-tokens and the sentences that express them.

My main point in using my examples is that what goes on in a speaker’s head, the content available to the speaker, may well deviate from the semantic content of the sentence the speaker uses to express her content, her thought-token. Whether semantic contents of sentences are holistic or not, my point holds. Should semantic content be holistic, my point is even stronger since, as in the quark case, the semantic relations that hold between the terms ‘quark,’ ‘up,’ and ‘color’ would then be essential to their semantic content and the speaker is ignorant of these relations. So, her thought-token could not possibly map one–one to the sentence she utters. On the other hand, if semantic content is not holistic, there still are cases where the speaker would be unaware of the exact semantic contents of the sentences she would use to express her thought-tokens. This may be due to either ignorance, as in the quark case, or to the speaker’s association of collateral information or misinformation on the occasion of utterance (as in the example below of the use of the ordinary term ‘elephant’). Therefore, my argument for many–one mappings between thought-tokens and the sentences used to express them is independent of the question whether semantic meaning is holistic.

The quark example is meant to expose two fundamental facts. First, as just indicated, two agents’ thought-tokens may significantly differ in content, while the same sentence may appropriately be used to express them. Henceforth, when I speak of ‘a sentence appropriately expressing a thought-token,’ this is to be understood to mean that doing so is in accord with our ordinary linguistic practices of thought ascription. Second, the semantic content of a sentence so used to express a thought-token may diverge from the content of the thought-token it expresses. While the content of the expert physicist’s thought-token may smoothly map to the semantic content of the sentence used to express it, my thought-token decidedly would not.

When these factors are overlooked, the single sentence, with its ordinary dictionary meaning, used to express the various thought-tokens washes out the individual variation; consequently, attention is focused on the sentence uttered by or ascribed to the thinker and its content, or the proposition expressed, while ignoring possibly significant differences that may be manifest in the respective thought-token’s content. The content of the sentence used to express the thought-token then serves as a proxy—an inaccurate proxy—for what the thinker is actually thinking. As a result, we illegitimately assume that the content of the agent’s intentional state, her “thought,” maps one–one to the sentence that expresses the thinker’s thought-token. This, despite the fact that the agent’s thought-token and the sentence contents may radically differ. It is, therefore, a mistake to willy-nilly identify the content of the agent’s thought-token with that of the sentence she uses to express it, as is exemplified in the quark case.

The distinction between how an agent understands what goes on in her head—her thought-token—and, hence, how she expresses it, on the one hand, and the semantics of the sentence she appropriately uses to express her thought-token, on the other, plays a crucial role in my theory.

Loar (Reference Loar, Grimm and Merrill1988) early on was aware of some such distinction, as he recognized the potential for difference in what he called ‘psychological content’ and ‘social content’ of a belief. The work the former locution does in his paper somewhat corresponds to that of ‘thought-token’; his ‘social content’ does similar work as that to which I put the semantic content of a sentence used to express an agent’s thought-token. He argued, “… psychological content is not in general identical with what is captured by oblique that-clauses, that commonsense constraints on individuation induce only a loose fit between [psychological] contents and that-clauses … ,” which express the agent’s belief (Reference Loar, Grimm and Merrill1988, 102). My point regarding the many–one mappings between thought-tokens and the sentences that express them is similar to Loar’s “loose fit.”

There are, however, some important differences with the current work, as Loar does not treat belief/thought as a facon de parlor (explained below), nor does he embed his ideas in a general theory of intentionality though he sketches features for one. He does say, “The point is that conceptual roles of one’s thoughts determine how one conceives things, and it is difficult to see how one can consider how one oneself conceives things without that in some sense involving what one’s thoughts are ‘about’” (Reference Loar, Grimm and Merrill1988,108). This goes directly to my argument that the question as to what an agent’s thought is about is inherently ambiguous between a subjective and an objective reading of the question and that both readings are necessary for a proper understanding of intentionality. In marked contrast, Loar claims that “internal intentionality,” or the subjective reading, “is a sort of illusion” (136).

The variation in the contents of different thought-tokens that are expressed by the same sentence is widespread. It is not restricted to cases that involve sophisticated or arcane terminology, such as ‘up quark.’ The same variation in thought-tokens can be manifested even when an ordinary term (e.g., ‘elephant’) appears in a sentence expressing a thought-token. For, in the first place, a great deal of the words we commonly and correctly use, many of us would be hard put to provide the standard dictionary definition. This is sufficient reason to deny that the actual contents of our thought-tokens always match the standard meanings of the words used to express our thought-tokens. Second, even when one has mastery of the standard meanings of the words used to express one’s thought-token, this would not prevent some collateral information or misinformation regarding the referent of whatever term is used from intruding into the actual thinking of the agent on any given occasion and thereby partially constitute her thought-token’s content.

Armed with these kinds of cases, I maintain that although different agents’ differing contents of their thought-tokens may be appropriately expressed by one and the same sentence, the differences in their understanding from one another are still manifested in differences in each of their actual thought-tokens, and either or both contents of these thought-tokens may differ from that of the content of the common sentence uttered. Thus, in such cases, we may conclude that whatever the component contents of the different thinkers’ thought-tokens are, they do not map one–one to the contents of components of sentences that express them. Quite generally, there is a many–one relation between thought-tokens and the sentences that express them.Footnote ¹⁴

Once we recognize: (1) the potential differences between the contents of thought-tokens and the sentences which express them, and (2) the proposed account of thought is accepted, any puzzlement regarding the plausibility of both yes and no answers to questions such as, “Does the expert physicist and the novice have the same thought?”—as in the quark example—are substantively explained. Any initial puzzlement regarding this question is based on the mistaken assumptions that what one thinks is a thought, and its content maps one–one to the sentence that expresses it. When in fact, what one thinks is best described as a thought-token that typically does not map one–one to the sentence expressing it.

The point made here is that often, even typically, there is a divergence between the content of what a speaker understands and the public meaning of the expressions she utters to express her thought-tokens. My further claim, which I will argue for more fully below, is that accommodation of such divergences is of the utmost importance in considering various philosophical problems, problems whose adequate resolution depend on such accommodation.

The divergence between the contents of thought-tokens and the sentences that express them is typically neglected, but I have been gratified to learn that I am not alone in the recognition of its importance. Consider what Michael Dummett has to say on this point in the preface to his Reference Dummett1981:

Thus, no less an authority on Frege’s philosophy than Dummett had come to maintain that an accurate evaluation of Frege’s philosophy of language must wait until the problematic nature of the relation between a speaker’s understanding of the terms of her language community and the public meaning of those terms is addressed, something of which Frege explicitly says nothing.Footnote ¹⁵

A speaker’s variation in understanding from the public meaning of words is captured by my concept of thought-token and the relation of that to the public language is accommodated by the many–one mappings. However, more must be said of these many–one mappings. Two other factors relevant to these many–one relations is the agent’s understanding of her circumstances and required restrictions on many–one mapping. (Both of which will be explained below. The first factor is addressed in the Al, Bob, and Sam cases, and the second, when I consider substitution in propositional attitudes.) But before developing these conditions and restrictions, let us consider what thought is on my theory.

What exactly is my sense of ‘thought,’ as distinct from that of ‘thought-token’? I maintain that though we undeniably have thought-tokens, we do not have thoughts. The locution ‘thought’ is just a facon de parler. The expression ‘thought’ is, of course, useful, but on my theory, it should be understood simply as the ordinary-dictionary-meaning-of-a-sentence-appropriately-and-accurately-used-to-express-an-agent’s-thought-token-on-some-occasion.Footnote ¹⁶ To say that someone “has” the thought, say, that some apples are red, when not entertaining it, is just to say that under appropriate conditions the agent would have a thought-token that she would appropriately and accurately express by uttering or assenting to the sentence ‘Some apples are red.’

Frege also denied that we have thoughts; he held that thoughts are abstract objects that we grasp. Consequently, there is a question for him concerning just how we can grasp abstract thoughts, which exist apart from us. There is no such problem for the current theory in ‘grasping’ either thought-tokens or thoughts, for the former are episodic occurrences in us, which directly prompt the sentences we utter to express them. I have been careful not to talk of thought-tokens in an ontological vein. For me, they are simply subjective concepts that play a role in an analysis of intentionality. From a subjective point of view—something going on in us prompts the sentences we utter. Nor is there a problem on my theory of grasping thoughts, since ‘thought’ is simply a facon de parler. (Compare the earlier definition of ‘thought.’)

My contention is that to have a satisfactory theory of intentionality, we must accommodate the distinctions between thought, thought-token, and the sentence used to express thought-tokens, as well as the many–one mappings that obtain between the latter two. Toward that end, on my theory the question as to what an intentional state is about is viewed as inherently ambiguous between what an agent’s intentional state is about (a) from her own subjective perspective, what she herself is thinking, her thought-token and (b) from the perspective of an objective observer of the agent.Footnote ¹⁷ These different points of view frequently, even typically, result in the same answer. However, when the answers diverge and the question of aboutness pertains to what the agent herself is thinking, the subjective reading is paramount. Generally, which answer is relevant would depend on the context in which the question is raised.

Ignoring the ambiguity when such divergences occur is a deep source of philosophical conundrums; the ambiguity is importantly manifested in the problems of substitution in propositional attitudes, distinguishing speaker’s referent from semantic referent, Donnellan’s alleged referential/attributive distinction, and Kripke’s puzzle about belief, among others.Footnote ¹⁸ One need not reflect too deeply to realize that a central source of these issues is the discrepancy between the content of an agent’s thought-token or how she understands either the sentences she utters or the circumstances prompting her thought-token and, thus, what words she chooses to express it, on the one hand, and what an objective observer would ascribe to the agent, on the other. In such cases an objective observer might well express what the agent’s “thought” is about differently than the agent herself. (Scare quotes here used because of my different accounts of thought-token and thought.)

In brief, what has emerged in this section is that intentional states are ambiguous between subjective and objective readings. The intentional content of an intentional state, therefore, has two readings also. The subjective reading of intentional content is identified by the agent’s thought-token, while the objective reading is identified by the public sentence that appropriately expresses the agent’s thought-token.

6. Ignorance of circumstances as a qualification on specifying intentional states

To get clearer on the ambiguity at issue, I start with a simple example used by Kripke, though he uses it to a different end.Footnote ¹⁹ The quark case brings out both the ambiguity in the question as to what a thought is about and the importance of many–one mappings. The Kripke case highlights the alleged ambiguity, though it does so in a different way since it does not require many–one mappings. Instead, it demonstrates a further requirement for some types of cases: the consideration of the conceived circumstances of utterance. The new qualification applies in mundane cases of agents mistaking their circumstances.

Suppose Al and Bob see a man in the distance who happens to be Smith, but they mistake him for Jones, and Al asks, “What is Jones doing?” Bob responds, “He is raking leaves.” Now if you were to ask Al or Bob who they were talking about, they would unhesitatingly say, ‘Jones.’ Just as clearly, we, who are in the know—that is, objective observers—would say their thoughts are about Smith; after all, he is the guy who is raking leaves. We might even correctly say, though Al and Bob think they are talking about Jones, they are mistaken.

But focus again on Al and Bob for a moment and put aside what they said. Is it Smith who they themselves are thinking about? Do they have Smith in mind on this occasion? Clearly, if asked “Are you thinking of Smith?” each would unhesitatingly say “No, I’m thinking about Jones and what he is doing,” or some such. The subject or topic of each of their respective thought-tokens is not Smith but Jones. They have Jones in mind; neither of them has Smith in mind. This certainly provides some reason for holding that who their thought-tokens are about is Jones, not Smith.

Consider another simple example: suppose Sam is stimulated in such a way that he utters, “The orange on the table is small.” Clearly, as Sam conceives the situation, he is talking about an orange. Part of the intentional contents, satisfaction conditions, to his mind specifies an orange. Suppose in fact that what is on the table is not an orange but a kumquat. Again, an objective observer would rightly hold that Sam is talking about a kumquat, though Sam himself mistakenly takes it to be an orange. Neither answer should be faulted, but which is appropriate depends on the context in which the question “What is Sam thinking about?” is raised. Does the question concern what Sam, himself, is thinking, or does it concern what is objectively prompting Sam’s utterance?

These simple examples are unlike the quark case, where both the physicist and I intend to speak of the same thing—up quarks. That case exemplified the importance of recognizing many–one mappings of thought-tokens to sentences expressing them. The recent examples further exemplify the importance of accommodating the stated ambiguity in what a thought is about, while at the same time recognizing how the agent conceives her circumstances may be also relevant. They illustrate that what an agent is actually thinking about—what she has in mind—may well differ from what is prompting what she thinks on a given occasion and, hence, what an objective observer would say her intentional state is about. Of course, where there is such divergence, the agent is factually mistaken, but her thought-token—that which she herself is entertaining at the time, what she thinks it is about—cannot be appropriately characterized by a term for the objectively prompting object; that is why she is mistaken. While it may seem platitudinous to point out such divergences, not only can the consequences of such differences prove to be of great importance, but they also go to the very heart of the question as to what an agent’s intentional state is about. The question of aboutness has been widely thought to have a univocal answer; I take such cases to illustrate that this is mistaken.

7. Required concepts for the new theory: minimal content and objective content

My extension of Searle’s theory of intentionality accommodates this ambiguity. As we have seen, a fundamental idea of the extended theory is that of the distinction between thought-token and thought. A further crucial idea of my theory concerns a fine-grained component of thought-token, the subject or topic component of thought-token: I call it ‘minimal content’:

Were the agent to express her minimal content she would use a noun or noun phrase which, being part of a public language, would have its standard meaning or reference. While the agent deploys a linguistic expression that appropriately expresses her minimal content, her understanding of it may not coincide with the public meaning or reference of the expression, as in the quark case.

The objective correlate to minimal content of the inherent ambiguity is indicated by what I call the ‘objective content’:

I stress that the concept ‘objective content,’ as here used and as in the case of minimal content applies only to the topic or subject of the agent’s intentional state, not the entire thought-token. A noun or noun phrase, which has public meaning or reference, would be used to express the objective content.

Both the agent and an objective observer may appropriately use the same linguistic expression to express the subject of the intentional state. Even here, however, the agent’s understanding of the linguistic expression used may deviate from its standard meaning, as in the quark case. The variation that gives rise to the ambiguity in what an intentional state is about may have a different source though. In the Al, Bob, and Sam cases the referent of each noun or noun phrase used is clear to the speakers and is in accord with the public referents of the names; Al and Bob just misidentify who is raking leaves. They misidentify the circumstances. Either inadequate understanding of terms or misidentifying the circumstances of utterance may give rise to the ambiguity. We will soon see a third source (section 10).

Searle’s modeling of his theory of intentionality on speech acts represented an intentional state as S(r), where S is the propositional attitude and r is the representative or intentional content (Reference Searle1983, 6). The distinction between minimal and objective content forces a modification in his theory. Accordingly, S(r) should have two decompositions: S(r(m)) and S(r(o)), where ‘r(m)’ represents the subjective intentional content of the agent’s intentional state, her thought-token, and ‘r(o)’ represents the objective intentional content of the agent’s intentional state.

8. Objectual intentional states?

While I am sympathetic to the idea that there are merely objectual intentional states, intentional states that are about an object or kind of object, rather than states of affairs, e.g., wishing to ride Pegasus or fearing lions, as opposed to, say, believing that it will rain, respectively. I do not argue for that here. Nor does anything that I have said here commit me to either affirming or denying such a claim. Nevertheless, I sketch how one would represent objectual intentional states within the view presented here.

When the intentional content is not propositional, the two readings of the intentional state would be represented simply as S(m) and S(o). When there is an intentional object to an objectual intentional state, it is the state’s objective content. As I will presently make clear, in states directed simply at an object, the ‘m’ and ‘o’ in ‘S(m)’ and ‘S(o)’ involve features of the purported intentional object and are not propositional in form.

There are two issues to be addressed here: (1) Searle’s theory and my extension of it hold that intentional content specifies satisfaction conditions: How could that work when the intentional content is not propositional but objectual? (2) If the subject of an objectual intentional state does not exist as, say, in wishing to ride Pegasus or Macbeth wanting to clutch the dagger he hallucinates, then wouldn’t the extended theory require nonexistent intentional objects? The answer to (1) will, together with the theory, provide the basis to negatively answer (2).Footnote ²²

Can one be in an objectual intentional state without having any feature or purported feature of it in mind? Can I think of Pegasus without, ipso facto, thinking of a winged horse or some such? Similarly, can I think simply of Jones without thinking of any feature of his, say, that he is my friend? Can one think of such without thinking of it “under some aspect,” as Searle would put it? In objectual intentional states, are the individuals we think about to be featureless, bare particulars? I doubt that anyone would wish to answer the latter affirmatively and while the same responses to the other questions may not be quite as evident as it is for the last, they are the same.

Even in the extreme, when we have virtually no information about an existing object, as when I am not hallucinating (to be considered presently), and I think about the object over there but have no clue what it is or what features it has, I am aware, at the very least, that it is a spatial object and has a certain spatial relation to me.

In the case of fictious objects, the aspects under which they are thought provide the satisfaction conditions which constitute the intentional content. Satisfaction conditions are also required to distinguish one fictitious object from another—say, Pegasus from a centaur. The features of Pegasus and a centaur are different, and they provide distinct satisfaction conditions for any intentional state regarding either. (Compare the criticism of Crane in section 3.)

What was just said applies as well to hallucinations. When Macbeth hallucinated a dagger before him, there were certain dagger-features he was aware of; they were not, say, axe-features. His minimal content would be indicated by ‘dagger.’ The intentional content constituted by dagger-feature satisfaction conditions are eminently sufficient. Thus, whether an objectual intentional state is directed at an actual, fictitious, or hallucinated object satisfaction conditions apply. Objectual intentional states in the latter two cases would lack an intentional object, though all three would have intentional contents, satisfaction conditions.

Importantly, intentional states regarding fictions or hallucinations fail to have any intentional object. For according to Searle’s and the extended theory, though all intentional states, objectual or otherwise, must have intentional content to secure their aboutness, there is no need for them to have mysterious nonexistent intentional objects. Indeed, the latter are proscribed.

9. Further explanation of the new concepts and brief recap of recent examples

Let me try to clarify what I mean by the concepts of thought-token and the content of a thought-token. The concept of thought-token is a theoretical construct introduced to indicate certain complex empirical facts related to what one is episodically thinking: whatever is going on in the head when one expresses what one thinks on a given occasion. Whatever that is, it is not something the agent internally inspects; instead, they are brain events of which the agent is unaware that directly prompt utterances of sentences.

Consider what happens when engaged in a discussion: one utters a temporal sequence of sentences. These are not first internally and serially inspected as mental sentences or neural representations in the head that the agent then “translates” into public sentences. Rather, the uttered sentences flow in consequence of the agent’s understanding or misunderstanding of whatever is being discussed. I assume this flow of sentences is s result of (1) various dynamical processes occurring in one’s head, (2) these processes are constrained by one’s understanding or misunderstanding of whatever is being discussed at the time, (3) such dynamical processes can change over time as one’s understanding increases or one gets confused, (4) they can prompt utterances of sentences.

The concept thought-token is meant to capture the agent’s understanding on an occasion, which is manifested in such dynamical processes. As such, thought-token is an idealization of a brain event at a point in time that directly prompts or would prompt an agent to utter a public sentence. The concept thought-token factors in the agent’s understanding as shaping such dynamical processes. It thereby serves to mark what she is actually thinking on a given occasion as a function of her degree of understanding and accounts for how its content may vary from the content of the sentence she may utter.

There is, then, an intrinsic vagueness in the theoretical construct ‘thought-token,’ as the dynamical processes are left wholly unspecified. Still, I think that there can be very little, if any, doubt that such processes occur which directly prompt our utterances; this is sufficient for my purposes.

When I speak of the ‘content’ of a thought-token, this is to convey what the agent is actually thinking and, to repeat, what the agent is actually thinking is determined by the extent of her understanding or misunderstanding of the topic of which she is thinking. It is the subjective reading of the intentional content. Cases like that of the quark example, clearly indicate differences between what agents may be thinking, and yet they may still appropriately use the same sentence to express their thought-tokens. Hence, the possible variation in the thought-token content from the publicly identifiable sentence content uttered to express it.

Even so, the subject or topic of a thought-token is rather determinate from the agent’s subjective point of view: she directly knows what she thinks she is talking about. (Of course, the agent can be wrong as to what is actually prompting her thought-token, as in the Bob and Al case.) The direct knowledge that the agent has of the subject of her thought-token is one reason why I introduced the concept of minimal content as conceptually distinct from, though a conceptual element of, the concept of thought-token. A more important reason for singling it out from the concept thought-token is because of the central role this aspect of thought-token, the concept of minimal content, plays in the application of my theory to other philosophical problems to which I allude at various points in this paper.

Highlighting some key features of the current theory, intentional states have both subjective and objective readings, similar to the intentional contents of intentional states. A thought-token is the subjective reading of intentional content, the subject of which is the minimal content. The objective reading of the intentional content of an intentional state is what an objective observer would ascribe to the agent, the subject of which is the objective content.

Lacking mental telepathy, to communicate, we express with words what we or others are thinking about. The earlier examples (Al/Bob and Sam) have been offered as an attempt to make it plausible that sometimes it would be in accord with our linguistic practices to give two different linguistic expressions, which designate different objects, to specify what is the subject of an agent’s intentional state. This allows us to characterize the agent’s intentional state from her own subjective perspective—what she is actually thinking on a given occasion—on the one hand, and what it is about from an objective perspective, on the other.

As we have seen, Al and Bob would express the subject of their intentional states with the proper name ‘Jones,’ as it would more accurately represent their minimal content—who they were thinking about. Note that in this case, the subjective reading of the subject of their intentional states would fail to have an intentional object, as Jones is not in the vicinity. Whereas on the objective readings of the subject of their intentional states there is an intentional object, viz., Smith raking leaves. Thus, the objective content, the objective subject of their intentional states, would be more accurately expressed by the name ‘Smith,’ as this is the individual (and what he is doing) actually prompting their thought-tokens, even though Smith is not who they had in mind. Similarly, Sam would express the minimal content of his thought-token, the subject component, using the expression ‘orange,’ not ‘kumquat,’ though the latter would be a more accurate expression of the objective content of the subject of his intentional state, what was prompting it.

As Al/Bob and Sam cases illustrate, an agent in linguistically expressing what he himself is thinking about might well use a nonequivalent expression to that which an objective observer would use. When there is a divergence between the subjective and objective reading of the subject of an intentional state, which linguistic item more accurately expresses the intentional content would depend on which reading of the question is at issue: Is the point of the question to get at what the agent actually has in mind and, therefore, might govern what she might further say or do, or is it to get at what is actually prompting her? In cases such as those considered, there is no correct univocal answer. The answer depends on which reading of the question “What is the agent’s intentional state about?” is at issue.

The Al/Bob and Sam cases presented were cases where different linguistic expressions were required to appropriately express the subject part of the intentional contents, depending on whether the subjective or objective reading was at issue; moreover, the expressions were not extensionally equivalent. Importantly, the same points sometime apply even when the differently linguistically expressed sentences are extensionally equivalent. A familiar example of this occurs with substitution in propositional attitudes, to which I now turn.

10. A restriction on many–one mappings and the substitution problem

Frege famously proposed a solution to the substitution problem based on his sense/reference distinction. For Frege, what mattered to thought’s content is not its reference but its sense—how its reference is thought about. My way of solving this problem relies not on sense, but on my concept of thought-token and permissible mappings from thought-tokens to sentences which express them. What an agent expresses is her thought-token, not abstract senses or propositions. Unlike Fregean senses, thought-tokens are decidedly subjective; they may vary considerably from one agent to another and still map to the same sentence to express them, as in the quark case. In such cases, many–one mappings are crucial. In the Al/Bob and Sam cases, what was central is not the many–one relation between thought-tokens and the sentences that express them, but how the agent’s understood their respective circumstances. My treatment of the substitution problem turns on a yet different condition on mapping, viz., the previously mentioned restriction on permissible mappings.

Consider Sally who believes that Mark Twain was a great American author though she does not know that ‘Mark Twain’ is a pseudonym for Samuel Langhorne Clemens, nor has she ever heard the name ‘Clemens.’ Now familiar ground. We may express this by saying:

1. (1) Sally believes that Mark Twain was a great American author.

Since the sentence ‘Mark Twain = Samuel Langhorne Clemens’ is true, substitution in (1) yields:

1. (2) Sally believes that Samuel Langhorne Clemens was a great American author.

Now while (1) would be clearly true in the described circumstances, (2) must give pause.Footnote ²³ If Sally were asked “When did you learn Samuel Clemens was an author?”, she would be perplexed, likely saying, “Who’s he?”, or some such, as Sally has never heard the name ‘Samuel Clemens’ prior to this occasion. She herself would never express or assent to the subordinate clause of (2), while readily assenting to or asserting the subordinate clause in (1). Though the two subordinate clauses are extensionally equivalent, the subordinate clause in (1) accurately and appropriately expresses Sally’s intentional content as she conceives the subject of her thought—her minimal content—while the subordinate clause in (2), though objectively correct, fails to either accurately or appropriately express the subject of her thought-token as she conceives it—her minimal content—and, thus, fails to express what she is actually thinking.Footnote ²⁴ It is clear that were Sally not ignorant of the relevant fact, either of the subordinate clauses would be appropriate expressions of her subjective intentional content. But if some such ignorance is operative, we must accommodate this limitation, if we are to express accurately and appropriately what the agent herself thinks.

In short, when the candidates for a linguistic expression to be used to express an agent’s intentional state are extensionally equivalent, but the agent is unaware of this, we must recognize the ambiguity in the question as to what an agent is thinking about and accommodate her limitation; we must restrict the range of mappings to sentences that do not mislead regarding the content of her actual thought-token. It is evident that the source of problems generated by substitution in propositional attitudes is the disconnect between how an agent subjectively conceives the subject of what she thinks—minimal content—on the one hand, and what would be a perfectly acceptable objective characterization of it—objective content—on the other.Footnote ²⁵

11. Does the new theory presuppose a theory of proper names?

How one understands proper names is obviously relevant to the recent example. My claims regarding the minimal content associated with a proper name might well lead one to suspect that I am sympathetic to a descriptivist’s view of proper names.Footnote ²⁶ The suspicion is correct, and this may cause some readers to think that what I have presented must be flawed, for it is widely held that Kripke’s Reference Kripke1972 refuted descriptivist theories of proper names. First, it is fair to say that Kripke refuted Frege’s and Russell’s theories; I think Searle’s theory of proper names is significantly different and is not subject to the same objections.Footnote ²⁷ That aside and more importantly for the current discussion is the fact that no matter how effective or ineffective Kripke’s objections are against Searle’s theory proper names, or descriptivists generally, what I want to stress is that they are irrelevant to what I have said regarding substitution in propositional attitudes. For my discussion of the problem of substitution in propositional attitudes did not rely on any theory of proper names; instead, it relied on what an agent is thinking, how she understands her thought-token and, thus, is independent of this or that theory of the semantics of proper names; indeed, it is also independent of how the referent of a name is determined.

This independence from any particular theory of proper names is a consequence of the crucial role played by the distinction I draw between the contents of thought-tokens and contents of sentences that express them, together with the many–one relation between them. Contents of thought-tokens are determined by what is going on in the agent’s head, her understanding, no theory of proper names here. A theory of proper names, or more generally, a semantic theory, bears on sentence content. Whatever might be the content of a sentence as delivered by, let us say, a correct semantic theory, it may well significantly diverge from the content of an agent’s thought-token notwithstanding the fact that the sentence is appropriately used to express an agent’s thought-token.

In the next section, we will see that Searle himself does not separate his theory of proper names from his theory of intentionality. Nevertheless, he tries to show why his theory of intentionality can resolve the problem of substitution in propositional attitudes while denying that descriptions that are associated with a proper name (or, as in his Reference Searle1983, intentional contents associated with a name) give the name’s meaning. I attempt to provide additional support for my alternative theory by arguing that though Searle was on the right track, his account fails in this attempt and fails precisely because he does not explicitly accommodate the inherent ambiguity in the question as to what an intentional state is about.

12. Subjective and objective interpretations of intentional content: Searle’s implicit but unsupported reliance

As just explained, my theory of intentionality is completely independent of any semantic theory of proper names. Searle, in contrast, joins the two: his Reference Searle1958 theory of names is merged with his theory of intentionality in his Reference Searle1983. The connection is clear since, for him, names have associated descriptions and intentional contents generally can be expressed using descriptions. Thus, he frequently speaks of intentional contents associated with a name in his Reference Searle1983, as well as descriptions.

Searle’s Reference Searle1958 explicitly argues against the claim that descriptions associated with names constitute a name’s meaning; his Reference Searle1983 similarly denies that intentional content gives the meaning of a proper name.Footnote ²⁸ Despite this, he maintains that the intentional content associated with a name can contribute to the propositional content of a statement made using a name. More precisely, he says:

If this is true, then it would appear that Searle can provide answers to the classic problems along Fregean lines, including the problem of substitution, even while denying that he is giving a theory of meaning of proper names. If correct, this would refute Kripke’s (Reference Kripke1972) claim that any apparent advantage the descriptivist has in solving the classic problems requires that associated descriptions give the meaning of names.Footnote ²⁹

As promising as [*] would seem to be to provide a solution to the problem of substitution along Fregean lines while denying that descriptions associated with a proper name give its meaning, there is a serious problem with it—it seems impossible for something to be part of the propositional content of a statement but not part of its meaning. Consequently, clauses [1] and [2] would appear to be irreconcilable. I think the source of this difficulty is Searle’s implicit vacillation between subjective and objective readings of ‘intentional contents.’ Let me explain.

Clause [1] arguably pertains to an objective interpretation of ‘intentional content.’ I say this for two reasons. First, if intentional content is univocal, as it is for Searle, there should not be any subjective variation in the intentional contents of two individuals who would express their respective intentional contents with the same linguistic expression. This is reinforced by the fact that the phrase “the propositional content of a statement made by a speaker” (as in clause [1]) normally does not depend on the speaker, on what content the speaker might associate with the name used in the statement made.Footnote ³⁰

My second reason for my claim that Searle sometimes treats ‘intentional content’ objectively is that he argues that there are parasitic uses of proper names. (Reference Searle1969, 170–71; he quotes this in his Reference Searle1983 [242–43, 252, passim]) A parasitic use of a proper name occurs when a speaker knows very little about referent of a name she uses. She is unfamiliar with many of the objective intentional contents (descriptions) associated with the name and may associate some which do not apply; nevertheless, she can still successfully refer to the referent with the proper name but only because objective intentional contents are associated with the name. She succeeds although she is not privy to these and, hence, they are not in her mind. Her use of the name in expressing her intentional content is parasitic on those who know the objective intentional contents associated with the name.

In contrast, clause [2] rather clearly requires the subjective interpretation of ‘intentional content’ since it is “the speaker’s associated intentional content” (italics added) that is appealed to there.Footnote ³¹ The possible variation in a speaker’s associated intentional content for a given proper name from that of others in her community is, of course, what makes it subjective.

Thus, Searle requires and implicitly deploys subjective and objective concepts of intentional content in his discussion. The shortcoming in his theory is that he fails to explicitly recognize and accommodate the distinct roles each of these two concepts play. This lacuna undercuts his claim (*), viz., that the intentional content associated with a name can be both part of the propositional content of a statement using that name and yet not be part of the meaning of the statement made by the speaker. The agent’s subjective understanding of her thought-token, her subjective intentional contents, is, I think, what Searle was trying to get at when he says that the intentional content associated with a name can be part of the propositional content without being part of the meaning of the sentence used to express her thought. His mistake was attempting to locate that intentional content in the propositional content of the statement made. This made [1] and [2] irreconcilable.

His dilemma is that, on the one hand, he needs the subjective intentional content to indicate how the agent conceives the subject of her thought, which is crucial to the solution of the substitution problem but, on the other hand, this cannot be part of the objective content of the statement which expresses her intentional content. On my theory, the subjective intentional content constitutes the agent’s thought-token, while the objective intentional content constitutes the semantic content of the sentence that appropriately expresses her thought-token, even while the two may differ in content. What Searle should have said is that the speaker’s subjective intentional content is what she was thinking—her thought-token—rather than its being part of the propositional content of the statement made, as [1] in the quotation maintains. This distinction is one my extended theory insists upon. Showing that thought-tokens map many–one to sentences that express them avoids the problem confronting Searle, since while the content of an agent’s thought-token is appropriately expressed by some public, objective sentence, the latter’s content may differ from the former’s. Thus, my theory explicitly incorporates subjective intentional content implicit in Searle’s theory, while also accommodating the objective content.

13. Concluding remarks

A theory of intentionality must incorporate the concepts of subjective intentional content and objective intentional content because of the inherent ambiguity in the question as to what intentional states are about. From the perspective of my theory, the context in which the question of aboutness arises is what determines whether it is one or the other or both that is relevant to the discussion at hand. The distinction between the contents of thought-tokens, on the one hand, and that of sentences which express them, on the other, is fundamental to my theory, as is the distinctively different roles they play. It is the many–one mappings between thought-tokens and the sentences that express them that enables their playing these different roles. It also proved important to distinguish the subject or topic of an intentional state according to whether the objective or subjective reading is at issue. This is accomplished with the concepts of objective content and minimal content, respectively.

An individual’s understanding both of her thought-tokens and her circumstances is, indisputably, central to what she is thinking; they are determinants of the content of her thought-token. On my theory: (1) subjective intentional content of intentional states is explicitly addressed by focusing on thought-tokens; (2) how the contents of thought-tokens may differ from the sentences which express them is accounted for via the concept of many–one mapping. Importantly, the way this second feature is secured does not wash out what the agent is actually thinking.

These two features are in accord with our ordinary linguistic practices of thought ascription, and together they explain the relation between one’s understanding of the expressions of her language, on the hand, and their public meanings in her language community, on the other. As the passage from Dummett quoted earlier indicates, I am not alone in recognizing the utmost importance of having a clear view of this relation. I have provided an account of this relation and it is incorporated in my theory of intentionality. I maintain that failure to explicitly accommodate it is at the core of many problems in the philosophy of language and philosophy of mind. I also contend that the proposed theory has the resources to recast and resolve many of these problems.Footnote ³²