Propositional Attitudes

The Philosophical Problem of Propositional Attitudes

A person can believe that a river has flooded, hope that the water will recede, fear that the bridge will collapse, and wonder whether a road remains open. Such cases exhibit a common structure: the agent is oriented toward something represented as being, becoming, or possibly being the case. Their differences concern how the agent takes that content. A belief presents a possibility as obtaining; a hope relates to its desired realization under uncertainty; a fear anticipates an unwelcome possibility; a question holds the matter open. Propositional-attitude theory asks what makes these descriptions true, what their contents are, how such attitudes figure in cognition and action, and why the language used to attribute them differs from ordinary extensional descriptions.

The philosophical problem is multidimensional. At the level of ontology, we ask whether there are distinct mental states with the specified content and attitude type. At the level of content, we ask what a proposition is and how a cognitive state comes to be about the world. At the level of psychology, we ask what causal, dispositional, representational, or functional role the state has. At the level of language, we ask how sentences such as “Mara believes that the road is open” acquire truth conditions. At the level of epistemology, we ask how beliefs and doubts answer to evidence, how thoughts can be mistaken, and when an interpretation is warranted. These questions interact, yet no answer to one automatically settles all the others.

Consider a city map used by a traveler. A map can depict a road without the traveler believing the road exists. Conversely, the traveler may believe the road exists without having an explicit map or an occurrent sentence in mind. A public inscription, a structured proposition, a neural state, and a dispositional pattern are distinct candidates for explaining the thought. Confusing the vehicle with its content, or the content with the attitude taken toward it, creates spurious solutions to the problem of mentality. This distinction underlies the major debates documented in the Internet Encyclopedia of Philosophy's survey of propositional attitudes and the Stanford Encyclopedia's analysis of attitude reports.

A research program must also resist a common asymmetry. Some philosophers treat statements about belief and desire as straightforward reports of real psychological kinds; others interpret them as useful predictions or interpretive practices. Neither a successful prediction of behavior nor a successful compositional semantic analysis is sufficient to identify a single underlying psychological entity. Conversely, the absence of one expected kind of internal symbol would not establish that no pattern deserves the name belief. The question is which commitments follow from which evidence and arguments.

Definition, Scope, and Constitutive Distinctions

A propositional attitude is, in its central analytic use, a kind of intentional state or orientation attributed to a bearer with respect to content expressible by a proposition. We may represent the ascription schematically as A(s, p), where A identifies an attitude type, s an attitude bearer, and p the content. This notation is an instrument of analysis. Its two-place surface grammar does not establish that the bearer literally stands in a primitive relation to an abstract object. The contested nature of that metaphysics is itself part of the research subject.

The adjective “propositional” identifies the content's susceptibility to representation as what is the case. The adjective “attitude” identifies the different orientation a subject may take toward such content. Hope that it rains and belief that it rains may agree in content yet differ in what would count as satisfaction. Belief that it rains and belief that it snows may agree in kind yet differ in content. There can also be mixed, conditional, interrogative, and nonassertoric attitudes whose precise logical analysis remains controversial. A belief about an individual may be reported by a nominal phrase without an explicit that-clause; the report's grammar is therefore a diagnostic rather than a necessary constitutive criterion.

Four distinctions structure the field. First, the act of entertaining a thought differs from a standing state that continues when unattended. Second, the truth of a proposition differs from the truth of an attribution of an attitude toward it: false content may be genuinely believed. Third, a mental representation differs from the proposition or circumstance it represents. Fourth, the explanatory use of attitude discourse differs from a theory of the ultimate constituents of the thinker. The proper subject includes these relations and the reasons given for treating them as interdependent.

Belief, desire, hope, fear, intention, supposition, acceptance, doubt, imagining that, and knowing that are the central cases. Some are straightforwardly propositional in familiar applications; others have objectual, practical, emotional, or interrogative forms that demand further analysis. Knowledge also differs from bare belief through its success conditions. The range of cases guards against treating belief as the only pattern and then redefining every other attitude as a minor variation of it. The present dossier concentrates on the general architecture of propositional attitudes. The epistemology of belief, full theories of mental representation, and the logic of particular attitude operators receive their own independent studies.

Ancient Foundations: Opinion, Judgment, and Desire

The modern technical phrase does not belong to ancient Greek authors. Its philosophical antecedents do. Plato's Theaetetus examines knowledge in relation to true opinion and the possibility of false judgment (187b–200d). The inquiry forces a question of representation: if a thinker judges falsely, what is the object of thought, and how can a cognitive act present something otherwise than it is? False judgment cannot be explained simply by identifying thought with contact with an existing external thing, because falsity requires the possibility of misrepresentation. The dialogue explores mistaken identification, the wax-block analogy, and the aviary of acquired knowledge as provisional models. Each fails in a revealing way to explain all cases of false judgment. These are historical problems about doxa, knowledge, and judgment; describing them as debates about contemporary propositional-attitude semantics is an interpretive reconstruction.

Aristotle's De Anima distinguishes perception, imagination (phantasia), thought, and judgment. Book III, chapter 3 (427b–429a) treats imagination as capable of falsity and distinguishes it from assertoric opinion, which involves conviction. Aristotle thus identifies differences among cognitive capacities that later theories would distinguish by attitude type. In De Anima III.10 (433a–434a), practical movement requires an interaction of cognition and appetite; what an agent apprehends and what the agent wants jointly contribute to action. The structure anticipates the explanatory power of belief–desire analyses without committing Aristotle to a modern theory of propositional objects.

Stoic philosophy offers another distinct historical resource through accounts of assent to impressions (phantasiai) and their sayables (lekta). A representation may be presented to a rational agent who assents, withholds assent, or rejects it. The separation of impression, assent, and expressible content helps explain how an attitude may be assessed independently of the truth of what is presented. Our surviving access is largely mediated by later reports and collections of fragments, a transmission constraint that prevents easy identification of a single canonical Stoic statement as a modern theory of belief. The important conceptual gain is the difference between receiving an appearance and taking responsibility for affirming it.

Ancient sources accordingly establish three durable problems: how judgment can be false; how cognition differs from apprehension or imagination; and how what is judged interacts with what is desired. These are functional ancestors, while the subsequent semantics of propositional attitude ascriptions arises from different logical and linguistic pressures. Historical continuity is strongest at the level of recurring problems, rather than anachronistic terminological identity.

Medieval Intentionality and Mental Language

Medieval philosophical psychology provides a bridge between questions about cognition, the objects of intellect, and the relation between mental acts and linguistic signs. In discussions influenced by Aristotle, mental acts can be directed toward universal natures, individual things, and objects absent from immediate perception. The distinction between the cognitive act and the object cognized becomes decisive when an object is nonexistent, fictional, or incorrectly represented. An account of intentionality must explain how thinking of a nonexistent island can be a genuine episode without requiring the island to exist as an ordinary geographical object.

Thomas Aquinas describes intellectual understanding as the acquisition of intelligible form and distinguishes the act of understanding from the external thing understood (Summa theologiae, I, q. 85, arts. 1–2). The philosophical task is not exhausted by calling the content an inner picture: doing so risks making the mental species the final object of cognition and obscuring the intellect's directedness toward things. His account belongs to an Aristotelian theory of cognition and its causal orientation; later frameworks that cast the problem as one of subject–proposition relations offer a different analysis of the same basic phenomenon.

William of Ockham develops a theory of mental terms and propositions in the Summa logicae, part I, opening chapters. Mental language supplies a model in which concepts are signs naturally significant of things and combine into proposition-like structures. The proposal matters because linguistic form can be located at the level of thought, making inference and conceptual composition systematic. Yet calling thought a language leaves further questions about the metaphysics of signs, how they acquire significance, and whether their combinatorial properties map onto modern syntax.

Medieval discussions of intentionality are internally diverse. A sign theory, an account of intelligible species, and an account of mental acts constitute different explanatory projects. Their collective contribution is to make clear that relation to an object can involve cognitive form, internal signification, and active assent; each may be present even when the object is remote or the represented state of affairs does not obtain. A responsible history thus identifies partial conceptual continuities instead of assigning the same doctrine to the scholastic tradition as a whole.

Early Modern Thought: Ideas, Assent, and Passion

Early modern philosophy places the relation between ideas, judgment, and will at the center of disputes about error and certainty. Descartes's Meditations on First Philosophy (1641), especially Meditation III, distinguishes ideas from the judgments made concerning them and argues that error arises in part when the will extends affirmation beyond clear intellectual perception. In Meditation IV the investigation of error concerns the relation between the finite intellect and the freedom of judgment. The doctrine embeds an account of attitude kind within a metaphysics of mind and a program of epistemic justification; it is not merely a classification of reports introduced by verbs.

Locke's Essay Concerning Human Understanding (1690), Book IV, particularly chapter 15, distinguishes probability from certainty and treats assent as admitting degrees. This introduces a spectrum between unqualified belief and suspension. The significance for the modern subject is that an attitude's degree of commitment cannot always be captured by the binary question whether a person believes that p. Later probabilistic approaches formalize degrees of confidence, but the historical argument concerns the proper proportion between assent and available grounds rather than a technical Bayesian probability calculus.

Hume's Treatise of Human Nature (1739–1740), Book I, part III, section 7, asks what distinguishes believing an occurrence from merely conceiving it. Belief, on his account, involves a characteristic vivacity or felt manner of conceiving rather than the addition of a further representational object. This proposal makes attitude type dependent on the mode of a cognitive presentation. It also introduces a systematic problem: can differences of conviction be identified with phenomenological force when many standing beliefs exert influence without any present feeling?

Spinoza's Ethics (1677), Part II, proposition 49, explores the relation of ideas and affirmation, questioning whether an independent volition is required for judgment. His argument bears on the possibility of treating belief as a separable act of will directed toward a neutral proposition. The differences among Descartes, Locke, Hume, and Spinoza show why contemporary accounts require a distinction between what is entertained, the degree or mode of assent, and the psychological mechanism of endorsement. They also show that the dependence of belief upon voluntary choice is a substantive thesis rather than a definition.

Brentano, Husserl, and Intentional Directedness

Brentano's Psychology from an Empirical Standpoint (1874), Book II, chapter 1, makes intentional directedness central to the distinction between mental and physical phenomena. A mental episode includes a relation to something as its object, whether or not that object has straightforward existence outside the episode. The celebrated doctrine of intentional inexistence has generated extensive interpretive dispute: it should not be read immediately as the assertion that a miniature object is physically contained in consciousness. Its philosophical role is to describe the characteristic aboutness of mental life and to identify something a material or purely formal psychology must explain.

Brentano's classification of mental phenomena into presentations, judgments, and phenomena of love and hate is especially important. It denies that every psychological attitude is merely a relation to a full proposition. Judgment is in a significant sense an acceptance or rejection built upon presentation; affective orientation has its own form. The historical view therefore challenges a purely propositional generalization even while contributing an indispensable theory of intentionality.

Husserl's Logical Investigations (1900–1901), especially the Fifth Investigation, distinguishes an intentional act's quality from its matter, enabling different modes of intentionality to concern the same object or meaning. The doctrine belongs to a phenomenological analysis of conscious acts and ideal meaning; it must be distinguished from a computational account of data structures. The act-quality/matter distinction is nevertheless an instructive parallel to the contemporary attitude/content distinction, particularly where belief, wish, questioning, and perception have related subject matter.

The phenomenological tradition raises a critical objection to purely external attribution: a description of behavior might miss how an object is experienced or given in a particular intentional orientation. A reply must explain whether first-person manner of presentation is part of attitude content, part of a mode of access, or a separate experiential feature. Brentano and Husserl show why there may be more in intentional directedness than the truth conditions of an embedded linguistic clause. Their work also cautions against assuming that all intentionality must have propositional content.

Frege: Sense, Reference, and Cognitive Significance

Frege's “Über Sinn und Bedeutung” (1892), commonly translated “On Sense and Reference,” supplies a central logical motivation for distinguishing an attitude's subject matter from its mode of presentation. Two names can designate the same object while informing a thinker differently. A person can recognize one identity claim as informative while treating another as trivial. If cognitive significance were completely determined by external reference, the difference would be obscure.

The difficulty becomes especially sharp under verbs of belief. An agent can believe that the evening star is visible without recognizing that the morning star is visible, even though both astronomical descriptions refer to Venus. The truth of an ascription can thus change when a co-referring expression is substituted within an embedded clause. Frege proposes that in indirect contexts an expression can refer to its customary sense. Attitude reports accordingly involve a shift in semantic contribution relative to the ordinary extensional use of terms. This account connects facts about language to the way a thinker apprehends an object.

The strength of Frege's solution lies in its explanatory precision. It explains why two reports can differ in truth value without attributing logical incompetence to the thinker. It also recognizes that a content adequate to account for cognitive significance may be more finely individuated than an external referent. The difficulty is to develop a principled account of senses and their individuation and of how nested attitude contexts behave. Once indirect reference is admitted, an account must specify what expression denotes in reports embedded within further reports, and how that semantic hierarchy is governed.

The issue is not resolved by declaring that modes of presentation must be mental pictures. Frege's senses have a logic and objectivity distinct from private psychological imagery. The Stanford Encyclopedia's account of attitude reports reconstructs the resulting dispute between Fregean and Russellian approaches. The lasting thesis is that identity of external reference and identity of cognitive significance cannot simply be equated.

Russell, Wittgenstein, and Logical Form

Russell's analyses of denoting, acquaintance, and propositional structure give a different explanation of the relationship between thought and objects. “On Denoting” (1905) treats definite descriptions as expressions to be analyzed through quantificational structure rather than simply as names of special descriptive entities. The distinction between primary and secondary occurrences permits the apparent object of an attitude to be placed within or outside the scope of an attitude operator. Russell's broader theory of acquaintance and description, especially as developed in The Problems of Philosophy (1912), also asks which objects can figure directly in a subject's thought and which are cognitively accessed through descriptions.

For instance, the assertion that Nora believes that the leader of an expedition is brave may report a belief about whoever fills that description or a particular individual's bravery under some mode of identification. These readings have different truth conditions. The scope of a description relative to “believes” is therefore not mere typographical detail: it determines what the report commits its speaker to. Russell's analysis supplies a formal resource for separating what a person thinks from how a reporter chooses to identify the person or object thought about.

In the Tractatus Logico-Philosophicus (1921 German publication; 1922 English–German edition), propositions 5.541–5.5422, Wittgenstein questions the simple appearance that a proposition is related to a psychological object or an isolated subject in the form “A believes p.” His discussion treats the logical form of representing and saying as involving coordination among facts. The text addresses the requirement that a judgment be a judgment with sense; it criticizes views that permit the subject to stand in a judging relation to nonsense as though this were a genuine proposition.

This argument yields a challenge to primitive relational accounts. If the proposed belief relation can hold between any subject and any syntactic string, then a relation to a nonsensical sequence might count as a belief. The theory requires an account of the conditions under which the second relatum counts as a proposition, and an account of the cognitive capacity to grasp it. Wittgenstein's reasoning does not in itself disprove that subjects believe propositions. It forces theories to explain the logical structure of understanding that makes meaningful belief possible.

Later Wittgenstein's Philosophical Investigations (published 1953), especially §§243–271, develops investigations of criteria, language, and the intelligibility of mental ascription. The private-language discussion bears on the conditions for attributing stable inner meanings and psychological states. Read together, these works oppose the idea that adding a mental object automatically solves the problems of understanding and communication. The logical and public standards that make an attribution intelligible are themselves philosophical data.

The Ontology of Propositions and Attitude Objects

The apparent form “subject S believes proposition p” invites a relational ontology. If a proposition is an object distinct from a particular thought or sentence, what kind of object is it? Among prominent answers are structured propositions composed of objects and properties, Fregean thoughts constituted through senses, abstract intensions characterized by possible-world truth conditions, and content individuated by inferential or cognitive roles. These are substantive alternatives. A proposition as a set of worlds groups together every thought true at exactly those worlds; a structured account can distinguish some thoughts with the same truth conditions through their constituent structure.

The possible-worlds model is elegant: if a proposition p corresponds to the set of worlds in which p is true, belief can be modeled as excluding worlds incompatible with the agent's information. Yet this model is coarse-grained. Necessarily equivalent propositions have identical extensions over possible worlds, even when they differ dramatically in mathematical complexity or cognitive accessibility. A subject may understand a basic arithmetic identity while being ignorant of a sophisticated theorem with the same necessary truth value. Thus a state description that identifies all necessary truths as one content loses distinctions important for psychological attribution.

Structured propositions can distinguish components and their order, allowing the content that Alice admires Ben to differ from the content that Ben admires Alice. Yet structure alone is insufficient when a person thinks of one object in two different ways. A Russellian singular proposition containing the same individual may still be an object of distinct cognitive attitudes. A theory may therefore need to supplement public propositional content with representational modes, ways of grasping, or states of access.

A third option identifies attitudes primarily with relations to mental or linguistic representations. Here the representation serves as a vehicle and its semantic properties supply content. This addresses cognitive accessibility and causal implementation, but it generates another question: by what facts does a symbol possess its content? If the answer invokes the belief or intention of an interpreter, the explanation risks presupposing the intentionality it was meant to naturalize. Informational, teleological, inferential, and social accounts offer different responses, each with its own difficulties.

The choice of propositional ontology affects theories of identity, learning, inference, and disagreement. If the content objects are too coarse, distinct states collapse; if too finely specified, common knowledge or disagreement between speakers becomes unexpectedly difficult to explain. A theory must therefore state which distinctions answer to linguistic truth conditions, which to cognitive processing, and which to the social practice of interpretation. No single criterion of individuation can be assumed adequate for all purposes before its explanatory role is established.

Opacity, Substitution, and Quantification

Attitude reports are paradigm cases of intensional or opaque contexts. Outside such contexts, replacement of an expression with a co-referential term ordinarily preserves the truth of an extensional claim. If one object has two names, a true assertion about it can normally be restated using either. Inside a belief report, a substitution may change the accuracy of the attribution because the believer may recognize the object under one description but not another.

Let m and n co-refer, and let B(s, F(m)) mean that s believes that m is F. The inference from m = n and B(s, F(m)) to B(s, F(n)) is not generally valid as a principle governing ordinary reports. The failure concerns the relation between the speaker's identifications and the subject's cognitive perspective. It does not imply that identity itself is relative or that two numerically identical objects diverge in their actual properties.

Quine's “Quantifiers and Propositional Attitudes” (1956) explores the resulting difficulties of quantifying into opaque contexts. What is asserted when one says that there is somebody whom a subject believes to be a spy? Such a report may require an interpretation on which an individual is identified independently of the believer's description. Another report might state only that the subject believes someone or other is a spy. Apparent existential import changes with scope and interpretation. Quine's analysis makes vivid that a grammar of variable binding over object positions cannot be applied mechanically to every expression embedded in attitude discourse.

The apparent paradox can be expressed in a simple example. A historian believes that the author of a certain manuscript lived in exile, but has no idea who authored it. An external researcher identifies the author as a particular general. We may accurately say, in one contextualized sense, that the historian believes of the general that he lived in exile, while a report that the historian believes the general lived in exile might misleadingly suggest recognition under that title. The report must specify the intended relation between the subject's perspective and the interpreter's identifying practices.

Two responses are available in principle. One locates a semantic shift in embedded expressions, as Frege does. Another keeps directly referential content while positing context-sensitive ways of reporting or hidden modes of presentation. The second retains a coarse object-directed content but must justify the additional contextual parameters. Neither answer should be chosen solely because it produces a tidy logical notation. The account must preserve independently attested facts about judgments, recognition, reference, and reporting practices.

De Dicto, De Re, and De Se Attitudes

The distinction between de dicto and de re interpretations tracks an important difference in what an attitude report says. A de dicto attribution concerns a proposition under a description or mode supplied within the reported thought. A de re attribution concerns a particular thing to which the attitude is directed, even if the subject would not identify that thing under the reporter's terms. These are families of interpretive analyses; they are not guaranteed to correspond to separate psychological compartments in an agent's mind.

Suppose Lea hopes the museum director will receive a scholarship. If Lea knows the director only by that institutional role, her hope may concern whoever occupies it. If Lea has a particular person in mind, an attribution of hope about that person may remain true regardless of how the reporter identifies the individual. In a logically regimented language, the positioning of a definite description or existential quantifier relative to the attitude operator can model the difference. Natural-language context and presupposition may require more than a scope distinction alone.

Indexical and self-locating attitudes present another layer. John Perry's “The Problem of the Essential Indexical” (1979) describes cases where a person's realization “I am the person making this mess” changes behavior despite the availability of general information describing the same individual. David Lewis's “Attitudes De Dicto and De Se” (1979) develops the idea that an attitude can concern the subject's location within a possibility, not merely which uncentered possible world obtains. The issue is especially clear in the amnesiac who knows a complete objective description of a building yet does not know which person in the description is herself.

One formal treatment models contents by centered worlds: a possible world together with a specified individual and time, or an equivalent index. Knowing all ordinary world facts is then insufficient for knowing one's own location among centers. A competing view locates the special feature in the subject's way of thinking or in a mode of presentation attached to a more ordinary proposition. The dispute concerns whether the explanatory difference belongs to content itself, to a manner of apprehending it, or to the functional role of indexical thought. See the Stanford Encyclopedia's reconstruction of self-locating belief.

De se cases provide a stringent test for any account of attitudes. A theory that identifies content solely with a possible-world truth set must explain why acquiring no new uncentered information can nevertheless change rational action. A theory that assigns a unique subjective content to every such episode must explain how communication and shared reference remain possible. A successful account must preserve both the agent's perspectival orientation and the common world about which agents reason.

Contemporary Semantics of Attitude Reports

The contemporary semantics of attitude reports is a laboratory for competing theories of meaning. Fregean accounts place senses or modes of presentation at the center; Russellian accounts emphasize direct reference and structured propositions; possible-world semantics analyzes attitudes through sets of alternatives compatible with the agent's informational state; and contextualist and hidden-indexical approaches seek to explain differences among reports by additional parameters. Each answers some data but confronts constraints arising from substitution, quantification, multiple embedding, and the interpretation of names and descriptions.

Saul Kripke's “A Puzzle about Belief” (1979) deepens the challenge. The puzzle shows that belief ascription can become paradoxical even without ordinary replacement of one coreferential name by a different one. A bilingual speaker may assent to a sentence in one language and deny a sentence in another, unaware that the names designate the same place. Applying apparently innocent principles of disquotation and translation may produce inconsistent reports about what the subject believes. The problem illustrates the difficulty of aligning sentence assent, semantic content, and rational belief with a simple general rule.

Mark Crimmins and John Perry's “The Prince and the Phone Booth” (1989) develops a way of capturing differences among attitudes and reports through cognitive ideas or modes of presentation while retaining singular propositional content. Their proposal is valuable because it differentiates the referent involved in thought from the agent's way of accessing it. Critics must ask whether the added parameter belongs to the assertion's truth conditions, conversational pragmatics, or the believer's cognitive architecture. A hidden parameter that can be freely adjusted risks immunizing the account from test.

A further issue concerns reports involving desire and imagination. A sentence such as “Nora wants to find a unicorn” need not imply that an existing unicorn is among the objects desired. The report may specify satisfaction conditions for an unrealized future possibility. Likewise, nested reports—“Eva suspects that Nora believes the rumor”—require a compositional explanation of successive attitude embeddings. Recent discussions of propositionalism examine whether all such constructions can be rendered as relations to truth-evaluable content or whether objectual and pictorial contents demand richer semantics. The 2024 Annual Review discussion of intensionality and propositionalism is one useful guide to the live debate.

A key methodological conclusion follows. The semantics of a report is about the truth conditions of what an interpreter says. The metaphysics of the reported state concerns the bearer, content, and realization that make the attribution accurate. A semantics may be technically successful even if different psychological theories realize the same reported pattern. Conversely, an empirical theory may illuminate processing without resolving which natural-language reports are true in a given context. Their legitimate interaction requires an explicit bridge principle connecting linguistic interpretation to cognitive evidence.

Belief, Desire, Intention, and Directions of Fit

Belief and desire illuminate why attitude type matters independently of content. Believing that a meeting will occur tomorrow commits the believer, in an epistemic sense, to the meeting's occurrence. Wanting the meeting to occur has a practical orientation toward bringing it about or welcoming its occurrence. Philosophers characterize this by different directions of fit: belief is assessed against the world; desire makes a world-directed demand. The distinction is structural rather than a psychological law that beliefs are always accurate or desires always motivate immediate action.

J. L. Austin and Elizabeth Anscombe supply important resources for distinguishing descriptions, intentions, and action. Anscombe's Intention (1957) examines the special sense in which an agent knows an intended action by engaging in it, and the role of “Why?” questions in distinguishing intentional from merely observed behavior. The investigation shows that practical understanding includes a form of orientation to acting that cannot be inferred from an isolated descriptive belief. John Searle's Intentionality (1983) develops conditions of satisfaction and direction of fit as central analytical resources, while also relating intentionality to perception, action, and linguistic acts.

Intentions differ from ordinary desires through commitment to a course of action, practical planning, and coordination over time. Michael Bratman's Intention, Plans, and Practical Reason (1987) argues that plans and intentions help stabilize deliberation and organize future activity; they are not simply strong desires paired with beliefs. A person can desire mutually incompatible outcomes while intending one feasible course. Plans also provide norms of means–end coherence, which may operate even when desire intensity fluctuates.

Fear, hope, and expectation introduce further complexities. Fear can be directed toward an outcome regarded as possible without taking its occurrence as established. Hope combines an evaluative orientation with an unsettled possibility, but its degree of optimism varies. Suspicion is not simply low-confidence belief in every context; it can carry an investigative disposition and an attitude toward potential evidence. Imagining that something is the case differs from believing it because its imaginative deployment does not ordinarily require epistemic endorsement. Treating each of these merely as belief plus a numerical parameter would erase practical and expressive distinctions.

The category of propositional attitudes is thus unified by directedness toward representable content, not by one norm or causal trajectory. Each kind is characterized by distinctive satisfaction conditions, transitions, behavioral roles, and interrelations with other kinds. A philosophical taxonomy must be justified by explanatory and normative differences among attitudes, and must remain responsive to cases where the usual distinctions overlap.

Rationality, Normativity, and the Explanation of Action

Reasons explanations link belief and desire to intentional action. Suppose a traveler enters a station because she wants to catch the last train and believes the departure platform is there. The explanation identifies what made the action intelligible from her point of view. Yet an explanation that merely states a desire and belief leaves open a familiar problem: a person may possess those attitudes without acting on them, or may act for a different reason. A successful account must identify the reason that actually explains the particular action.

Donald Davidson's “Actions, Reasons, and Causes” (1963) argues that an agent's primary reason, involving a pro-attitude and a relevant belief, can both rationalize and cause an action. The central point is that displaying an action as intelligible under a description and explaining why it occurred are connected tasks. Davidson's discussion challenges a sharp separation between causal and reason-giving explanation. The original article, Journal of Philosophy 60(23), 685–700, is available in a university-hosted scan.

The causal thesis faces a problem of deviant causal chains. An agent's fear and belief may cause nervousness that accidentally produces the intended outcome, without the agent's action being intentional in the relevant way. Showing that the correct attitudes causally contributed is therefore insufficient; the causal pathway must sustain appropriate practical control. This difficulty motivates accounts of intention and planning that distinguish rationalizing causes from incidental psychological antecedents.

Beliefs are also normatively governed by relations of evidence, consistency, and inferential support. If an agent believes p and believes that p implies q, there is a prima facie rational pressure to accept q, but it does not follow that ordinary believers invariably possess all logical consequences of their beliefs. The familiar logical omniscience problem emerges when formal models idealize belief as closed under logical consequence. A human agent's limited memory, attention, and computational capacity show why norms of reasoning should be distinguished from descriptions of actual doxastic state.

Normative pressure differs across attitude kinds. Belief answers to considerations of truth; intention to coherence, feasibility, and means; desire to practical evaluation under some theories; fear to perceived danger, salience, and susceptibility to harm. Some pressures are constitutive of a type, others rational ideals or social standards. A complete theory should distinguish the conditions for possessing an attitude from the standards under which that attitude is judged reasonable. Otherwise, irrational belief would cease to count as belief by stipulation, making a central feature of psychology inexplicable.

Content Internalism, Externalism, and Social Meaning

The individuation of propositional content has a deep consequence for the ontology of belief. Internalist views privilege characteristics intrinsic to the thinker or accessible within an individual's cognitive system. Externalist views hold that the contents of thoughts can depend partly on physical surroundings, social conventions, or historically established links with objects. An attitude's content may therefore be broader than its immediate psychological realization.

Hilary Putnam's “The Meaning of ‘Meaning’” (1975) develops Twin Earth cases concerning natural-kind words. Two intrinsically similar speakers may employ “water” in environments with different underlying substances. Their reference differs even though familiar internal associations are very similar. Putnam's argument primarily addresses meaning and reference; its consequences for thoughts arise when the mental concepts involved are connected to public-language reference and environmental relations.

Tyler Burge's “Individualism and the Mental” (1979) adds a social dimension. His cases of incomplete understanding of a term such as “arthritis” suggest that a thinker may entertain a concept whose accurate application is partly established by a surrounding linguistic practice. Believing something false about the term need not prevent the person from having a genuine belief involving it. Burge's article in Midwest Studies in Philosophy gives a fundamental argument against treating all intentional content as wholly determined by narrow individual conditions.

A major objection appeals to causal explanation. If intrinsically similar individuals display similar behavior because their internal states are alike, what explanatory role do their different broad contents play? One possible response distinguishes the role of internal implementation in immediate processing from broad content in explaining environmental success, knowledge, and reference. This proposal remains incomplete until it specifies the bridge between those explanatory descriptions.

Externalism also challenges simplistic claims of introspective omniscience. An individual can know that she believes a claim while remaining ignorant of the conditions fixing the content of some of its terms. First-person authority over an attitude does not imply exhaustive knowledge of its semantic determinants. The task is to identify what self-knowledge actually guarantees, rather than to choose between infallible introspection and a complete loss of first-person knowledge.

Functionalism, Dispositions, and Multiple Realization

Functionalism identifies mental kinds through organized relations among inputs, other states, and outputs. The belief that a storm approaches, for example, can be related to perception of dark clouds, expectations of rain, a desire to remain dry, and decisions to find shelter. The relevant state may be realized through different physical mechanisms while sustaining a comparable role in the system's economy.

David Lewis's “Psychophysical and Theoretical Identifications” (1972) proposes understanding psychological kinds by their place within interconnected commonsense generalizations and identifying their actual realizers. This view moves beyond an isolated definition of belief and desire. It also faces the difficulty of specifying a role network with enough precision to distinguish attitudes while permitting individual variation, error, and irrationality.

Gilbert Ryle's The Concept of Mind (1949) is an important historical source for dispositional analyses. In a contemporary development, Eric Schwitzgebel's “A Phenomenal, Dispositional Account of Belief” (2002) models belief through resemblance to a characteristic pattern of dispositions. Such accounts handle standing attitudes that persist without current conscious episodes, as well as inconsistent cases in which a subject avows a conviction but repeatedly acts against it.

The cost of flexibility is a problem of thresholds. If possession of a belief means matching a sufficient cluster of dispositions, what determines sufficiency? A rigid checklist misclassifies agents with limited capacities, while a vague resemblance criterion makes disagreement about attribution difficult to adjudicate. An adequate theory should connect the pattern to relevant explanation and rational appraisal, specifying when the case is borderline and when a different attitude kind explains it better.

Multiple realization is best understood as a possibility constrained by functional organization. It neither proves that every outwardly similar device has the same belief nor requires that realizers share a single biological substrate. The question is whether a proposed role tracks a genuinely stable explanatory structure across counterfactual situations.

The Language of Thought and Computation

Jerry Fodor's The Language of Thought (1975) and “Propositional Attitudes” (1978) argue that structured mental representations can explain systematic and productive cognition. If someone can entertain the thought that Ada admires Ben, she will normally possess related capacities involving these concepts in different arrangements. A compositional representational system offers a clear explanation: reusable constituents are arranged by rules that affect whole-thought content.

The representational theory of mind distinguishes vehicles from content. A Mentalese expression, as a hypothetical inner vehicle, can be manipulated computationally according to structural properties. Its intentional interpretation supplies what the expression is about. The second part of this account is indispensable. A purely syntactic manipulation of symbols does not yet explain why they mean what they do.

The positive argument is an inference to explanatory adequacy. Systematic relations among thoughts require explanation; compositional representational architectures explain many such relations; therefore the hypothesis deserves serious consideration. The conclusion is defeasible. Distributed, imagistic, spatial, and other cognitive architectures might explain some systematic capacities without literally employing a sentence-shaped internal language. The target should be specified by cognitive task and level of description.

Fred Dretske's information-based semantics and Ruth Millikan's teleosemantic work seek to explain representational content through information and proper functions rather than treating content as primitive. These views address misrepresentation and the relation between proximal signals and distal environmental conditions, while confronting questions about content indeterminacy. Teleological Theories of Mental Content surveys these difficulties and the distinction between producer and consumer mechanisms.

Computationalism is also distinct from the claim that every mental state is functionally individuated. The implementation of a computational theory, the semantics of its representations, and the normative assessment of inference are separate explanatory commitments. The Stanford Encyclopedia's Computational Theory of Mind develops this distinction in connection with recent research. The philosophical task is to show how causal computation, semantic interpretation, and rational thought fit together without allowing any one vocabulary to substitute for the others.

Interpretationism, Holism, and the Intentional Stance

Interpretationism investigates the conditions under which beliefs and desires can be attributed as part of an intelligible pattern of behavior. Donald Davidson's work on radical interpretation places meaning, belief, and rationality in an interdependent structure. One cannot simply read an isolated belief from a behavior without some account of the subject's other commitments, circumstances, language, and reasons. Rational interpretation offers constraints on the assignment of mental content, even where the person being interpreted makes mistakes.

This holism is both an insight and a challenge. The same utterance can express different beliefs against different backgrounds, and understanding one attribution may require many others. But if the identity of every belief depends on an entire personal theory, two people with different histories might never share a belief. A theory therefore needs principles for local stability of content amid widespread partial disagreement. Interpreters must be able to distinguish a subject's false beliefs from a wholesale failure to identify any subject matter at all.

Daniel Dennett's The Intentional Stance (1987) proposes treating a system as a rational agent whose beliefs and desires can make its behavior predictable. This practice is an important explanatory level even when no one-to-one internal sentence corresponds to each attribution. The success of the stance depends on robust patterns of counterfactual behavior rather than the verbal elegance of an invented story. A prediction that succeeds only because a reporter freely revises the attributed beliefs has limited explanatory force.

An intentional description may be useful for a simple control system and much richer for a person planning a complicated journey. Such usefulness alone does not establish that both have identical psychological architectures. Interpretationist realism must state what makes the pattern objective, how to measure explanatory success, and which attributes of rationality the interpretation presupposes. Its strongest contribution is to remind theories of internal realization that content-laden explanation serves social, practical, and communicative functions.

Eliminativism, Connectionism, and Embodied Cognition

Paul Churchland's “Eliminative Materialism and the Propositional Attitudes” (1981) challenges the assumption that a mature scientific psychology will preserve the familiar categories of commonsense mental explanation. Stephen Stich's From Folk Psychology to Cognitive Science (1983) similarly questions whether a semantic taxonomy of beliefs is appropriate for cognitive science. Their argument is significant precisely because familiar attitude terms can be experimentally or theoretically revised, just as other explanatory categories have been.

The central eliminativist inference is conditional: if ordinary belief–desire psychology posits a network of kinds whose essential explanatory commitments fail, and a more successful theory does not preserve corresponding states, then retention of those kinds becomes questionable. A critic must address both premises. The alleged failures may concern only a narrow representational picture, while the replacement framework may preserve higher-level stable dispositions or patterns of intentional interpretation.

Connectionist systems demonstrate why the distinction matters. A distributed pattern of activation can encode task-relevant information without storing one independently identifiable sentence per proposition. Such an architecture challenges a literal model of private language without refuting all theories that individuate belief functionally or interpretively. A state description and its physical realization need not exhibit the same granularity.

Embodied and ecological approaches analyze cognition through agent–environment dynamics, motor capacities, and context-sensitive interaction. Anthony Chemero's Radical Embodied Cognitive Science (2009) offers a prominent argument for reducing the explanatory centrality of internal representations in certain cognitive tasks. The proper comparison must be task-specific: explaining continuous sensorimotor coordination is not automatically equivalent to explaining counterfactual historical inference or the revision of a mathematical hypothesis.

A further position treats some attitude ascriptions as useful fictions or pragmatic instruments rather than literal reports of inner states. Contemporary discussions of mental fictionalism distinguish this interpretation from the eliminativist claim that an ordinary realist theory makes false predictions. The debate reveals the necessity of stating exactly which ontological commitment is under discussion: discrete symbolic vehicles, intentional functional patterns, publicly warranted attributions, or the broader reality of reason-guided behavior. The Stanford Encyclopedia's revised entry on eliminative materialism maps these distinct positions.

Self-Knowledge, Unconscious Attitudes, and Social Attribution

A person may avow a belief without observing her own actions as an outsider would. This first-person relation to thought is a major datum for theories of attitudes. It is equally evident that people can rationalize decisions, contradict themselves, and fail to understand motives that affect their conduct. The problem is to explain a special kind of access without making every self-ascription infallible.

An avowal may express an ongoing commitment rather than describe the result of introspective inspection. Saying “I intend to leave” can be part of forming and communicating an intention; its significance cannot be exhausted by observing a private sensation. Still, the sincerity of a declaration and the persistence of a plan require assessment through a wider network of conduct and circumstances. This difference between expressive authority and complete cognitive transparency helps reconcile first-person and interpretive evidence.

Some cognitive orientations are standing and unattended. Others are action-guiding without being incorporated into reflective convictions. Tamar Gendler's account of alief (2008) addresses affective and behavioral patterns that can diverge from what a person sincerely believes. A visitor standing on a transparent but structurally safe platform may recoil despite carefully justified confidence in its safety. The explanation may require distinct representational, affective, and sensorimotor states rather than two contradictory full beliefs.

A social attribution also depends on testimony, shared linguistic norms, the interpreter's evidence, and the subject's epistemic opportunities. Disagreement over an attitude can reflect divergent criteria for concept possession, different accounts of sincerity, or competing explanations for an action. This makes attitude ascription defeasible while allowing it to remain rationally grounded. Epistemic authority is distributed among an agent's expressed commitments, observed performance, and the broader practices that establish content.

Artificial Systems and the Conditions of Attribution

Artificial cognitive systems invite reconsideration of which features of an attitude are essential and which belong to familiar human realizations. Some systems represent propositions in explicit data structures; others produce stable informational and goal-directed patterns through distributed computation. Observers can describe these performances using the vocabulary of belief, uncertainty, preference, and intention. The philosophical question concerns the warrant for these ascriptions and the demands imposed by stronger readings of them.

A useful analysis differentiates design-level states, realized functional organization, and interpretive descriptions. An explicit variable labeled belief is a programmer's representation; it becomes evidence for an attitude only through the system's operation and explanatory role. A persistent pattern of evidence-sensitive revision supplies a different kind of evidence. A natural-language assertion such as “I believe p” is an interpretable output, and its force depends on the system's consistency, context, mechanisms, and control of that assertion.

The distinction applies across possible bearers. An outward pattern can warrant a limited predictive interpretation while leaving unanswered questions about durable integration, conceptual competence, self-location, or personal practical commitment. Likewise, different materials and architectures may implement comparable abstract cognitive roles. An assessment of artificial agents should use explicit criteria for each attitude kind rather than relying on mere resemblance to human expressions or presuming that every criterion requires biological realization.

The Artificial Era extends the stakes of this inquiry to nonbiological forms of participation in reasoning, authorship, and coordinated intellectual activity. The research question is which forms of semantic stability, inferential competence, corrective sensitivity, and publicly assessable commitment support warranted attribution. This monograph's methodological conclusion remains general: interpreting a performance, identifying a realizer, and establishing a norm-governed attitude are connected but distinct achievements. Their convergence must be argued in each case.

A Critical Synthesis of Theories

The philosophical study of propositional attitudes supports a multidimensional account of their identity. Six distinguishable questions organize the evidence. What attitude is attributed? What content does it concern? How is that content fixed? By what organization or process is the state realized? Which standards govern its rational assessment? Under what evidential and interpretive conditions is the attribution warranted? This is an analytical map of the research problem, not an additional canonical theory or a claim to historical priority.

The first dimension concerns type. Belief, desire, intention, fear, and hope differ in their satisfaction conditions and characteristic relations to inquiry, action, and emotion. An account that identifies every attitude merely with entertaining a proposition loses these differences. The second concerns content. Contents require some account of truth, satisfaction, or relevant possibility, but the resolution appropriate to truth conditions may be too coarse for cognitive explanation. The third concerns determination. A subject's individual organization and her historical, environmental, and social relations may jointly contribute to what an attitude is about.

The fourth dimension concerns realization. Inner representational tokens, distributed processes, functional states, and patterned dispositions compete or cooperate as explanations of particular cognitive capacities. The fifth concerns normativity. Beliefs can be epistemically justified or unjustified; intentions can be coherent or incoherent; desires can be appropriately or inappropriately integrated with practical deliberation. These standards are connected to the kind of attitude attributed, while their application must permit genuine cases of error. The sixth concerns attribution. Whether an interpreter is justified in saying that another agent believes something depends on language, behavior, contextual background, counterfactual evidence, and sometimes first-person avowal.

This map permits several precise evaluations. A purely possible-worlds semantics may be adequate for modeling information change while insufficient to individuate every mathematically distinct thought. A cognitive architecture may explain symbolic inference without determining what natural-language belief reports mean. An interpretationist account may explain public rational patterns while leaving the physical mechanism underdescribed. A strict inner-symbol realism may explain local computational transitions while failing to explain why a particular token refers to a historical figure rather than to a merely correlated stimulus. Such limitations are relative to specific explanatory tasks.

The strongest version of the relational view claims that attitudes involve genuine content-bearing relations, but it must supply a theory of grasping, reference, and logical competence. The strongest functionalist proposal identifies explanatorily stable roles that can vary in material realization, but must explain their content and boundaries. The strongest interpretive view offers principled constraints on public ascription, but must show why the constraints identify more than convenient fictions. The strongest eliminativist view demands that ordinary categories answer to a successful scientific account; it must also establish that the relevant kind fails at all appropriate levels rather than only in one overly literal architecture.

A synthesis does not require an ungrounded choice of a single level as philosophically fundamental. It requires explicit connections among levels. In favorable cases, one can identify a representation's content through stable external relations, show how the system uses it in inferences and planning, specify the norm under which its use counts as success or error, and justify public attribution through a robust behavioral pattern. The case then warrants a stronger judgment than an attribution supported by a single utterance. The resulting explanatory convergence is a substantive achievement that can be investigated, disputed, and revised.

The history of the subject shows a recurring temptation to make one distinction do the work of all the others. The difference between concept and object does not automatically resolve the difference between belief and desire. The difference between an expression's reference and its sense does not decide how thought is neurally realized. The difference between reasons and physical causes does not by itself show that they are incompatible explanatory categories. Distinguishing these levels is valuable precisely because it allows arguments to be assessed at the level to which their premises apply.

Open Philosophical Questions

The first open question concerns the individuation of content. How fine-grained must a proposition be to distinguish rationally different thoughts while remaining genuinely shareable between subjects? A possible-worlds intension can model the alternatives an agent excludes, but struggles with distinct necessary truths. A highly individualized cognitive mode can account for differences in access while threatening an account of common disagreement. The unresolved task is to connect public truth conditions, cognitive significance, and shared conceptual competence.

A second question concerns the relation between de se content and general thought. Does knowledge of who one is require a genuinely different semantic object, or does its practical force derive from a special route of apprehension? Centered-world models and modes-of-presentation accounts make different predictions about what information must change when the agent learns an indexical truth. Evaluating them requires attention to learning, memory, action, and communication, including cases of multiple self-identifications.

A third question concerns psychological ontology. Are beliefs and desires natural, functional, interpretive, or socially constituted kinds, and at what levels can their identity be tested? Distributed cognitive architectures invite skepticism about one-to-one symbolic vehicles, yet robust patterns of rational response remain available for interpretation. Determining which kinds a mature cognitive science should retain requires a theory of explanatory continuity across levels and scales.

A fourth concerns the relationship between normativity and causal implementation. The recognition that some inference is rationally warranted is distinct from describing a process that produces its conclusion. How can an attitude be subject to standards of truth and coherence while being realized by mechanisms that operate causally? Accounts of computational representation, teleosemantics, inferential roles, and social practices offer partial answers, with different commitments regarding error and meaning.

A fifth question concerns social and artificial bearers. When does a group, institution, animal, or artificial system have attitudes in a literal explanatory sense, and when is intentional vocabulary an instrument of interpretation? The answer cannot be a mere stipulation based on ordinary speech. It must specify the type, persistence, content, integration, and corrective responsiveness of the states attributed. The diversity of potential realizers makes explicit attributional standards especially important.

A sixth question concerns first-person authority. Can a subject know her attitude directly when the content depends partly on relations she does not understand, and when unconscious dispositions diverge from explicit endorsement? A theory must preserve the distinctive epistemic place of avowal while accommodating error, social learning, and fragmented cognitive control.

Conclusion

Propositional attitudes constitute a foundational problem in the philosophy of mind and language because they connect the structure of thought with the intelligibility of action, the truth conditions of reports, and the causal organization of cognition. The historical record shows a continuous concern with judgment, assent, representation, and desire, while the modern subject becomes sharply articulated through the logical difficulties of indirect discourse and the systematic analysis of intentional states.

The philosophical results are determinate. Attitude kind must be distinguished from content; the object represented from its mode of presentation; the truth of content from the accuracy of attribution; and rational standards from the causal processes that implement reasoning. Frege's semantic distinction, Russell's quantificational analyses, Wittgenstein's logical criticisms, and later debates over self-location each expose limits of a simple picture of a subject related to an unqualified proposition. Functional, computational, interpretive, and eliminativist theories then dispute how the relevant structures are realized and whether ordinary categories survive explanatory revision.

The most defensible general framework identifies a series of connected problems with independently assessable constraints. Adequate theories must explain content and error, distinctions of cognitive significance, relations among attitude kinds, inferential and practical integration, and the evidential conditions of attribution. They must state which kinds of facts their central claims concern. This approach leaves room for plural explanatory models while preserving rational criteria for adjudicating their disagreements.

The future of the subject lies in systematic connections among semantics, epistemology, psychology, cognitive science, and the philosophy of action. New realizations of content-sensitive reasoning sharpen rather than dissolve its questions. Propositional attitudes remain the principal philosophical site at which the world's intelligibility for a thinker intersects the thinker's intelligibility to others.