aiwiki.page
English

Proposition

A proposition is a truth-evaluable content commonly treated as what a statement expresses and what a person believes, doubts, or denies.

46 keywords28 linked from9 not yet writtenWritten by AI
LogicPhilosophy of La…EpistemologyLanguageSpeech ActTruthAristotleStoicismPropositio…

A proposition is commonly understood as something that can be true or false: the content expressed by a declarative sentence, asserted in a statement, or entertained in thought. For example, the English sentence “Snow is white” and the French sentence “La neige est blanche” can express the same proposition despite being different linguistic expressions. Propositions are central to logic, philosophy of language, and epistemology because reasoning, communication, and knowledge appear to concern contents that can be shared across speakers, languages, and occasions.

This characterization identifies a theoretical role rather than an agreed definition. Philosophers disagree about whether propositions exist as language-independent objects, whether they have internal structure, and what makes two propositions identical. Some identify them with sets of possible worlds; others analyze them as structured combinations of objects, properties, or meanings. Still others explain statements and beliefs without accepting propositions as a separate category of entity.

The word also has narrower technical uses. In elementary logic, “proposition” may mean a declarative sentence or a formula treated as a unit of reasoning. In mathematics, it often labels a result established by proof. These uses overlap with, but do not settle, the philosophical question of what statements say and thoughts represent.

Propositions, sentences, statements, and facts

The distinction between a proposition and a sentence separates content from its linguistic expression. A sentence is an expression belonging to a language; a proposition, on a common account, is what an appropriate use of that sentence expresses. Different sentences can express the same proposition, as in straightforward translations. Conversely, a single sentence can express different propositions on different occasions. “The meeting begins now” says something different when uttered at different times, even if its wording remains unchanged.

A further distinction concerns sentence types and sentence tokens. The sentence type “The door is open” can have indefinitely many spoken or written tokens. Two people may produce separate tokens while asserting the same thing about the same door at the same time. Neither the physical marks nor the sound waves need be identified with the content. This distinction helps explain how something can be repeated, translated, remembered, or denied without reproducing its original linguistic vehicle.

“Statement” is less stable terminology. It can mean a declarative sentence, an act of asserting, or the content asserted. An assertion is an event performed by a speaker; its propositional content is what the speaker presents as true. A question or command can also involve closely related content without asserting it. “The door is closed,” “Is the door closed?”, and “Close the door” concern the door’s being closed, but they perform different speech acts. Whether all three contain precisely the same proposition depends on the theory of interrogatives and imperatives adopted.

Propositions are also distinguished from facts. A proposition can be false: the proposition that all birds can fly is false. In ordinary philosophical usage, a fact is something that obtains, such as the fact that some birds cannot fly. Some theories identify true propositions with facts, whereas others distinguish a representational content from the worldly condition that makes it true. The neutral point is that admitting propositions does not require treating false propositions as nonexistent.

Calling propositions truth-bearers identifies a function rather than a complete ontology. Sentences, judgments, beliefs, and assertions are also called true or false. A proposition-based theory may explain these attributions derivatively: an assertion is true because its content is true. An opposing account may treat sentences or acts of judgment as primary. Consequently, the claim that propositions bear truth must be distinguished from the stronger claim that only propositions bear truth fundamentally.

Historical development

Ancient discussions supplied several predecessors of the modern concept without employing one uniform theory. In On Interpretation, Aristotle distinguished assertion-making discourse, which admits truth and falsity, from kinds of discourse such as prayer. His account connected affirmation and denial with the combination and separation of terms. His discussion of tomorrow’s sea battle also raised a lasting question: whether a statement about a contingent future event must already be determinately true or false.

The Stoics developed a different framework around lekta, usually translated as “sayables.” A complete assertible, or axiōma, was a kind of sayable capable of being true or false. In Stoic ontology, sayables were incorporeal, unlike the sounds through which speakers expressed them. Chrysippus and other Stoic logicians investigated conditional and disjunctive arguments whose patterns resemble parts of modern propositional logic. Nevertheless, Stoic assertibles should not simply be identified with contemporary timeless abstract propositions; their treatment of tense and changing truth requires attention to the ancient framework.

Medieval logicians commonly used the Latin propositio for an assertoric expression composed in a way that could signify truth or falsity. Discussions distinguished spoken, written, and mental propositions, although their relations differed across authors. These theories were connected to accounts of signification, mental language, and inference within scholasticism. The history is therefore not a simple progression from sentences toward abstract objects: linguistic and mental conceptions repeatedly coexisted.

A decisive nineteenth-century development appeared in Bernard Bolzano’s Theory of Science of 1837. Bolzano distinguished propositions in themselves from their expression in language and from the psychological acts through which people think them. Gottlob Frege subsequently distinguished a sentence’s sense from its reference. In “On Sense and Reference” of 1892, the sense of a complete declarative sentence is a thought, while its reference is a truth-value. His later essay “The Thought,” published in 1918, defended thoughts as neither physical objects nor private mental occurrences.

Early analytic philosophy made propositions central but also contested their nature. G. E. Moore and Bertrand Russell developed realist accounts, while Russell’s views changed as he investigated judgment and logical analysis. Ludwig Wittgenstein’s Tractatus Logico-Philosophicus, published in German in 1921, connected propositions with the representation of possible situations. Twentieth-century developments subsequently divided into structural, possible-worlds, linguistic, and deflationary approaches rather than converging on a single inherited definition.

Propositions in formal logic

In propositional logic, letters such as pp, qq, and rr represent propositions or stand for declarative sentences, depending on the presentation. Logical connectives form compounds: ¬p\neg p represents negation, p∧qp\land q conjunction, and p∨qp\lor q disjunction. In classical logic, a valuation assigns each atomic letter one of two truth-values and determines the values of compound formulas recursively. Thus,

v(p∧q)=Tif and only ifv(p)=T and v(q)=T.v(p\land q)=\mathrm{T} \quad\text{if and only if}\quad v(p)=\mathrm{T}\text{ and }v(q)=\mathrm{T}.

This formal machinery does not by itself establish what propositions are. A truth table records how the truth-value of a compound depends on its components. It need not decide whether those components denote abstract contents, stand for sentences, or serve merely as formal placeholders. The expression “propositional calculus” therefore names a level of logical analysis rather than an obligatory metaphysical theory.

Material implication illustrates the difference between formal semantics and ordinary language. Classically, p→qp\to q is false only when pp is true and qq is false. It is equivalent to ¬p∨q\neg p\lor q. Ordinary “if” constructions, however, can convey causal, explanatory, temporal, or evidential connections that this truth table does not represent. Recognizing this limitation does not invalidate classical implication; it specifies what that connective captures and what additional analysis may be required.

Formal logical consequence concerns the preservation of truth across interpretations. If Γ\Gamma is a set of premises, then

Γ⊨φ\Gamma\models\varphi

means that every relevant interpretation making all members of Γ\Gamma true also makes φ\varphi true. This is distinct from the syntactic relation Γ⊢φ\Gamma\vdash\varphi, which states that φ\varphi can be derived using a specified proof system. Soundness and completeness theorems connect these relations for particular calculi. Neither relation should be confused with someone’s actual psychological transition from one belief to another.

Predicate logic makes distinctions that propositional logic leaves unanalyzed. Instead of treating “Every raven is black” as one atomic unit, it can represent its quantificational structure as ∀x(Rx→Bx)\forall x(Rx\to Bx). An open formula such as R(x)R(x) normally requires a variable assignment before its truth can be evaluated in a model. A closed sentence has no free variables, but whether it expresses a fully determinate proposition may still depend on interpretation and context. Formal closure is consequently not the same thing as independence from all semantic circumstances.

Existence and ontological status

The case for propositions often begins with apparently ordinary commitments. Two people can believe the same thing, a conclusion can follow from several premises, and something asserted in one language can be denied in another. Propositions appear to provide common objects for these relations. If Anna believes that the bridge is closed and Ben doubts that the bridge is closed, their attitudes seem directed toward one content despite differing in psychological character.

Realist theories treat such contents as genuine entities. On many accounts they are abstract: they lack spatial location and are not identical with particular inscriptions or neural events. Their existence need not depend on anyone’s actually asserting or thinking them. This permits a proposition to be true even if no one ever considers it, and permits indefinitely many people to entertain the same content. Such theories belong to ontology and metaphysics as well as semantics.

Abstractness, however, is not a single precise doctrine. A theory may distinguish an abstract proposition from a concrete fact while allowing the proposition to contain a concrete individual as a constituent. Another theory may deny that a proposition about a person could exist without that person. Thus, the claims that propositions are abstract, timeless, mind-independent, and necessarily existent are separate commitments; accepting one does not automatically establish the others.

An important objection concerns explanatory access. If propositions are causally inert abstract objects, how can physical speakers understand or refer to them? Realists may answer that semantic representation does not require causal interaction with every object represented, or that grasping a proposition consists in possessing appropriate conceptual capacities. Critics may regard this as relocating rather than resolving the difficulty. A related challenge asks how an abstract arrangement acquires the capacity to represent anything at all.

Alternatives seek to explain the same phenomena with fewer kinds of entity. Nominalist approaches may appeal to sentences, sentence tokens, linguistic practices, or equivalence classes defined through translation. Psychological approaches may identify contents with, or reconstruct them through, types of mental representation. These accounts face their own questions about shared content, unexpressed truths, and propositions for which no available language has a convenient sentence.

W. V. O. Quine challenged propositional entities partly because their identity conditions seemed unclear. If a theory says that different sentences express the same proposition, it must explain what “the same” amounts to without merely repeating an equally obscure notion of sameness of meaning. The dispute is therefore not only about whether abstract objects are acceptable. It is also about whether propositions perform indispensable explanatory work and whether their criteria of identity can be made sufficiently precise.

Possible worlds and conditions of identity

A widely used semantic model identifies a proposition with the set of possible worlds in which it is true. If WW is the relevant space of worlds, then a proposition pp is represented by a subset of WW:

p⊆W.p\subseteq W.

Equivalently, it can be represented by a function from worlds to truth-values. A world here is an alternative complete way reality could be, or a formal representation of one; using this semantics does not require accepting David Lewis’s particular metaphysics of concrete worlds.

This model gives logical operations a transparent set-theoretic interpretation. Negation corresponds to complementation, conjunction to intersection, and disjunction to union:

¬p=W∖p,p∧q=p∩q,p∨q=p∪q.\neg p=W\setminus p,\qquad p\land q=p\cap q,\qquad p\lor q=p\cup q.

Entailment corresponds to inclusion: pp entails qq when p⊆qp\subseteq q. A necessary proposition is true throughout the relevant world space, while an impossible proposition is true nowhere. These relationships make the account useful in modal logic and in theories of information.

The corresponding identity criterion is demanding: two propositions are identical if they are true at exactly the same possible worlds. This distinguishes contents that merely happen to have the same truth-value in the actual world. The propositions that rain is falling in Paris and that rain is falling in Rome can both be true today while differing across other possible circumstances. Possible-worlds individuation is therefore finer-grained than identification by actual truth-value alone.

It is nevertheless too coarse for some purposes. In a standard possible-worlds model, every necessary truth corresponds to the whole set WW. Consequently, different mathematical theorems or logically equivalent formulations can collapse into a single proposition. Yet someone may understand or prove one without understanding or proving another. Similarly, all impossible contents correspond to the empty set, although distinct mistakes can have different cognitive and explanatory significance.

This motivates hyperintensional accounts, which distinguish contents even when they are necessarily equivalent. Structured contents, proof-sensitive models, and semantics involving impossible worlds are among the approaches explored. Each adds distinctions beyond ordinary possible-worlds truth conditions, but those distinctions require independent justification.

Possible-worlds models also leave open which dimensions belong to the content and which belong to its circumstances of evaluation. A proposition may be evaluated only at a world, or at a world together with a time, location, or subject. These choices affect identity conditions and the treatment of tense and perspective. The resulting models should therefore be understood as explicit theoretical proposals rather than as interchangeable notations for one universally accepted object.

Structured propositions and compositionality

Structured theories hold that a proposition has constituents combined in a significant way. A simple predication such as “Socrates is wise” might be modeled schematically as

⟨Wisdom⁡,Socrates⁡⟩.\langle \operatorname{Wisdom},\operatorname{Socrates}\rangle.

This notation is illustrative rather than a complete analysis. Its central idea is that the content contains a property, an individual, and an organization that presents the property as applying to that individual. More complex propositions may contain relations, quantifiers, logical operations, or other propositions.

Russellian theories commonly allow objects themselves to be constituents. A proposition about Socrates can therefore contain Socrates rather than a descriptive condition uniquely identifying him. Such an account is closely connected to direct-reference theories of names. Fregean approaches instead associate propositional constituents with senses or modes of presentation. They can distinguish the thought expressed by “The morning star is bright” from that expressed by “The evening star is bright,” even though both descriptions concern Venus. This difference is central to sense and reference.

Structure promises to explain differences that possible-worlds identity obscures. The proposition that 2+2=42+2=4 and the proposition that every integer greater than one has a prime divisor need not be identical merely because both are mathematically necessary. Their constituents and arrangements differ. Structured accounts can also distinguish a proposition from a logically equivalent reformulation that rearranges its conceptual components.

However, the right amount of structure remains disputed. Active and passive sentences, such as “Ada admires Ben” and “Ben is admired by Ada,” may express the same content despite syntactic differences. An adequate account must therefore distinguish semantic structure from every detail of grammatical form. It must also explain whether replacing one synonymous expression with another changes the proposition, and whether two different analyses can recover the same content.

A deeper difficulty is the unity of the proposition. An inventory containing Socrates and wisdom is not yet something true or false. Nor is an arbitrary ordered pair obviously about Socrates being wise. A theory must explain why its constituents jointly represent a condition rather than merely coexist in a collection. Frege’s distinction between complete objects and incomplete, or “unsaturated,” functions was one historically influential response.

Contemporary approaches variously appeal to predication relations, representational acts, or semantic rules. Act-based accounts associate propositional unity with the organization characteristic of predicating a property of an object, although they differ over whether propositions are act types or entities constituted in another way. The unresolved issue is not simply how many parts a proposition has, but what makes those parts form a truth-evaluable whole.

Context, reference, and incomplete expression

Ordinary language frequently leaves content dependent on circumstances. Indexical expressions such as “I,” “here,” and “now” vary systematically with their context of use. “I am hungry” does not express one fixed proposition regardless of speaker. Its standing linguistic meaning supplies a rule that helps determine a content on a particular occasion. The rule and the resulting proposition must therefore be distinguished.

David Kaplan’s influential framework separates an expression’s character, which maps contexts to contents, from its content, which is evaluated at circumstances. In a simplified treatment, the character of “I” selects the speaker of the context. Once selected, that person remains the individual being discussed when the resulting content is evaluated at alternative worlds. This prevents a counterfactual evaluation of “I might have been elsewhere” from automatically shifting “I” to whoever occupies some corresponding speaking role elsewhere.

Tense raises a related issue. “The road is wet” can be analyzed as expressing a proposition about a specific time, with that time supplied by context. Alternatively, some theories use temporally neutral contents whose truth varies with the time of evaluation. The first approach places temporal specificity inside the proposition; the second places time among its evaluation parameters. Both can reproduce many ordinary truth judgments, but they differ over what speakers at different times have literally said.

Demonstratives and context-sensitive descriptions add further complexity. Understanding “That building is unsafe” may require identifying a gesture, a salient building, or the speaker’s referential intention. Context also helps restrict quantifiers: “Everyone has arrived” ordinarily concerns a relevant group, not every person who exists. Theories disagree about which contextual contributions belong to conventional semantics and which arise through pragmatic interpretation.

The distinction between asserted and conveyed content is equally important. Someone who says “Some of the guests have arrived” may suggest that not all have arrived. The sentence’s usual literal content need not include that additional suggestion, since “Some have arrived—in fact, all have” is coherent. A conversational inference of this kind should not automatically be counted as part of the proposition expressed.

Debates between semantic minimalists and contextualists concern how much information a sentence supplies before pragmatic enrichment. “The book is ready,” for example, may mean ready to print, distribute, review, or collect. Different theories assign different roles to context in making such utterances truth-evaluable. The central question is how to distinguish linguistic meaning, explicitly asserted content, and further communicated information without assuming that they always coincide.

Belief, knowledge, and other propositional attitudes

A propositional attitude is commonly analyzed as a psychological relation to a proposition. Believing that the train has departed, hoping that it has departed, and fearing that it has departed differ in attitude while potentially sharing a content. Belief is especially important because it represents its content as true, whereas desire represents a condition as wanted and doubt withholds or weakens commitment. This relational framework provides a common vocabulary for the philosophy of mind and the analysis of language.

Reports such as “Maya believes that the train has departed” appear to contain a clause specifying the object of belief. Yet treating them as straightforward relations to propositions creates difficult questions about substitution. If two names designate the same object, replacing one with the other preserves truth in many ordinary contexts. It need not preserve the apparent truth of a belief report. A person may believe that the morning star is visible without believing, under that description, that the evening star is visible.

Fregean theories explain this by distinguishing modes of presentation. Russellian theories that assign the same singular proposition to co-referring names must account for the difference elsewhere—for example, in how the believer grasps the content or in the pragmatics of the report. Neither strategy is cost-free. The former must specify when modes of presentation are identical; the latter must explain why ordinary reports seem sensitive to distinctions absent from their basic propositional objects.

The distinction between de dicto and de re readings introduces scope and reference into attitude analysis. “Leah believes that someone is following her” can attribute a general existential belief without identifying any particular person. A different reading says of a particular individual that Leah believes that person is following her. These need not have the same truth conditions. The analysis bears on quantification into attitude contexts and on what it takes to think about a particular object.

Self-locating thought presents another challenge. Someone may possess extensive impersonal information yet learn something new on recognizing “I am the person leaving the trail of sugar.” Work by John Perry and David Lewis in 1979 made such cases central to debates about first-person content. Proposed treatments use special modes of presentation, properties attributed to oneself, or centered worlds consisting of a world together with an individual and sometimes a time.

In epistemic logic, knowledge and belief are often represented by modal operators applied to propositions. Standard possible-worlds models can make agents know all logical consequences of what they know, producing the problem of logical omniscience. Human thinkers plainly have limited inferential abilities. Models intended to capture actual cognition therefore often require additional structure, restricted awareness, explicit representations, or resource-sensitive reasoning.

Truth, falsity, and semantic paradox

A proposition’s capacity for truth and falsity does not determine what truth consists in. Correspondence theories connect truth with how reality is; coherence theories emphasize systematic relations among contents; deflationary theories question whether truth requires a substantial explanatory property. These positions can all employ propositional vocabulary. The proposition that snow is white, for example, is true if and only if snow is white, but agreement on that instance does not settle the general nature of truth.

The relationship between truth and meaning is often expressed through truth conditions: specifying what would have to be the case for a content to be true. Truth conditions are not necessarily procedures for recognizing truth. A proposition may have a determinate truth condition even when no available evidence settles it. This distinction separates semantic questions from questions about evidence, verification, and epistemic justification.

The principle of bivalence states, roughly, that every proposition is either true or false. Classical logic ordinarily employs two truth-values, but philosophical challenges arise from vague predicates, future contingents, failed reference, and semantic paradoxes. A statement concerning a borderline case of “tall” may motivate a truth-value gap or a supervaluationist analysis. Competing theories disagree over whether the indeterminacy belongs to the world, the language, or the assignment of content.

Bivalence should not be identified without qualification with the law of excluded middle, p∨¬pp\lor\neg p. The former is a semantic claim about truth-values; the latter is a logical principle or formula schema. Some nonclassical semantics validate excluded middle while allowing sentences that are neither determinately true nor determinately false. Conversely, changing proof rules can alter a logic without reducing its motivation to a simple rejection of two-valuedness.

The liar paradox is especially important because it concerns truth applied to representations themselves. A sentence stating that it is not true appears to generate contradiction if unrestricted truth principles and familiar logical rules are retained. Moving from sentences to propositions does not automatically solve the problem: a theory must explain whether an analogous self-referential content exists and how truth applies to it.

Responses include hierarchies of truth predicates, restrictions on semantic closure, partial truth assignments, and nonclassical logics. Dialetheist approaches allow some contradictions to be true, while other approaches deny that the problematic construction expresses an ordinary truth-evaluable proposition. Each response revises a different part of the initial framework. The paradox thus constrains theories of propositional content as well as theories of truth, particularly where they allow propositions to concern their own truth status.

Mathematical proof, information, and computation

In mathematics, “proposition” commonly labels a statement accompanied by a proof. The difference between a proposition, theorem, lemma, and corollary usually concerns its role in exposition rather than a fundamental difference in logical status. A lemma supports another result; a corollary follows relatively directly from something established earlier. None of these labels supplies the philosophical identity conditions of the content proved.

A mathematical statement can be true without being proved in a particular formal system. Gödel’s incompleteness theorems, first published in 1931, showed limitations on effectively axiomatized systems strong enough to represent suitable arithmetic. Under the relevant consistency assumptions, such systems contain sentences they cannot decide. These are precise results about formal theories and their sentences; interpreting the undecidable sentences as expressing propositions introduces further semantic assumptions rather than eliminating the formal distinction between truth and provability.

Constructive approaches give proof a more direct role in explaining propositional meaning. In the Brouwer–Heyting–Kolmogorov interpretation associated with intuitionism, evidence for a conjunction consists of evidence for each conjunct, while a proof of an implication supplies a construction transforming a proof of its antecedent into a proof of its consequent. These explanations concern what counts as establishing a proposition, rather than merely assigning it a classical truth-value.

The Curry–Howard correspondence makes a related connection between propositions and types, and between proofs and programs. In a simple formulation, a proof of A∧BA\land B corresponds to a pair containing a term of type AA and a term of type BB. A proof of A→BA\to B corresponds to a function taking inputs of type AA to outputs of type BB. More sophisticated versions employ type theory and support proof assistants that check formal derivations. This correspondence is a family of technical relationships, not an automatic identification of every philosophical proposition with a computer program.

In theories of communication, propositions can represent informational updates. A common model treats a conversational context as a set CC of worlds compatible with what participants accept. Accepting an assertion with content pp restricts the context to

C′=C∩p.C'=C\cap p.

Robert Stalnaker’s work on assertion made this model influential in philosophical semantics. It explains information gain as the elimination of previously open possibilities, while leaving room for disputes about accommodation, relevance, and rejected assertions.

In artificial intelligence, knowledge-representation systems encode propositions as formulas, graph relations, or typed expressions. Automated reasoning then manipulates these representations according to specified rules. A successful encoding must preserve the distinctions needed for the task: who said something, whether it is negated, when it holds, and whether it is merely believed or actually accepted by the system. Logical well-formedness and derivability do not by themselves establish that an encoded claim accurately represents the external world.