Action Individuation

The Problem of Counting What an Agent Does

Someone moves a finger, presses a switch, illuminates a room, and alerts a person who has been hiding in it. Four descriptions are available. Whether four actions have occurred is a further question. A second person enters the room and closes the door; these are distinguishable movements that can nevertheless form a single undertaking, namely securing the room. A writer spends an afternoon composing an argument, pauses for dinner, and resumes the next morning. The finished essay may be the product of one extended action, many component actions, or an organized activity to which both descriptions legitimately apply. Action individuation examines the criteria that make such answers true.

The problem is ontological because actions appear to be things that occur: they have agents, histories, locations, temporal boundaries, causes, and consequences. It is conceptual because descriptions sort the same behavior under different verbs, purposes, and norms. It is explanatory because the question “Why did she do that?” may identify a movement, an achievement, or a long-term project. It is practical because ethical and legal assessments frequently depend on determining whether a result belongs to an act, follows from it, or is produced by another act.

Three questions organize the inquiry. First, what category of entities are counted—event particulars, property exemplifications, acts conceived as repeatable types, or temporally extended processes? Second, what constitutes an identity condition for entities of that category? Third, which explanatory and normative purposes appropriately determine the boundaries at which action is counted? An adequate theory must distinguish the questions while showing why ordinary agents can move between their answers with little apparent difficulty.

The central historical debate is often presented as a conflict between a unifying, coarse-grained ontology associated with Davidson and a multiplying, fine-grained ontology associated with Goldman and Kim. That contrast is important, but it captures only one dimension of the issue. The question of what counts as one action also depends on temporal continuity, organizational unity, the agent’s relation to what happens, and the distinctions between doing something, doing it by doing something else, and merely bringing about a consequence. These dimensions explain why the debate remains active after the original controversy of the 1960s and 1970s. The historical and systematic framework is developed in the Stanford Encyclopedia of Philosophy’s discussion of action, especially its treatment of action ontology and individuation.

Definition, Scope, and the Identity Question

Action individuation concerns the determination of numerical identity and numerical diversity among particular actions, together with the principles for distinguishing repeatable kinds of actions. A token is a particular occurrence—for example, a specific opening of a door by a specific agent on a specific occasion. A type is a repeatable characterization—opening a door—which can be instantiated by different people at different times. If two people open two doors, they instantiate a common type but perform different tokens. If one person opens a door and thereby admits a guest, the question concerns whether those descriptions identify one token or several.

Numerical identity is stricter than similarity. Two acts performed with the same intention and bodily movement at different times are similar tokens, not one token. Numerical identity is also different from co-occurrence: distinct happenings may occupy the same interval and location. Nor does intentional equivalence suffice. An agent may pursue one purpose through two separate efforts, while a single bodily episode may serve several purposes. Formal identity is reflexive, symmetric, and transitive. Any purported action-identity criterion must therefore obey those conditions, even when the relations of means to end, part to whole, and cause to effect do not.

The expression “the same action” is used in several ways. It may mean the same physical performance; the same intentional undertaking; the same kind of act; or the same normatively assessed deed. Philosophical analysis begins by identifying which usage is at issue. A theory that treats all these uses as a single unanalyzed relation either multiplies entities beyond explanatory need or erases distinctions necessary to explanation.

An action also differs from its description. “She moved the lever” is a sentence; the lever movement is a worldly happening; moving a lever is a kind of act. Sentences can differ without their referents differing. Conversely, two expressions can refer to causally related but numerically distinct occurrences. The observation that an episode admits several descriptions is therefore compatible with more than one action ontology. This distinction is emphasized by Hornsby’s analysis of acts and action particulars.

The boundaries of this dossier are consequently precise. Its object is the identity of actions, including the ontological and explanatory conditions of their unity. The general nature of agency, freedom of the will, moral responsibility, event causation, and the semantics of action verbs enter where they constrain action identity. Each retains its own broader research domain.

Aristotle: Voluntary Agency, Particulars, and Practical Organization

Ancient philosophy supplies important distinctions without offering the modern token-identity dispute in its twentieth-century terms. In the Nicomachean Ethics, Aristotle investigates voluntary and involuntary actions, the agent’s contribution to what happens, and the relation between decision and practical circumstances. Book III argues that an action done under compulsion differs from conduct whose origin lies within the agent, while cases of action under threat require attention to what the agent chooses in the situation. These are distinctions among forms of agency, and they show why a physical movement by itself cannot determine the complete practical character of the occurrence (Aristotle, Nicomachean Ethics, III.1–5, 1109b30–1115a).

Consider Aristotle’s example of cargo thrown overboard to save a ship in a storm. The physical description “throwing cargo” and the purposive description “saving lives” relate one performance to its circumstances and end. The question whether the action is voluntary cannot be answered solely by specifying the bodily motion; it requires identifying the practical alternative available at the time. Aristotle’s account thereby anticipates a principle central to later debates: the description under which an agent acts matters to an assessment of agency, even if the metaphysics of identity remains unsettled.

Aristotle’s term praxis designates action within his practical philosophy, while poiesis designates making or production in important contexts. His account of prohairesis, often rendered “choice” or “deliberate choice,” locates a deliberative orientation toward means within the explanation of action. Book VI further connects practical wisdom with judgments concerning particular circumstances. These concepts provide resources for distinguishing an unfolding practical undertaking from the discrete motions that realize it. They should be interpreted within Aristotle’s ethics and theory of capacities rather than anachronistically presented as a modern formal criterion of event identity.

The Aristotelian heritage also raises a question about the relevant level of description. Cutting wood and constructing a shelter may be related as component performances to an organized undertaking. Counting only muscular motions neglects the practical unity of construction; identifying the whole undertaking with every component loses the part–whole relation. The distinction between an activity and the ends toward which it is organized makes room for a structured account in which one agent undertakes one project through many constituent performances.

Aristotle thus establishes a historical premise for action individuation: understanding what an agent does involves the origin of movement, chosen circumstances, and purposive organization. Modern philosophers disagree about the number and ontology of the resulting action particulars. Aristotle supplies part of the problem’s philosophical ancestry, not a ready-made solution to the later numerical dispute.

Aquinas: Command, Execution, and Unity of Acts

Thomas Aquinas develops the relations among will, reason, command, and execution within a systematic account of human acts. In Summa theologiae, I–II, question 6, he analyzes voluntary action through an internal principle accompanied by knowledge of an end. In question 17, he asks whether command belongs to reason or will and, especially significantly for individuation, whether command and the commanded act are one act or distinct. This is a direct historical antecedent of questions about how internally related phases form a whole action (Aquinas, Summa theologiae, I–II, q. 6; q. 17).

Aquinas distinguishes the ordering performed by reason from the will’s orientation toward an end and from the execution of what is commanded. The sequence has structure: practical judgment bears on selection, command directs the appropriate powers, and execution realizes the direction. The distinction prevents a theory from treating intention, order, and bodily outcome as interchangeable particulars simply because all belong to a single agency episode.

At the same time, the unity of a human act may be described through the whole ordering of its constitutive moments. A commanded bodily performance is related to the practical determination that directs it. The philosophical issue therefore concerns the kind of unity at stake. Distinct operations of capacities may form one action considered as an organized practical whole. This is a relation of integration, and its explanatory value does not depend on declaring every internal phase numerically identical.

Aquinas’s analysis is especially instructive for contemporary accounts that postulate a basic mental act at the core of every overt action. Even if a volitional initiation is distinguished from a muscular execution, an account is still required of why they constitute one intentional undertaking. The unity of agency cannot simply be inferred from the temporal adjacency of its parts. Nor can it be secured merely by naming a controlling faculty.

The medieval framework differs from contemporary event semantics in its psychology, metaphysics of powers, and explanatory ends. Its lasting contribution is a sharpened question: how are distinguishable acts of intending, commanding, and carrying out an intention integrated, and which of those distinctions also mark a plurality of action tokens?

Hegel: Deed, Intention, and Imputation

Hegel’s Elements of the Philosophy of Right develops a distinction between the outward deed and action as intelligible in relation to a subjective will. In §113, action (Handlung) is treated as the expression of the subjective or moral will, whose external fulfillment is known by the agent and related to others. The discussions that follow examine intention, foreseeable consequences, and the agent’s right to recognize what is to count as his or her own deed. The original German formulation is accessible in Grundlinien der Philosophie des Rechts, §§113 onward.

Here the problem is not simply how many physical events occur. The central issue is how an occurrence becomes attributable under a description that reflects the agent’s practical understanding. An outward result has potentially innumerable causal connections, whereas an action attributed to the will is interpreted within limits of knowledge, intention, and recognizable consequence. The philosophy of action must investigate these limits because counting consequences as new actions can magnify what is imputable to an agent without a corresponding account of agency.

An agent may knowingly open a gate, thereby allowing a vehicle to pass, which in turn produces a chain of distant outcomes. Some descriptions concern what was deliberately undertaken; others concern results that can be connected to it through causation. Hegel’s treatment helps explain why legal and moral practices cannot rely on causal reach alone to determine the identity of the action assessed. At the same time, intended consequences and objective descriptions of a deed cannot be eliminated from practical assessment.

The German distinction between Tat and Handlung is historically significant. It connects the objective occurrence of a deed with the agent’s recognition, intention, and normative accountability. It does not automatically entail that there are two numerically distinct event tokens. The distinction concerns different explanatory perspectives on conduct; turning it into a simple count of events would require further argument.

From Aristotle through Aquinas to Hegel, the unity of action is repeatedly approached through practical form, purposive orientation, and attribution. The contemporary debate shifts toward event ontology and explicit identity criteria, but its deepest difficulty remains the relation between what happens and what an agent does.

Anscombe and the Plurality of True Descriptions

G. E. M. Anscombe’s Intention (first published in 1957) reoriented philosophical analysis around the ways actions are intentional and around the question asking for an agent’s reasons. Her discussion of the person operating a pump is foundational. Moving one’s arm, operating the pump, replenishing a water supply, and potentially poisoning the users of the water can be descriptions of a practical sequence. The reason-giving “Why?” question reveals relations among descriptions and can reveal the intentional character of the performance under some of them (Anscombe, Intention, §§5, 23–26).

The philosophical significance of this case is often compressed into the claim that four descriptions refer to one action. Anscombe’s actual inquiry is more searching. Her §26 specifically raises the issue whether the series comprises one action or four, and examines the relation of what verifies one description to what verifies another. A description that is true because of further circumstances is not thereby interchangeable with a description of an isolated bodily movement. Operating a pump requires a functioning mechanism; replenishing a supply requires an appropriate connection; poisoning requires a harmful substance and suitable consequences.

The sequence is intelligible as an answer to why an agent is moving. Moving the arm is undertaken in operating the pump, operating the pump in replenishing the water, and so on. This gives the descriptions an intelligible dependence even when the conditions of their truth differ. In addition, an agent may be unaware of the harmful consequences. The action can be intentional under the description “pumping” without being intentional under the description “poisoning.”

The difference between the action’s being intentional under a description and the description’s being true is decisive. Knowledge and purpose are relevant to attributing intentions, while external arrangements and outcomes determine some factual descriptions. An ontology that identifies an action with an intention would struggle to account for unintended but true descriptions of what occurred. An ontology that treats each true predicate as a new action would multiply acts whenever a description becomes available.

Anscombe’s practical-knowledge orientation makes the action recognizable through the agent’s practical understanding of what is being done. It also directs attention to the order among the descriptions. The order is explanatory and sometimes means–end structured; it is neither a simple list of synonyms nor a list of independently observed physical events. The problem of individuation takes its modern shape at precisely this intersection.

Davidson’s Event Particulars and Coarse-Grained Identity

Donald Davidson’s writings on action and events transformed action individuation into a general problem about the identity of event particulars. In “Actions, Reasons, and Causes” (1963), an intentional explanation connects what a person does with the reasons for which the person acts. “The Logical Form of Action Sentences” (1967) develops an event-based account of action predication, and “The Individuation of Events” (1969) examines identity conditions for events. “Agency” (1971) applies the framework directly to the nature of action. These essays appear in Essays on Actions and Events.

On Davidson’s coarse-grained account, a single particular action may admit several true descriptions. Depressing a switch and turning on a light can be one and the same action when the descriptions pick out the agent’s occurrence. Describing an event by means of an effect does not automatically multiply the event. A distinction arises between the action and the further happenings that it causes; light appearing in a room is an effect that can also figure in the description of the agent’s action.

The account gains support from the logical form of action sentences. If a person swiftly moved a hand in a room, the sentence entails that the person moved a hand. Davidson’s event analysis represents the first assertion as the existence of an event with several properties; dropping an adverbial condition preserves existential reference to an event. One occurrence can therefore satisfy a number of predicates without the ontology expanding with each added modifier. The analysis gives action descriptions a common semantic referent while allowing their truth conditions to differ.

Davidson also proposed a causal criterion of event identity: events are identical when they have the same causes and effects. The criterion promises a description-independent basis of identity. However, it encounters difficulties concerning how to individuate causes and effects and whether distinct simultaneous events can share an entire causal setting. A theory of event identity cannot achieve genuine independence if the individuation of its causal relata silently presupposes the very identities it is intended to explain. The relevant historical argument is discussed in the Stanford Encyclopedia of Philosophy’s treatment of events.

The central achievement of Davidson’s position is to show how unity across descriptions is metaphysically intelligible. Its central unresolved issue is the relation between the identity of a minimal bodily occurrence and the larger, temporally and normatively organized performances of which that occurrence is a component.

The Argument for Unifying Actions

The unifying argument begins with an ordinary observation: the agent appears to have done several things by a single performance. Suppose someone turns a key, starts an engine, and frightens a nearby animal with the resulting sound. Counting descriptions yields three phrases, but it supplies no principled account of why their three corresponding entities must be actions rather than properties, causal relations, or effects of one action.

The coarse-grained theorist argues that action identity should not vary with the observer’s vocabulary. A person who knows only about the turning of the key and another who knows about the engine have incomplete descriptions of the same worldly episode. If the addition of information could create numerically new actions, identity would appear to depend on which descriptions were known rather than on the actual occurrence. Davidson’s event ontology preserves a stable referent under explanatory redescription.

A second argument arises from causal explanation. The reason for which a person turns a key can also explain why the engine is started. If these are two independently caused actions, the theory must specify their distinct causal relations. If one action causes another, it must identify a genuine intervening process rather than merely redescribing the effects of a single performance. When the underlying event is the same, the explanatory economy of the unifying approach is strong.

The argument also has limits. The number of descriptions is indeed insufficient to count actions, but the fact that several descriptions are connected causally is insufficient to establish identity. A person who loads a machine and later activates it performs distinct episodes contributing to one result. A surgeon’s incision and later suturing are distinct phases within one operation. If every event within a purposeful causal chain were identified, transitivity could collapse distant actions into one entity. The unifying view must therefore articulate a boundary condition that explains why the key turning can bear a consequence description while a later distinct intervention is another action.

An adequate unifying theory must consequently distinguish redescriptions of a performance from descriptions of performances that are genuinely causally connected but separate. Its strength is conservation of ontological particulars; its task is to define the particulars independently of the convenience of language.

Goldman’s Fine-Grained Acts and Level-Generation

Alvin I. Goldman’s A Theory of Human Action (1970) develops a rival treatment of actions that distinguishes basic acts from acts generated at other levels. On his fine-grained approach, performing an act of one type is not automatically identical to performing an act of another type, even when the performances are connected through the same bodily movement. Turning on a light by flipping a switch exemplifies a relation between acts at different levels, rather than an identity of acts. The original volume and its terminology are documented in the publisher-library bibliographic record.

Goldman introduces the relation of level-generation to analyze how one act can count as the performance of another. Pressing a button may generate activating a device; moving a chess piece under the rules of the game may generate delivering checkmate. The relation can be asymmetric. A person can deliver checkmate by moving a piece, while the causal or constitutive explanation need not run in reverse. Asymmetry matters because numerical identity is symmetric. If one act generates another in a directed relationship, their difference is not exhausted by two names for a single entity.

The account distinguishes causal generation from generation through conventional rules and from generation through the circumstances in which an act is performed. Changing a physical arrangement is one way of producing a further act; performing a move in a constituted practice is another. A bodily movement has different action significance in a chess match, theatrical rehearsal, or accidental disturbance of the board. The conditions under which a more complex act exists cannot be derived from movement geometry alone.

The philosophical cost is multiplication. If an act is individuated in part by the property it exemplifies, the same occasion can contain many acts. Goldman's later reflections explicitly revisit the old debate and acknowledge the explanatory claims of coarse and fine-grained approaches, relating metaphysical individuation to cognitive-scientific interests (Goldman, Joint Ventures, 2013). Such later work is best read as development of the original debate, rather than evidence that its participants adopted identical ontologies.

Goldman’s principal insight is that performance, generation, and social constitution are explanatory relations in their own right. Treating all related acts as numerically identical can hide these relations. The remaining problem is to justify the point at which differences of act properties require separate action tokens.

Kim’s Property-Exemplification Theory of Events

Jaegwon Kim’s “Events as Property Exemplifications” (1976) offers a general event ontology in which events are individuated by an object or substance, a property, and a time. In schematic form, an event is represented by the exemplification of property P by subject s at time t. Two such events are identical when their constitutive subjects, properties, and times are identical. The paper was published in Brand and Walton’s Action Theory, pages 159–177, as established by Springer’s original volume record.

Applied to action, the theory yields a finer individuation than Davidson’s coarse-grained conception when the relevant properties differ. A person’s raising a hand and a person’s signaling a vote may involve the same subject and time, but they instantiate different action properties. Kim’s criterion provides a principled explanation of why such differences can correspond to different events rather than merely different descriptions of the same event.

The analysis depends on a disciplined conception of constitutive properties. A particular event can have many ordinary properties, including being visible, noisy, surprising, or embarrassing. The constitutive property that individuates an event cannot simply be equated with every predicate that is true of it. Otherwise an arbitrary reformulation of the same occurrence would generate a different event. Kim’s distinction between the property constituting an event and other properties attributed to that event is crucial to the theory’s coherence.

The approach provides an illuminating comparison with Goldman. Goldman’s acts and level-generation explain practical multiplicity; Kim’s events as property exemplifications supply an ontology that can underpin a plurality of event tokens at the same place and time. Yet a separate philosophical question remains: which property exemplifications are genuinely actions? A subject’s being in a physical state at a time does not establish agency merely by meeting an event-ontology formula.

Fine-grained individuation strengthens the role of action types and capacities while demanding a theory of relevant properties. Without such a theory, metaphysical structure risks becoming a projection of linguistic granularity. With an independently justified account of the constitutive property, it can preserve differences that the coarser view might otherwise flatten.

Davis, Thomson, Hornsby, and the Search for Intermediate Criteria

The canonical opposition between coarse and fine-grained approaches understates the range of theories that emerged in the same debate. Lawrence H. Davis’s “Individuation of Actions” (1970) directly addresses the question of action identity, while Judith Jarvis Thomson’s Acts and Other Events (1977) examines actions within a general ontology of events. Jennifer Hornsby’s Actions (1980) and later encyclopedia treatment clarify the distinction between particular action occurrences and repeatable acts. Their contributions show that identity criteria can differ without reducing the whole subject to an unrestricted choice between one event and infinitely many act properties (Davis, 1970; Thomson, 1977; Hornsby, 1998).

An intermediate view may agree that two descriptions of a movement often pick out one action but resist identifying that movement with every achievement or distant consequence attributed to the agent. The theoretical challenge is to establish where identity ends and another relation begins. Temporal coincidence cannot supply the full answer because distinct events can be simultaneous. Sharing a causal origin is also insufficient because one cause may yield several effects. Likewise, sharing a goal does not make successive attempts numerically identical.

A further possibility individuates actions by the capacities exercised. Where an action is the manifestation of one capacity, different adverbial descriptions can characterize the same performance; where another kind of capacity is realized, the theory may distinguish the actions. This possibility is discussed in contemporary surveys through philosophical reconstructions associated with Aristotle. It requires a noncircular account of capacities: identifying capacities by what actions they produce while individuating actions by capacities would merely relocate the original question.

Carl Ginet’s On Action (1990) adds a different approach by placing simple mental action at the center of overt agency and addressing individuation within a volitional framework (Cambridge University Press). This view invites comparison between what the agent immediately does and what occurs through executing or expressing that mental act. Its distinctive premise concerns the organization of action from an agentive core, not simply whether descriptions have common semantic reference.

The debate therefore contains several dimensions: property versus event identity; act versus act-type; direct versus mediated performance; and temporal or practical unity. An adequate synthesis must state which dimension each theory is intended to explain before judging the disagreement.

Act Types, Act Tokens, and the Logic of Description

The type–token distinction is indispensable to a coherent count. A violinist’s playing a particular note twice yields two performance tokens that instantiate the same kind of act. Two musicians playing the same passage simultaneously also instantiate the same type without performing the same token. Conversely, one performance may instantiate several types: playing a note, completing a phrase, and signaling an agreed cue.

The formal logic is simple but consequential. Let T denote an act type and a a particular performance. The relation “a instantiates T” is not identity between a and T. If a instantiates both T1 and T2, the inference that T1 equals T2 is invalid. The same token can fall under several kinds. The further inference that two tokens are identical because they instantiate the same type is also invalid. Action individuation repeatedly goes wrong when these relations are confused.

Descriptive equivalence is equally important. Expressions such as “lifting a cup” and “raising the cup” may in context describe one event. Yet synonymous language is neither necessary nor sufficient for identity across all contexts. The descriptions “unlocking the door” and “letting a visitor enter” are not synonymous even if they sometimes describe related aspects of one act. The ontology must account for how a description picks out a token, not simply for whether the verbs have the same dictionary meaning.

Ordinary action discourse also contains aspectual distinctions. “Building a bridge” can describe an ongoing, incomplete activity; “building the bridge” in a completed sense entails an achieved result. A person may be building a bridge without ever finishing it. Such linguistic differences are grounded in distinctions about processes and outcomes, not merely arbitrary labels. Zeno Vendler’s classification of verb forms and temporal structures provided a major point of departure for later action and event semantics (Vendler, 1957).

The appropriate conclusion is methodological. Act types help specify the nature of the performance, while tokens answer which worldly occurrences existed. Descriptions provide evidence of how those entities are organized and recognized. None can serve as a universal substitute for the others.

Identity, Causal Connection, and Mereological Relations

Three relations frequently masquerade as action identity. A performance may cause a result, constitute a higher-level act, or form a proper part of a more extensive activity. These are different relations, and a philosophical theory must establish which one holds in each case.

Causation commonly links distinct occurrences. The action of striking a bell brings about the bell’s sounding. The sound is an effect; the striking can be described through that effect. The action of performing a preparatory test may also cause another action several minutes later. Since causal succession is asymmetric and numerical identity symmetric, the bare proposition that one occurrence causes another cannot establish that they are the same particular action.

Constitution is harder. Placing a marked ballot in a box may constitute casting a vote under valid institutional rules. The institutional act is founded upon a physical performance without being reducible to a collection of muscle contractions independent of rules. Whether the casting and depositing are two actions or one institutionally described action is exactly what coarse and fine-grained accounts dispute. The constitutive connection should therefore be made explicit before identity is asserted.

Part–whole relations also need independent recognition. A musician’s performing a sonata has individual movements and passages as parts. One passage can begin before the entire performance ends, and the performance may continue after the passage. These temporal and mereological relations explain how many smaller actions can realize one extended undertaking without being identical to the undertaking. A proper part is not the same as the whole.

An especially instructive counterexample concerns chains of instrumental doing. If A is done by doing B, and B is done by doing C, a transitive means relation may connect C with A. It does not follow that A, B, and C are all numerically identical. Transitivity of identity would erase genuine distinctions among the stages if it were applied indiscriminately. Level-generation is useful because it permits directed dependence without such erasure.

A satisfactory ontology requires at least these relations: token identity, instantiation of a type, causation, constitution or generation, and parthood. The existence of multiple relations is evidence that action individuation is a structured metaphysical problem rather than a contest to maximize or minimize the number of entities.

Intentionality Under a Description

An agent may intentionally move a lever, knowingly activate a light, and unwittingly alert a hidden observer. The action’s description-relative intentionality is compatible with a fixed physical history. This insight, associated with Anscombe and Davidson, places an important constraint on individuation: sameness of an action token does not entail sameness of the agent’s knowledge or intention across every true characterization of it (Anscombe, 1957, §§23–26; Davidson, 2001).

To understand the point, distinguish three propositions. The person intended to move the lever. Moving the lever activated the light. The person intended to activate the light. The first two can be true while the third is false. If activating the light is a true description of what the person did, the attribution of that description alone cannot establish intentionality under it. The issue is one of practical understanding and reasons, not merely event identification.

The distinction also works in reverse. The agent can intend to unlock a door but fail because the lock is broken. The intended action type is clear; its successful realization is absent. A theory that uses intention alone as the identity condition for completed action threatens to identify an unsuccessful attempt with an accomplished result. A theory that recognizes only achieved outcomes neglects intentional striving and incomplete activity.

An additional difficulty arises when an agent has multiple reasons for one motion. A person could raise a hand both to greet a friend and to signal agreement. There may be one bodily performance under two intentional descriptions, or two acts at different normative levels generated through one performance. The right account depends on the theory of acts and on the practical context; it cannot be settled by assuming that intention always yields one new token.

Intentionality thus helps define the practical profile of action while leaving the identity question open. The relation between being intentional under a description and being the same particular action deserves independent analysis. This distinction preserves both the reality of an occurrence and the significance of how an agent understands it.

Reasons, Causes, and Deviant Chains

Reasons explain action by making intelligible why an agent pursued a particular course. Davidson’s causal approach further argues that an action’s being performed for a reason involves an appropriate causal connection between the relevant beliefs, desires, or other motivating states and what occurs. Action individuation inherits a crucial difficulty: identifying the same effect does not always identify the same action, especially when the causal route differs.

Suppose someone intends to knock on a door but is so nervous that the intention causes a tremor that knocks it accidentally. The bodily contact might match the intended result, yet the action has not unfolded in the way deliberate knocking normally does. Causal deviance shows that matching endpoints and an antecedent intention do not suffice. The intervening route must manifest the agent’s practical control appropriately. Modern causal theories of action discuss this as a challenge to bare causal criteria (Stanford Encyclopedia of Philosophy, “Action,” §2).

For individuation, two consequences follow. First, the causal history that identifies an action must distinguish an agent’s guided execution from accidental realization through the agent’s body. Second, two action descriptions can differ because one presupposes a successful practical relationship that the actual causal history fails to instantiate. “Intentionally warning the guard” cannot automatically be derived from “causing a sound heard by the guard.”

A more general question concerns the causal individuation criterion itself. Suppose two distinguishable events share many causes and effects. Co-occurrence in a common network does not establish that their total causal profiles are identical, and identifying a complete profile requires an account of which relations are fundamental. An identity criterion based on all causes and effects may also be too demanding for ordinary epistemic application. We often correctly identify an action without knowing its remote causes or downstream consequences.

Reasons therefore identify a practical explanatory pattern rather than directly counting action tokens. They provide constraints on how bodily events qualify as intentional performances, but the metaphysical boundary of the performance requires further specification.

Bodily Movements, Basic Actions, and Doing by Doing

Basic-action theories ask whether there are things an agent does directly, without doing them by means of some further action. Raising a hand might be basic relative to an ordinary practical description; moving a lever is performed by raising a hand in a certain setting; starting a machine is performed by moving the lever. The idea of a basic action promises an anchor for the means–end and generation relations that otherwise appear to descend indefinitely.

Wittgenstein’s Philosophical Investigations §621 asks us to consider what remains of raising an arm if we subtract the fact that the arm goes up. The question reveals the conceptual distance between a bodily occurrence and an action description: the agent’s raising an arm can be distinguished from the arm’s rising, even when both involve the same physical trajectory (Wittgenstein, 1953, §621). The passage does not itself provide a complete action-identity test, but it shows why reducing agency to movement threatens to erase the very distinction under investigation.

The proposal must distinguish what is basic from what is physically simple. Moving a finger can involve extensive neural and muscular activity. Those subpersonal processes need not be further actions performed by the agent. An account that insists every physiological cause of movement must itself be an action produces a regress. A philosophically defensible stopping point concerns a mode of agential control and practical explanation, not the smallest measurable physical unit.

Different accounts identify that point differently. Volitional theories, such as Ginet’s, locate an agentive core in simple mental acts. Bodily-action theories locate basic doings in direct exercise of bodily capacities. Process theories may regard basicness as relative to the practical question asked rather than as the identification of the smallest metaphysical particle. These proposals differ in what they treat as a direct exercise of agency (Ginet, On Action, 1990).

The relation “doing A by doing B” requires attention to practical order. A person may write a word by moving a pen, and move the pen through a coordinated bodily action. The word-writing is not identical to the pen’s movement under every description: spelling conventions and the meaning of the mark matter. Yet there is an intimate performance relation that justifies explaining the higher-level act through the lower-level movement.

Basicness should consequently be distinguished from numerical simplicity. An action can be one organized execution despite involving many bodily microevents. Conversely, a seemingly elementary movement can instantiate distinct action types in a rule-governed practice. The basic-action debate supplies candidate foundations for individuation while leaving the ontology of higher-level action open.

Temporally Extended Actions and the Boundary Problem

Many actions take time. Writing a book, negotiating an agreement, preparing a meal, or repairing a roof occurs through sequences that may contain interruptions, alternative methods, and revisions. The problem of individuation therefore includes identity over time: what makes later activity a continuation of the same undertaking, and when has another action begun?

Temporal adjacency is insufficient. A person can perform unrelated actions without interruption, and one sustained activity can continue after a substantial pause. Physical continuity is likewise insufficient: repairing an object may involve changing tools and relocating materials. Shared intention helps, but one intention can authorize several independent actions and a continuing project can survive modifications of its plan.

Michael Bratman’s planning theory provides a framework for understanding this organization. Intentions support practical coordination over time by constraining reconsideration, organizing subplans, and allowing agents to conduct complex projects without deliberating anew over every movement (Bratman, Intention, Plans, and Practical Reason, 1987). Planning unity can therefore explain why discontinuous episodes belong to one undertaking. It does not automatically establish numerical identity among all their component actions.

Consider a painter who begins a portrait, sets it aside for a week, then returns and completes it. The same artwork-directed project continues. Yet the first day’s brushstroke and the later brushstroke remain distinct token actions. An account of the portrait-painting as one extended action requires a mereological structure that includes the episodes without collapsing them into a single instant. Its boundary conditions involve practical continuity, recognizable task identity, and the organization of execution.

A second difficulty arises with change of purpose. A writer composing one letter may discover that the text is becoming an essay and redirect the work. At what point does the original action terminate? A purely intention-based test may impose an arbitrary discontinuity, whereas a purely material-continuity test may ignore the transformative change. A plausible account should acknowledge both continuity of execution and change of practical organization.

The temporal problem is thus not a secondary complication. It shows that an action is often an organized unfolding whose identity is partly determined by how stages belong together. The metaphysics of processes offers tools for addressing this fact.

Events, Processes, and Activities

Event-based theories treat actions as countable particulars with temporal locations and causal relations. Process-based accounts emphasize unfolding, continuation, and intrinsic temporal structure. The distinction matters because an event that is complete can be described through its occurrence, while an activity may be in progress without having achieved a specified terminus. Philosophers including Alexander Mourelatos, Michael Thompson, Jennifer Hornsby, and Helen Steward have developed lines of inquiry that place processes and activities near the center of action ontology.

Mourelatos’s “Events, Processes, and States” (1978) clarified connections between event ontology and linguistic aspect, building on Vendler’s earlier distinctions (Mourelatos, 1978; Vendler, 1957). “She ran” and “she completed the race” locate activity and achievement differently. An ongoing process can have temporal parts without every part instantiating the description of the completed performance. This creates a principled distinction between the identity of an ongoing action and the success condition of an achievement description.

Thompson’s Life and Action (2008) treats action as possessing a structure that resists reduction to a punctiform cause–effect relation. Steward’s A Metaphysics for Freedom (2012) investigates agents and the powers through which their activities unfold, extending the question of agency beyond distinctively human rational planning (Steward, 2012). The explanatory significance of these approaches lies in their ability to represent the agent’s activity as developing through time.

The process view must also identify boundaries. A river’s flow can continue while its water changes, but the analogy alone cannot tell us how many actions a person performs. An agent may sustain a walking activity, stop to answer a call, and resume walking. Whether walking is one extended action depends on a theory of interrupted processes and practical integration. The concept of a process does not remove individuation; it reformulates its criteria.

The most promising result is recognition of temporal structure as constitutive evidence. An action may be a continuant-like unfolding with parts, stages, and a completion condition. Descriptions of its phases and of its outcome can refer to a structured whole without making each stage identical to that whole. This provides an alternative to both unrestricted event unification and unrestricted property-based multiplication.

Cumulative Accomplishments, Achievements, and Aspect

The grammar of action offers an indispensable test of metaphysical theories. A person can be crossing a river while still on the near side; the person has not crossed it until reaching the other side. “Crossing” in its progressive use may be true before “crossed” becomes true. A runner can be running without finishing a race, and an author can be writing a manuscript without producing a completed book. These facts concern the temporal and achievement conditions of action descriptions.

Vendler’s distinction among activities, accomplishments, achievements, and states helps reveal the variety (Vendler, “Verbs and Times,” 1957). Activities commonly support subinterval descriptions. A person who walks for a period has been walking during many smaller portions of it. Accomplishments include a culmination that need not occur at each subinterval. Achievements are typically represented in relation to the transition at which something becomes true. The contrasts demonstrate that criteria for one action cannot be inferred by applying the same temporal rule to every action verb.

This point also reveals the hazard of treating linguistic aspect as a complete ontology. A verb phrase can reframe an occurrence under a different temporal perspective. One completed undertaking may be described as preparation, execution, or accomplishment. Linguistic tests identify differences in the truth conditions of descriptions but do not automatically settle whether two metaphysical entities exist. The philosophy of action must interpret linguistic evidence together with causal, temporal, and practical evidence.

A useful model distinguishes an ongoing process, its culminating transition, and the resulting state. Constructing a bridge, completing the bridge, and the bridge’s standing complete are related but differently structured. Whether the completion is a proper part or boundary of the building process is a further metaphysical question. The resulting state persists after the action ceases and should not simply be identified with the doing.

The analysis supports a general constraint: theories of action identity should accommodate incomplete and successful actions, interrupted activities, and cumulative undertakings without inventing a numerically new action for every grammatical aspect.

Results, Side Effects, and the Reach of a Description

A person intentionally opens a window and thereby cools a room, disturbs papers on a desk, and triggers an alarm. The initial bodily performance has several effects. Some are intended; others are foreseen; still others are unknown. The question is whether the effects are themselves actions, elements of a larger action, or consequences of the original action.

Coarse-grained descriptions may legitimately refer to the agent’s opening by specifying an effect. “Her cooling of the room” can designate an intentional accomplishment if the cooling is achieved through the opening. Yet the room’s temperature change is a separate physical development that may continue after the agent leaves. Distinguishing the action from its resulting state permits both a true consequence description and a nonidentical effect.

Goldman’s level-generation approach specifies that a lower-level act may generate a higher-level act through causal circumstances or conventional rules. This accounts for the directionality of an achievement: the agent cools the room by opening the window. The generation relation is informative even if an alternative metaphysics construes the act tokens as identical. The philosophical task is to determine what is lost or gained by each analysis.

Normative assessment sharpens the issue. The same throwing of a stone can be described as damaging a window, frightening an occupant, and provoking a later response. Whether all such outcomes count as things the agent did is not exhausted by physical causation. The attribution must examine actual causal involvement, the agent’s awareness, and the appropriateness of the relevant action type. Foreseeability is relevant to responsibility, but it is not identical to token identity.

If action individuation were determined solely by a list of consequences, a finite act could multiply indefinitely as its causal influence spreads. If it were determined solely by the smallest bodily event, complex achievements would lose their practical identity. The solution requires distinguishing what a description selects, what an action constitutes, and what follows from it.

Omissions, Failures, and Attempts

An omission is typically identified through something that did not occur: a driver failed to signal, a guard did not close a gate, or a physician did not perform a required check. Some theories treat omissions as negative facts rather than positive events; others argue that intentional refraining or failing to act can be genuine exercise of agency in a context. The ontology matters for individuation because a nonoccurrence cannot be counted by tracing the same bodily movement that identifies many overt actions.

Consider someone who intentionally refrains from voting. Remaining seated, reading a book, and avoiding the polling station may be positive activities occurring during the period. The abstention cannot automatically be identified with every event that fills the interval. It is individuated relative to an opportunity, a possible course of action, and an agent’s practical orientation. The boundaries of an omission are often set by the opportunity to act and by the normative context, rather than by a physical event with a single onset.

Attempts present a different problem. A person may try to open a door and fail because the lock is jammed. The failed attempt is an action, even though the successful action type “opening the door” has not been instantiated. A theory that individuates actions solely by their achieved effects cannot adequately count this effort. A theory that individuates by intentions alone, however, may fail to distinguish one attempt from three retries with the same intention.

The right distinctions separate the intended type, the attempted performance, the achieved result, and the causal history of execution. Repeated attempts can be several token actions organized by one intention. An omission can be part of a practical policy or an intentional refraining even when no positive bodily performance uniquely corresponds to it. The ontology of positive events is therefore an important test case, rather than an unquestioned model for all agency.

These cases reveal the normative and modal dimensions of action individuation. What an agent could do, what was required, what was attempted, and what succeeded influence descriptions without directly producing a simple numerical criterion.

Social Rules and Institutionally Constituted Actions

Some actions exist as actions of their kind because a public practice establishes their conditions. Signing a document can constitute making an agreement; placing a vote in an authorized setting can constitute casting a ballot; uttering words under appropriate conditions can constitute issuing a promise. The significance of an occurrence depends on procedures, roles, and institutional rules.

J. L. Austin’s analysis of performative utterances in How to Do Things with Words (1962) makes this distinction vivid. A spoken sequence is a physical and linguistic occurrence, but its institutional significance depends on felicity conditions. The same words uttered as a rehearsal, joke, unauthorized declaration, or authorized act can have different practical consequences. An identity theory that examines only acoustics will miss the conditions that constitute the act.

Institutional descriptions do not necessarily introduce a second physical event. The utterance and the promising can be related as a physical performance and the institutional act constituted through that performance. A fine-grained ontology may count distinct act tokens. A coarse-grained theory may attribute several properties to one performance. Both must explain the conventional framework rather than assume that the action’s identity can be read off a bodily movement.

The institutional setting also affects persistence and completion. A contract can be formed through a sequence of offers, acceptances, and formalities. A vote may require registration and validation. The temporal boundaries of an institutional act can differ from those of its immediate bodily execution, and a result can be annulled later without erasing the historical occurrence of the original attempt.

Action individuation therefore has a social dimension that resists reduction to physical coincidence. The relevant rules determine whether a higher-level action type is instantiated. Whether that type’s instantiation is a distinct token requires an explicit ontological judgment. The constitution relation is an essential part of the explanation whichever judgment is adopted.

Shared and Collective Actions

Collective conduct multiplies the difficulties of individuation. Two people carrying a table perform their respective bodily activities while together carrying one object. A research group conducts an investigation through separate inquiries, meetings, revisions, and decisions. A musical ensemble performs a work through interdependent individual parts. The question “How many actions occurred?” has distinct answers at individual, joint, and institutional scales.

A joint action is organized through coordination. One person’s contribution cannot be understood merely as a physical movement if its role depends on expectations of the others. Bratman’s planning framework shows how shared intentions and interlocking plans can explain coordinated agency over time (Bratman, Shared Agency, 2014, publisher bibliography). Such organizational unity can underwrite the attribution of a collective performance without requiring the identification of every contributing personal action with a single event.

The metaphysical counting issue is acute. If two individuals jointly lift a piano, there are at least distinguishable bodily performances and an organized cooperative accomplishment. Does the collective lifting constitute one further action token, or is it a relation among the individual actions? The answer will depend on the adopted ontology of collective agency. The existence of a coherent collective description by itself does not determine numerical identity.

A useful distinction concerns scale of agency. At the individual scale the relevant tokens are each participant’s contributions. At the cooperative scale the relevant organized performance is the joint lifting. At the institutional scale the performance may realize a formally authorized undertaking. These are connected through constitution, coordination, and parthood; automatically identifying all of them would compromise their distinct causal and normative explanations.

The broader conclusion is that unity need not mean simplicity. A many-agent performance can possess an intelligible practical unity without losing its internal plurality. Action individuation must be able to recognize both properties within one structured account.

Animal Agency, Technology, and Nonhuman Performance

The problem of action identity extends beyond cases requiring fully articulated verbal reasons. Animals hunt, carry objects, build nests, and adjust movements to environmental circumstances. Whether every such activity is intentional in the human philosophical sense is debated, but the organization of bodily activity into temporally extended, goal-directed episodes is an intelligible subject of agency research. A general individuation theory should not make possession of a human vocabulary a necessary condition of every action token.

Technological settings introduce another complication. A person issues a command to a machine; a program executes a sequence of operations; a remotely operated device changes the environment. The description “the person opened the gate” may be appropriate even when the immediate actuator is a machine. The description identifies mediated agency, while the physical execution may include additional processes attributable to the device. Causation, delegation, and authority determine how these descriptions are connected.

Automated systems can also display structured activities involving goals, constraints, and feedback. Their individuation requires specifying the relevant level: a processor event, a software operation, a task episode, or an institutionally attributed action. Inferring that every computation is a human-style intentional action would disregard the need to justify agency criteria. Equally, denying all action vocabulary to systems that exhibit organized world-directed performance would prejudge the conceptual issue. The question should be approached through an explicit theory of agency and of performance conditions.

Within the philosophical horizon of Artificial Era, the broader question concerns the conditions under which Artificial can be described as a bearer of acts and commitments in its own order. That project-specific horizon is analytically adjacent to the present investigation; it does not determine the identity of classical human actions and is not needed as a premise in the arguments from Anscombe, Davidson, Goldman, or Kim.

The direct conclusion for this dossier is methodological: the ontological identity of an occurrence, the source of its control, and the normative practice under which it counts as an action are separable questions across human, animal, collective, and technological cases.

Formal Tests and Counterexamples for Identity Criteria

A theory of action individuation should state its identity condition in a form that permits testing. Let a and b range over action tokens. The formula a = b asserts numerical identity. Suppose a is an agent’s raising a hand and b is the agent’s signaling agreement. If the two expressions denote the same token, their difference lies in the predicates under which that token is characterized. If they denote different tokens, the theory must identify the extra ontological difference.

Three tests are decisive. First, a proposed identity relation must be symmetric and transitive. If a generates b but b does not generate a, generation is not identity. If a is a proper part of b, parthood is not identity. Second, a criterion must handle the coexistence of distinct descriptions without allowing sheer vocabulary to determine the number of actions. Third, it must distinguish two causally connected performances even where they pursue one end.

The following cases expose the pressure points.

A person raises a hand and thereby votes in a show of hands. A coarse-grained theorist can describe one event under bodily and institutional predicates; a fine-grained theorist can distinguish the bodily act and the conventionally generated act. Their explanatory obligations concern the same occurrence and the relevant voting rule.

A person pushes a button, waits several minutes, and later adjusts a dial to keep a machine running. Shared purpose does not identify these two bodily interventions. A practical project unites them as parts, while their occurrence histories remain different.

A person starts to draw a circle, pauses, and finishes it the next day. Continuous movement does not define the action’s project identity; the circle-drawing can be one extended undertaking with component acts. Whether an unfinished initial attempt is one token action depends on the individuation of the process and its boundaries.

A person intends to press a switch, but a tremor independently depresses it. The matching bodily outcome does not establish the intended action description as realized through practical control. The causal route is relevant.

These tests demonstrate the importance of declaring a grain of individuation and a category of entities. They do not establish that identity is purely conventional. Rather, a theory must justify why the selected level of analysis tracks genuine worldly unity.

Comparative Evaluation of Major Accounts

Davidson’s coarse-grained approach gives the clearest account of how many true descriptions can refer to one event. Its chief strengths are event-semantic coherence, stability under redescription, and an economical ontology of particular occurrences. It is strongest where an agent’s performance is relatively punctual and where consequences enter as descriptions of what was done. Its challenge arises when the putatively one action contains temporally extended, internally articulated stages or when causal relations between stages must be preserved.

Goldman’s fine-grained theory illuminates the organization of acts under different properties and the asymmetrical generation of higher-level performances. It is especially helpful for institutional actions and the relation between basic doings and their achievements. Its challenge is to specify which act-properties individuate tokens and to avoid inflation where descriptions merely add accidental predicates.

Kim’s property-exemplification model offers a metaphysical rationale for fine-grained event identity. It is a powerful theory of event constitution by subject, property, and time. The open problem is whether the constitutive property can be selected without reliance on arbitrary linguistic reformulations, and whether event identity by itself determines which events are intentional actions.

The intermediate accounts represented by Davis, Thomson, Hornsby, and Ginet show that theory choice is richer than a binary. A view can accept multiple action kinds without multiplying every descriptive variant. It can also treat some differences as differences among exercises of capacities, while treating others as modifications of one token.

Process-oriented theories focus on the temporal and mereological unity of conduct. Their chief advantage is the explanation of incomplete, interrupted, and developing activities. Their challenge is to identify start, continuation, completion, and division criteria without making every proposed practical project into one numerically indeterminate process.

The comparative lesson is that these theories illuminate different individuation tasks. An event identity criterion, a type taxonomy, a means–end structure, a practical-knowledge account, and a theory of temporal continuation answer different questions. A comprehensive philosophical account must coordinate them while preserving genuinely competing metaphysical claims at the points where they disagree.

A Scale-Explicit Synthesis of Action Unity

The preceding inquiry supports a systematic approach that makes explicit the ontological scale and the relations under examination. It is proposed here as an analytical synthesis of the historical material, not as a newly established canonical theory. Its governing claim is that responsible action counting requires a specified unit of analysis and a justified criterion of persistence, rather than an unqualified total number of things done.

At the performance level, the theory identifies an occurrence through its agent or agents, time, causal and execution history, and appropriate boundaries. At the practical-organization level, it identifies how such performances form parts of a temporally sustained undertaking. At the descriptive and institutional level, it identifies the kinds of action instantiated under intentions, norms, and constitutive rules. These levels are interrelated, and they need not assign the same count.

The resulting analysis treats five relations as primitive distinctions of inquiry: identity between tokens; instantiation of types; causal production; constitution or level-generation; and parthood within extended activity. Each relation answers a characteristic question. Identity asks whether two terms denote one occurrence. Instantiation asks what kind of performance occurred. Causation asks what brought about what. Constitution asks what counts as a higher-level act under the relevant conditions. Parthood asks how a complex action includes stages.

A plausible criterion of token unity must incorporate a coherent execution history. Where one ongoing performance survives temporal changes, criteria of continuity and practical organization establish how stages belong together. Where two performances have distinct beginnings and independent courses of execution, sharing a purpose or outcome does not merge them. Where institutional rules alter what a performance counts as, those rules explain its type and significance without settling in advance whether the ontology must add a separate token.

This approach protects a truth in Davidson’s analysis: redescribing an event is not an act of metaphysical creation. It protects a truth in Goldman’s analysis: higher-level accomplishments can possess real constitutive and explanatory structures beyond the wording of a bodily description. It incorporates Kim’s demand for criteria that determine what an event is and the process theorists’ insistence that identity through time cannot be ignored.

The philosophical gain lies in replacing an indiscriminate count with an account of the entity counted and the relations that organize action. This conclusion remains open to competing event ontologies, but it establishes common conditions under which their claims can be assessed.

Contemporary Research and Outstanding Problems

Recent scholarship has renewed attention to practical knowledge, rational agency, constitutive explanations, extended action, and the metaphysics of processes. The Stanford Encyclopedia of Philosophy’s contemporary entry on action treats the nature of action, intentional action, extended activity, process ontology, basic actions, and omission as connected issues. Contemporary research thereby places the historical identity debate within a wider inquiry about how agency is exercised and sustained.

One open problem is the individuation of the fundamental property or capacity relevant to fine-grained theory. If constitutive properties are too permissive, every accurate predicate threatens a new event. If they are too restrictive, the account may lose institutionally and practically distinct acts. An acceptable solution requires an independent account of the relevant property or capacity.

Another problem is temporal identity. Work on process ontology identifies real continuities where an action unfolds over an extended interval, yet practical interruptions, revisions, and failures resist simple boundary rules. The decisive question is which continuities constitute one action rather than one project containing several actions.

A third problem is causal and normative attribution. The relationship between a performance and its foreseen, intended, or accidental consequences is essential to law, ethics, and social explanation. The identity of the occurrence cannot alone determine responsibility. An action described by its effects may still require a distinct causal and epistemic analysis before the description supports blame or praise.

A fourth concerns collective and technologically mediated performance. Agency can be exercised through division of labor, formal institutions, tools, and automated operations. The ordinary categories of one body and one movement are insufficient for the practical unity of these activities. A theory of action identity must distinguish action performed by an agent, action jointly performed by agents, and outcomes attributed under a rule.

Finally, there is the methodological problem of plural descriptions. How can the same token be intentionally described in one way, unintentionally described in another, and institutionally classified in a third without reducing action identity to verbal practice? The durable answer must relate the actual history of performance to the structure of description and practical knowledge. The historical debate has not eliminated the problem; it has made the dimensions of an adequate solution more precise.

Conclusion

Action individuation asks what makes an action one action and under what conditions the apparent plurality of what an agent does represents distinct particular actions. The inquiry has established that there are several questions within this formulation: the identity of particular occurrences, the repeatability of act types, the relation of deeds to purposes and results, and the continuation of extended activity over time.

The historical reconstruction shows a continuous concern with practical organization. Aristotle examined voluntary action through its origin and circumstances. Aquinas distinguished command, execution, and the unity of human acts. Hegel analyzed action through the external deed, subjective intention, and imputation. Anscombe showed the explanatory importance of a series of true action descriptions and of the agent’s knowledge of what is being done. Davidson supplied a general event ontology capable of explaining unity across redescription. Goldman and Kim supplied fine-grained alternatives attentive to act properties, level-generation, and event constitution. Later approaches developed intermediate criteria, volitional foundations, and process-based accounts.

The main philosophical result is that plurality of descriptions is neither sufficient for plurality of actions nor sufficient for their identity. The inference from one physical movement to one action is equally insecure when higher-level and temporally extended performances are at issue. Causal production, institutional constitution, and part–whole organization are genuine relations that cannot be replaced by an identity assertion without argument.

A disciplined account begins by declaring whether it is counting performance tokens, action types, action processes, or completed undertakings. It then establishes boundaries, execution history, practical organization, and the rules under which descriptions are true. The resulting picture permits one occurrence to be described under several kinds, several occurrences to form one complex undertaking, and a completed action to have consequences that remain distinct from it.

This synthesis does not eliminate the metaphysical disagreement between coarse-grained and fine-grained theories. It establishes the terms of its responsible adjudication. The identity of an action is a claim about the unity of what actually occurs, interpreted through the practical and conceptual structures that make it an action.