The four independent sub-features of agentivity (@cite{cruse-1973} §pp.17–21).
Cruse argues that "agentivity" is not a single binary feature but decomposes into at least four independent components, each with distinct linguistic diagnostics.
- volitive : AgentivityFeature
Volitive: an act of will is stated or implied (Cruse p.18). "John deliberately drifted downstream."
- effective : AgentivityFeature
Effective: force deriving from position, motion, or kinetic energy (Cruse p.19). "The bullet smashed the collar-bone."
- initiative : AgentivityFeature
Initiative: initiating action by command or instruction to another agent (Cruse p.19). "The warder marched the prisoners."
- agentive_ : AgentivityFeature
Agentive: use of one's own body's internal energy source (Cruse p.20). "John ran." The prototypical agent feature.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
An agentivity profile assigns Prop-valued feature predicates over entity–event pairs for each of Cruse's four sub-features.
Each field hasF x e means "entity x exhibits feature F in event e".
Does x exhibit an act of will in e?
Does x exert force (from position/motion/energy) in e?
Does x initiate action by command/instruction in e?
Does x use own body's internal energy in e?
Instances For
The do-test (@cite{cruse-1973} §pp.13–14): "NP VP" entails "NP did something" iff at least one agentivity sub-feature is present.
This is the disjunction of all four features.
Equations
- Semantics.Events.Agentivity.passesDoTest x e profile = (profile.hasVolitive x e ∨ profile.hasEffective x e ∨ profile.hasInitiative x e ∨ profile.hasAgentive x e)
Instances For
The do-test is equivalent to the 4-way disjunction (definitional).
Axiom class witnessing that the four agentivity features are logically independent: each can be present without the others.
volitive_without_agentive: "John deliberately drifted downstream" (will present, no own-energy expenditure)effective_without_volitive: "The bullet smashed the collar-bone" (force present, no will)initiative_without_agentive_: "The warder marched the prisoners" (command present, prisoners do the marching)agentive_without_initiative: "John ran" (own energy, no command to another)
- volitive_without_agentive : ∃ (x : Entity) (e : Ev Time), profile.hasVolitive x e ∧ ¬profile.hasAgentive x e
Volitive without agentive: "John deliberately drifted downstream"
- effective_without_volitive : ∃ (x : Entity) (e : Ev Time), profile.hasEffective x e ∧ ¬profile.hasVolitive x e
Effective without volitive: "The bullet smashed the collar-bone"
- initiative_without_agentive_ : ∃ (x : Entity) (e : Ev Time), profile.hasInitiative x e ∧ ¬profile.hasAgentive x e
Initiative without agentive: "The warder marched the prisoners"
- agentive_without_initiative : ∃ (x : Entity) (e : Ev Time), profile.hasAgentive x e ∧ ¬profile.hasInitiative x e
Agentive without initiative: "John ran"
Instances
Link between Parsons' agent role and Cruse's agentive_ sub-feature.
The Parsonian agent(x,e) captures specifically the own-energy
sub-feature: an agent uses its own body's internal energy source.
This is strictly narrower than the full do-test.
Parsons' agent implies Cruse's agentive_ feature.
Instances
Parsons' agent(x,e) entails passesDoTest(x,e), since agentive_ is one of the four disjuncts.
In any model where agent → hasAgentive, the result follows immediately from the fact that agentive_ is the fourth disjunct.
Parsons' agent role captures specifically Cruse's agentive_
sub-feature (own energy, dynamic), not the full do-test notion.
This is stated as: any model satisfying AgentAgentiveLink and
ThematicAxioms has agent entail agentive_ (from the link) and
agent entail action (from the axioms). Together these characterize
the prototypical "own-energy + dynamic" combination.
Map agentivity features to CoerciveImplication's ActionType.
- volitive → Volitional (Cruse: act of will ↔ N&L: volitional action)
- effective → NonVolitional (force without will)
- initiative → Volitional (initiating by command is volitional for initiator)
- agentive_ → Ambiguous (own energy can be volitional or habitual)
Equations
- Semantics.Events.Agentivity.AgentivityFeature.volitive.toActionType = NadathurLauer2020.CoerciveImplication.ActionType.Volitional
- Semantics.Events.Agentivity.AgentivityFeature.effective.toActionType = NadathurLauer2020.CoerciveImplication.ActionType.NonVolitional
- Semantics.Events.Agentivity.AgentivityFeature.initiative.toActionType = NadathurLauer2020.CoerciveImplication.ActionType.Volitional
- Semantics.Events.Agentivity.AgentivityFeature.agentive_.toActionType = NadathurLauer2020.CoerciveImplication.ActionType.Ambiguous
Instances For
Volitive maps to Volitional.
Effective maps to NonVolitional.
Initiative maps to Volitional.
Agentive maps to Ambiguous.
Coercive implication (@cite{nadathur-lauer-2020}) arises exactly when the causee's action is volitional — i.e., when the causee has at least the volitive sub-feature.
This bridges Cruse's agentivity decomposition to N&L's coercion analysis: "X made Y do Z" implies coercion when Z is volitional for Y, which Cruse would analyze as Y having the volitive feature.
Cruse's initiative feature (initiating action by command) corresponds to causative constructions where the subject causes an agentive action by the object.
"The warder marched the prisoners" → the warder has initiative,
the prisoners are the agentive doers. This pattern is lexicalized by
CausativeBuilder.make and .force: the subject (causer) initiates,
the object (causee) performs the action.
The bridge: initiative ↔ {make, force} builders, where the causer has initiative and the causee has agentive_.
- causer_has_initiative (causer causee : Entity) (e : Ev Time) : profile.hasInitiative causer e → ¬profile.hasAgentive causer e → profile.hasAgentive causee e
In a causative construction, the causer has initiative.
Instances For
The force builder is coercive — it lexicalizes coercion of a volitional causee, which is precisely the initiative pattern where the initiator overrides the causee's will.
The make builder asserts sufficiency — the initiator's action is sufficient for the causee to act.
There exist stative eventualities that pass the do-test.
Witness: "John is standing" — volitive (John can stop standing) but stative (no change over the interval). The do-test passes via the volitive feature, even though e.sort =.state.
This shows the do-test is strictly broader than Parsons' agent
role, which requires e.sort =.action.
Parsons' agent_selects_action is NOT contradicted by stative
do-verbs, because agent captures the narrower agentive_
feature (own energy → dynamic), while the do-test also detects
volitive which can apply to states.
Formally: given ThematicAxioms (which assert agent → action),
and AgentAgentiveLink (which assert agent → agentive_), if an
entity is agent of a stative event we get a contradiction — so
no stative event has a Parsonian agent. The do-test still passes
for statives via other features (volitive, effective).