Burning Facts: Thick and Thin Causatives #
@cite{embick-2009} @cite{martin-rose-nichols-2025} @cite{rose-nichols-2021} @cite{wolff-2003}
Two concepts of CAUSE underlie lexical causative verb semantics:
- P-CAUSE (production): Energy transfer from cause to effect. Physical, direct, requires concrete causer. Thick causatives (burn, break, melt) preferably encode P-CAUSE.
- D-CAUSE (dependence): Counterfactual dependence of effect on cause. Abstract, allows absences/facts/degrees as causes. Thin causatives (kill, destroy, damage) are compatible with D-CAUSE.
Key Property #
P-CAUSE asymmetrically entails D-CAUSE: if x produces y (energy transfer), then y counterfactually depends on x. But not vice versa: absences can be d-causes without producing anything.
Bridges to Existing Infrastructure #
| Concept | Maps to | Module |
|---|---|---|
| P-CAUSE (deterministic) | causallySufficient + directness | Sufficiency.lean |
| D-CAUSE (deterministic) | causallyNecessary | Necessity.lean |
| Thick → P-CAUSE preference | production constraint | (this file) |
| Thick → strong ASR | resultative compatibility | Causation/Studies/Semantics.Causation.ProductionDependence.lean |
Builder .make | sufficiency = P-CAUSE in deterministic limit | Builder.lean |
Causation Type #
The two concepts of causation that lexical causative verbs can encode.
This is orthogonal to the Causative (which classifies periphrastic
causatives by force-dynamic mechanism) — it classifies how lexical
causatives encode the causal relation itself.
Two concepts of CAUSE operating in lexical causative semantics.
production: Energy/force transfer from causer to causee (P-CAUSE). Requires a concrete, physical causer. Thick causatives preferably encode this. Entailsdependence.dependence: Counterfactual dependence of effect on cause (D-CAUSE). Compatible with abstract causes (absences, facts, degrees). Thin causatives and overt cause encode this.
- production : CausationType
Production-based: energy transfer, requires concrete causer. burn, break, melt in their physical sense.
- dependence : CausationType
Dependence-based: counterfactual, allows abstract causers. kill, destroy, damage, overt cause.
Instances For
Equations
- Semantics.Causation.ProductionDependence.instDecidableEqCausationType x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Thick vs Thin Classification #
The core empirical distinction: thick causatives encode manner of causing (either via an event property like break or a state property like bury), while thin causatives specify only the result state.
Whether a lexical causative verb encodes manner of causing.
thick: Encodes manner — restricts subjects to productive causes. Subdivided by HOW manner is encoded (event predicate vs state property).thin: Result-only — silent on manner, compatible with any cause type.
- thickManner : ThickThinClass
Thick via event predicate: root is a predicate of the causing event. break, burn, melt, cut — @cite{embick-2009} break-class. Compatible with strong adjectival resultatives (burn clean).
- thickState : ThickThinClass
Thick via state property: result state reveals production process. bury (buried → covered with earth). Not compatible with strong ASR.
- thin : ThickThinClass
Thin: result-only, no manner specification. kill, destroy, damage, change, activate.
Instances For
Equations
- Semantics.Causation.ProductionDependence.instDecidableEqThickThinClass x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Is the verb thick (encodes manner of causing)?
Equations
Instances For
Is the verb a causative manner verb (@cite{embick-2009} break-class)? These are the thick verbs whose root is an event predicate, compatible with strong adjectival resultatives.
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Asymmetric Entailment: P-CAUSE → D-CAUSE #
In the deterministic limit, production-based causation (P-CAUSE) entails dependence-based causation (D-CAUSE): in a single-pathway causal model (no overdetermination), a sufficient cause is also necessary. Under overdetermination, sufficiency and necessity come apart — exactly when P-CAUSE and D-CAUSE diverge.
The qualitative claim is witnessed concretely via V2 BoolSEM models
in study files — e.g., the overdetermination case is witnessed by
Lewis1973.Overdetermination.overdetermination_no_butfor_a (the
chronologically-canonical owner of the symmetric-overdetermination
scenario; see also the analogous V2 sufficient-but-unnecessary
divergence in NadathurLauer2020.Bus.cause_infelicitous_for_bus).
Production Constraint #
Thick causatives preferably convey P-CAUSE (production-based causation). This is a pragmatic constraint arising from competition between the lexical (covert) causative and the periphrastic (overt) cause.
The production constraint: thick causatives prefer production causation.
When a thick causative is used in its physical sense, the CAUSE operator preferably receives a production-based interpretation. This is because the manner information makes P-CAUSE a salient alternative, and the more specific lexical form specializes in the more specific meaning.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Thick verbs prefer production causation.
Thin verbs default to dependence causation.
Bridge to Causative #
P-CAUSE maps to causallySufficient (sufficiency): production causes are sufficient.
D-CAUSE maps to causeSem (necessity): dependence causes are necessary.
The overt verb cause encodes D-CAUSE and uses Causative.cause.
Thick lexical causatives encode P-CAUSE and align with Causative.make.
Note: lexical causatives don't literally use Causative (which
classifies periphrastic verbs), but their internal CAUSE operator has the
same truth conditions as causallySufficient when P-CAUSE applies.
Map causation type to the analogous Causative.
This is the structural bridge: P-CAUSE's truth conditions correspond
to sufficiency (causallySufficient), D-CAUSE's to necessity (causeSem).
Equations
Instances For
Production causation is analogous to the .make builder.
Dependence causation is analogous to the .cause builder.
P-CAUSE's analogous builder asserts sufficiency.
D-CAUSE's analogous builder asserts necessity.
Bridge to Resultatives #
Thick causative manner verbs (break-class) are compatible with strong
adjectival resultatives (break open, burn clean). This connects to
the resultative infrastructure in ArgumentStructure/Studies/TheoryComparison.lean,
where the constructional CAUSE uses Causative.make.
Causative manner verbs (thickManner) are compatible with strong ASR. This is the @cite{embick-2009} generalization formalized as a derived property.
Thin verbs and thick-state verbs (bury) are NOT compatible.
Equations
Instances For
Manner verbs are ASR-compatible.
Thin verbs are NOT ASR-compatible.
Thick-state verbs (bury) are NOT ASR-compatible. bury is thick but not a causative manner verb.
Production entails directness.
When a verb encodes P-CAUSE (energy transfer), the causal relation is necessarily direct: for energy to transfer, there must be physical contact or at least no intervening causer at the same level of granularity. This is why the directness constraint holds specifically for thick causatives.
Equations
Instances For
Bridge to structural queries #
The structural causal model's directness and necessity directly determine the dominant causation type. This connects model-level structural properties to the production/dependence distinction without going through any study-specific representation.
Map directness/necessity to the dominant causation type.
direct = true→ P-CAUSE (production): a direct causal law implies energy/force transfer.necessary = true(without directness) → D-CAUSE (dependence): counterfactual dependence without direct interaction.- Neither → no causal involvement.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Production type iff direct causal connection.
Dependence type iff necessary but not direct.
Bridge to CC-Selection #
CausationType determines which CC-selection mode applies:
P-CAUSE (production) → completion (the cause directly completes a
sufficient set). D-CAUSE (dependence) → membership (any necessary
condition qualifies).
Map causation type to CC-selection mode.
production→completionOfSufficientSet: energy-transferring causes complete the sufficient set directlydependence→memberOfSufficientSet: counterfactual causes are any necessary member of a sufficient set
Equations
- Semantics.Causation.ProductionDependence.CausationType.production.selectionMode = Semantics.Causation.CCSelection.CCSelectionMode.completionOfSufficientSet
- Semantics.Causation.ProductionDependence.CausationType.dependence.selectionMode = Semantics.Causation.CCSelection.CCSelectionMode.memberOfSufficientSet
Instances For
Production uses completion selection.
Dependence uses member selection.
selectionMode_roundtrip removed in Phase D-G2 — the legacy bridge
between CCSelectionMode.toSemantics (now deleted) and
Causative.toSemantics (now polymorphic V2-shaped, different signature).
The three encodings (CausationType, CCSelectionMode, Causative) remain
consistent at the enum level (production_selects_completion etc.); the
semantic-dispatch consistency follows from the V2 hub structure.