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/MartinRoseNichols2025.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 CausativeBuilder (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
- One or more equations did not get rendered due to their size.
Instances For
Equations
- MartinRoseNichols2025.instBEqCausationType.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
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
- One or more equations did not get rendered due to their size.
Instances For
Equations
- MartinRoseNichols2025.instBEqThickThinClass.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
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
Asymmetric Entailment: P-CAUSE → D-CAUSE #
In the deterministic limit (our CausalDynamics), production-based
causation (P-CAUSE) entails dependence-based causation (D-CAUSE).
Formally: in a single-pathway causal model (no overdetermination), if the cause is sufficient for the effect, it is also necessary. This captures: if x produces y via direct energy transfer, removing x would prevent y.
In a single-pathway model (one causal law, no background alternatives), sufficiency implies necessity.
This is the deterministic formalization of "P-CAUSE entails D-CAUSE": when there is exactly one causal pathway and no overdetermination, a sufficient cause is also necessary.
Note: This fails under overdetermination (see builders_truth_conditionally_distinct),
which is exactly when P-CAUSE and D-CAUSE can come apart.
Concrete instance of the single-pathway entailment, fully proved.
For specific variables a, b, the entailment P-CAUSE → D-CAUSE holds: sufficiency implies necessity in a single-pathway model. This is verified by computation.
Under overdetermination, sufficiency does NOT imply necessity.
This is when P-CAUSE and D-CAUSE come apart: a cause can transfer energy (sufficient) but the effect would have occurred anyway from another source (not necessary).
Reuses the existing witness from builders_truth_conditionally_distinct.
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
Instances For
Thick verbs prefer production causation.
Thin verbs default to dependence causation.
Bridge to CausativeBuilder #
P-CAUSE maps to makeSem (sufficiency): production causes are sufficient.
D-CAUSE maps to causeSem (necessity): dependence causes are necessary.
The overt verb cause encodes D-CAUSE and uses CausativeBuilder.cause.
Thick lexical causatives encode P-CAUSE and align with CausativeBuilder.make.
Note: lexical causatives don't literally use CausativeBuilder (which
classifies periphrastic verbs), but their internal CAUSE operator has the
same truth conditions as makeSem when P-CAUSE applies.
Map causation type to the analogous CausativeBuilder.
This is the structural bridge: P-CAUSE's truth conditions correspond
to sufficiency (makeSem), 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 CausativeBuilder.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 (§6).
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 CausalProfile #
The structural causal model's CausalProfile directly determines the
dominant causation type via its directness and necessity fields. This
connects model-level structural properties to the production/dependence
distinction without going through any study-specific representation.
Map a structural causal profile 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.