Solstad & Bott (2022): Implicit causality and consequentiality for psych verbs

@cite{solstad-bott-2022} @cite{dowty-1991} @cite{kehler-2002}

Experimental data on implicit causality (I-Caus) and implicit consequentiality (I-Cons) for German psych verbs, with proto-role analysis and cross-study bridges.

Verb classes

• StimExp (Stimulus-Experiencer): frighten, annoy, amuse — NP1 bias
• ExpStim (Experiencer-Stimulus): admire, like, fear — NP2 bias
• AgentEvocator (Agent-Evocator): criticise, congratulate — NP2 bias
• AgentPatient (Agent-Patient): kick, chase, hit — NP1 bias

Key empirical findings

1. Exp 1 (sentence continuation): I-Caus and I-Cons biases mirror each other for psych verbs. STIM-EXP: 87.4% NP1 with weil; EXP-STIM: 96% NP2 with weil.
2. Exp 2 (coherence relations): Explanations dominate over consequences for both classes; consequence rate differs by class.
3. Exp 3 (forced coreference): Asymmetry Hypothesis confirmed — even bias-incongruent continuations produce explanations.
4. Exp 4 (explanation types): Explanatory specifications appear almost exclusively in congruent explanations; consequence specifications are never produced — supporting verb-semantic I-Caus (Empty Slot Theory) vs. discourse-structural I-Cons (Contiguity Principle).

Two-Mechanism Account (Asymmetry Hypothesis)

I-Caus is verb-semantic: the verb's meaning contains an underspecified causal slot (Empty Slot Theory) that the continuation fills. I-Cons is discourse-structural: the Contiguity Principle prefers temporal continuation, defaulting to the endpoint of the described eventuality.

Proto-role analysis (@cite{dowty-1991})

IC bias tracks the stimulus argument: explanations in because-continuations target the participant whose entailment profile includes causation, regardless of grammatical position.

Class	Subject profile	P-Agent entailments
StimExp	C + IE (stimulus/causer)	causation, indep.exist.
ExpStim	S + IE (experiencer)	sentience, indep.exist.
AgPat	V + S + C + M + IE	all five

inductive SolstadBott2022.VerbClass : Type

Verb classes from the IC bias literature.

• stimExp : VerbClass
• expStim : VerbClass
• agentEvocator : VerbClass
• agentPat : VerbClass

@[implicit_reducible]

instance SolstadBott2022.instDecidableEqVerbClass : DecidableEq VerbClass

Equations

• SolstadBott2022.instDecidableEqVerbClass x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

def SolstadBott2022.instReprVerbClass.repr : VerbClass → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance SolstadBott2022.instReprVerbClass : Repr VerbClass

Equations

• SolstadBott2022.instReprVerbClass = { reprPrec := SolstadBott2022.instReprVerbClass.repr }

inductive SolstadBott2022.ICBias : Type

Implicit causality bias direction.

• np1 : ICBias
• np2 : ICBias

@[implicit_reducible]

instance SolstadBott2022.instDecidableEqICBias : DecidableEq ICBias

Equations

• SolstadBott2022.instDecidableEqICBias x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

@[implicit_reducible]

instance SolstadBott2022.instReprICBias : Repr ICBias

Equations

• SolstadBott2022.instReprICBias = { reprPrec := SolstadBott2022.instReprICBias.repr }

def SolstadBott2022.instReprICBias.repr : ICBias → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.VerbClass.predictedBias : VerbClass → ICBias

Predicted IC bias direction for each verb class.

The IC bias tracks the STIMULUS argument, not the subject per se:

• StimExp (stimulus = subject) → NP1 (explanation about subject)
• ExpStim (stimulus = object) → NP2 (explanation about object)
• AgentEvocator (evocator = object) → NP2
• AgPat (agent = subject) → NP1 (default)

Equations

• SolstadBott2022.VerbClass.stimExp.predictedBias = SolstadBott2022.ICBias.np1
• SolstadBott2022.VerbClass.expStim.predictedBias = SolstadBott2022.ICBias.np2
• SolstadBott2022.VerbClass.agentEvocator.predictedBias = SolstadBott2022.ICBias.np2
• SolstadBott2022.VerbClass.agentPat.predictedBias = SolstadBott2022.ICBias.np1

inductive SolstadBott2022.ExpConnective : Type

Connective conditions in @cite{solstad-bott-2022}. German connectives weil (because) and sodass (and so).

• weil : ExpConnective
• sodass : ExpConnective

@[implicit_reducible]

instance SolstadBott2022.instDecidableEqExpConnective : DecidableEq ExpConnective

Equations

• SolstadBott2022.instDecidableEqExpConnective x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

@[implicit_reducible]

instance SolstadBott2022.instReprExpConnective : Repr ExpConnective

Equations

• SolstadBott2022.instReprExpConnective = { reprPrec := SolstadBott2022.instReprExpConnective.repr }

def SolstadBott2022.instReprExpConnective.repr : ExpConnective → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

structure SolstadBott2022.CorefDatum : Type

Subject coreference proportion from Exp 1, Table 1 of @cite{solstad-bott-2022}. These are real data from 52 German participants with 20 STIM-EXP and 20 EXP-STIM verbs (gefallen excluded).

• verbClass : VerbClass
• connective : ExpConnective
• subjectCorefPct : ℚ

@[implicit_reducible]

instance SolstadBott2022.instReprCorefDatum : Repr CorefDatum

Equations

• SolstadBott2022.instReprCorefDatum = { reprPrec := SolstadBott2022.instReprCorefDatum.repr }

def SolstadBott2022.instReprCorefDatum.repr : CorefDatum → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.exp1_stimExp_weil : CorefDatum

Equations

• SolstadBott2022.exp1_stimExp_weil = { verbClass := SolstadBott2022.VerbClass.stimExp, connective := SolstadBott2022.ExpConnective.weil, subjectCorefPct := 874 / 10 }

def SolstadBott2022.exp1_expStim_weil : CorefDatum

Equations

• SolstadBott2022.exp1_expStim_weil = { verbClass := SolstadBott2022.VerbClass.expStim, connective := SolstadBott2022.ExpConnective.weil, subjectCorefPct := 40 / 10 }

def SolstadBott2022.exp1_stimExp_sodass : CorefDatum

Equations

• SolstadBott2022.exp1_stimExp_sodass = { verbClass := SolstadBott2022.VerbClass.stimExp, connective := SolstadBott2022.ExpConnective.sodass, subjectCorefPct := 48 / 10 }

def SolstadBott2022.exp1_expStim_sodass : CorefDatum

Equations

• SolstadBott2022.exp1_expStim_sodass = { verbClass := SolstadBott2022.VerbClass.expStim, connective := SolstadBott2022.ExpConnective.sodass, subjectCorefPct := 779 / 10 }

theorem SolstadBott2022.stimExp_opposes_expStim : VerbClass.stimExp.predictedBias ≠ VerbClass.expStim.predictedBias

StimExp and ExpStim have opposite predicted IC bias.

theorem SolstadBott2022.stimExp_weil_np1_bias : exp1_stimExp_weil.subjectCorefPct > 50

I-Caus (weil): StimExp has strong NP1 bias (87.4% > 50%).

theorem SolstadBott2022.expStim_weil_np2_bias : exp1_expStim_weil.subjectCorefPct < 50

I-Caus (weil): ExpStim has strong NP2 bias (4.0% < 50%).

theorem SolstadBott2022.icons_mirrors_icaus : exp1_stimExp_sodass.subjectCorefPct < 50 ∧ exp1_expStim_sodass.subjectCorefPct > 50

I-Cons (sodass): Biases mirror I-Caus — StimExp → NP2, ExpStim → NP1. (@cite{solstad-bott-2022}, §2.3: "almost perfect negative correlation" r = −0.94)

inductive SolstadBott2022.ContinuationType : Type

Coherence relation types produced in continuations.

• explanation : ContinuationType
• consequence : ContinuationType

@[implicit_reducible]

instance SolstadBott2022.instDecidableEqContinuationType : DecidableEq ContinuationType

Equations

• SolstadBott2022.instDecidableEqContinuationType x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

def SolstadBott2022.instReprContinuationType.repr : ContinuationType → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance SolstadBott2022.instReprContinuationType : Repr ContinuationType

Equations

• SolstadBott2022.instReprContinuationType = { reprPrec := SolstadBott2022.instReprContinuationType.repr }

inductive SolstadBott2022.Congruence : Type

Congruence: whether the forced coreference target matches the verb's predicted IC bias direction.

• congruent : Congruence
• incongruent : Congruence

@[implicit_reducible]

instance SolstadBott2022.instDecidableEqCongruence : DecidableEq Congruence

Equations

• SolstadBott2022.instDecidableEqCongruence x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

@[implicit_reducible]

instance SolstadBott2022.instReprCongruence : Repr Congruence

Equations

• SolstadBott2022.instReprCongruence = { reprPrec := SolstadBott2022.instReprCongruence.repr }

def SolstadBott2022.instReprCongruence.repr : Congruence → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

structure SolstadBott2022.ForcedCorefDatum : Type

Exp 3 data: proportion of explanations (vs consequences) under forced coreference. @cite{solstad-bott-2022} Table 3.

The key finding: even when forced to refer to the non-biased argument (incongruent condition), participants STILL produce explanations rather than consequences. This supports the Asymmetry Hypothesis — I-Caus (explanation-seeking) is the default mechanism driven by verb semantics, while I-Cons only emerges when discourse structure demands it.

• verbClass : VerbClass
• congruence : Congruence
• explanationPct : ℚ

@[implicit_reducible]

instance SolstadBott2022.instReprForcedCorefDatum : Repr ForcedCorefDatum

Equations

• SolstadBott2022.instReprForcedCorefDatum = { reprPrec := SolstadBott2022.instReprForcedCorefDatum.repr }

def SolstadBott2022.instReprForcedCorefDatum.repr : ForcedCorefDatum → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.exp3_stimExp_congruent : ForcedCorefDatum

Equations

• SolstadBott2022.exp3_stimExp_congruent = { verbClass := SolstadBott2022.VerbClass.stimExp, congruence := SolstadBott2022.Congruence.congruent, explanationPct := 879 / 10 }

def SolstadBott2022.exp3_stimExp_incongruent : ForcedCorefDatum

Equations

• SolstadBott2022.exp3_stimExp_incongruent = { verbClass := SolstadBott2022.VerbClass.stimExp, congruence := SolstadBott2022.Congruence.incongruent, explanationPct := 876 / 10 }

def SolstadBott2022.exp3_expStim_congruent : ForcedCorefDatum

Equations

• SolstadBott2022.exp3_expStim_congruent = { verbClass := SolstadBott2022.VerbClass.expStim, congruence := SolstadBott2022.Congruence.congruent, explanationPct := 837 / 10 }

def SolstadBott2022.exp3_expStim_incongruent : ForcedCorefDatum

Equations

• SolstadBott2022.exp3_expStim_incongruent = { verbClass := SolstadBott2022.VerbClass.expStim, congruence := SolstadBott2022.Congruence.incongruent, explanationPct := 870 / 10 }

theorem SolstadBott2022.asymmetry_hypothesis_exp3 : exp3_stimExp_incongruent.explanationPct > 80 ∧ exp3_expStim_incongruent.explanationPct > 80

Asymmetry Hypothesis: explanations dominate regardless of congruence. Even in bias-incongruent conditions, explanation rate stays above 80%. This shows I-Caus (explanation) is the default mechanism.

theorem SolstadBott2022.congruence_does_not_reduce_explanations : exp3_stimExp_congruent.explanationPct - exp3_stimExp_incongruent.explanationPct < 1 ∧ exp3_expStim_incongruent.explanationPct ≥ exp3_expStim_congruent.explanationPct

Congruence does NOT significantly affect explanation rate — incongruent and congruent conditions produce similar proportions. This is the core prediction of the Asymmetry Hypothesis: if I-Caus were simply the mirror of I-Cons, incongruent coreference should force consequence relations, but it doesn't.

inductive SolstadBott2022.ExplanationSubtype : Type

Subtypes of explanation continuations from Exp 4 annotation. @cite{solstad-bott-2022} distinguishes three explanation categories:

• specifying: fills the verb's causal slot (Empty Slot Theory prediction)
• mentalBackground: provides the experiencer's mental state
• nonmentalBackground: provides non-mental context

• specifying : ExplanationSubtype
• mentalBackground : ExplanationSubtype
• nonmentalBackground : ExplanationSubtype

@[implicit_reducible]

instance SolstadBott2022.instDecidableEqExplanationSubtype : DecidableEq ExplanationSubtype

Equations

• SolstadBott2022.instDecidableEqExplanationSubtype x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

@[implicit_reducible]

instance SolstadBott2022.instReprExplanationSubtype : Repr ExplanationSubtype

Equations

• SolstadBott2022.instReprExplanationSubtype = { reprPrec := SolstadBott2022.instReprExplanationSubtype.repr }

def SolstadBott2022.instReprExplanationSubtype.repr : ExplanationSubtype → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

structure SolstadBott2022.ExplanationTypeDatum : Type

Exp 4 explanation subtype frequencies. Table 4 of @cite{solstad-bott-2022}.

• verbClass : VerbClass
• congruence : Congruence
• specifyingPct : ℚ
• mentalBgPct : ℚ
• nonmentalBgPct : ℚ

@[implicit_reducible]

instance SolstadBott2022.instReprExplanationTypeDatum : Repr ExplanationTypeDatum

Equations

• SolstadBott2022.instReprExplanationTypeDatum = { reprPrec := SolstadBott2022.instReprExplanationTypeDatum.repr }

def SolstadBott2022.instReprExplanationTypeDatum.repr : ExplanationTypeDatum → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.exp4_stimExp_congruent_expl : ExplanationTypeDatum

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.exp4_stimExp_incongruent_expl : ExplanationTypeDatum

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.exp4_expStim_congruent_expl : ExplanationTypeDatum

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.exp4_expStim_incongruent_expl : ExplanationTypeDatum

Equations

• One or more equations did not get rendered due to their size.

theorem SolstadBott2022.empty_slot_congruent_specifications : exp4_stimExp_congruent_expl.specifyingPct > 80 ∧ exp4_expStim_congruent_expl.specifyingPct > 80

Empty Slot Theory prediction: explanatory specifications dominate in congruent conditions (where the continuation fills the verb's causal slot).

theorem SolstadBott2022.empty_slot_incongruent_no_specifications : exp4_stimExp_incongruent_expl.specifyingPct < 5 ∧ exp4_expStim_incongruent_expl.specifyingPct < 10

Empty Slot Theory prediction: explanatory specifications nearly vanish in incongruent conditions (the "wrong" argument cannot fill the slot).

def SolstadBott2022.consequenceSpecificationRate (vc : VerbClass) : ℚ

Two-Mechanism Account: consequence specifications are never produced (0.4% STIM-EXP, 0.0% EXP-STIM in Table 4). The specification strategy is not available for consequences — only for explanations. This is because I-Cons derives from the Contiguity Principle (discourse-structural), not from an underspecified slot in verb meaning (verb-semantic).

We encode this as: consequence-specification is not a viable strategy.

Equations

• SolstadBott2022.consequenceSpecificationRate SolstadBott2022.VerbClass.stimExp = 4 / 10
• SolstadBott2022.consequenceSpecificationRate SolstadBott2022.VerbClass.expStim = 0
• SolstadBott2022.consequenceSpecificationRate vc = 0

theorem SolstadBott2022.consequence_specification_negligible : consequenceSpecificationRate VerbClass.stimExp < 1 ∧ consequenceSpecificationRate VerbClass.expStim < 1

def SolstadBott2022.stimExpSubjectProfile : Semantics.ArgumentStructure.EntailmentProfile.EntailmentProfile

Stimulus-experiencer verb subject profile: causation + independent existence. The subject is a stimulus/cause (B&R Class II, Levin 31.1 amuse class). @cite{solstad-bott-2022}: STIM-EXP verbs show NP1 I-Caus bias.

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.stimExpObjectProfile : Semantics.ArgumentStructure.EntailmentProfile.EntailmentProfile

Stimulus-experiencer verb object profile: sentience + independent existence. The object is an experiencer.

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.expStimSubjectProfile : Semantics.ArgumentStructure.EntailmentProfile.EntailmentProfile

Experiencer-stimulus verb subject profile: sentience + independent existence. The subject is an experiencer (B&R Class I, temere class). @cite{solstad-bott-2022}: EXP-STIM verbs show NP2 I-Caus bias.

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.expStimObjectProfile : Semantics.ArgumentStructure.EntailmentProfile.EntailmentProfile

Experiencer-stimulus verb object profile: causation + independent existence. The object is a stimulus (cause of the experience).

Equations

• One or more equations did not get rendered due to their size.

def SolstadBott2022.agPatSubjectProfile : Semantics.ArgumentStructure.EntailmentProfile.EntailmentProfile

Agent-patient verb subject profile: full agent (all 5 P-Agent). Identical to existing kickSubjectProfile.

Equations

• One or more equations did not get rendered due to their size.

theorem SolstadBott2022.stimExp_subject_eq_expStim_object : stimExpSubjectProfile = expStimObjectProfile

StimExp subject profile = ExpStim object profile (both are stimulus/C+IE).

theorem SolstadBott2022.expStim_subject_eq_stimExp_object : expStimSubjectProfile = stimExpObjectProfile

ExpStim subject profile = StimExp object profile (both are experiencer/S+IE).

theorem SolstadBott2022.theta_reversal_at_proto_level : stimExpSubjectProfile = expStimObjectProfile ∧ stimExpObjectProfile = expStimSubjectProfile

This is the B&R theta-role reversal expressed at the proto-role level: Class I and Class II swap the same two profiles between subject and object.

theorem SolstadBott2022.stimExp_subject_passes_doTest : Semantics.ArgumentStructure.EntailmentProfile.PassesDoTestFromProfile stimExpSubjectProfile

StimExp subjects pass the do-test (they have causation). "What the noise did was frighten John" is grammatical because the subject has the causation entailment (Dowty's P-Agent (c)).

theorem SolstadBott2022.expStim_subject_fails_doTest : ¬Semantics.ArgumentStructure.EntailmentProfile.PassesDoTestFromProfile expStimSubjectProfile

ExpStim subjects fail the do-test (experiencers lack volition, causation, and movement). "??What Mary did was admire John" is marginal.

theorem SolstadBott2022.agPat_subject_passes_doTest : Semantics.ArgumentStructure.EntailmentProfile.PassesDoTestFromProfile agPatSubjectProfile

AgPat subjects pass the do-test (full agents).

theorem SolstadBott2022.stimExp_profile_matches_stimulus : stimExpSubjectProfile = ThetaRole.stimulus.canonicalProfile

StimExp subject profile matches ThetaRole.stimulus's canonical profile.

theorem SolstadBott2022.expStim_profile_matches_experiencer : expStimSubjectProfile = ThetaRole.experiencer.canonicalProfile

ExpStim subject profile matches ThetaRole.experiencer's canonical profile.

theorem SolstadBott2022.agPat_profile_matches_agent : agPatSubjectProfile = ThetaRole.agent.canonicalProfile

AgPat subject profile matches ThetaRole.agent's canonical profile.

theorem SolstadBott2022.because_triggers_causal_search : Core.Discourse.Coherence.CoherenceRelation.explanation.selectsCause = true

The Explanation relation (triggered by "because") selects for causes.

def SolstadBott2022.predictICBias (subjProfile : Semantics.ArgumentStructure.EntailmentProfile.EntailmentProfile) : ICBias

IC bias prediction: under Explanation (because), the continuation targets the STIMULUS argument — the participant whose entailment profile includes causation.

• StimExp: subject has causation → explanation about subject → NP1
• ExpStim: subject has sentience only → explanation about object → NP2
• AgPat: subject has causation (+ volition, etc.) → NP1

Equations

• One or more equations did not get rendered due to their size.

theorem SolstadBott2022.stimExp_predicted_np1 : predictICBias stimExpSubjectProfile = ICBias.np1

StimExp predicted as NP1 (stimulus subject has causation).

theorem SolstadBott2022.expStim_predicted_np2 : predictICBias expStimSubjectProfile = ICBias.np2

ExpStim predicted as NP2 (experiencer subject lacks causation).

theorem SolstadBott2022.agPat_predicted_np1 : predictICBias agPatSubjectProfile = ICBias.np1

AgPat predicted as NP1 (agent subject — default).

theorem SolstadBott2022.predictions_match_data : predictICBias stimExpSubjectProfile = VerbClass.stimExp.predictedBias ∧ predictICBias expStimSubjectProfile = VerbClass.expStim.predictedBias ∧ predictICBias agPatSubjectProfile = VerbClass.agentPat.predictedBias

The prediction matches the empirical data for all tested classes.

theorem SolstadBott2022.thematic_role_not_position_determines_bias : predictICBias stimExpSubjectProfile = ICBias.np1 ∧ predictICBias expStimSubjectProfile = ICBias.np2 ∧ Features.AccessibilityLevel.rank (RosaArnold2017.transferNextMention ThetaRole.goal).predictedForm > Features.AccessibilityLevel.rank (RosaArnold2017.transferNextMention ThetaRole.source).predictedForm ∧ ¬RosaArnold2017.kehlerRohdeIndependence (fun (role : RosaArnold2017.TransferRole) (gram : RosaArnold2017.GramRole) => match role, gram with | RosaArnold2017.TransferRole.goal, RosaArnold2017.GramRole.subject => 64 | RosaArnold2017.TransferRole.source, RosaArnold2017.GramRole.subject => 37 | RosaArnold2017.TransferRole.goal, RosaArnold2017.GramRole.nonsubject => 31 | RosaArnold2017.TransferRole.source, RosaArnold2017.GramRole.nonsubject => 18) RosaArnold2017.GramRole.subject

The IC reversal (StimExp→NP1, ExpStim→NP2) and the transfer verb goal bias are both instances of the same deeper pattern: swapping which argument carries a discourse-prominent thematic role reverses the discourse bias direction.

For IC: swapping stimulus between subject (StimExp) and object (ExpStim) reverses the IC bias from NP1 to NP2. For transfer: swapping goal between subject and nonsubject doesn't eliminate the goal bias — goals still get more pronouns in BOTH positions.

The IC reversal is the stronger demonstration: it shows the bias direction is ENTIRELY determined by the thematic role, not the grammatical position. @cite{rosa-arnold-2017}'s data corroborates this by showing that thematic role affects form even when grammatical role is held constant, violating @cite{kehler-rohde-2013}'s independence hypothesis.

theorem SolstadBott2022.coherence_selects_complementary_roles : Core.Discourse.Coherence.CoherenceRelation.explanation.selectsCause = true ∧ RosaArnold2017.occasionResult_interaction.significant = true ∧ RosaArnold2017.other_interaction.significant = false ∧ Core.Discourse.Coherence.CoherenceRelation.explanation.toClass ≠ Core.Discourse.Coherence.CoherenceRelation.occasion.toClass

Coherence relations select for COMPLEMENTARY thematic roles in the two phenomena, demonstrating that the coherence-role interaction is systematic rather than accidental:

Explanation (because) → selects CAUSE → stimulus in psych verbs Occasion/Result → selects ENDPOINT → goal in transfer verbs

This complementarity is predicted by the semantics of the coherence relations: Explanation asks "why did this happen?" (→ cause), while Occasion/Result asks "what happened next?" (→ endpoint).

theorem SolstadBott2022.ic_instantiates_KR_explanation_bias : KehlerRohde2013.cr_explanation.cr.selectsCause = true ∧ KehlerRohde2013.cr_explanation.sourceGivenCR > 50 ∧ predictICBias stimExpSubjectProfile = ICBias.np1 ∧ predictICBias expStimSubjectProfile = ICBias.np2

@cite{kehler-rohde-2013}'s Table 2 establishes that Explanation coherence relations are Source-biased (80% Source for transfer verbs). This study's IC data instantiates the same mechanism for psych verbs: Explanation (triggered by "because") selects for causes, and IC bias tracks whichever argument carries the causation entailment — the stimulus.

theorem SolstadBott2022.because_collapses_KR_mixture : Core.Discourse.Coherence.Connective.because.toRelation = Core.Discourse.Coherence.CoherenceRelation.explanation ∧ Core.Discourse.Coherence.CoherenceRelation.explanation.selectsCause = true ∧ predictICBias stimExpSubjectProfile = ICBias.np1 ∧ predictICBias expStimSubjectProfile = ICBias.np2

@cite{kehler-rohde-2013}'s key structural claim is that coherence relations and referential form contribute to DIFFERENT terms in Bayes' rule: P(referent) vs P(pronoun|referent). The IC data provides the strongest evidence for the P(referent) side:

• K&R: P(referent) = Σ_CR P(CR) × P(referent | CR)
• IC context: "because" sets P(Explanation) ≈ 1
• Therefore: P(referent) ≈ P(referent | Explanation)
• P(referent | Explanation) = whichever argument has causation