Huijsmans (2025) @cite{huijsmans-2025} #
Timing of evidence and epistemic modal claims. Natural Language Semantics 33: 207–254.
The future morphemes səm (ʔayʔaǰuθəm; Central Salish) and will (English) pattern with strong epistemic modals like ćε (ʔayʔaǰuθəm inferential) and must, but their distributions diverge in inferences about non-future eventualities. Huijsmans argues the difference is temporal, not modal strength: səm/will presuppose the modal base is established prior to the earliest moment the prejacent is claimed to hold, while ćε requires at least one modal-base proposition to hold at or following that moment.
Structural claim #
The temporal-presupposition partition is structurally identical to the
EP-condition partition in Theories/Semantics/Tense/Evidential.lean,
but indexed on MBT (modal-base time) rather than EAT (evidence
acquisition time). EPCondition is reused under the slot reinterpretation
MBT ↦ acquisitionTime, PrejT ↦ eventTime. The bridge is
MBTProfile.licenses_iff_EP.
Empirical wedge #
Most of Huijsmans's stimuli (cooking, Daniel-home, poison) satisfy
MBT < PrejT ∧ EAT < ET — both analyses agree. The Felipe-rain stimulus
(ex. 46-49) and clam-digging stimulus (ex. 48-49) have MBT < PrejT but
ET < EAT — Huijsmans's MBT analysis correctly licenses will/must; an
EAT analysis (with the same partition reinterpreted) does not.
MBT_vs_EAT_diverge_on_felipe is the load-bearing theorem.
At-issue content #
The at-issue content is Kratzer strong necessity over the modal base,
shared across all four lexical entries (modulo modal-base type). It is
not formalized in this file; consumers wanting full denotations should
combine Modal.flavor with Theories.Semantics.Modality.Kratzer.necessity.
A stimulus context, supplying the four temporal anchors needed to state both Huijsmans's MBT predictions and the EAT predictions she is contrasted against. The MBT/EAT split lives in which pair of anchors a profile reads.
- earliestMBT : Time
⌜MBT⌝ — the maximum over
q ∈ MBofEARLIEST(SHIFT q), i.e. the latest "earliest moment" required by the modal base. - earliestPrejT : Time
⌜PrejT⌝ —
EARLIEST(p)for the prejacentp. - eat : Time
The (Cumming/Hirayama-Matthewson) evidence-acquisition time.
- et : Time
The event time of the prejacent.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
The three MBT-presupposition profiles Huijsmans's four lexical items
instantiate. A domain-flavored selector over the abstract
Core.Time.Relation partition (cf. EPCondition in
Tense/Evidential.lean, which selects from the same partition for
the EAT/ET slot pair). Only three of the five abstract relations
name a Huijsmans modal; the others are unused here.
- strictPrior : MBTProfile
MBT < PrejT. Modal base established strictly before the prejacent's earliest moment. səm, will. Eq. (35). - nonProspective : MBTProfile
PrejT ≤ MBT. At least one MB proposition becomes true at or after the prejacent's earliest moment. ćε. Eq. (36). - unrestricted : MBTProfile
No temporal restriction. must.
Instances For
Equations
- Phenomena.Modality.Studies.Huijsmans2025.instDecidableEqMBTProfile 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
Underlying point-relation: the slot pair is (earliestMBT, earliestPrejT),
and each profile picks one shape from Relation. Mirrors
EPCondition.toRelation in Tense/Evidential.lean.
Equations
- Phenomena.Modality.Studies.Huijsmans2025.MBTProfile.strictPrior.toRelation = Core.Time.Relation.before
- Phenomena.Modality.Studies.Huijsmans2025.MBTProfile.nonProspective.toRelation = Core.Time.Relation.notBefore
- Phenomena.Modality.Studies.Huijsmans2025.MBTProfile.unrestricted.toRelation = Core.Time.Relation.unrestricted
Instances For
The MBT-licensing predicate, derived from the abstract partition.
Equations
- c.licenses s = c.toRelation.eval s.earliestMBT s.earliestPrejT
Instances For
Equations
- Phenomena.Modality.Studies.Huijsmans2025.instDecidableLicensesOfDecidableEq c s = id inferInstance
A modal in the Huijsmans typology: a label, a (possibly underspecified) modal-flavor, and an MBT-presupposition profile. All four entries in this paper are necessity modals; force is omitted.
- label : String
- flavor : Option Core.Modality.ModalFlavor
Modal flavor, or
nonefor underspecified (Huijsmans p. 218). - profile : MBTProfile
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
ʔayʔaǰuθəm future clitic səm. Underspecified for modal base (epistemic and circumstantial uses; Huijsmans §2.3).
Equations
- Phenomena.Modality.Studies.Huijsmans2025.sem = { label := "səm", flavor := none, profile := Phenomena.Modality.Studies.Huijsmans2025.MBTProfile.strictPrior }
Instances For
English future auxiliary will. Underspecified for modal base (epistemic in non-future-eventuality uses, circumstantial in future-eventuality uses; Huijsmans §2.4).
Equations
- Phenomena.Modality.Studies.Huijsmans2025.will = { label := "will", flavor := none, profile := Phenomena.Modality.Studies.Huijsmans2025.MBTProfile.strictPrior }
Instances For
ʔayʔaǰuθəm inferential clitic ćε. Epistemic only.
Equations
- One or more equations did not get rendered due to their size.
Instances For
English necessity modal must. Underspecified for modal base; this paper only treats epistemic uses.
Equations
- Phenomena.Modality.Studies.Huijsmans2025.must = { label := "must", flavor := none, profile := Phenomena.Modality.Studies.Huijsmans2025.MBTProfile.unrestricted }
Instances For
(40) Cooking. The fish-cooking stimulus: MB props (fish in oven, typical cook time) hold prior to PrejT (it is cooked). Both MBT and EAT analyses predict səm/will/must ✓, ćε ✗.
Equations
- Phenomena.Modality.Studies.Huijsmans2025.cooking = { earliestMBT := 0, earliestPrejT := 2, eat := 0, et := 2 }
Instances For
(46) Felipe-rain. The wedge case. MB props (it is raining, Felipe has
no rain jacket) become true earlier than PrejT (Felipe is wet), but
the speaker only acquires the evidence later (sees the rain at 6pm,
after Felipe has already gotten wet at 5pm). So MBT < PrejT but
ET < EAT. Huijsmans correctly licenses will/must; an EAT analysis
incorrectly rejects will.
Equations
- Phenomena.Modality.Studies.Huijsmans2025.felipeRain = { earliestMBT := 0, earliestPrejT := 1, eat := 3, et := 1 }
Instances For
(50) Smell-of-fish. The smell coincides with PrejT: at least one MB proposition (the fish smells cooked) holds at the moment the prejacent holds. ćε/must ✓, səm/will ✗.
Equations
- Phenomena.Modality.Studies.Huijsmans2025.smellOfFish = { earliestMBT := 1, earliestPrejT := 1, eat := 1, et := 1 }
Instances For
The EAT-indexed analogue of MBTProfile.licenses — applies the same
partition shape to (eat, et) instead of (earliestMBT, earliestPrejT).
Defined inline rather than imported from a (not-yet-existent)
Hirayama-Matthewson study file.
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
The empirical wedge (Huijsmans (46)–(49)).
On the Felipe-rain stimulus the MBT and EAT analyses disagree on
will: the MBT-analysis licenses it (matching the judgment), the
EAT-analysis rejects it (mispredicting). This is the entire empirical
content of Huijsmans's argument compressed to a single decide.