Kamp (1975): Two Theories about Adjectives #
Theory 1 classifies adjective meanings (functions from properties to
properties) by meaning postulates — the classification lives in
Semantics/Modification/Classification.lean. Theory 2, building on
[VF69]'s supervaluations, derives the comparative from
quantification over completions of a partial model; we formalize its
comparative core, leaving the model apparatus abstract.
Main results #
kleene_dilemma: no truth-functional conjunction is both idempotent at the borderline value and false on borderline contradictions.kampPreorder: the completion comparative, definition (12).kampMeasureLe: the measure comparative, definition (13);kampMeasureLe_totalis Kamp's objection to it andclever_incomparablehis Smith/Jones witness.kampPreorder_le_iff_kampMeasureLe: for one-dimensional adjectives, (12) and (13) coincide.grayAdj,fakeAdj,skillfulAdj,allegedAdj: witnesses for the four classes, shared byPartee2010.leanandDelPinal2015.lean.
Bridge to single-world predicates #
The classification (Modifier.isIntersective, .isSubsective, …) is
one order-theoretic definition instantiated at two carriers: the
intensional Property W E = W → E → Prop and the single-world
E → Prop. The bridge theorems below show that fixing a world sends
the first instance to the second.
Single-world specialization: given a fixed world, the intensional
instance of Modifier.isIntersective reduces to the E → Prop
instance on the rigidified single-world view N ↦ adj (fun _ => N) w.
Single-world specialization of Modifier.isSubsective.
The many-valued dilemma #
No truth-functional conjunction is both idempotent at the borderline
value and false on borderline contradictions: with neg indet = indet,
both demands constrain the same input pair. This is the dilemma of
[Kam75], pp. 130–131, stated there for every linearly ordered
n-valued logic; Trivalent is the minimal witness.
Kamp's completion comparative, definition (12) #
Definition (12) (paper § 4): u₁ is at least as A as u₂ iff every
admissible completion that puts u₂ in the extension also puts u₁ in it.
[Kle80] § 5.3 states the strict comparative existentially over
comparison classes; the bridge is
Klein1980.kleinPreorder_eq_kampPreorder.
Kamp's completion comparative (definition (12), paper § 4) as a
Preorder: le u₁ u₂ iff every completion in S that puts u₂ in
the extension also puts u₁ in — le reads "u₁ is at least as A as
u₂", Kamp's ≥. The S-restricted analogue of kleinPreorder in
Delineation.lean. Kamp credits (12) to Lewis (1970), where it is
attributed to Kaplan.
Equations
- Kamp1975.kampPreorder ext S = { le := fun (u₁ u₂ : E) => ∀ c ∈ S, ext c u₂ → ext c u₁, le_refl := ⋯, le_trans := ⋯, lt_iff_le_not_ge := ⋯ }
Instances For
(12) vs (13): definite vs measured comparatives #
Kamp's second candidate, definition (13) (paper § 4), compares the measures of the completion sets rather than the sets themselves. His § 5 argues against (13) for multi-criteria adjectives: it makes any two entities comparable, while (12) leaves Smith and Jones incomparable in cleverness — for Kamp the right verdict. For one-dimensional adjectives (heavy, tall, hot) the two provably coincide.
[Kam75] definition (13) (paper § 4): the measure-based
comparative — u₁ ≤ u₂ iff the measure of completions putting u₂
in the extension is at most that putting u₁ in (kampPreorder's
orientation). Kamp's probability measure over a field of subsets is
specialized to atomic ℚ weights over a finite completion set; only
the ordering matters, so weights need not sum to 1.
Equations
- Kamp1975.kampMeasureLe ext S p u₁ u₂ = (∑ c ∈ S with ext c u₂, p c ≤ ∑ c ∈ S with ext c u₁, p c)
Instances For
(13) is total: it makes any two objects comparable. This is Kamp's
§ 5 objection to (13); (12) does not share the property
(clever_incomparable).
Definite comparison entails measured comparison: (12) implies (13) for nonnegative weights.
The Smith/Jones incomparability witness (§ 5) #
Two criteria for clever — problem-solving and quick-wittedness — as
two completions; Smith passes one, Jones the other. Under (12) the two
are incomparable, which Kamp argues is correct; (13) must issue a
verdict (kampMeasureLe_total). Kamp's own scenario is asymmetric
(Smith much better at problems, only slightly worse in wit, so (13)
wrongly makes Smith cleverer); this symmetric toy witnesses the
incomparability and the forced verdict, not that specific outcome.
Equations
- Kamp1975.instDecidableEqCrit x✝ y✝ = if h : Kamp1975.Crit.ctorIdx✝ x✝ = Kamp1975.Crit.ctorIdx✝ y✝ then isTrue ⋯ else isFalse ⋯
Equations
- Kamp1975.instDecidableEqP2 x✝ y✝ = if h : Kamp1975.P2.ctorIdx✝ x✝ = Kamp1975.P2.ctorIdx✝ y✝ then isTrue ⋯ else isFalse ⋯
Under (12), Smith and Jones are incomparable in cleverness — Kamp's argument that (12) "captures the comparative correctly" for multi-criteria adjectives, against (13)'s forced totality.
One-dimensionality #
One-dimensional adjectives ([Kam75] § 5: heavy, tall, hot):
any two entities' completion-sets are ⊆-comparable, so the
extensions form a chain (threshold structure). The formal condition
is Kamp's (18), stated in § 6 where it grounds the adjective/noun
asymmetry.
Equations
- Kamp1975.OneDimensional ext S = ∀ (u₁ u₂ : E), (∀ c ∈ S, ext c u₁ → ext c u₂) ∨ ∀ c ∈ S, ext c u₂ → ext c u₁
Instances For
For one-dimensional adjectives with strictly positive weights, the measured comparative (13) collapses to the definite comparative (12) — Kamp's § 5 observation that "for this special case the two definitions are equivalent", with strict positivity rendering his "provided p has been correctly specified".
Concrete Witnesses for Each Class #
Each class in the hierarchy is non-empty: explicit denotations that
provably satisfy each definition from Classification.lean, modeling
the classic examples from the literature — "gray" (intersective), "fake"
(privative), "skillful" (subsective but not extensional), "alleged"
(non-subsective/modal).
[Par10] argues that the privative class should be eliminated
in favor of subsective + noun coercion. The witness fakeAdj below
models the traditional analysis; see Partee2010.lean for the
coercion reanalysis.
Two worlds suffice to distinguish extensional from non-extensional.
Instances For
"gray": an intersective adjective ([Kam75] definition (4), "predicative") — a fixed property conjoined with the noun, so "gray cat" entails both "gray" and "cat".
Equations
- Kamp1975.grayAdj N w Kamp1975.E3.a = (True ∧ N w Kamp1975.E3.a)
- Kamp1975.grayAdj N w x = (False ∧ N w x)
Instances For
"fake": a privative adjective ([Kam75] definition (5); fake and
false are his examples) — "fake gun" entails "not a gun". Kamp
doubts any English adjective is privative "in all of its
possible uses", anticipating [Par10]'s subsective-plus-coercion
reanalysis; see Partee2010.lean.
Equations
- Kamp1975.fakeAdj N w Kamp1975.E3.b = (True ∧ ¬N w Kamp1975.E3.b)
- Kamp1975.fakeAdj N w x = (False ∧ ¬N w x)
Instances For
"skillful": subsective ([Kam75] definition (6), "affirmative") but not extensional — "skillful surgeon" entails "surgeon", yet skill depends on the noun's intension, not just its current extension (Kamp's example, crediting the cobblers/darts-players case to David Lewis).
Equations
- Kamp1975.skillfulAdj N w Kamp1975.E3.a = (N w Kamp1975.E3.a ∧ N Kamp1975.W2.w₁ Kamp1975.E3.a)
- Kamp1975.skillfulAdj N w x = (N w x ∧ False)
Instances For
"alleged": a non-subsective (modal) adjective — [Kam75]'s opening example (1), "Every alleged thief is a thief" is no logical truth. No meaning postulate relates the modified and unmodified extensions (likewise "potential", "putative").
Equations
- Kamp1975.allegedAdj _N x✝ Kamp1975.E3.a = True
- Kamp1975.allegedAdj _N x✝ x = False
Instances For
"alleged" ignores the noun entirely, so it is trivially extensional —
with skillful_not_extensional and skillful_subsective, this
witnesses that extensionality is orthogonal to subsectivity.
"alleged N" does not entail "N" (not subsective).
"alleged N" does not entail "not N" (not privative).