Bhadra 2024: Verb roots encode outcomes #
The reversative prefix un- and the restitutive prefix re- are result-state
modifiers sensitive to the outcome set a verb root lexically encodes
([Bha24], §5). Every verb root carries a set of outcomes O — the
states its object can be in after the action — and a contextual set of
thresholds T of states it can start from.
The two affixes split along two independent axes:
- un- is driven by outcome-set cardinality (eq. 66): it needs
O.Nontrivial(|O| > 1, the potential-for-change / PFC profile) plus inverse equivalence between the base result and the un-event's boundaries. Singleton or empty outcome sets block it. - re- is not cardinality-sensitive (eq. 72); it needs result equivalence with a prior event (eq. 68 i) and that the object can re-enter a threshold state (eq. 68 ii). Consumption/destruction and impingement (IE) verbs block it ((25), (48)) because their outcome leaves the object outside every threshold.
One idealization: each VerbOutcomes fixes its outcome and threshold sets once
per root, freezing the [verb + object] unit that Bhadra's Level-2 story varies
with the object (✓rebreak a limb vs. #rebreak a sewer, her (73)).
Main definitions #
VerbOutcomes— a root's base predicate (EventRel), outcome set, threshold setresState/preState— state at the right / left event boundary (eqs. 64–65)unSem/reSem— the un- and re- denotations (eqs. 66, 68)LevinClass.outcomeCard— the single Levin-class → outcome-cardinality bridge (ℕ∞; eq. 62 hierarchy multi > singleton > empty)
Main results #
subsingleton_blocks_un—¬ O.Nontrivial → ¬ unSem(eq. 67), with the singleton and empty corollaries*_un/*_reper-verb predictions, derived compositionally and bridged from Fragment entries viaVerb.levinClass
References #
The compositional verb root (eqs. 53–60) #
A verb root lexically encodes its base predicate P (the ⟨v,⟨e,t⟩⟩ meaning the
affixes modify), its outcome set O (states at the right boundary, eq. 56a), and
its threshold set T (left boundary, eq. 56b) — the substrate VerbOutcomes
(Semantics/Verb/Root/Outcomes.lean), with resState/preState the eq. 64–65 boundary
operators and StateFunction the paper's state k : t ↦ l(x) (eq. 53).
APPLIES (eq. 59) is folded into verb.
un- and re- as result-state modifiers (eqs. 66, 68) #
Reversative un- (eq. 66). unSem k vro e x holds iff there is a prior
base event e' whose result is the un-event's start state
(res(e') = pre(e)), the outcome set is multi-membered (O.Nontrivial,
|O| > 1), and the un-event undoes it (res(e) = pre(e')). The vacuous
∃Q. Q(e)(x) — which sits in the assertion of eq. 66, not its
presupposition — is dropped.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Restitutive re- (eq. 68). reSem k vro e x holds iff there is a prior
base event e' with the same result (res(e) = res(e'), eq. 68 i), the
re-event starts from a valid threshold state (pre(e) ∈ T), and the base
predicate holds of the re-event (P(e)(x)). The threshold clause is a
rational reconstruction — not a transcription — of eq. 68 ii's negative
existential (no threshold state may render the re-action undefined; the
cyclic-re- condition, eq. 72). Unlike unSem, this places no cardinality
demand on O.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Cardinality blocks un- (eq. 67) #
The core asymmetry: a verb whose outcome set is not multi-membered cannot
host un- ([Bha24], eq. 67). The multi-membered presupposition of
unSem fails.
Singleton outcome sets (IE and COS verbs) block un-.
Empty outcome sets (no-change-specified verbs) block un-.
un-'s cardinality demand (eq. 67, |O| > 1) is exactly the substrate
OutcomeCardinality.multi tier (eq. 62, empty < singleton < multi): hosting
un- forces the verb root's outcome set into the top tier. Wires Bhadra's
outcome cardinality to the OutcomeCardinality substrate.
The Levin-class → outcome-cardinality bridge (eq. 62) #
The single per-class datum: the cardinality of a verb root's lexical outcome
set, valued in ℕ∞. PFC roots have open-ended outcome sets (eq. 60:
O = {k | k = APPLIES(e)(x)}, unbounded), idealized as ⊤; IE and COS roots
have a single lexically-specified result (eqs. 61a–g); no-change roots have
none (eq. 61h). Only 1 < · is load-bearing (it drives un-); the three tiers
realize the eq. 62 hierarchy multi > singleton > empty.
Outcome-set cardinality of a Levin verb class ([Bha24], eq. 62).
Equations
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.coil = ⊤
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.bend = ⊤
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.conceal = ⊤
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.dress = ⊤
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.hit = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.swat = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.wipe = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.touch = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.break_ = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.destroy = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.cooking = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.otherCoS = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.entitySpecificCoS = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.calibratableCoS = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.color = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.imageCreation = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.build = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.create = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.grow = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.knead = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.turn = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.cut = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.carve = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.eat = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.devour = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.murder = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.poison = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.mix = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.amalgamate = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.separate = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.split = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.clear = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.sprayLoad = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.pushPull = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.remove = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.throw = 1
- Bhadra2024.LevinClass.outcomeCard Semantics.Lexical.LevinClass.put = 1
- Bhadra2024.LevinClass.outcomeCard x✝ = 0
Instances For
The eq. 62 hierarchy, witnessed on a representative of each tier: multi-membered (PFC) > singleton (IE, COS) > empty (no change).
un- prediction is derived from cardinality: PFC classes have
1 < outcomeCard, every other class does not.
Worked examples (eqs. 60–61, 66–71) #
fold: a PFC root (multi-membered outcomes), un- and re- both attach #
States of a parchment under folding (eq. 54: a multi-membered outcome set).
- flat : ParchmentState
- slightlyCreased : ParchmentState
- folded : ParchmentState
- tightlyFolded : ParchmentState
Instances For
Equations
- Bhadra2024.instDecidableEqParchmentState 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
Equations
- Bhadra2024.instReprParchmentState = { reprPrec := Bhadra2024.instReprParchmentState.repr }
fold (Levin bend class): a multi-membered outcome set.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Positive: un- IS satisfiable for fold (eq. 66 worked example, "Veena unfolded the parchment").
Positive: re- IS satisfiable for fold — the multi-membered (PFC) tier's re- cell, completing the Fig. 5 distribution: re- attaches across the multi and singleton tiers (eq. 72), where un- takes the multi tier only. Reuses the root of her worked unfold derivation (74)–(75).
break: a COS root (singleton outcome), un- blocked, re- allowed #
Equations
- Bhadra2024.instDecidableEqLimbState x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Bhadra2024.instReprLimbState = { reprPrec := Bhadra2024.instReprLimbState.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
break (Levin break_ class): a single lexically-specified result
(eq. 61a).
Equations
- Bhadra2024.breakVRO = { verb := fun (x : Event ℤ) (x_1 : Unit) => True, outcomes := {Bhadra2024.LimbState.broken}, thresholds := {Bhadra2024.LimbState.intact} }
Instances For
unbreak is blocked: the singleton outcome set fails eq. 67.
Positive: re- IS satisfiable for break (rebreak a limb): result
equivalence holds and the re-event starts from the intact threshold.
hit: an IE root (singleton outcome), un- and re- both blocked #
- unaltered : SurfaceState
- surfaceAltered : SurfaceState
Instances For
Equations
- Bhadra2024.instDecidableEqSurfaceState x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Bhadra2024.instReprSurfaceState = { reprPrec := Bhadra2024.instReprSurfaceState.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
hit (Levin hit class): impingement, a single surface-alteration result
(eq. 61g).
Equations
- Bhadra2024.hitVRO = { verb := fun (x : Event ℤ) (x_1 : Unit) => True, outcomes := {Bhadra2024.SurfaceState.surfaceAltered}, thresholds := {Bhadra2024.SurfaceState.unaltered} }
Instances For
Negative: re- is blocked for hit — IE verbs block re- as well as un- ((25), (48)): the impingement outcome never re-enters the pre-state (threshold) set, so the cyclic-re- condition (eq. 68 ii) fails.
load vs shatter: the re- minimal pair (eqs. 69–71) #
Both are singleton-outcome (so both block un-); they differ on re- purely through the threshold condition (eq. 68 ii). After loading, the truck can be loaded again (it re-enters a threshold); after shattering, the mirror cannot.
- empty : TruckState
- partlyLoaded : TruckState
- full : TruckState
Instances For
Equations
- Bhadra2024.instDecidableEqTruckState x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Bhadra2024.instReprTruckState = { reprPrec := Bhadra2024.instReprTruckState.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
load (degree achievement, eq. 70): a single contextually-salient result; the object stays loadable.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Positive: re- IS satisfiable for load (eq. 69a "reloaded the truck").
- intact : MirrorState
- shattered : MirrorState
Instances For
Equations
- Bhadra2024.instDecidableEqMirrorState x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Bhadra2024.instReprMirrorState = { reprPrec := Bhadra2024.instReprMirrorState.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
shatter (eq. 71): a single result that leaves the object outside every threshold — it cannot be shattered again.
Equations
- Bhadra2024.shatterVRO = { verb := fun (x : Event ℤ) (x_1 : Unit) => True, outcomes := {Bhadra2024.MirrorState.shattered}, thresholds := {Bhadra2024.MirrorState.intact} }
Instances For
Negative: re- is blocked for shatter (eq. 69b "#reshattered the
mirror"): the cyclic-re- threshold condition (eq. 68 ii) fails — the
shattered mirror is never in the intact threshold the re-event needs.
End-to-end: Fragment entry → Levin class → prediction #
The Fragment verb's levinClass drives the outcome cardinality, which drives
the un- prediction.
kick (Levin 18.1 hit) → singleton outcomes → un- blocked.
bend (Levin 45.2) → multi-membered outcomes → un- predicted.
break (Levin 45.1) → singleton outcomes → un- blocked.
Bridges to the substrate #
The outcome-cardinality classification is finer than the change-of-state label
([Bha24]'s central move): bend is CoS per [Lev93] yet has a
multi-membered (PFC) outcome set.
[Bha24] reclassifies bend from a flat change-of-state verb to a
potential-for-change root: CoS per [Lev93]'s meaning components, but
multi-membered (and hence un--compatible) by outcome cardinality.
The outcome-cardinality classification is orthogonal to event-template
result states: bend and coil share the PFC profile but differ on
whether their template carries a result state.
Affectedness bridge ([Bha24] §2.4.2, §6.1): the affectedness
projection cannot separate PFC roots from IE roots — the PFC object
profile (causally affected, no entailed change) and kick's object
profile both land on Beavers's potential-for-change degree, whose
exemplars include surface-contact hit/kick. Bhadra carves the IE
class (kick is named in her (25)) out of exactly that cell (§2.4.2),
and her §6.1 point is that the split needs outcome-set structure a bare
latent-scale existential cannot see: outcome cardinality separates the
two (⊤ vs 1) where the affectedness degree does not.
RootTypology bridge: result roots entail change and lack the restitutive again reading; property-concept roots are the reverse.