Aspect × Temporal Connective Interaction Bridge #
@cite{alstott-aravind-2026} @cite{moens-steedman-1988} @cite{rett-2020}
Connects three layers:
- Fragment field:
TemporalExprEntry.embeddedTelicityEffect : Bool - Theory:
VendlerClass.telicity,INCHOAT,COMPLEToperators - Data:
AspectInteractionjudgments from @cite{moens-steedman-1988}
What This File Proves #
embeddedTelicityEffect = trueholds exactly for connectives whose truth conditions change depending on the Vendler class of the embedded clause (before, after). This is because they reference boundary points (onset/telos) that depend on aspect.embeddedTelicityEffect = falseholds for connectives that use overlap or containment semantics (while, when, until, since), which are insensitive to the internal temporal structure of the embedded clause.The
triggeredCoercionfield is grounded in the theory'sINCHOATandCOMPLEToperators: INCHOAT extracts the onset point (=CoSType.inception), COMPLET extracts the telos (=CoSType.cessation).The
satisfiesDurativeRestrictionpredicate (from Data) coincides withVendlerClass.toProfile.isHomogeneous: exactly states and activities satisfy the durative requirement for until and since main clauses.
The Explanatory Chain #
VendlerClass.telicity TemporalExprEntry.embeddedTelicityEffect
.telic /.atelic ──────────► true (before, after: reading depends on telicity)
false (while, when, until, since: insensitive)
VendlerClass.duration *until*/*since* selectional restriction
.durative ─────────────────► satisfiesDurativeRestriction = true (OK)
.punctual ─────────────────► satisfiesDurativeRestriction = false (BAD)
INCHOAT / COMPLET Fragment.triggeredCoercion
INCHOAT(stative) = onset ───► "INCHOAT" for within_, after_ (coerced)
COMPLET(telic) = telos ─────► "COMPLET" for at_punct, before_ (coerced)
embeddedTelicityEffect = true holds exactly for connectives that
reference boundary points (onset/telos) of the embedded clause.
These are before and after: their truth conditions change depending
on whether the embedded clause is telic or atelic, because:
- Telic → natural endpoint exists → COMPLET/INCHOAT not needed
- Atelic → no natural endpoint → coercion needed for endpoint reading
The telicity sensitivity split corresponds to the ordering direction:
connectives with order ∈ {before, after} are telicity-sensitive;
overlap/persistence connectives are not.
The Fragment's embeddedTelicityEffect for before is consistent with
the data: before + stative is acceptable without coercion (default
before-start reading), but before + accomplishment has a coerced
alternative (before-finish via COMPLET).
The Fragment's embeddedTelicityEffect for after is consistent with
the data: after + accomplishment is default (after-finish), but
after + stative has a coerced alternative (after-start via INCHOAT).
The satisfiesDurativeRestriction predicate from the data layer is
equivalent to the theory's isHomogeneous predicate on the canonical
VendlerClass profile. This grounds the data-level predicate in the
theory-level aspectual feature system.
The until interaction data matches VendlerClass predictions: acceptable iff homogeneous (state or activity).
INCHOAT extracts the onset of an atelic/stative denotation.
The theory proves this equals the start point — connecting the
Fragment's triggeredCoercion = "INCHOAT" to concrete behavior.
COMPLET extracts the telos of a telic denotation.
The theory proves this equals the finish point — connecting the
Fragment's triggeredCoercion = "COMPLET" to concrete behavior.
The Fragment entries that specify triggeredCoercion are exactly those
that need aspectual adjustment for their complement type:
within_triggers INCHOAT (needs onset of duration)at_puncttriggers COMPLET (needs telos of telic event)- before/after have
triggeredCoercion = nonebecause they have both default and coerced readings (specified viacoercedReading).
When accepts states and achievements without coercion: states are homogeneous (any subinterval works for overlap), and achievements are already punctual. Activities and accomplishments require coercion to achievement because when selects a single reference point.
When's coercion need correlates with the Vendler class: classes that are either stative or punctual need no coercion. This is because when selects a reference point, and statives trivially provide one (any subinterval) while achievements ARE a single point.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Moens & Steedman's "strip-process" coercion (accomplishment → achievement)
corresponds to the theory's telicize-then-punctualize, but more precisely
it is the inverse of duratize: it removes duration.
Moens & Steedman's "add-result" coercion (activity → accomplishment)
corresponds to the theory's telicize operation:
adding a natural endpoint to an atelic predicate.
Moens & Steedman's "iterate" coercion (achievement → activity)
corresponds to the theory's duratize ∘ atelicize:
stretching a punctual event over time and removing the endpoint.
M&S's unified when-semantics agrees with the empirical coercion
pattern: when needs no coercion iff whenCompatible is true.
The M&S analysis explains the pattern: states are compatible
because they're homogeneous; achievements because they already ARE
culmination points. Processes and accomplishments require coercion
because when must access a culmination they don't directly provide.
M&S's coercion type matches the data layer: inception coercion targets processes (= Vendler activities), completion coercion targets culminated processes (= Vendler accomplishments).