before-clauses @cite{beaver-condoravdi-2003} @cite{sharvit-2014} #
@cite{beaver-condoravdi-2003}'s semantics for before, as adopted by
@cite{sharvit-2014} for the cross-linguistic before-clause typology.
The B&C semantics (@cite{sharvit-2014} (23)-(24), p. 271) #
before^{B&C} takes a body p (a partial time-predicate) and an
evaluation interval t. Its definedness presuppositions are:
t ⊆ C(containment in the contextually-supplied restrictorC)EARLIEST_C({t' ∈ C | p t'})is definedMIN(C) < EARLIEST_C(...)(a strict predecessor inCexists)
When defined, [[before^{B&C}]]^{C,g}(p)(t) is true iff
t < EARLIEST_C(...).
Inherent Presupposition Failure (IPF, @cite{sharvit-2014} p. 272-273) #
The central cross-linguistic prediction. Presupposition (2) fails
inherently whenever the body p itself contains an existential quantifier
over a dense time axis: there is no "first" time t such that some
t'' < t' satisfies the embedded predicate, because density supplies
arbitrarily-late witnesses below any candidate minimum.
This blocks past-under-past in before-clauses when the language has
quantificational tenses (Japanese), and is the technical core of the
pronominal/quantificational distinction in @cite{sharvit-2014}'s typology.
A pronominal past, by contrast, denotes a single time g k (no embedded
existential), so its body in before^{B&C} is constant in t and IPF
does not arise.
@cite{sharvit-2014} (24) presupposition (ii), p. 271: EARLIEST_C is
defined for body p iff the set of C-times where p holds has a
least element (mathlib's IsLeast).
Equations
- Semantics.Tense.TemporalConnectives.Before.hasEarliest C p = ∃ (t : Time), IsLeast {t' : Time | t' ∈ C ∧ p t'} t
Instances For
IPF (@cite{sharvit-2014} (27), p. 272): when the body of
before^{B&C} is the quantificational past [[PAST]]^{K,g}(q), and
the restrictor C is order-dense (interval-like) with K ⊆ C, the
EARLIEST presupposition fails. The proof is by contradiction:
a witness t_q < t_min with q t_q lifts via density to a strictly
smaller body-witness t_mid in C.
This is the technical core of @cite{sharvit-2014}'s thesis that only
languages with pronominal tenses license past-under-past in
before-clauses. The q-witness is not a separate hypothesis: it
falls out of hasEarliest's witness for t_min, which is why no
∃ t ∈ K, q t premise appears.
The Bool-valued IPF dispatch on tense lexical type, for use in cross-linguistic typology theorems. Quantificational tenses trigger IPF in before-clauses; pronominal tenses do not.
Equations
Instances For
@cite{sharvit-2014}'s prediction ((27), p. 272): a language's
past tense is well-formed under before^{B&C} iff its tense lexical
type does not trigger IPF — i.e., iff it is pronominal.