Pre-Existence Theory vs. Empirical Data #
@cite{white-2014}
Connects the pre-existence theory (from PreExistence.lean) to
empirical data about forget's presuppositions (from
ForgetPresuppositions.lean) and to the English verb Fragment.
What This File Tests #
Pre-existence predictions match data —
needsModalInsertioncorrectly predicts which frames get modal vs. non-modal presuppositionsMCA overprediction — the Modalized Complement Analysis wrongly predicts a modal presupposition for the gerund case
Fragment consistency — the two Fragment entries for forget (implicative and factive/rogative) align with the theory
The pre-existence theory predicts: modal presupposition iff the complement type does NOT satisfy pre-existence. We verify this against each judgment from @cite{ippolito-kiss-williams-2025}.
Finite CP: pre-existence satisfied → non-modal presupposition. Matches: "forgot that she stopped" presupposes she stopped.
Gerund: pre-existence satisfied → non-modal presupposition. Matches: "forgot stopping by" presupposes stopped. This is the case that refutes the MCA's overprediction.
Plain infinitive: pre-existence NOT satisfied → modal presupposition. Matches: "forgot to stop by" presupposes was supposed to stop.
Aggregate: the pre-existence theory matches all data points. Modal content arises iff modal insertion is needed.
The Modalized Complement Analysis predicts modal presuppositions for ALL non-finite complements. This overgenerates for gerunds: "forgot stopping by" has a non-modal presupposition.
MCA wrongly predicts a modal presupposition for the gerund case.
MCA correctly handles the finite case.
MCA correctly handles the infinitival case.
MCA matches only 2 of 3 data points; pre-existence matches all 3.
The English Fragment has two entries for forget:
- forget (implicative, infinitival complement): "forgot to VP"
- forget_rog (factive/rogative, finite complement): "forgot that p"
Un@cite{ippolito-kiss-williams-2025}, these are NOT two distinct lexical items but
one verb with uniform factivity. The Fragment's split is a practical
choice: it separates the implicative entailment pattern (forgot to VP
→ didn't VP) from the rogative/factive pattern (forgot that p / forgot
whether p). Williams's pre-existence analysis explains why these two
uses surface differently without positing lexical ambiguity.
The factive entry (forget_rog) has factivePresup = true.
The implicative entry takes infinitival complements.
The factive entry takes finite complements.
The factive entry's primary complement satisfies pre-existence.
The implicative entry's complement does NOT satisfy pre-existence. This is why the modal is inserted, yielding the obligation reading.
Basic verification that ComplementType.isFinite classifies correctly. These hold independently of any theory of factivity.