We are happy to announce a talk by Pascal Hohmann (Frankfurt) in the Semantics Colloquium.
The talk will take place on campus in IG 4.301.
Date: July 16, 2026
Time: 4 pm – 6 pm c.t.
Title: Before what happens? NPIs and expletive negation in German bevor-clauses
Abstract:
German ‚bevor‘ (‚before‘) clauses are an unusual environment in which two superficially opposite, negation-related phenomena surface in one and the same syntactic frame. On the one hand, ‚bevor‘-clauses license negative polarity items (NPIs) — such as ‚mit der Wimper zucken‘ ‚bat an eyelid‘, ‚(einen) Mucks machen‘ ‚make a peep‘, ‚ein Wort sagen‘ ’say a word‘, or ‚je‘ ‚ever‘. On the other hand, they host expletive negation (EN): a ’nicht‘ that, in the right context, does not reverse the truth conditions of the clause it appears in (Krifka 2010):
(1)
Ich gehe nicht nach Hause, bevor das (nicht) fertig ist.
‚I won’t go home before that is (not) finished.‘ — both variants truth-conditionally equivalent
The talk pursues a semantic question: are NPI licensing and expletive negation freely available inside ‚bevor‘-clauses, or are they conditioned — and if so, by what?
The point of departure is Tahar’s (2021) typology of before-clauses, which distinguishes three uses: consecutive (a factual temporal sequence), apprehensive (preventive; the main clause aims at avoiding the before-event — a negative purpose), and frustrative (the main clause is an ‚active inaction‘ motivated by the expectation that the before-event will not occur — a negative reason). The hypothesis under investigation is that the two phenomena stand in complementary distribution along this divide: NPIs surface in apprehensive (and consecutive) ‚bevor‘, expletive negation only in frustrative ‚bevor‘:
(2)
a. Noch bevor die Frau einen Mucks machen konnte, hielt ihr der Fremde den Mund zu. (apprehensive; NPI)
b. Kein Spieler darf zwei Elfmeter ausführen, bevor (nicht) alle anderen Mitspieler geschossen haben! (frustrative; EN)
Empirically, I test the hypothesis on web-scale corpus data (deTenTen23, 1.17B sentences; ~2.25M ‚bevor‘ tokens), extracting NPI and EN candidates and classifying random samples of 100 hits each by Tahar’s diagnostics. The data support the complementary distribution: EN occurs exclusively in frustrative contexts, and NPIs are absent from frustrative ones. At the same time, a robust consecutive/apprehensive ambiguity in the NPI data indicates that the three-way typology is under-determined for German.
Theoretically, the result sharpens a puzzle for the existing accounts: if EN is licensed by matrix negation together with a semantically uniform ‚bevor‘, why is it barred from consecutive ‚bevor‘ even when the main clause is negated?