Negation reveals preferences in long distance dependeny formation
Speaker: Zeynep Ilkin
Institution: University of Edinburgh
Abstract:
Humans show a preference for forming long-distance dependencies at the first potential gap position (Frazier&Clifton,1989) However in head final structures like; [Filler...Gap1?...[...Gap2?...V2]...V1] the embedded verb (V2) precedes the main verb (V1), so Gap2 is the first gap position that will allow a thematic role assignment of the filler, although it is structurally the most distant. Here we report 2 reading time studies and a sentence completion test exploring the time course of gap projection. In Turkish a neg(ative) NP ("Nobody") needs neg morphology on its licensing verb. If a neg NP is an embedded verb argument, there will be processing difficulty if the embedded verb lacks neg morphology, but if the NP is interpreted as a main clause argument, a similar effect should be found at the main verb. We orthogonally manipulated the ambiguous dative NPs, ("Nobody" vs. "Somebody"), and the form of the embedded and main verbs (if subordinate verb was neg, the main verb was affirmative, and vice-versa). In exp 1 the ambiguous dative NP was sentence initial, in exp 2 it followed the matrix subject but preceded the embedded subject.
In exp1 and exp2 at the embedded verb, there was no hint of an interaction so no evidence for the embedded clause interpretation. However, the main verb showed an interaction; (affirmative verbs were slow following "Nobody" and vice versa for neg verbs): clear preference for the main-clause interpretation. Again, in the off-line completion data, dative NPs were overwhelmingly associated with the matrix predicates rather than the embedded predicates, regardless of the negative NP's position. However there was also some evidence of negation priming on the embedded verbs, despite clear matrix-clause disambiguation.
In similar constructions in Japanese, Aoshima et al (AeA) (2004, JML) found that sentence-initial dative wh-NPs were associated with the embedded verb on-line, in contrast to our results. One interpretation of AeA's results might be that the parser tries to satisfy morphological requirements as soon as possible, leading to an embedded clause interpretation (Japanese has an obligatory scope-marking question-particle, and the embedded verb is the first opportunity to place it). However, our results don't support this: the Turkish requirement for neg verbal morphology didn't lead to an embedded clause interpretation. It is possible that AeA's embedded clause preference reflects some special property of wh-questions. Future work with Turkish wh-questions (which don't have question-particles) will examine this issue further.
