Retour vers liste

Détail de la contribution

Auteur: Carlo CECCHETTO

Co-Auteur(s): Caterina DONATI, Sapienza University of Rome Mirta VERNICE, University of Milan-Bicocca

Titre:
Rethinking adjunction and complementation. Making sense of an unexpected garden path


Abstract/Résumé: A Relative Clause (RC) like ‘the claim that John made (is false)’ is temporarily ambiguous with a structure where a noun takes a Clausal Complement (CC) like ‘the claim that John made (a mistake)’. A natural expectation is that the CC analysis should be chosen over a RC by the parser because CCs involve less nodes, (cf. Frazier’s Minimal Attachment). We monitored eye-movement while 29 Italian participants read 24 subject-RCs and the corresponding CCs (cf. 1). CCs and RCs differed only in critical area 3, where the sentence is disambiguated (convinse/convincessero). (1) Alla fine l’ordine che CONVINSE/ CONVINCESSERO l’ufficiale a dare l’allarme fu cruciale “All in all, the order that convinced / they-should-convince the officer to give the alarm was critical’ At critical area 3, there were significantly more regressions in CCs. This suggests that readers initially go for the RC interpretation and are temporarily ‘in trouble’ with CCs (a garden path effect arises). Similar results emerge from a second experiment contrasting CCs and object-RCs. Why is the RC analysis preferred? Answer 1: RCs are more frequent than CCs and frequency wins over syntactic complexity. However, we doubt that answer 1 is satisfactory (we will report data showing that the RC analysis is preferred even with nouns that have a frequency bias in favor of CCs). Answer 2: the observed garden path effect makes sense at the light of [1]’s view of relativization and noun complementation. First, under the raising analysis, RCs are not adjuncts. Furthermore, according to [1]’s version of the raising analysis, RCs are really akin to head-complement configuration at the right level of abstraction. Under their account, the head of the relative clause is a lexical item which relabels the structure, hence nominalizing it, by virtue of its word status. The fact that the head provides the label when it is internally merged with the RC makes relativization very similar to the configuration where a head provides a label when it is externally merged with its complement. [1] also argue extensively that nouns do not take complements. If so, CCs are not complements despite their name, but late inserted adjuncts. So [1] reverse standard wisdom: they see complementation where other approaches see adjunction (RCs) and they see adjunction where other approaches see complementation (CCs). Adopting this perspective, the observed garden path effect makes sense, if the parser chooses the basic head-complement configuration over an adjunction configuration. [1] C. Donati e C. Cecchetto (2011), Relabeling Heads. Linguistic Inquiry, 42.4, 519-560.