Abstract :
[en] In a traditional view of language processing, language users fully analyze a sentence to its underlying syntactic structure during interpretation, and compose each sentence from scratch during production. It has been pointed out, however, that a faster ‘pseudo-parse’ is usually sufficient for successful communication (Ferreira, Bailey and Ferraro 2002; Ferreira and Patson 2007). In order to speed up language processing, language users may store frequently reoccurring ‘chunks’ of language in memory as unanalyzed wholes, as to not having to repeatedly compose and decompose them (Bybee 2010; Dąbrowska 2014; Diessel 2015). In (1) and (2), you may find two Dutch utterances that look deceptively similar, yet are structurally unrelated. In (1), iets ‘something’ forms a separate noun phrase, while verkeerd ‘wrongly’ constitutes an adverb. Meanwhile, in (2), iets verkeerd forms a noun phrase in a partitive genitive construction (Broekhuis and Strang 1996; Booij 2010: 223–228). Such a construction may appear both with and without an -s ending on the adjective, as shown in (3), in both Belgium and the Netherlands (Pijpops and Van de Velde 2014). (1) dat iets verkeerd geïnterpreteerd wordt? (ConDiv) that [something]NP [wrongly]Adv interpreted gets? ‘that something gets wrongly interpreted?’ (2) iets verkeerd gegeten (ConDiv) [something wrong]NP eaten ‘eaten something wrong?’ (3) Ik had iets verkeerd-s gegeten (ConDiv, Grondelaers et al. 2000) I had something wrong-S eaten ‘I had eaten something wrong’ Assuming that utterances like (1) and (2) are always fully analyzed during interpretation, we do not expect these constructions to interfere. Still, we will show that the construction in (1) has a contaminating influence of the construction in (2), solely on the basis of superficial similarity. We call this effect constructional contamination, and will argue that it is a by-product of exemplar chunking. In this talk, we will evaluate several measures to quantify this effect.
