Abstract

Over the last years several results demonstrated that context-based expectations on both word-class and concepts influence the word processing at very early stages, namely at sensory analysis level. Given that these early effects are modulations of the process of stimulus analysis they depend on physical and orthographical properties of critical words in interaction with linguistic expectations. This evidence on early effects is in contrast with a syntax-first approach for which the cognitive system builds at first the syntactic structure by exploiting word-class information only. This strong syntax-first assumption pushed forward by Friederici (2002) model is based on a very early ERPs effect with latency around 150 ms that is elicited by word-class violations (eLAN: early left-anterior negativity). I studied three linguistic violations with an ERPs sentence processing paradigm. In two studies in Italian word-class violations on prepositions and verbs were implemented, overcoming the more important methodological limitations of previous studies on word-class violations. In a third study we investigated determiner-noun gender agreement in Italian using nouns for which grammatical gender is expressed unambiguously by a long derivational morpheme, that is very salient at orthographic and visual level. ERPs results show a LAN (300ms latency) followed by a P600 (500ms latency) for all the conditions. The lack of replicability of eLAN, already discussed in the literature, makes Friederici (2002) model difficult to be maintained. The ERPs elicited by gender disagreeing nouns also show an effect on the amplitude of the N250 (200ms onset), an effect specific to morphological processing since a previous study with no control on how gender was expressed reported a LAN+P600 pattern only (Molinaro et al., 2008). The latter result shows that gender agreement can affect word recognition (at least the morphological parsing) during sentence processing earlier than violation detection indexed by the LAN. This result enrich the evidence about early context top-down effects that are different from syntagmatic structural processing.