Does the ending matter? The role of gender-to-ending consistency in sentence reading.

In many languages, during language comprehension the cognitive system needs to recover grammatical gender features in order to identify agreement dependencies established between different sentence constituents. A two-route model proposes that gender can be retrieved either lexically or computing its correlations with the word-form. However, evidence supporting this model has been collected thus far only with metalinguistic tasks on isolated nouns or word pairs. The present ERP study was aimed at testing whether the system is sensitive to gender formal cues within a sentence context. Specifically, we investigated the time course of gender processing in sentence reading where the target nouns could show a reliable gender-related ending (i.e., transparent nouns) or an ambiguous ending (i.e., opaque nouns). The results showed a greater central-anterior negativity for transparent nouns than for opaque nouns between 200 ms and 500 ms, suggesting that the system can rapidly detect reliable formal cues to gender. In addition, gender agreement violations showed a LAN-P600 pattern that was not modulated by the gender-to-ending consistency. Taken together, these results confirm that also during sentence comprehension, distributional gender cues conveyed by noun endings can be detected. This finding is compatible with the existence of a form-based route. The formal cues to gender are detected at an early stage, this probably being part of the word recognition process. Whereas this distributional information does not seem to be crucial in computing agreement dependencies within a sentence context.