Brocher, A. & von Heusinger, K.
University of Cologne
We conducted an ERP story comprehension experiment, testing to what extent referent introduction in discourse involves (a) activation of a concept and (b) activation of a referent. Critical referents always appeared in object position and were inferred/unique (gym... the trainer), inferred/non-unique (gym... a trainer), brand-new/unique (gallery... the trainer), or brand-new/non-unique (gallery... a trainer). We found that, at referent encounter (trainer), unique referents elicited more negative ERPs than non-unique referents in frontal channels between 250 ms - 450 ms post onset, p = .054. Furthermore, larger late positivities obtained frontally for inferred than brand-new referents between 700 ms - 850 ms, p = .004. Disambiguation of a subsequent pronoun (subject (e.g., Philip) vs. critical object) led to larger N400s for inferred/unique than inferred/non-unique referents, with no differences between brand-new referents, p = .026. We suggest that retrieval processes recruit a referent's unique identifiability, while integration processes recruit a referent's concept, which may or may not be pre-activated through context. Data from disambiguation show that inferred/unique referents were better competitors to subject referents than inferred/non-unique referents (cf. larger N400): When a referent could be inferred but not uniquely identified, readers activated a concept rather than a referent, yielding a weak referential representation.