The anaphoric resolution process, and more precisely the retrieval of the relevant prior information has been extensively studied. Several studies showed an effect of physical distance between an anaphor and its antecedent even when close and early antecedents were backgrounded (O'Brien, Albrecht, Hakala & Rizella, 1995, Journal of Experimental Psychology: Learning, Memory and Cognition, 21: 626-634; O'Brien, Plewes & Albrecht, 1990, Journal of Experimental Psychology: Learning, Memory and Cognition, 16: 241-249). However, the advantage of proximity should be reduced when the intervening sentences explicitly or implicitly refer to the antecedent. This assumption is consistent with the memory based text processing conception. This conception (see Lorch, 1998, Discourse Processes, 26: 213-221) assumes a bottom-up access in which concepts and propositions derived from the text representation resonate in response to related concepts in the current processed sentence. The interaction between the text processing and the retrieval of information from long term memory results in a highly connected representation providing an elaborated context for understanding the current sentence.

We conducted an eyetracking experiment to test the hypothesis that an elaborated context provides a rapid access of the pronoun referent despite the distance between the anaphor and its antecedent. This was achieved by writing texts in two versions. In one texts version, the topic was kept and concepts semantically connected to the referent were mentioned, while in the other version the topic shifted and no concepts were related to the referent. The referential coherence was maintained throughout text by the mention of the protagonist in each sentence so that there was no break of local coherence. Each version of texts contained potential candidates on the basis of number and gender agreement and they could be located either in the Middle or at the End of the sentence. The last sentence contained an unheralded pronoun which referred to a concept evoked in the common first sentence of texts (see McKoon, Gerrig & Greene, 1996, Journal of Experimental Psychology: Learning, Memory and Cognition, 22: 919-932). The remaining of that sentence contained disambiguating information that was supposed to point to the relevant antecedent of the pronoun. The subjects' task was to read each sentence while their eye movements were monitored and then to answer to a comprehension question.

We investigated eye movement behaviour during the initial processing and re-processing of the anaphoric sentence. This was achieved by dividing the sentence into several zones. Thus, we tracked precisely the time course of the anaphoric resolution processes. Eye movement data showed three main results. First, the anaphoric resolution begins in the region (i.e., a word) immediately before the pronoun, suggesting a parafoveal pre-processing. During this early processing, the degree of Context Elaboration interacted with Position of potential candidates. As a consequence, the Pronoun zone was skipped more often supporting the parafoveal pre-processing hypothesis. Finally, when a non elaborated context was used, resolution processing persisted until the re-processing of the zone following the pronoun immediately (i.e., a verbal phrase).