SEMANTIC FACTORS IN IMPLICIT ANAPHORA RESOLUTION

Implicit anaphora, pronouns without explicitly mentioned antecedents, are used commonly in everyday speech. Reading experiments have identified two semantic properties that may facilitate the resolution of implicit anaphora: 1) number (plural pronouns are easier to process than singular), and 2) conceptual centrality (central referents easier than peripheral). We assumed a multimodal approach to language processing and used visual-world eye-tracking to investigate whether participants benefit from a visual context when processing pronouns without antecedents in speech. Participants listened to dialogues that contained either ambiguous 3^rd-person singular pronouns ("it") or plural pronouns ("they"), followed by a disambiguating verb. The dialogues were manipulated such that the pronoun was either explicit (e.g. My tooth was hurting terribly/The drill made a loud noise) or implicit (e.g. I was in terrible pain) and the intended referent was conceptually either central (e.g. tooth) or peripheral (e.g. drill) to the topic (e.g. the dentist) of the dialogue. Four images were shown on a screen while the participants listened to a dialogue: the protagonist (i.e. the topic), the central referent, the peripheral referent, and a distractor image. The number of looks to each image was measured at the pronoun onset and verb onset. A linear regression analysis showed a significant interaction between explicitness, number, and centrality, confirming that explicit plural central referents were most readily processed (t = 2.300, df = 1374.0, p < 0.05). Our results from the singular condition are consistent with previous findings demonstrating central antecedents are more easily resolved than peripheral ones, as indicated by a larger amount of looks sooner after the pronoun and verb onsets. Our plural results, however, showed implicit pronouns remained unresolved, regardless of the antecedent’s centrality. These results, along with previous findings that plural pronouns are read faster than their singular equivalents, suggest shallow processing for 3^rd-person plural pronouns in English. Importantly, visual context did not facilitate uncovering the referent(s) of implicit pronouns when the referents were peripheral or plural, suggesting that visual context cannot compensate for a diminished conceptual-semantic link.