Kaiser, E. , Bhide, S. & McMahon, C.
University of Southern California
In halfalogues, only one speaker is audible (e.g. overheard phone conversations). Halfalogues may trigger automatic filling-in of missing information (Norman/Bennet'14). However, it is not known how detailed this filling-in is or what guides it. In Example(1), would ovearhearers try to guess who "he" is, and what the inaudible part was?
EXAMPLE(1): SPEAKER1: "I heard that Bob helped Tim move into his new apartment." SPEAKER2: <inaudible/quiet mumble> SPEAKER1: "Oh wow, he did?"
EXPERIMENT (n=72): We used halfalogues like EXAMPLE(1). Half the trials contained two additional sentences about the subject (e.g. Bob) before Speaker2's inaudible response, marking the subject as salient/topical. The NEUTRAL_GROUP listened to halfalogues and answered a question about who the ambiguous final-sentence pronoun refers to. The GUESS_GROUP guessed at Speaker2's inaudible turn before answering the questions. The RECALL+GUESS_GROUP additionally retained random letter-sequences in memory during each halfalogue. NEUTRAL and GUESS-GROUPS tend to interpret pronouns as referring to subjects, especially when subjects are topical. The RECALL+GUESS_GROUP shows mostly chance-level performance in pronoun interpretation and no topicality effects. This suggests the mind fills in missing information, even when not asked to do so, and uses discourse/topicality cues, but this is cognitively costly and may not occur if resources are limited.