[PS-2.90] When what you see trumps what you know: Superficial object features drive listeners' computation of referential contrast

Klammer, C. 1 , Mozuraitis, M. 1, 2 & Chambers, C. . 1

1 University of Toronto
2 Cancer Care Ontario

Research has shown the underlying 'essence' of visual objects is dominant during real-time referential interpretation, even when objects' appearance is misleading and when attending to superficial features would benefit listeners (i.e., conversational scenarios where speakers are unaware of visually-misleading objects' identity). In the current visual-world study, listeners followed instructions referring to one of four real objects. The target object (small box) always occurred alongside a size-contrasting alternative (big box). In the Two-Contrasts condition, the remaining two objects were also a size-contrasting pair (small, big lightbulb), entailing (predictably) slow target saccades, measured from adjective onset. In the One-Contrast condition, the remaining two objects were from distinct categories (small lightbulb, big candle), entailing fast target identification. Of interest was the Visually-Deceptive condition, where the remaining objects resembled those in the Two-Contrasts condition but 'conceptually matched' the One-Contrast condition because listeners fully knew about deceptive objects' actual identity (small lightbulb, candle resembling a big lightbulb). Here, eye-movements resembled the Two-Contrasts case. We speculate that size adjectives, which denote surface properties, may lead listeners to emphasize appearance over reality when categorizing contrasting objects. Together with previous findings, the results highlight context-dependent variability in how visual and conceptual knowledge are weighed during referential processing.