Chow, W. Y. & Chen, D.
University College London
A key question about the brain's predictive nature is what happens when the input conflicts with our predictions. Previous research has shown that humans can quickly detect cues that conflict with their predictions during language comprehension, and some have proposed that prediction errors can aid learning over time. In the present study we turn our focus to ask how quickly our brains can use unexpected information to revise an existing prediction.
In Mandarin Chinese, a noun is preceded by a grammatical element known as a 'classifier'. Since Mandarin has over 100 classifiers and different classifiers are associated with nouns with different semantic-conceptual properties like shape and function (e.g., 'zhi' for pens and tree branches, 'ding' for hats and crowns), an unexpected classifier can not only disconfirm listeners? predictions about an upcoming noun, but also inform them what the correct noun is likely to be.
We tracked participants' eye-movements as they listened to Chinese sentences with an English equivalent like (1), where the target noun is preceded by a classifier and an adjective. Crucially, the classifier was either compatible with the most expected noun coffee, or incompatible with coffee but indicative of another contextually suitable noun like cake. A general classifier (e.g., 'ge', 'xie') that is compatible with both nouns (and therefore is uninformative) was used in the control conditions.
(1) Anna went to Starbucks to buy a {CLcup /CLpiece/ CLgeneral} of very tasty {coffee/cake}.
Results (n=24) showed that listeners were initially more likely to look at the expected object (coffee). Further, starting from 460ms after the onset of an unexpected classifier, they would redirect their eye gaze to the previously unexpected object (cake). These results extend previous evidence for listeners' sensitivity to prediction errors and show that they can also use unexpected information to revise their predictions rapidly.