888 - Eun-kyoung Rosa Lee / A divergence between comprehension and production measures reveals underlying mechanisms in real-time verb prediction
888 - Eun-kyoung Rosa Lee / A divergence between comprehension and production measures reveals underlying mechanisms in real-time verb prediction
Friday October 6, Rosa defends her 888 work, "A divergence between comprehension and production measures reveals underlying mechanisms in real-time verb prediction," supervised by Colin. The abstract is below.
The present study explores the potential mechanisms involved in real-time verb prediction. Previous EEG findings suggest that people initially fail to respect argument role constraints when predicting upcoming verbs in comprehension. Conversely, studies using speeded cloze measures have found rapid use of role information in production. The current study examined whether the divergence arises from experimental factors or underlying mechanisms. The N400 and speeded cloze measures from three experiments revealed systematic variability in whether role-sensitivity was observed: task-related and item-related differences. Participants showed immediate use of argument roles when the task involved quickly producing next-word continuations and when there was a tight relation between the preceding context and a candidate (role-appropriate) verb. A model of verb prediction is proposed to capture the variability through potential mechanisms. In the model, contextual information is actively used to access appropriate event representations in memory which inhibit the processing or production of a role-inappropriate verb. Initial lack of role-sensitivity arises in measures when no strong event memory is retrieved before the verb position. The model offers potential mechanisms that extend to prediction more generally, describing how bottom-up and top-down information affect the processing of subsequent input in real-time sentence processing.