Poster B77, Sunday, March 26, 8:00 – 10:00 am, Pacific Concourse
Precursors and Processes of Prediction: A Word-Stem Completion ERP Study
Ryan J. Hubbard1, Kara D. Federmeier1; 1Beckman Institute, University of Illinois, Urbana-Champaign
Prior behavioral and electrophysiological research has demonstrated that individuals generate predictions about upcoming information during language processing. Although there is a large body of work examining the consequences of prediction when an expected or unexpected word is encountered, much less is known about the processes that occur as predictions are being formed. Here, we used a novel paradigm to investigate the neural processes associated with encountering more and less constraining language cues. First, over 100 individuals completed 330 3-letter word stems with the first word they could think of. This gave us probability distributions for each completion for each stem, allowing us to calculate entropy for the word stems. We could then manipulate both the constraint (entropy) of the stem and the predictability (probability) of the completion. In a separate ERP experiment, subjects were shown a stem, told to generate a completion, and then shown either a high probability completion, a low probability completion, or a pseudoword that started with the same letters as the stem; participants reported whether the presented completion was the one they had generated. N400 responses to the completion were graded with probability (even for words that were not predicted), such that high probability < low probability < pseudoword, aligning with results manipulating cloze probability in sentences. Importantly, at the presentation of the word stem, the N400 was modulated based on entropy, with low entropy stems eliciting smaller N400s. Thus, prior to any confirmation of predictions, the N400 was affected by the predictability of upcoming information.
Topic Area: LANGUAGE: Semantic