Understanding prediction in language processing: Quantifying the degree to which a word's meaning and orthographic form expected on the basis of the preceding context assist in word processing efficiency (G011617N)

Language comprehension is incremental. This means that words are predicted to some extent on the basis of the preceding context. The prediction is both semantic (the word "oak" is processed faster when the word "palm" is expected) and orthographic (the initial response to the unexpected word "dish" is less disrupted when the word "wish" is expected). Up to now, however, all the evidence is limited to yes/no situations (expected vs. unexpected) and to the violation paradigm (how much disruption is caused by a word that does not fit into the context). In the present research we want to investigate how strong prediction effects are in the processing of continuous text, where predictions are rarely very strong and where no really unexpected words are encountered. To do this, we will determine the predictability of each word in a text on the basis of newly developed semantic vectors that have proven their usefulness in our previous research. Next we will investigate the impact of the semantic vectors on the processing efficiency in eye movement data and in ERP results. This will allow us to quantify the extent to which word processing in text reading depends on the fit with the preceding context and how the exact relationship looks like across a wide range of predictability values. The present research is part of the current movement to supplement factorial, experimental designs with analyses of big data from large-scale corpora and megastudies.