Testing predictions about the processing of word stress in reading using event-related potentials

Both computational models of English reading that generate word stress predict a processing advantage for words with initial syllable stress. They differ, however, on whether they process words incrementally and learn nonlinear spelling-stress relationships. Two experiments using event-related potentials were used to investigate these predictions. The first examined trisyllabic stimuli. Differences found on P200 and N400 components suggested a processing advantage for words with initial syllable stress. The second examined root morphemes within words that have high frequency suffixes that are stress predictive. A processing advantage on the N400 component was found with root morphemes that typically have initial syllable stress, even when the whole-word stress pattern differed. This provides evidence that stress is generated incrementally, where it is assigned to parts of words as they are processed, and that stress assignment is not necessarily affected by high frequency nonlinear relationships.