Use of fMRI for cross-validation of EEG neural source modeling

We conducted a study of sentence comprehension, using ERP (brainwave) and fMRI (brain imaging) methods to examine the brain's response to anomalous vs. unexpected words, two types of semantic violation that are often confounded. Consider the following sentence: "The barber trimmed the man's..." People are faster to make semantic judgments judgment if "beard" appears as the final word, versus "cuticles" (an unexpected ending) or "clock" (an anomalous ending). To dissociate neural responses to these two kinds of linguistic stimuli, we recorded ERP and fMRI data as participants viewed sentences that ended with a word from one of three categories: Expected (meaningful completions that were highly expected based on the context), Unexpected (meaningful completions that were unexpected), or Anomalous (words that did not fit the context). Analysis of fMRI data revealed greater activity in the left angular gyrus, left middle temporal gyrus, and left inferior prefrontal cortex for Unexpected versus Anomalous endings. ERP results showed a classical "N400 semantic effect," as well as a late, right frontal pattern that was unique to the anomalous endings and peaked at around 700ms. Neural source modeling suggested that this later effect was explained by sources in right inferior temporal and right inferior parietal cortex. We discuss implications for cross-modal studies of language, and the role of right temporal cortex in meaning integration (the N700).