How necessary are the stripes of a tiger?: Diagnostic and characteristic features in an fMRI study of word meaning

Abstract

This study contrasted two approaches to word meaning: the statistically determined role of high-contribution features like striped in the meaning of complex nouns like “tiger” typically used in studies of semantic memory, and the contribution of diagnostic features like parent's brother that play a critical role in the meaning of nominal kinds like “uncle.” fMRI monitored regional brain activity while participants read complex noun descriptions consisting of statistically high-contribution and low-contribution features; and nominal kind descriptions consisting of diagnostic and characteristic features. We found different patterns of activation depending on the type of noun and the type of feature contributing to the noun. Complex nouns recruited significantly greater bilateral superior temporal and left prefrontal activation compared to nominal kind nouns, while nominal kind nouns activated bilateral medial parietal and right inferior parietal regions more than complex nouns. Moreover, features making a statistically high contribution to complex noun meaning activated right inferior frontal cortex relative to low-contribution features, while diagnostic features of nominal kinds activated left dorsolateral prefrontal and right parietal regions more than characteristic features. These findings are consistent with the hypothesis that at least two different neural mechanisms appear to support word meaning: one driven by a statistically determined approach to feature knowledge, and the other sensitive to the qualitatively critical role that a specific diagnostic feature plays in word meaning.