Grammatically relevant and grammatically irrelevant features of verb meaning can be independently impaired

Abstract

According to the 'grammatically relevant semantic subsystem' hypothesis proposed by Pinker (1989), the meanings of verbs contain two different types of information: (1) a restricted set of semantic features that are visible to grammar and that influence the syntactic argument structure possibilities of verbs; and (2) an open - ended set of idiosyncratic perceptual and conceptual features that are not visible to grammar and that enable verbs to encode all kinds of subtle distinctions involving manner of motion, change of state, and so forth. If these two components of verb meaning reside at different levels of mental representation, as the hypothesis maintains, then it is possible that they could be independently disrupted by brain damage. This paper reports a series of experiments that confirmed this prediction. Three brain - damaged subjects exhibited different performance profiles on two tests that evaluated their knowledge of grammatically relevant and grammatically irrelevant components of verb meaning. 1978JB and 1962RR performed well, but 2011SS performed poorly, on a word - picture matching test that required them to discriminate between verb triplets that differed only with respect to subtle, grammatically irrelevant perceptual and conceptual features–e.g., spill–pour–sprinkle. In contrast, 1978JB and 1962RR performed poorly, but 2011SS performed well, on a second test that required them to judge the grammaticality of sentences containing the very same verbs–e.g., Sam spilled beer on his pants vs. *Sam spilled his pants with beer. Linguistic analyses have shown that the grammaticality of these sentences depends on the compatibility between certain semantic features of the verbs and the inherent semantic content of the grammatical constructions. Finally, all of the subjects performed well on a third test that required them to judge the grammaticality of sentences that have purely syntactic violations. Taken together, these results suggest that 1978JB and 1962RR have a disorder that selectively involves the level of grammatically relevant semantic structure, whereas 2011SS has a disorder that selectively involves grammatically irrelevant aspects of verb meaning. This study therefore supports the hypothesis that grammatically relevant and grammatically irrelevant components of verb meaning are segregated in the mind/ brain.