A bird is a bird is a bird: Making reference within and without superordinate categories

Adults with localized cerebral insult often err in their use of a word to refer to an object or an idea. It is important to assess, then, the patients' appreciation of what a word can refer to, and the way in which they may violate the borders around a referential field. Nonfluent aphasics, fluent aphasics, and nonaphasic patients with insult to the right hemisphere were asked to provide names of items which could be referred to by familiar superordinate terms like “bird.” The principal results revealed that the nonfluent aphasics are anchored to the central portions of a superordinate's referential field (naming items like “robin” and “sparrow,” for instance). While fluent aphasics often violate the borders around a referential field (e.g., providing “beaver” in response to “birds”), it was nevertheless possible to characterize some limits to their choice of a superordinate's referents. These findings were independent both of the absolute number of responses provided and the frequency of occurrence of the response. Further, the patients with insult to the left hemisphere produce few consecutive items whose referents hold attributes in common. When these clusters are produced by aphasics, they consist primarily of subordinates whose referents exhibit many overlapping features (e.g., “bald eagle, black eagle, golden eagle” in response to “bird”). The right-hemisphere-damaged subjects, in contrast, produce many clusters of related items. These consist of less central, basic object level words whose referents hold less obvious features in common (e.g., “albatross, crane, gull” in response to “bird”). Aphasics, then, may be limited in their ability to analyze referents for critical features. Taken together, these data contribute to a more precise characterization of the nonfluent aphasics' and the fluent aphasics' referential deficits, and lend support specifically to the notion that neither group of aphasics relies on definition-like features to determine what a word can refer to.