One hypothesis about implicit personality theory holds that laypersons' beliefs about how personality traits covary are based on semantic overlap or linguistic similarity among the trait words involved. Words that can describe both people and things (Asch's physicalistic metaphors, “warm,” “cold,” “hard,” “soft,” “sweet,” “sour,” “bright,” “dull”) were employed in a test of this hypothesis. In learning and memory tasks, these words were paired either with names of people or with nonsense syllables or paralogs. Subjects' patterns of errors when names were used as stimuli resembled trait inference patterns derived from other methods. When nonsense syllables or paralogs were used as stimuli, the error patterns did not resemble trait inference patterns. Instead, the most frequent confusions were intradimensional (opposites) confusions and intramodality (somesthetic) confusions. It is argued that trait inferencing is a fundamental process in person cognition, spontaneous, coercive, and persistent; and that implicit personality theory is, in fact, a theory about people and is not reducible to linguistic similarity among trait terms.