Three experiments investigated antonym comprehension, that is, the judgment that two words are or are not antonyms. In Experiment 1, the latency of yes responses to antonyms decreased with degree of antonymy. In Experiment 2, no responses to nonantonym pairs were slower for pseudoantonyms (e.g., popular-shy) than for synonyms and unrelated words that did not differ from each other. In Experiment 3, subjects made either antonym judgments or synonym judgments for the same set of antonyms, synonyms, pseudoantonyms, pseudosynonyms, and unrelated words. The no responses were slowest for pseudoantonyms in the antonym task and for pseudosynonyms in the synonym task. Categorization models can explain these results if it is assumed that processing is a function of relationship similarity, the degree to which the relationship between members of a stimulus pair exemplify the relationship being judged by the subject.Comprehension of category relationships has been investigated extensively in recent years, and several models have been proposed that provide fairly good accounts of category comprehension findings (Smith, 1977). The progress made in investigating category comprehension makes it appropriate to consider what kind of model would provide a general account of the comprehension of semantic relationships. A general model should, for example, be able to explain comprehension of relationships like antonymy and synonymity in addition to categorical relationships. Although categorization models were not explicitly designed to provide a general account of semantic relationships, the task of developing a general model may reasonably begin by applying categorization models to other semantic relationships. The general purpose of the present research is to determine whether categorization models can accountThe authors thank Christopher Peterson for valuable comments on an earlier draft of this article.