An analysis of part-of-speech membership was made utilizing certain mechanisms that have been proposed to explain the nature of word classes. It was proposed that words of the same form class constitute "verbal habit families" on the basis of either a common grammatical meaning response, a common affix, or a common label (e.g., "verb"). One implication of this model was that sentences may be regarded as sequences of grammatical habit families. Three experiments tested two derivations from the model. Experiments one and two indicated that, under certain circumstances, words classified as verbs do indeed constitute a word response class. Experiment three tested the hypothesis that a novel word will gain verb properties by being paired with other verbs. The results of this experiment indicated that the greater the number of times a nonsense syllable was paired with verbs, the less often it was used as a verb in a sentence completion task. This result, contradictory to the hypothesis, did not appear incompatible with the model developed, but further research is needed for an adequate evaluation. (Author/LH)