Nonsense forms with nonsense labels were utilized in a match to sample task to observe whether children's word extensions are based on static form characteristics or on functional action characteristics. The stimuli consisted of: (1) a single model object which performed an action; (2) a selection array consisting of one object similar in form to the model, but with no demonstrated action; and (3) another object of a very different form than the model's, but which performed a similar action to that of the model's. Subjects were 31 normal preschool children ranging in age from 2;3 to 5;9. When form and action were in competition for the basis of the children's selection responses, action was the more frequently noted basis. More data are being collected from older children to observe when a shift in basis of extension occurs. Longitudinal data are also being collected on four children, two of whom are presently action based, and two of whom are presently form based in their word extensions. Implications for contrastive investigations with language impaired children are noted. (Author/SW)