The stimuli were generated as in Experiments 1 and 2. The 16 less compact ``adjective-like'' categories were defined in terms of ranges of 2/3, 2/3, 1/3, and 1/12 of the possible values along the four input dimensions used in this experiment. That is, one input dimension, the one for which the possible within-category range was 1/12th of the input dimension, was much more relevant than the other three in defining the category. Each of the four dimensions played this role for four of the adjectives. Each of the more compact noun-like categories was defined in terms of a range of 1/3 of the possible values along each input dimension. The extensions of both the noun-like and adjectives-like categories encompassed the same volume (1/81 of the space). The noun-like and adjective-like categories overlapped in the space. As in Experiment 2, ``noun-like'' categories were associated with a linguistic context input specifying noun targets and all the adjective-like categories were associated with one linguistic input specifying adjective targets.