Most experiments in psycholinguistics use association strength obtained by group-level production frequencies to control experimental stimuli or to define experimental condition. Researchers assume that group-level production frequencies reflect association strength in the mental lexicon of each individual, and the validity of the above approach depends on this assumption. It is assumed that associates are produced by spreading activation from the cue to its targets as a function of their association strength. Two questions are warranted: (1) How does the cognitive system choose a response among the activated associates? (2) Why do people produce different associates if they share the same associative network? I present a simple connectionist model that explains how the same associative network might give rise to different responses, whose frequencies approximate the underlying individual association strengths.