Three experiments examined whether people spontaneously generate evaluations of target individuals under circumstances in which they are also known to generate spontaneous trait inferences (STIs). The first experiment used a standard savings-in-relearning paradigm to explore whether exposure to trait-implicative behavior descriptions facilitates the learning of evaluatively-congruent, as well as behavior-implied, personality traits. Evidence for the facilitated learning of evaluatively-congruent traits was not obtained. This led to a second experiment in which the savings-in-relearning paradigm was altered to directly assess participants’ relearning of evaluative words (good/bad). The results demonstrated that the same trait-implicative behavioral stimuli can produce both spontaneous trait inferences and spontaneous evaluations when both are measured correctly. Both of these outcomes were replicated in a third study using a false recognition paradigm. The implications of these findings for impression formation processes and for the possible independence of semantic information and evaluative information are discussed.