In an associative learning preparation, the participants were given partial reinforcement (PRF) with two different cues. For one of the cues, the nonreinforced presentations consisted of pairings of the cue with a neutral outcome, whereas these presentations consisted of pairings with an aversive outcome for the other cue. The results showed that PRF training produced strong responding to the cue paired with the neutral outcome on the nonreinforced trials, whereas responding to the cue paired with the aversive outcome on the nonreinforced trials was strongly suppressed. The present results are problematic for current theories of learning (e.g., Rescorla & Wagner, 1972), but can be explained by classical theories involving motivational mechanisms (e.g., Konorski, 1967), as well as by recently developed model, in which incompatible outcome expectations compete for their expression into behavior (i.e., Pineno & Matute, 2003). (Contains 2 figures and 4 footnotes.)