A two-alternative choice procedure was used with rats to examine preference for discriminative stimuli (correlated with the occurrence of reinforcement) versus uncorrelated cues. Choice of discriminative stimuli was below chance, despite the use of very low levels of reinforcement (12.5% for some rats) known to produce a preference for discriminative cues. Levers that had been retracted during the stimulus presentation in Experiment 1 were extended in Experiment 2 to assess the role of response elicitation during the observed stimuli. In Experiment 3, the impact of cue-duration behaviors was further evaluated by lengthening the duration of the cues and providing an opportunity for an escape response. The preference for nondiscriminative stimuli was reversed only with increased exposure to the cues. The present results point to fundamental differences in the prevailing context of the choice procedure as compared to the usual free-operant observing design.