Using recorded movements of the arena (together with the heating schedule) from a previously trained fly (master) during open loop training periods in a second fly (replay-experiments, Wolf and Heisenberg, 1991) fails to elicit a pattern preference comparable to that observed in the paradigms used in this study. The recorded sequence of patterns associated with heat sufficed to produce a significant learning score in the master-fly, so the operant component clearly is required. The classical component of pattern sequence associated with the reinforcer is not sufficient to explain learning in the standard paradigm. However, it is not clear how the important operant component might exert its effect. Maybe the control of the reinforcer facilitates a stimulus-reinforcer association; it might be easier to perceive contingencies with one's own behavior than among some stimuli totally independent of oneself.
If there is a 'classical' (learning about contingencies in the world) association formed in the standard experiment, the assumptions made in the beginning have to be questioned. Either one can not consider an experiment 'purely operant' as soon as a single sensory stimulus is contingent with the reinforcer and the subject learns about this stimulus. Or operant conditioning is accomplished by the formation of two (or more) associations. Obviously, the view of singular associations that are formed in the investigated learning tasks was oversimplified.