Dynamic cue use in pigeon mid-session reversal.

The systematic anticipation and preservation errors produced by pigeons around the reversal point in midsession reversal (MSR) learning experiments suggest that an internal time estimation cue, instead of a more efficient external cue provided by reinforcement, controls behavior over the course of a session. The current experiments examined the role and effectiveness of other external cues in the MSR task. In Experiment 1, providing differential outcomes based on response key location produced fewer errors prior to, but not after, the reversal as compared with a non-differential outcomes condition. Experiment 2a used alternating differentially colored ITIs (cued sessions) or dark ITIs (un-cued sessions) during each half of the session. The ITI cues improved switch efficiency both prior to and after the reversal. Experiment 2b introduced probe trials around the reversal, testing ITI color cues added to un-cued sessions or removed from cued sessions. Results showed control by the ITI cues when they were available and control by the time-based cue when they were unavailable. This suggests both cues were being simultaneously processed when available and that the cues could also independently provide sufficient information about future reinforcement. In Experiment 2c, ITI cues were inserted as probe trials in the opposite half of the session (miscues). The closer such miscue trials were to the reversal, the more the ITI cues exerted control over behavior. Together, these results indicate that as the utility of internal temporal cues is reduced, the use of external visual cues increases. These results have implications for the way in which cues dynamically shift in controlling behavior over time based on their relative rates of utility, and are discussed in light of an occasion setting perspective.