Within-trial contrast: when you see it and when you don't

Within-trial contrast occurs when a discriminative stimulus that is preceded by a relatively aversive event is preferred over another that is preceded by a less aversive event. Recent failures to replicate (Arantes & Grace, 2008; Vasconcelos, Urcuioli, á Lionello-DeNolf, 2007) may allow us to identify factors that may be responsible. In the case of Vasconcelos et al., it is likely that insufficient training was provided (often 35–65 sessions are required). In the case of Arantes and Grace (Experiment 2), these pigeons had been involved in prior experiments involving lean schedules of reinforcement, and we find that prior experience with lean (relatively aversive) schedules appears to reduce the presumed aversiveness of the many-peck requirement, thus obviating the contrast effect. Finally, in the case of Vasconcelos and Urcuioli (2008), although the contrast effect with a simultaneous discrimination was not reliable, it was not reliably smaller than with a successive discrimination that did show a reliable effect, and the contrast effect was also similar in magnitude to a reliable effect reported by Kacelnik and Marsh (2002). Thus, although there have been several failures to replicate the original effects reported by Clement, Feltus, Kaiser, and Zentall (2000), insufficient training, prior history with lean schedules of reinforcement, and low statistical power may have been responsible for those failures.