Abstract
Using a Pavlovian conditioning procedure, we recently showed that previously irrelevant visual signals can be recruited to act as new cues that control the appearance of bistable stimuli (Haijiang et al., 2005). This learning was positive in sign (unlike most contingent adaptation aftereffects) and long lasting. In group average data, the learning was incremental. However, the average may not faithfully reflect the individual time courses, which could be either truly incremental or step-like with the abrupt changes that occur at different times (Gallistel et.al. 2004). Here we conducted additional experiments to study the time course of cue recruitment in individuals, using the rotating Necker cube stimuli of Haijiang et al. (2005). In training trials the cube's position (top and bottom screen locations) was made contingent on the trusted cues (stereo and occlusion) that forced the direction of perceived rotation, while the test trials contained the new position cue but not stereo or occlusion. After the first block of training, which caused an initial bias in the observer's perception, signal contingency was reversed for the remainder of the session. Data from each individual subject show a gradual transition from the initial bias to its opposite over a 30–40 minute period. There was also a slow oscillating component with a period of approx. 15 min. We conclude that cue recruitment (at least for learning after reversal) is incremental in some cases. This is consistent with a Bayesian probabilistic decision-making framework, and with models of incremental learning such as the Rescorla-Wagner model.
This work is supported by NIH R01-EY-013988