Abstract
The visual system should rely heavily on a cue that has high ecological validity, and less on a cue that has low validity. We used the cue recruitment paradigm of Haijiang et al. (2006) to measure the ability of a new cue to control perceptual appearance after training in which the new cue was paired with a long-trusted cue with different levels of contingency. We predicted that the new cue would have greater subjective reliability when paired consistently with long-trusted cues. Stimuli were rear-projected movies of a rotating wire frame Necker cube. The new cue was stimulus position (above or below fixation) and the long-trusted cue was binocular disparity coupled with occlusion (which effectively disambiguated apparent rotation direction). Each experimental session consisted of 240 training trials (containing both new and trusted cues) and 240 test trials (new cue only) in a counterbalanced pseudorandom sequence. Thirty two trainees participated in four groups of eight. The probabilistic relationship between the new and trusted cues on training trials was 0.5, 0.7, 0.85, or 1.0, for the different groups, respectively. Each trainee ran three sessions on consecutive days. On Days 1 and 2 cue probability was fixed at one of the four levels. On Day 3 the probability on training trials was 1.0 for all groups but contingency was reversed. Subjective reliability of the new cue increased with cue probability, as indicated by inter-group differences in the effect of the new cue on apparent rotation direction on Day 3: those trainees exposed to high contingency on Days 1 and 2 responded according to that training even after contingency was reversed on Day 3. Analysis of data from the zero-contingency group (probability 0.5) revealed that the first few trials of the experiment had a surprisingly large effect on the subjective reliability of the new cue.