Short-term dynamics of perceptual bias for bistable stimuli

The perception of a bistable stimulus is influenced by prior presentations of that stimulus. This influence has been characterized as priming, and more recently as a memory process that biases perception on independent multiple timescales on the order of seconds to minutes (e.g. Brascamp et al., 2008; Pastukhov Braun, 2008). Cue recruitment studies show that learned bias can last for days, and that reversing it requires extensive retraining (Haijiang et al., 2006). These very-long-term biases are measured using trial sequences of disambiguated and ambiguous stimuli, so it is important to understand how recent stimulus presentations interact with such biases on a given trial. We investigated short-term temporal influence using rotating Necker cubes. Observers viewed a randomly permuted sequence of 480 trials, each consisting of a 2-sec movie depicting a rotating cube just above fixation. The sequence included 320 “Training” trials (rotation direction disambiguated by stereo and occlusion cues) randomly interleaved with 160 “Test” trials (no disambiguating cues). Observers reported whether a dot moving right-to-left or left-to-right through fixation on each trial moved in the same direction as the front or the back of the cube. Reported rotation on Test trials correlated positively with reported rotation for both recent Test trials and recent Training trials. Correlations were highest with the immediately preceding trial, diminishing for earlier trials. This temporal pattern of correlations indicates an active influence from recent trials. The short-term influence of the Test trials was at least as great as that of the Training trials. It therefore seems likely that a perceptual decision itself can affect perceptual decisions on subsequent trials. Some observers developed a bias to perceive all Test trials in one direction, suggesting an interaction of recent history with an internal state variable that governs a long-term perceptual bias