Abstract
Our visual environment is relatively stable from one moment to the next. Even if the current visual input is noisy, it is generally likely to be similar to the previous input. Incorporating prior and present information is therefore advantageous. Studies of serial dependence reveal attractive biases towards previously attended stimuli that stabilize noisy visual input, but their strength scales with the uncertainty in the stimulus: With greater uncertainty comes greater serial dependence. More recent work shows that ignored visual information in the scene creates repulsive biases that can also serve to optimize perception. Here, we investigate the role of uncertainty on both types of bias. In previous studies, attractive biases have been found to scale with uncertainty through manipulation of the sensory noise in the stimulus, but two questions remain unanswered: Firstly, how does integration noise (here distractors) influence the attractive bias? Secondly, does the repulsive bias introduced by ignored information also scale with uncertainty? We designed two experiments in which either sensory noise (contrast) or integration noise (orientation variability of distractors) was manipulated. This creates high and low uncertainty trials. For a set of 2 or 3 trials, observers searched for an oddly-oriented line and were subsequently asked to adjust a single line so that it matched the previous target. Uncertainty was either kept constant during search trials or changed on the last search trial. Afterwards observers were asked to judge their confidence about their adjustment response. Preliminary results show that not only sensory noise, but also integration noise, affected the strength of the attractive bias while these effects were not found for the repulsive bias. The repulsive bias therefore appears not to be influenced by uncertainty. Confidence judgments were reduced when uncertainty changed during search trials, independent of the type of noise.