Abstract
How do people make perceptual decisions? One critical issue in answering this question is understanding how stimulus information is combined with non-perceptual information. For example, the perceptual information from a noisy grating can be combined with the non-perceptual experimenter-provided information that left (vs. right) orientation is more likely on a particular trial. This type of combination has been extensively studied via pre-cuing paradigms in which subjects are provided with a predictive cue followed by the stimulus itself. According to modeling work using drift diffusion, the subject incorporates the predictive cue by changing the starting point of accumulation. According to signal detection theory, the subject incorporates the predictive cue by shifting her decision criterion. To test the plausibility of these models, we compared the standard pre-cuing paradigm with a post-cuing design where the predictive cue comes after the stimulus. Surprisingly, post-cuing had a much larger influence on subjects' decision than pre-cuing. This effect was large (criterion shift of 0.46 for pre cues vs. 0.82 for post cues; p = 0.00002) and robust (larger effect for post cues in 25/30 subjects). Further, reverse correlation analyses demonstrated that when subjects were prevented from giving an immediate response, both pre- and post-cuing did not alter the timecourse of stimulus processing. These data falsify the drift diffusion model (which has no proper way of dealing with post cues and predicts timecourse changes for pre cues), as well as both signal detection theory and probabilistic population codes (both of which predict equal criterion shift for pre and post cues). Instead the data are readily explained by a model sensitive to the order of information integration where newer information is given higher weight. This model further accounts for the ubiquitous finding that subjects in pre-cuing paradigms shift their criterion too little and thus fall short of optimality.
Meeting abstract presented at VSS 2017