TY - JOUR T1 - Quantifying the importance of guessing, decision noise, and variable precision in explaining behavioral variability in visual perception JF - bioRxiv DO - 10.1101/153650 SP - 153650 AU - Shan Shen AU - Wei Ji Ma Y1 - 2017/01/01 UR - http://biorxiv.org/content/early/2017/06/22/153650.abstract N2 - When presented with the same sensory stimuli and performing the same task, people do not always make the same response. Such behavioral variability can have different causes, including sensory noise, decision noise, and guessing. In addition, recent work has proposed that the precision of sensory encoding is itself variable, both driven by the stimulus (heteroskedasticity) and independent of the stimulus. We analyzed data of 3 published and 8 new visual decision-making tasks with a single relevant feature, orientation. In modeling each experiment, we considered four factors: guessing (lapses), decision noise, orientation-dependent variable precision (oblique effect), and orientation-independent variable precision (inspired by visual working memory models). Modern computational power allows us to test all combinations of these factors; in a given model, each factor could be present or absent. To quantify the importance of each factor in explaining the data, we introduce three metrics: factor knock-in, factor knock-out and factor posterior probability. Across all experiments, we found strong evidence for guessing and for orientation-dependent variable precision. We found evidence for decision noise in only one experiment, and for orientation-independent variable precision only when distractors are variable across trials. On a methodological note, the factor importance metrics can be applied widely in factorial model comparison. ER -