Disentangling Perceptual and Process-Related Sources of Behavioral Variability in Categorization

People often categorize the same object variably over time. Such intra-individualbehavioral variability can be difficult to identify, because it can be confused with abehavioral bias in binary data and it can originate in different stages of the cognitivecategorization mechanism. This article provides an overview of possible sources ofbehavioral variability in categorization, focusing on perceptual and cognitive processes. Todisentangle various sources of behavioral variability, a simulation with a similarity-basedcategorization model was run. The simulation shows that noise during perceptual andcognitive processes can lead to considerable misestimations of a response determinismparameter that tries to capture the variability in the responses. However, the binarycategory responses cannot discriminate between the different sources of variability, as theyall lead to similar response patterns across the stimuli. The continuous model predictions,in turn, can identify the source of the variability: Noise during feature perception leads tovariable predictions for central stimuli close to the category boundary. Noisy attention tothe features, in turn, increases the prediction variability for stimuli differing from eachcategory on another feature. Finally, more general noise during similarity computationincreases the prediction variability for the stimuli with moderate predictions. Measuringpeople’s category beliefs in a continuous way may therefore help disentangle perceptual andprocess-related sources of behavioral variability. Ultimately, this can inform interventionsaimed at making human categorizations more precise in specific domains by indicatingwhich steps of the categorization mechanism to target.