Perceptual decision making with dynamic stimuli

People routinely categorise objects, images and other people’s movements. Research in perceptual decision making investigates the processes underlying such categorization based on continuous sensory input. Recent experimental and theoretical work presented evidence that neurons in lateral intraparietal area in monkeys accumulate evidence until a threshold is crossed and a decision is made. One key assumption about this accumulation is that the evidence is constant across time with additional perturbations by noise. However, this assumption excludes modelling of many real-life stimuli which display considerable variation across time. For example, the goalie in a penalty kick initially receives very little evidence for his decision to either jump left or right, or stay in the centre, but as the penalty taker approaches the ball his movement increasingly reveals the future flight path of the ball. Here I propose a novel computational approach which can appropriately model perceptual decision making using such highly dynamic stimuli. This method can be used to infer decision-relevant internal variables of subjects, such as their certainty about the stimulus. We demonstrate the method by quantitative fitting of behavioural data from a decision making experiment. Furthermore, the method can be used to derive novel hypotheses for experiments involving dynamic stimuli, as e.g. used in somatosensory, action observation, or auditory speech experiments