Abstract
Following a fine direction-discrimination task, subjective reports of the motion direction can be systematically biased away from an oriented reference line. A decoding model that preferentially weights the responses of mechanisms tuned to directions away from the decision boundary can quantitatively account for such repulsion effects (Jazayeri & Movshon, 2007). However, an alternative possibility is that this repulsion effect is due to a bias in subjects' responses. We investigated this by manipulating two aspects of a global motion discrimination task: (1) the presence of a reference line and (2) the modality used to obtain reports of motion direction. In the first experiment, subjects performed a fine-discrimination task while a reference line was present. Subsequently, they estimated the perceived direction of motion by manually adjusting the orientation of a test line either in the presence or absence of the reference. When the reference was present during the estimation phase, reports of motion direction were systematically biased away from the reference. Importantly the same pattern of responding was found even when subjects were not asked to perform the initial fine-discrimination task. Furthermore, when subjects estimated motion direction in the absence of the reference line, their responses were veridical. In a second experiment, subjects reported the perceived direction of motion in the presence of a reference line by making a saccade. Surprisingly, under these conditions, repulsion from the reference line was greatly reduced. These findings are hard to reconcile with the idea of weighted decoding. As the repulsion effect depends on the presence of an explicit reference when a manual adjustment task is used, it most likely reflects a task-dependent bias in the subjects' responses.
Meeting abstract presented at VSS 2016