Abstract
3D information for perception and action is provided by multiple depth cues. Previous studies have shown that the combined estimate from conflicting cues can be changed by training where external feedback (e.g. haptic feedback) is constantly consistent with the same cue (the reinforced cue) across trials. Such training will shift the combined estimate toward the reinforced cue estimate. Maximum Likelihood Estimation (MLE) models of cue integration postulate that estimates from individual cues are unbiased and combined through a weighted average, where weights are proportional to the corresponding cue reliability. Such models interpret the results of the training as an increased weight of the reinforced cue. Hence the change is termed cue-reweighting. Here we studied whether these results are compatible with an alternative theory of cue integration, the Intrinsic Constraint (IC) theory, which does not assume unbiased depth estimates from individual cues. This theory instead predicts single-cue remapping from the training: the perceptual mapping between distal depth and estimated depth of the reinforced cue increases, and the perceptual mapping of the less trusted cue diminishes. To test this alternative hypothesis, we replicated previous studies with participants grasping curved surfaces along the depth dimension. These surfaces were defined by texture and disparity rendered with various levels of cue conflict (-8, -6, -4, 4, 6, 8 mm). Haptic feedback was always congruent with one of the cues. Differently from previous studies, participants also judged the perceived depth of the stimulus through Manual Size Estimation (MSE) before each grasp. This critical test showed results compatible with the predictions of the IC theory: the change of MSE during training suggests single-cue remapping, a result that cannot be explained by cue-reweighting. Conversely, we speculate that the cue-reweighting obtained with a traditional matching task, which was also adopted in this study, may instead indicate single-cue remapping.