Texture and haptic cues in slant discrimination:  reliability-based cue weighting without statistically optimal cue combination

Abstract

A number of models of depth-cue combination suggest that the final depth percept results from a weighted average of independent depth estimates based on the different cues available. The weight of each cue in such an average is thought to depend on the reliability of each cue. In principle, such a depth estimation could be statistically optimal in the sense of producing the minimum-variance unbiased estimator that can be constructed from the available information. Here we test such models by using visual and haptic depth information. Different texture types produce differences in slant-discrimination performance, thus providing a means for testing a reliability-sensitive cue-combination model with texture as one of the cues to slant. Our results show that the weights for the cues were generally sensitive to their reliability but fell short of statistically optimal combination-we find reliability-based reweighting but not statistically optimal cue combination.

Mass and density estimates contribute to perceived heaviness with weights that depend on the densities' reliability

“…While most authors agree that the weights depend on reliabilities, there is some controversy over the situations in which weights are set "optimally" [16]. Both optimal [10,15] and suboptimal reliabilitydependent weighting have been observed [16]. We suggest to model heaviness perception as a corresponding integration of mass and density information.…”

Exploratory pressure influences haptic shape perception via force signals

“…Oruç, Maloney, and Landy (2003) described in detail optimal integration for correlated estimates. In any case, both optimal (Alais & Burr, 2004;Ernst & Banks, 2002;Hillis et al, 2004) and suboptimal (Gepshtein, Burge, Ernst, & Banks, 2005;Rosas et al, 2005) reliability-dependent weightings have been observed.…”

“…Such a shift has been observed when external noise added to the display of one signal lowered both the reliability of that estimate and its weight (Backus et al, 1999;Young et al, 1993). Whereas most authors agree that an estimate's weight depends on its reliability, the circumstances that allow for weight choices that optimize the overall percept are still a topic of discussion (Rosas, Wagemans, Ernst, & Wichmann, 2005). According to the maximum likelihood estimation (MLE) model, the reliability R s of a percept is highest and the weights w i are optimal if the weights are proportional to the reliability of the signal estimates R i 1/ inconsistent signals could be predicted from a weighted average of the curvatures indicated by the position and force signals.…”

Bayesian reverse-engineering considered as a research strategy for cognitive science

“…Rosas et al 2005;Rosas et al 2007). Although an intuitive reaction may be to deny that the observed behavior is optimal, an alternative reaction is to argue that the task environment does not present itself in the same way to the subject as it does to the investigator.…”