Self-training of dynamic touch: Striking improves judgment by wielding

Abstract

In traditional theories of perceptual learning, sensory modalities support one another. A good example comes from research on dynamic touch, the wielding of an unseen object to perceive its properties. Wielding provides the haptic system with mechanical information related to the length of the object. Visual feedback can improve the accuracy of subsequent length judgments; visual perception supports haptic perception. Such cross-modal support is not the only route to perceptual learning. We present a dynamic touch task in which we replaced visual feedback with the instruction to strike the unseen object against an unseen surface following length judgment. This additional mechanical information improved subsequent length judgments. We propose a self-organizing perspective in which a single modality trains itself.

Self-trained perception need not be veridical: striking can exaggerate judgment by wielding and can transfer exaggeration to new stimuli

“…Judgments diminished further over Block as well as over Trial within blocks (Bs = −14.15 and −1.45, SEs = 5.33 and .71, both Ps < .05, respectively). As found before (Stephen & Arzamarski, 2009), discrepancy became more negative with greater LogI 1 (B = −3.91, SE = 1.03, P < .001), but this negative relationship diminished with Block and with Trial (Bs = 1.57 and .17, SEs = .44 and .07, Ps < .001 and .05). Also as found before (Stephen & Arzamarski, 2009), striking increased judgments of stimulus objects with greater LogI 1 (Strike × LogI 1 ; B = .79, SE = .30, P < .01).…”

“…Use of freely unconstrained wielding also replicated the method in from Stephen and Arzamarski (2009) as well as the general convention in dynamic-touch research to leave wielding behaviors wholly unconstrained (Arzamarski, Isenhower, Kay, Turvey, & Michaels, 2010). This difference presents a possible confound, but attempts to make free wielding and striking more comparable would also risk introducing new information beyond experimental control.…”

“…Previous research had found that the direction (i.e., left or right) of horizontal striking movements did not change participants' judgments (Stephen & Arzamarski, 2009), and so the experimenters kept only horizontal movements to the right for striking in this study. Use of freely unconstrained wielding also replicated the method in from Stephen and Arzamarski (2009) as well as the general convention in dynamic-touch research to leave wielding behaviors wholly unconstrained (Arzamarski, Isenhower, Kay, Turvey, & Michaels, 2010).…”

“…We might expect a roughly cubic trajectory of discrepancy between judged and actual length with discrepancies dropping to zero after initial transients of incorrect guesses and overcorrection. To begin with, we expected underestimation at the beginning (e.g., Stephen & Arzamarski, 2009), and then we expected regression to the mean (i.e., towards zero-discrepancy) as participants took advantage of striking the unseen surface. At a certain point, participants might overcorrect, past zero-discrepancy and into positive discrepancy, and then temper their overcorrection by settling back towards zero-discrepancy.…”

“…Prior work tested whether wielding and striking unseen objects would inform judgments of the objects' length (Stephen & Arzamarski, 2009). This work found that participants generally underestimated the veridical length, with discrepancies between judged and actual length following a negative linear relationship with the logarithm of the first moment of inertia (LogI 1 ).…”