Recent work on sensory cue integration suggests that given multiple sources of possibly noisy information about the world, human behavior can often best be described as near-optimal relative to an ideal Bayesian observer. That is, humans take into account the uncertainty of each source of information as well as prior information (e.g., knowledge of statistics of the world) and infer the most likely scene that gave rise to the sensory signals. Most recent work on sensory cue integration has centered on visual and auditory cues. I will describe several ongoing studies involving spatial localization using haptic (touch) and proprioceptive (body sense) cues. To interact with or direct gaze toward an object that has touched the skin requires haptic-proprioceptive integration. The haptic stimulus allows the observer to know which skin receptors were innervated. To localize the stimulus relative to the corresponding body part involves an internal body representation. To further localize the stimulus in external space requires integration with the body pose (estimated using motor corollary feedback or proprioceptive input). We describe several experiments in which the unseen hand is moved to a new location from trial to trial and haptic stimulation (a small “buzz”) is applied to one of several locations on the back of the hand. Participants estimate the location of the hand or the haptic stimulation relative to a visual scale projected on a platform above the hand. Ideally a vector sum is computed of the position of hand relative to the body plus the position of the haptic stimulus on the hand. We find evidence of integration of prior expectations with the two uncertain current stimuli: Location estimates are biased to the middle of the workspace and middle of the hand for weaker stimuli. These biases propagate to the integrated location estimate. Because the effect of the tactile prior is stronger, haptic location is underweighted in the vector sum. Additional experiments examine whether haptic/proprioceptive integration is mandatory.
http://www.cns.nyu.edu/~msl/papers/landyetal11.pdf(Citation: Landy, M. S., Banks, M. S. & Knill, D. C. (2011).
Ideal-observer models of cue integration. In Trommershäuser, J., Körding, K. & Landy, M. S. (Eds.), Sensory Cue Integration (Ch. 1, pp. 5-29). New York: Oxford University Press.)