Our perceptions occur only after the information we receive from our senses has been heavily preprocessed. The degree to which cognitive influences shape this processing has been debated for many years. Sensory saltation is a class of robust illusions that has been described as a shift in the perceived location of a stimulus toward the location of a rapidly delivered subsequent stimulus . Our studies show that this explanation of the illusion is flawed and demonstrate instead that the two perceived stimuli are both mislocalized, and are symmetrically shifted towards one another. The misunderstanding apparently arose because the powerful effect of attention on the illusion was not recognized. A new understanding implicates temporal integration as the mechanism of the cutaneous saltation illusion, and reveals a fundamental role for attention in the basic integrative processes underlying skin sensation.
Geldard and colleagues, who discovered sensory saltation, worked primarily on the tactile version . In their paradigm a tap, that serves as a reference point, is delivered 500 msec before the presentation of two experimental taps. The second tap is then delivered at the same site on the forearm followed by a third tap delivered at a site 10 cm away. The reference tap is separated in time such that it does not influence the mislocalization phenomenon. Geldard reported the second tap is perceived to be nearer to the third tap (which is properly localized) as a roughly linear function of interstimulus interval (ISI), as the ISI is decreased below 200 msec. With ISI's of 100 or 20 msec, the second tap is perceived to be halfway between the actual locations of the taps or at the same location as the third tap (i.e. perceptually displaced by 10 cm), respectively.
A fundamental problem with Geldard's explanation of sensory saltation has been debated for two decades: How could a decision be made about where to locate a tap if the tap that 'attracts' it has not yet occurred [3,4]? Our findings demonstrate that this apparent paradox arose from a misunderstanding of the illusion that resulted from the use of trained observers to eliminate the high degree of variability in the reports of the illusion made by naive subjects. In removing this variability, two important aspects of the illusion were eliminated.
In our experiments, a fourth stimulus is added at the second stimulus site, again well separated in time from the two experimental taps. In this simple variation of Geldard's paradigm, naive subjects are able to consistently and precisely estimate the perceived separations between the two rapidly delivered experimental taps. The illusion is not perceived by naive subjects as one tap mislocalized toward the second, but rather, as two taps perceived to be located closer to one another as the ISI is decreased. Thus, the perceptions of both points shift toward each other as a function of ISI (Fig 1a).
While there is a consistent relationship between ISI and perceived separation, the perceived locations of the two taps between the two reference points were quite variable. Figure 1b shows that the center points between the two perceived taps were distributed as a roughly Gaussian function centered at a point midway between the two actual stimulus locations.
The source of the substantial variability in locating the two events on the skin was apparent when eight naive subjects were instructed to "concentrate on the proximal [or distal] region of the forearm" (Fig 1a). When so instructed, the perceived separations between the two taps do not change significantly, but the perceived skin locations of the two taps are dramatically shifted toward the attended region. The average difference between the centers of the perceived locations when subjects attend to the proximal versus the distal forearm is 31mm ± 4.7 (mean ± s.e., p<.0001). It should be noted that this difference is the result of very simple instructions given to naive subjects and there is no way to ensure that the subjects are attending as instructed on every trial. Thus 31mm is likely to be an underestimate of the effect of attention. With practice the effect can result in a shift as large as 80 mm.
Thus, although the perceived separation is determined by the actual sparation and the ISI; the perceived location of the two stimuli is largely determined by a subject's selective attention. We conclude that in their training Geldard's subjects were effectively taught to concentrate on the second location to reduce response variability.
A better understanding of the cutaneous saltation illusion gives us insight into tactile perception. The mechanisms that generate cutaneous saltation (and possibly related phenomena in vision and audition) operate as a smoothing process by which the perceived distances between two events occurring within a short time domain are generated by symmetric convergence, with locations powerfully biased by ongoing cognitive estimates of where those events are most likely to have occurred
Michael P. Kilgard
Michael M. Merzenich
Keck Center for Integrative Neuroscience
University of California
San Francisco, California 94143 USA
1. Geldard, F.A. & Sherrick, C.E. Science 178, 178-179 (1972).
2. Geldard, F.A. Sensory Saltation (Lawerence Erlbaum Assoc. Publishers, NY, 1975).
3. Goodman, N. Ways of Worldmaking p. 74 (Hackett Pub Co., Indianapolis,1978).
4. Dennett, D. C. Consciousness Explained, (Little, Brown and Company, Boston, 1991).
FIG. 1 (a) Perceived stimulus separation as a function of ISI. The 13 naive observers were unaware of the experimental set-up, with the arm hidden from view. After ten repetitions of each ISI, subjects reported perceived stimulus locations by drawing the positions of the two experimental taps, with the locations of the two reference taps provided. Bars = s.e. R =.97 ISI's marked with an * are significantly different from each other with p< .01 (b) Distribution of center points of tap pairs. Pooled data for 30- 60 msec ISI's. Mean = 52.6mm, SD = 24.1mm. (c) Representative examples of the perception of the four stimuli with an ISI of 60 msec when subjects attend with or without specific instructions about attention. Filled circles represent actual stimulus locations; stippled circles represent perceived stimulus locations.