Touch Precision Modulates Visual Bias

ABSTRACT

The sensory precision hypothesis holds that different seen and felt cues about the size of an object resolve themselves in favor of the more reliable modality. To examine this precision hypothesis, 60 college students were asked to look at one size while manually exploring another unseen size either with their bare fingers or, to lessen the reliability of touch, with their fingers sleeved in rigid tubes. Afterwards, the participants estimated either the seen size or the felt size by finding a match from a visual display of various sizes. Results showed that the seen size biased the estimates of the felt size when the reliability of touch decreased. This finding supports the interaction between touch reliability and visual bias predicted by statistically optimal models of sensory integration.

Keywords:

• intersensory integration
• touch perception
• visual perception
• size perception

ACKNOWLEDGMENTS

Results were presented at the 89th Annual Meeting of the Midwestern Psychological Association (Chicago, IL). The authors thank S. R. Baird, F. P. Carpinelli, and E. I. Moreno for their helpful comments on an earlier version of the article.

Notes

1. A preliminary manipulation-check study examined whether haptic estimates of the felt size would be more accurate and less variable if the size were felt with the bare fingers rather than with the fingers through one of four intervening fabrics: thick wool (100%), plush felt (100% polyester), heavy corduroy (85% polyester and 15% nylon), and thin rayon (100%). Forty-five students, assigned to one of the five randomized haptic groups, used only touch to feel squares and to render their estimates. The plastic squares were all 5 mm thick but 40, 45, 50, 55, and 60 mm on a side. Each was suspended in midair in the inspection box so that its edges could be actively explored with all fingertips of the writing hand, and each participant practiced judging a randomized order of the five squares three times. After they explored each square for 5 s, they estimated its felt size by selecting a match from a set of haptic comparison squares, which ranged on a side from 20 to 75 mm in 5-mm increments. Participants reached into a draped box located next to the inspection box (see Figure 1A) to manually find a match from the horizontally arrayed sizes. Their matches varied with the size of the square (ω² = .93) but not with the type of fabric, F(4, 40) < 1, ω² = −.04, nor with practice (ω² = −.02). For example, the average size estimate for the bare fingers group was 49.85 mm (SD = 7.43 mm). Similarly, the group means for the thick wool, plush felt, heavy corduroy, and thin rayon fabrics were, respectively, 48.70 (SD = 8.46), 49.52 (SD = 8.01), 50.00 (SD = 7.58), and 49.23 (SD = 7.86). These findings suggest that Derrick and Dewar's (1970) fabric manipulation may have been too weak to affect haptic size perception (M_d = 0.07, SD_d = 0.05).

2. Van Doorn et al. (2010) showed that adding cutaneous to kinesthetic cues can shift the estimates of discordant haptic-visual size in favor of touch. Thus, another experiment assessed whether impairing both cues with rigid PVC tubes sleeved over the fingers would affect the haptic estimates of the felt size. Thirty-two students, assigned to one of two randomized groups, felt five squares (see Note 1) either with bare fingers or with rigid cylinders over the fingers (see Method). Each participant actively grasped the edges of each square with all fingertips for 5 s. Afterward, they estimated the felt size by finding the best match from a set of haptic comparisons (see Note 1). All matching responses were made with the naked fingers, and each participant practiced judging a randomized order of the five sizes three times. The estimates of the sizes felt with rigid tubes were consistently smaller and less precise than were those felt with bare fingers. Specifically, the mean size estimate for the bare fingers group was 50.65 mm (SD = 8.34 mm; 95% CI [49.58, 51.71]) and for the tubed fingers group was 46.58 mm (SD = 10.05 mm; 95% CI [45.30, 47.86]), F(1, 30) = 11.01, p = .002, ω² = .24. In effect, the average estimate of the tubed fingers group was below 67% of the estimates of the bare fingers group (Cohen's d = −0.44).

3. After completing the main experiment, 40 participants only viewed and only felt each individually presented square for 5 s. Then they found a match for the felt size and the seen size from the visual comparisons. These control estimates were subjected to a 2 Dummy Group × 2 Size × 2 Modality mixed ANOVA. It showed negligible statistical evidence of preexisting group differences, F(1, 38) = 1.08, p = .30, ω² < .01. This result raises confidence in the observed group by modality interaction as due to treatment effects rather than pre-existing response biases.