The vibrations of texture

Abstract

The Pacinian channel has been implicated in the perception of fine textures (Hollins et al., Somatosens Mot Res 18: 253-262, 2001a). In the present study, we investigate candidate codes for Pacinian-mediated roughness perception. We use a Hall effect transducer to record the vibrations elicited in the skin when a set of textured surfaces is passively presented to the index finger. The peak frequency of the vibrations is found to decrease systematically as spatial period increases. The power of the vibrations--weighted according to the spectral sensitivity of the Pacinian system--increases with spatial period for all but the coarsest surfaces. By varying the scanning velocity, we manipulate the temporal and intensive characteristics of the texture-induced vibrations and assess the effect of the manipulation on perceived roughness. We find that doubling the scanning velocity does not result in the substantial decrease in roughness predicted by a frequency theory of vibrotactile roughness perception. On the other hand, the effects of speed on roughness match those of speed on power. We propose that the roughness of a fine surface (spatial period<200 microm) is a function of the Pacinian-weighted power of the vibrations it elicits.

Figure 1

Experimental setup. As the belt moves the surface across the finger, texture-induced vibrations travel a short distance up the finger from the plane of contact with the surface, setting the magnet in motion. The Hall effect transducer (HET) produces a small current inversely proportional to the separation between itself and the magnet. At rest, the separation between the magnet and HET is approximately 5 mm. The signal is then amplified and digitized.

Figure 2

Mean scanning force as a function of spatial period for subjects S and M (bars show standard error of the mean). Subjects appeared to apply approximately the same amount of force regardless of which surface was presented.

Figure 3

Records obtained from the 124 and 416 μm surfaces (subject S). The peak-to-peak amplitudes of the vibrations are approximately 4 and 11 μm for the 124 and 416 μm records, respectively.

Figure 4

Significant differences between the mean Fourier spectrum of six of the textured surfaces and the null spectrum, divided by Pacinian threshold (arbitrary units). Data for subject M are shown. The frequency of the largest spectral components tends to decrease as the spatial period of the surface increases. Note that, since the peak amplitudes tended to increase with spatial period, the ordinates are scaled accordingly. Arrowheads denote the natural frequencies of the surfaces.

Figure 5

Peak frequency vs spatial period. Peak frequency tends to decrease as spatial period increases: for all but the finest surfaces, the observed peak frequency coincided with the natural frequency. Some evidence suggests that the exceptions are due to low signal-to-noise ratios for the finest surfaces. (The missing data correspond to conditions in which there were no significant differences between the texture spectra and the null spectra. The signal-to-noise ratio was too low for the differences between the two sets of spectra to be resolved by means of statistical inference.)

Figure 6

Median power and mean roughness vs spatial period. Power was normalized and corrected for baseline noise as described in the text. Power is plotted on a logarithmic scale and roughness on a linear scale because we find roughness to increase with the log of power (see General discussion).

Figure 7

Predictions derived from the frequency and intensive theories of roughness perception. The frequency theory predicts that the roughness of both surfaces will decrease as speed is doubled. The intensive theory predicts that the roughness of the coarse surface will increase while that of the fine surface will decrease as belt speed is doubled. In statistical terms, the frequency theory predicts a main effect of speed while the intensive theory predicts a surface×speed interaction.

Figure 8

Experiment 3 results. Median power and mean roughness vs belt speed.

Figure 9

Roughness vs P_adj from Experiments 2 and 3. Only data from surfaces with spatial periods<200 μm are included from the results of Experiment 2. The model (roughness = 1.3+0.76·log(P_adj)) accounts for 72% of the variance in the data (F=94.52, p<0.001).