Coding of Roughness, The
This review examines the way information about textures is captured, encoded, and processed by the somatosensory system to produce sensations of roughness/smoothness. Textures with spatial periods exceeding about 200 µm are encoded spatially, so roughness is nearly independent of the speed and direction of their movement across the skin. The information consists of spatial variations in activity among slowly adapting (SA1) mechanoreceptors, and appears to be extracted by specialized cortical neurons. Perception of the roughness of finer surfaces is mediated by detection, primarily by Pacinian afferents, of cutaneous vibrations generated when textures move across the skin. Movement is necessary to the perception of these textures, and vibrotactile adaptation interferes with it. The code is an intensitive one (i.e., the amount of activity in Pacinian afferents).
Résumé Dans cette revue, nous examinons comment le système somatosensoriel extrait, encode et traite l’information tactile pour produire des sensations de rugosité. Les textures dont la période spatiale dépasse 200µm sont encodées de manière spatiale: la rugosité est quasiment indépendante de la vitesse et de la direction du mouvement des textures sur la peau. L’information est contenue dans la variation spatiale de l’activité d’une population de mécanorecepteurs à adaptation lente (Type 1) et semble être extraite par des neurones corticaux spécialisés. La perception de rugosité pour les surfaces plus lisses dépend en revanche de la détection, principalement par les afférents Paciniens, des vibrations cutanées engendrées par le mouvement des textures sur la peau. Ce mouvement est nécessaire pour percevoir les textures, et l’adaptation vibrotactile interfère avec cette perception. Le code neuronal est un code d’intensité, c’est-à-dire qu’il exprime l’activité totale des afférents Paciniens.
Everyone uses the term roughness, but no one knows exactly what it means. For example, if we haptically examine a piece of fine sandpaper and a page of Braille, we cannot say for certain which is rougher until we are told what aspect of the surfaces to attend to, because the sandpaper and the Braille feel rough in different ways. This complexity of the roughness experience suggests that there is more than one physiological mechanism or code for using tactile information to create the perception of roughness.
It would not be surprising if this were so. Physical properties of tangible surfaces that correlate with their roughness (especially the size and spacing of texture elements) vary over wide ranges. In other sensory systems, it has been found that encoding mechanisms that work over one part of a stimulus dimension (such as light intensity) may not work over another part of the same dimension. In a number of cases, this problem has been solved by having separate mechanisms process information from different parts of a dimension (e.g., rods for low- and cones for high-light intensities). Other examples are the use of a time code and a place code for auditory pitch, and interaural time and intensity cues for binaural localization, over different parts of the sound frequency spectrum.
Recent research has shown that at least two codes exist for roughness as well: A spatial code that works for coarse and medium surfaces, and a vibrotactile code that works for very fine surfaces (with spatial periods – the centre-to-centre distance between texture elements – less than about 200 pm). These codes depend on signals that arise in different classes of mechanoreceptors and are processed using different cortical algorithms.
In the following sections, we first review the psychophysical evidence for the existence of these two roughness codes, and then explore the physiological mechanisms underlying each code. Finally, we ask whether the two codes can work together in some situations.
Psychophysical Evidence for Two Codes The Role of Movement
Rubbing a surface is the best way to examine its texture, and people spontaneously execute this exploratory procedure when asked to make textural judgments (Lederman & Klatzky, 1987). Yet some information about texture (including roughness) is often obtained when the finger is merely touched to a surface, without lateral movement. What information about a surface is eliminated by static contact, and what survives? To answer this question, Hollins and Risner (2000) used a two-interval, forced-choice (2IFC) procedure to measure the ability of subjects to discriminate textures. The subject lowered his/her finger onto each surface when instructed to do so by the experimenter, and raised it again 1 s later. The lowered finger, once in contact with a surface, was held motionless by the subject, but in some blocks of trials the experimenter introduced movement by drawing the surfaces along the fingertip. Subjects were asked to compare the two surfaces presented on each trial and indicate which was smoother.
On some trials, the two abrasive-paper surfaces had very fine textures, while on other trials, they were both coarse. The two papers making up each pair were slightly different, the one with the larger average spatial period feeling rougher when touched normally. In analyzing the data, we considered a correct response to be one in which the paper with a lower mean spatial period was called smoother.
With the coarse surfaces, performance was equivalent whether the surfaces moved across the skin, or were held in stationary contact with it. Movement enhanced the roughness of both surfaces, but by the same factor, so in accordance with Weber’s law they were equally discriminable under the two conditions of presentation.
This result differs from that reported by Morley, Goodwin, and Darian-Smith (1983), whose subjects were asked to discriminate coarse gratings, and who showed somewhat keener discrimination when lateral movement of the finger was allowed. Subjects were given up to 10 s per trial to examine the surfaces, and may have spent more time doing so in the movement condition, thereby giving it an unintended advantage. Still, the performance of Morley et al’s subjects was good even without lateral movement (Weber fraction of 10%), showing that it is not crucial to the discrimination of coarse surfaces.
Spatial Coding of Coarse Surfaces
That movement plays, at most, a secondary role in the discriminability of coarse surfaces is consistent with the spatio-intensive theory of Lederman and her colleagues, who demonstrated, in a series of now-classic studies (Lederman, 1974; Lederman, Loomis, & Williams 1982; Lederman & Taylor, 1972; Taylor & Lederman, 1975), that the roughness of gratings is determined by the spatial – rather than the temporal – aspects of their biophysical interaction with the hand. The dependence of roughness on groove width (strongly positive) and ridge width (weakly negative) can be well accounted for by a biophysical model in which roughness is proportional to the volumetric displacement of skin from its overall mean level (i.e., the combined amount of skin that is depressed by ridges, or that bulges into grooves) (Taylor & Lederman, 1975). When the finger moves laterally, this pattern of displacement shifts across the skin, but (within limits) is unaffected by speed, a property of the model that accounts for the fact that speed has little effect on roughness (Lederman, 1974; Meftah, Belingard, & Chapman, 2000).
Lederman and Taylor’s (1972) discoveries regarding the effects of grating spatial parameters have been confirmed by other labs (Cascio & Sathian, 2001; Sathian, Goodwin, John, & Darian-Smith, 1989; Yoshioka, Gibb, Dorch, Hsiao, & Johnson, 2001). Moreover, if groove width is considered a specific example of the spacing between texture elements, then the dependence of roughness on this variable has been found to apply also, within limits, to textures consisting of arrays of raised dots (Connor & Johnson, 1992; Meftah et al., 2000).
Interestingly, roughness is sometimes found to decline again when the spacing between adjacent dots becomes very large (i.e., greater than 2-3 mm) (Connor, Hsiao, Phillips, & Johnson, 1990; Klatzky, Lederman, Hamilton, Grindley, & Swendsen, 2003). Meftah et al. (2000), who did not find this secondary decline, suggested that it occurs only when texture elements are so short that the skin, bulging into the space between them, touches the smooth floor of the surface. This interpretation is consistent with research on indirect touch through a rigid probe (Klatzky & Lederman, 1999; Klatzky et al., 2003), a situation in which roughness is again an inverted-U (more specifically, a quadratic) function of interelement spacing; this work has shown that the peak of the function occurs just at the point where the probe is able to slip between texture elements and touch the floor of the surface.
A substantial and internally consistent body of evidence thus testifies to the centrality of spatial factors in determining the roughness of coarse textures. A somewhat different perspective, however, has been offered by Cascio and Sathian (2001), who re-examined the effect of movement speed on the roughness of gratings. They found that roughness slightly increases with speed, a result that could be attributed to changes in temporal frequency – the number of grating cycles moving over a spot on the skin every second. They tested this idea by asking subjects to discriminate between two gratings that had the same groove width but different ridge widths. Gratings were chosen that were moderately discriminable when they moved at the same speed. When the grating with wider ridges was moved slightly faster, thereby reducing the difference in temporal frequency between the gratings, discrimination grew worse. This result suggests that temporal frequency may have been the basis on which these gratings were discriminated. In contrast, even small differences in groove width between gratings made them easily discriminable, an effect that could not be cancelled out by changes in speed. Temporal factors thus have a modest effect on perceived roughness in some situations, but Cascio and Sathian point out that more research will be needed to determine whether this effect indicates the existence of a temporal code (i.e., the systematic extraction of useful information about surfaces). Alternatively, it may simply reflect the limits of the spatial code’s stability across a range of stimulating conditions.
In summary, results from several laboratories, and using a variety of methods, demonstrate that, for gratings or other textures with spatial periods greater than about 200 pm, discriminability is determined almost entirely by the spatial characteristics of the surface the size and spacing of elements. Lateral movement between the surface and the skin increases roughness magnitude, but is not necessary for texture discrimination.
Vibratory Coding of Fine Surfaces
A different picture emerged, however, when Hollins and Risner (2000) measured the discriminability of a pair of fine surfaces (mean spatial periods of 18 and 30 pm). Performance was at chance in the absence of movement, but nearly perfect when movement occurred. Something about lateral movement between the surfaces and the skin was crucial to the discrimination of fine textures, but irrelevant to the discrimination of coarse textures. Magnitude estimation data confirmed this distinction. With movement, roughness was a smooth power function of mean spatial period over the entire stimulus range (18-700 pm); without movement, the portion of the function below 200 pm became horizontal, implying that all fine surfaces felt the same. This experiment suggested that the transition from fine to coarse processing mechanisms occurs at a spatial period of about 200 pm.
How did movement enable discrimination of fine textures? We hypothesized that movement of the surface causes vibrations of the skin, and these, detected by the subject, form the basis of discrimination.
This idea of vibrotactile coding of fine surfaces is in fact an old one, going back at least as far as Katz (1925/1989), who conceived of vibrotaction as a distinct submodality of touch. Katz’s idea was viewed with skepticism for decades, but interest in it was renewed by the discovery (Bolanowski, 1998; Hollins, Bensmaïa, & Risner, 1998) that pre-exposure of the finger to an extended period of vibration could selectively reduce the roughness of fine surfaces. Perception of coarse surfaces was unaffected, a result consistent with earlier findings by Lederman, Loomis, and Williams (1982). The implication was that roughness-coding mechanisms were sensitive to (and could therefore be adapted by) vibration. But did this adaptation actually interfere with the perception of fine surfaces (i.e., reduce the precision with which textural information was extracted, so that roughness discrimination would be compromised) or simply adjust the “gain” of perception to optimize it for prevailing stimulus conditions?
To answer this question, Hollins, Bensma’ia, ^nd Washburn (2001) undertook a new series of experiments that combined adaptation with the 2IFC procedure. The surfaces used in this study were etched silicon wafers, with textures consisting of rectilinear arrays of truncated pyramids. Spacing and size of the pyramids were constant on any one surface, but varied proportionally from surface to surface so that the coarser surfaces had textures that were basically enlargements of the finer surfaces. Manufactured, regular surfaces have become the standard in roughness studies, but a unique feature of our stimuli was the great range of spatial periods represented: from 16 to 3,200 pm, a range of 200:1. This wide range was necessary for us to directly compare roughness codes for fine and coarse surfaces.
Hollins et al. (2001) began by selecting two sets of surfaces: a fine set, with spatial periods well below 200 pm, and a coarse set with spatial periods above this transitional value. We measured the ability to distinguish between pairs of surfaces in each set, and asked whether this ability was affected by pre-exposure to moderately intense (100 pm amplitude) 100-Hz vibration. Before each trial, the subject placed the index fingerpad on a vibrating (or, for the unadapted condition, stationary) contactor for 30 s, then moved it along each of two surfaces in turn, and indicated which of the two was smoother. Vibratory adaptation had no effect on performance with the coarse surfaces, but it rendered the fine surfaces indiscriminable (see Figure 1). The results support the view that perception of fine surfaces is mediated by systems that are highly sensitive to vibration.
Physiological Basis of Roughness
Physiological Mechanisms of Spatial Coding
In a series of studies pairing human psychophysics with macaque neurophysiology, Johnson and colleagues set out to establish the peripheral neural code underlying roughness perception (Blake, Hsiao, & Johnson, 1997; Connor & Johnson, 1992; Connor et al., 1990; Yoshioka et al., 2001). Their approach consisted of devising and testing a set of hypotheses linking the activity in populations of peripheral afferent fibers in macaque evoked by various textured surfaces to estimates of the perceived roughness of these same textures as measured in psychophysical experiments on humans. These estimates, obtained for a variety of dot patterns and gratings, were plotted against predictions derived from each putative neural code. A hypothesis was eliminated if it failed to account for roughness estimates in any given experimental condition. The putative neural codes for roughness included (but were not limited to): 1) the mean firing rate elicited in a given population of mechanoreceptive afferent fibers; 2) the temporal variability in the firing of a given population of mechanoreceptive afferent fibers; 3) the spatial variability in the firing of a given population of mechanoreceptive afferent fibers.
In the first study of the series, Connor et al. (1990) found, as explained earlier, that roughness estimates were an inverse U-shaped function of the centre-tocentre spacing between dots that peaked at spacings of about 3 mm. The mean rates evoked by these same stimuli in slowly adapting Type 1 (SAl), rapidly adapting (RA), and Pacinian (PC) afferents did not follow the same pattern as the roughness estimates, nor did simple measures of response variation (such as the variance in the firing rate). On the other hand, the temporal and spatial variation in SA1 and RA responses closely matched the psychophysical data.
Raised-dot patterns are potentially more complex than gratings in that they permit spatial parameters along and across the direction of movement to be manipulated independently. In a second study, Connor and Johnson (1992) made use of this additional degree of freedom to determine whether roughness depends on spatial rather than temporal variation. They used a set of textures in which the spacing of bumps in one direction was always the same, while spacing in the orthogonal direction varied from one surface to the next. In one set (the “horizontal” set), the dot spacing increased from surface to surface along the axis parallel to the scanning direction; in the second (“vertical”) set, the dot spacing increased along the axis perpendicular to the scanning direction. The two putative neural codes that had not been eliminated in the first study, namely, temporal and spatial variation in the afferent responses, yielded diverging predictions regarding the roughness of these surfaces. According to the temporal variation hypothesis, perceived roughness should increase with dot spacing for the horizontal set because the response of primary afferents waxes and wanes to a greater degree as dot-spacing increases. The same hypothesis, however, predicted that roughness would decrease as dot spacing increased for the vertical set, because an increasing proportion of the receptive fields of afferents would pass unstimulated between rows of dots. In contrast, the spatial variation hypothesis predicted that perceived roughness should increase for both stimulus sets; indeed, the spatial variation is computed over two dimensions and so should be affected to the same degree whether the stimulus configuration changes along one dimension or the other. Roughness estimates were found to increase with dot spacing for both stimulus sets, a pattern reflecting the spatial variation in SA1 and RA responses but not their temporal variation.
In a third study, Blake et al. (1997) set out to determine which of two populations of mechanoreceptive fibers, SA1 or RA, carried information about roughness (note that PC fibers had been eliminated in the first study). The stimuli consisted of embossed dots that varied in both height and diameter (in addition to centreto-centre spacing), as these two stimulus parameters had previously been found to modulate SA1 and RA responses differently (Blake, Johnson, & Hsiao, 1997). Specifically, SA1 responses are relatively sensitive to dot height and diameter while RA responses are not. Roughness estimates were found to be modulated by dot diameter and height in precisely the way predicted from the spatial variation in SA1 responses.
Using the method of elimination, then, Johnson and colleagues concluded that the perceived roughness of a texture is determined by the spatial variability in the SA1 response it evokes. Indeed, this neural code accounted for all the psychophysical data they collected over this series of studies, as well as those obtained in a subsequent study (Yoshioka et al., 2001).
Possible cortical mechanisms underlying the computation of spatial variability in the peripheral response were discovered in another set of experiments in which random dot patterns (DiCarlo & Johnson, 1999, 2000; DiCarlo, Johnson, & Hsiao, 1998) or spatio-temporal white noise stimuli (Sripati, Yoshioka, Denchev, Hsiao, & Johnson, 2006) were presented to the glabrous skin of the distal finger pads while responses evoked in SI neurons were recorded. The spatial and spatio-temporal receptive fields (SRFs and STRFs) of these neurons were computed from these measurements using reverse correlation. It was discovered that a subpopulation of neurons, whose SRFs or STRFs comprise excitatory and inhibitory subregions, effectively compute the spatial variability in the afferent input (see Figure 2). Furthermore, predictions derived from the SRFs of a subset of these neurons indicate that their responses to traditional embossed dot patterns would match the perceived roughness of those surfaces (A. Sripati, personal communication, December 5, 2004).
Although spatial variation seems to be computed in Area 3b, Area 1 has also been implicated in roughness perception (Randolph & Semmes, 1974), as have Area 5 (Roland, O’Sullivan, & Kawashima, 1998) and the second somatosensory cortex (SII) (Pruett, Sinclair, & Burton, 2001). However, it is not clear to what extent the roughness coding in these cortical areas is spatial in nature.
According to the spatial variation hypothesis, changes in force will impact perceived roughness to the extent that the spatial variation in the SA1 response changes. It has been shown that SA1 firing rates increase as local forces applied to the skin increase (Phillips & Johnson, 1981; Sripati, Bensma’ia, & Johnson, 2006). For many stimulus surfaces, this increased activation will lead to an increased spatial variation in the SA1 response and therefore to greater perceived roughness. However, it is important to keep in mind that these effects of increasing contact force will depend also on the spatial properties of the stimulus. For instance, a perfectly smooth stimulus, which evokes a spatially homogeneous response in SA1 fibers, will likely not be perceived as rougher if the contact force is increased because the spatial variation will not change. The fact that the perceived roughness of gratings increases as a function of contact force, and that this dependence becomes greater as groove width increases (Lederman & Taylor, 1972), is consistent with this hypothesis.
Physiological Mechanisms of Vibratory Coding
As we have seen, the roughness of fine surfaces is not spatially coded; instead, when judging a fine surface, we make use of our ability to detect the fine vibrations created when the surface moves across the skin (or vice versa). But what physiological mechanisms are involved in this? What receptoral channel or channels respond to the vibration, and what is the relationship between the properties of the vibration (such as frequency and amplitude) and the roughness of the perceived surface? To address these questions, Hollins et al. (2001) repeated their 2IFC discrimination experiment with the finely etched surfaces, using two different adapting vibration frequencies, selected for their ability to stimulate the RA channel (10 Hz), or the Pacinian channel (250 Hz). The higher frequency markedly compromised discrimination, but the lower frequency did not, results implicating the Pacinian system in texture perception.
A peculiarity of the Pacinian system is that it is virtually absent from the tissues of the lower face: Pacinian corpuscles are not found in the tissues here, and psychophysical detection thresholds for vibration show no evidence of Pacinian sensitivity (Barlow, 1987; Hollins, Delemos, & Goble, 1991; Verrillo & Ecker, 1977). It is not known what advantage this absence confers; perhaps the resulting low sensitivity of the orofacial region prevents the nervous system from being overloaded by the large vibrations generated by breathing, talking, and eating.
If vibrotactile coding of fine surfaces depends on Pacinian receptors, their absence from the face means that fine textures should be only crudely perceived here. We evaluated this prediction in an experiment reported here for the first time. Initially, we tried applying our etched surfaces to the face, but found them too harsh and unpleasant for repeated presentation. Instead, we used two pairs of brushes, one pair with thick bristles and one with thin. The two fine brushes were artist’s brushes with Taklon bristles having average diameters of 7 and 13 pm, respectively; when swept across the skin, they delivered flowing textures that were within the vibrotactile range, as defined earlier. The coarse brushes were toothbrushes with nylon bristles having average diameters of 159 and 234 pm, respectively, so that, allowing for some space between bristles, they produced textures within the spatial-coding range.
Brushes were trimmed so that in cross-section they measured 8 × 2 mm. The experimenter moved them across the skin for a distance of approximately 5 cm and with a speed of about 2 cm/s. Brushes were swept along the cheek from anterior to posterior, or (for comparison) from proximal to distal along the thenar eminence at the base of the thumb.
Each of the 10 subjects (university students) participated in four blocks of trials – fine brushes/face, fine/thenar, coarse/face, coarse/thenar – in a different random order. A block began with a few practice trials, followed by 25 archival trials, on each of which the two brushes being tested (i.e., either the two fine or the two coarse brushes) were presented in random order, with 5 s separating the two presentations. The subject then said whether the finer of the two had been presented first or second. Vision of the brushes was blocked.
Performance on the thenar eminence was comparable for the two sets of brushes, averaging 82.0% correct for the coarse pair, and 82.8% correct for the fine pair. On the face, however, performance was worse for the fine brushes (63.6%) than for the coarse ones (77.2%). An ANOVA showed that the main effect of region, the main effect of brush pair, and the interaction of these two factors were all significant (p
To further investigate the neural code underlying roughness perception for fine textures, we recorded the vibrations elicited in the skin when finely textured surfaces were scanned across the index finger. Vibrations were recorded using a Hall Effect Transducer (HET), a sensor that produces a small current proportional to the distance between it and a magnet. The magnet was glued to the skin of the index finger pad and the HET was placed a fixed distance away. As the stimulus was scanned across the finger, small vibrations were produced in the skin, which traveled a short distance up the finger and caused the magnet to vibrate (see Figure 3). The current produced in the HET was then amplified and stored to file (see Figure 4 for sample data). Our approach, then, consisted of recording the vibrations produced in the skin by various textured surfaces, and relating these vibrations to the resulting textural percepts.
In a first series of experiments (Bensmaïa & Hollins, 2003), we investigated whether perceived roughness was determined by the temporal or intensive properties of texture-elicited vibrations. To that end, subjects were presented with etched silicon surfaces (ranging in spatial period from 16 to 416 pm) moving at 2 cm/s, and were asked to produce magnitude estimates of roughness while we recorded the vibrations elicited in their skin by the stimuli. Several relationships were apparent in the data. First, the peak frequency of vibration was inversely proportional to the spatial period of the texture, as was to be expected given that, with scanning velocity held constant, the rate at which elements passed by a given point on the skin was inversely proportional to the distance between elements. Second, perceived roughness increased with spatial period (and was thus a negative function of peak frequency). Third, we found that the intensity of the vibrations increased with spatial period.
As the Pacinian system had been previously implicated in the perception of these same surfaces (Hollins et al., 2001), we adopted as an index of intensity Pacinian-weighted power, P^sub PC^, following Makous, Friedman, and Vierck (1995). Theoretically, P^sub PC^ gauges the efficacy with which a given vibratory stimulus excites the Pacinian system. Specifically, an individual spectral component of the vibratory stimulus contributes to P^sub PC^ to the extent that the Pacinian system is sensitive at that component’s frequency; thus, spectral components in the range from 100-300 Hz are weighted more heavily than components outside this frequency range. We found that perceived roughness increased as a logarithmic function of PPC. These data suggested, then, that perceived roughness may be a function of the peak frequency of the vibrations or of their intensity as indexed by P^sub PC^.
In order to discriminate between temporal and intensive theories of roughness perception for fine textures, we manipulated the speed (2 or 4 cm/s) at which surfaces with spatial periods of 80 and 184 µm were scanned along the fingerpad. The temporal hypothesis (i.e., that roughness is a negative function of peak frequency) made a clear prediction: that roughness would decrease as scanning speed increased for both surfaces. In contrast, according to the intensive hypothesis, the perceived roughness of the 80 µm surface would decrease while that of the 184 µm surface would increase as the scanning velocity doubled. The reasoning was as follows: Based on the relationship between velocity, spatial period, and peak frequency observed in the first experiment, the peak frequency of the vibrations elicited in the skin by the 80 µm surface would increase from 250 Hz at 2 cm/s to 500 Hz at 4 cm/s, while the peak frequency of the vibrations elicited by the 184 µm surface would increase from 108 Hz to 217 Hz. Because the Pacinian system is less sensitive at 500 than at 250 Hz, but more sensitive at 217 than at 108 Hz, the Pacinian response elicited by the 80 µm surface should decrease as scanning velocity doubles, while that of the 184 µm surface should increase, a pattern that should be reflected in judgments of roughness.
The psychophysical data matched these predictions derived from the intensive hypothesis, leading us to conclude that the perceived roughness of finely textured surfaces is determined by the efficacy with which these excite the Pacinian system (as gauged by P^sub PC^). In a separate study (Bensmaïa & Hollins, 2005), we confirmed, using a similar methodology, the finding that perceived roughness is a logarithmic function of P^sub PC^ using a stimulus set that included everyday textures such as fabrics, papers, wood, etc.
The hypothesis that vibratory power is the determinant of perceived roughness for fine textures was also explored using a discrimination paradigm (Bensmaïa, 2003). On each trial, subjects were presented with two etched silicon surfaces and judged which one was rougher. On half of the trials (“same” trials), both surfaces had the same spatial period (both 56 or 80 µm); on the remaining trials, the two surfaces had different spatial periods (56 and 80 µm). For each trial, the log(P^sub PC^) of the vibrations produced while scanning each surface (recorded using an HET) was computed, and the difference between the two (normalized by their mean), ΔP^sub PC^, was calculated; ΔP^sub PC^, then, was the (normalized) difference in the log Pacinian-weighted power of the vibrations elicited in the skin when scanning the surfaces presented in the first and second intervals. We wished to determine whether this difference was predictive of the subjects’ response. To that end, we performed a binary logistic regression with ΔP^sub PC^ as the predictor and the interval chosen by the subject as the dependent variable. The analysis was performed for all the data, and also on the subset of data obtained from “same” trials. A positive and significant regression coefficient would suggest that ΔP^sub PC^ is a reliable predictor of the subjects’ response. Indeed, subjects tended to judge the surface that elicited the vibrations with the higher Pacinian-weighted power to be rougher, as evidenced by a significant positive regression coefficient for ΔP^sub PC^ in the logistic regression model. Interestingly, this was also true for trials in which surfaces with the same spatial period were presented; in other words, when two nominally identical surfaces were presented, the surface that produced the more powerful vibrations was judged as rougher. We thus demonstrated, using a discrimination paradigm, that Pacinian-weighted power is a good predictor of perceived roughness.
We conclude, then, that the vibrotactile code for roughness is intensive in nature (i.e., the perceived roughness of a finely textured surface is determined by the intensity of the vibrations it produces in the skin during scanning). Furthermore, these data suggest that the peripheral neural code for perceived roughness is the total activity evoked in Pacinian fibers – the afferents already implicated in fine-texture perception by vibrotactile adaptation experiments (Hollins et al., 2001). Indeed, the index of vibratory intensity that was adopted throughout our vibration recording studies presupposes that the relevant receptor system acts as a critical band and has a specific spectral sensitivity profile peaking at 250-300 Hz (following Makous et al., 1995). Both of these properties are distinctive characteristics of the Pacinian system.
Can Spatial and Vibrotactile Codes Operate Together?
The adaptive significance of the duplex nature of roughness perception is that each code operates effectively over a range of texture scales where the other code is weak; in combination, they enable us to appreciate textures with elements so close together that they cannot be individually discerned, others with elements so far apart that we lose the sense that they constitute a texture, and everything in between.
The spatial code fails at very fine textures because the mechanical properties of the skin, the density of SAl receptors, and the receptive field structure of cortical cells set a lower bound on the ability to spatially represent these microscopic textures. The inability of the spatial mechanism to mediate the perception of very fine surfaces is shown by the fact that effective stimulation of the Pacinian channel is necessary over this range (Hollins et al., 2001).
The question of why the vibrotactile code makes little contribution to roughness for very coarse textures is more puzzling. After all, vibrations – sometimes very large ones (Bensma’ia & Hollins, 2003) – are created in the skin as coarse surfaces move across it. Part of the reason these have little perceptual impact is probably that the precision of the vibratory code is much lower than that of the spatial mechanism. Hollins et al. (2001) found that the Weber fraction for discriminating coarse surfaces was about 12%, and other investigators have found it to be even lower (Lamb, 1983; Morley et al., 1983); the Weber fraction for fine surfaces, however, averaged 38% (Hollins et al., 2001). When spatial and vibrotactile signals are both available, the latter may be ignored because they are less accurate.
Another possible explanation is that vibrotactile signals adapt out more rapidly than spatial ones, so that during the extended testing that typically prevails in psychophysical experiments, the spatial code will become predominant by default. While both Pacinian and SA1 afferents show clear adaptation (Bensmaïa, Leung, Hsiao, & Johnson, 2005), there is evidence, both physiological (O’Mara, Rowe, & Tarvin, 1988) and psychophysical (Gescheider & Wright, 1968) that the vibrotactile system shows, in addition, a profound central adaptation, which has no demonstrated equivalent in the central mechanisms mediating the spatial code. To test this explanation directly, Hollins, Lorenz, and Harper (2006) studied the effect on roughness perception of extended exposure to textures themselves, rather than to imposed vibration. They did this under conditions of direct touch, and also indirect touch through a rigid probe – a situation in which only vibratory cues can be operative, since the spatial pattern of the texture is never impressed onto the skin.
With indirect touch, it was found that adaptation to a texture in the coarse range (spatial period 416 µm) lowered the roughness of other surfaces, especially fine ones but to some extent coarse ones as well. The implication is that vibrotactile coding is potentially available for both fine and coarse surfaces, but is compromised by extended stimulation. With direct touch, however, Hollins et al. (2006) found that adapting to the same surface caused the roughness of fine surfaces to decline, while having no significant effect on the roughness of coarse surfaces. Hollins et al. concluded that texture adaptation is a specific case of vibrotactile adaptation, which weakens vibrotactile signals and therefore lowers the roughness of fine test textures; it has no effect on coarse surfaces because spatial coding mechanisms – predominant in any case – are not affected by adaptation.
A third possible reason for the apparently negligible role played by vibrotactile signals in coding the roughness of coarse textures (in direct touch) is that, as texture elements become more widely spaced, the “natural frequency” of vibration they produce will (if speed of movement is held constant) become lower and lower. That in some situations this is the largest component of the vibration was empirically demonstrated by Bensmaïa and Hollins (2003). Unless there are compensating increases in the amplitude of vibration, this drop in frequency will weaken the Pacinian system’s ability to respond to it.
This analysis suggests that the vibrotactile signal for roughness will grow less and less effective with increasing spatial period. It ignores two issues, however. One is that cutaneous vibrations are spectrally rich, including many frequencies higher than the fundamental (Bensmaïa & Hollins, 2003, 2005). These higher frequency components should retain the ability to excite the Pacinian system. The other issue is that the RA channel may contribute to roughness perception when coarse surfaces – hence low temporal frequencies – are used. Given their high density in glabrous skin, and their vigorous response to vibration in the 10-200 Hz range, it is surprising that no substantial role for RAs has been demonstrated in the perception of surfaces, beyond their ability to detect small, isolated asperities (LaMotte & Whitehouse, 1986) and to warn of slip (Johansson & Westling, 1984, 1987; Srinivasan, Whitehouse, & LaMotte, 1990). The potential role of RAs in texture perception is a promising subject for future investigation.
Still another possibility is that the vibration produced by examination of very coarse surfaces is actually too effective a vibrotactile stimulus, causing the response of Pacinian afferents to saturate. Saturated signals might add a fixed increment to perceived roughness, but would not contribute to – and might even compromise – the discriminability of coarse surfaces.
It is interesting, however, that none of these explanations for the seeming lack of involvement of vibratory channels in coarse texture perception makes a compelling case, either theoretical or empirical, that such involvement cannot occur. In fact, the observations of Cascio and Sathian (2001), described earlier, and of Gamzu and Ahissar (2001) may reflect such an involvement, albeit a modest one. The latter authors found that for some subjects, texture discrimination was better when coarse gratings were examined through a probe (with spatial cues eliminated) than when they were examined directly. Moreover, these subjects adjusted their rate of scanning in the probe condition so that temporal frequency of stimulation stayed between 15 and 30 Hz, a range that Gamzu and Ahissar point out is optimal for activation of RA fibers.
Further insight into the relationships among coarse textures, vibration, and roughness emerges from a careful study by Smith, Chapman, Deslandes, Langlais, and Thibodeau (2002). Their stimuli were surfaces embossed with rectangular arrays of truncated cones, examined with the middle finger; spacing between rows, in the direction of scanning, varied between surfaces, so that spatial period in this dimension ranged from 1.5 to 8.5 mm. Forces normal and tangential to the surface (in the direction of scanning) were continuously recorded. Smith et al. (2002) found that perceived roughness was not consistently related to the average tangential force between a surface and the skin, but was highly related to the root mean square (RMS) of the rate of change of tangential force. The authors point out that mechanoreceptors may respond not only to these force fluctuations themselves, but to vibrations that they generate.
Hollins, Fox, and Bishop (2000) put the question of whether vibrotactile signals can contribute to the roughness of macroscopic textures to a direct experimental test, by asking subjects to sweep a finger along identical surfaces, one vibrating and the other not, and to say which of the two was smoother. Subjects usually said the stationary surface was smoother, and did so increasingly as the Pacinian effectiveness of the vibration increased. Most subjects were unaware of the vibration; although it clearly influenced their responses, they apparently experienced it as a source of texture information. The results suggest that under carefully chosen conditions (the spatial period of the textures used was 298 µm, only moderately above the estimated transition point between mechanisms), vibrotactile and spatial signals can combine to determine roughness.
Perhaps the most compelling evidence that both codes can contribute to the roughness of the same surface comes from a study by Gescheider, Bolanowski, Greenfield, and Brunette (2005). When subjects were asked to examine embossed patterns and estimate their roughness, they gave responses that were not affected by adaptation to 250 Hz vibration. This indicates that the responses were not based on a vibratory code. When asked to judge the roughness of the individual texture elements, however, subjects’ estimates were substantially reduced by adaptation. In a final experiment, Gescheider et al. (2005) asked subjects to rate “overall roughness,” defined as a combination of element and pattern roughness. Their magnitude estimates were moderately reduced by the adapting stimulus, indicating that subjects were able not just to switch their attention between vibratory and spatial codes, but to combine these two types of signals into a unified percept.
The evidence reviewed here, taken as a whole, strongly supports the view that two coding mechanisms, mediated by two sets of tactile afferents and their central connections, mediate the perception of roughness for fine and coarse surfaces (empirically defined as those with predominant spatial periods less than or greater than 200 µm, respectively). Fine surfaces are perceived via a vibratory code mediated largely by the Pacinian system; coarse surfaces engage a spatial code that begins with stimulation of the SA1 afferent channel. The vibratory code is relatively imprecise, is easily deactivated by adaptation, and requires continuous movement of the stimulus across the skin; but it alone is capable of detecting microscopic textures.
Usually one or the other of the codes dominates the perception of the roughness of a particular surface, a lack of integration that is the result of a number of factors. Under some conditions, however, vibration can make a modest contribution to the perception of surfaces that are primarily encoded by the spatial mechanism.
Finally, it is important to keep in mind that there is more to texture perception than roughness. Even when some other salient properties of surfaces (notably hardness and slipperiness: Hollins, Bensmaïa, Karlof, & Young, 2000; Hollins, Faldowski, Rao, & Young, 1993) are controlled by making experimental surfaces out of a single material, they can still differ in ways other than roughness. This is shown by the simple fact that a raised-dot pattern and a grating can be readily distinguished, even if they are judged equal in roughness. The infinite variability of surface textures is no doubt what Katz (1925/1989) had in mind when he referred to “identifying characteristics” of surfaces that go beyond their “dimensional properties.” Like roughness, these unique characteristics are probably detected by vibrotactile mechanisms (Bensmaïa & Hollins, 2005; Bensmaïa, Hollins, & Yau, 2005) as well as spatial ones.
Barlow, S. M. (1987). Mechanical frequency detection thresholds in the human face. Experimental Neurology, 96, 253-261.
Bensmaïa, S. J. (2003). The vibrations of texture. Doctoral dissertation, University of North Carolina at Chapel Hill. Dissertation Abstracts International, 64, 1529B.
Bensmaïa, S. J., & Hollins, M. (2003). The vibrations of texture. Somatosensory & Motor Research, 20, 33-43.
Bensmaïa, S. J., & Hollins, M. (2005). Pacinian representations of fine surface texture. Perception & Psychophysics, 67, 842-854.
Bensmaïa, S. J., Hollins, M., & Yau, J. (2005). Vibrotactile information in the Pacinian system: A psychophysical model. Perception & Psychophysics, 67, 828-841.
Bensmaïa, S. J., Leung, Y. Y., Hsiao, S. S., & Johnson, K. O. (2005). Vibratory adaptation of cutaneous mechanoreceptive afferents. Journal of Neurophysiology, 94, 3023-3036.
Blake, D. T, Hsiao, S. S., & Johnson, K. O. (1997). Neural coding mechanisms in tactile pattern recognition: The relative contributions of slowly and rapidly adapting mechanoreceptors to perceived roughness. Journal of Neuroscience, 17, 7480-7489.
Blake, D. T., Johnson, K. O., & Hsiao, S. S. (1997). Monkey cutaneous SAI and RA responses to raised and depressed scanned patterns: Effects of width, height, orientation, and a raised surround. Journal of Neurophysiology, 78, 2503-2517.
Bolanowski, S. J. (1998). Tactile channels and their interactions. In S. Grondin & Y. Lacouture (Eds.), Fechner Day 98. Proceedings of the Fourteenth Annual Meeting of the International Society for Psychophysics (pp. 103-108). Québec, QC: The International Society for Psychophysics.
Cascio, C. J., & Sathian, K. (2001). Temporal cues contribute to tactile perception of roughness. Journal of Neuroscience, 21, 5289-5296.
Connor, C. E., Hsiao, S. S., Phillips, J. R., & Johnson, K. O. (1990). Tactile roughness: Neural codes that account for psychophysical magnitude estimates. Journal of Neuroscience, 10, 3823-3836.
Connor, C. E., & Johnson, K. O. (1992). Neural coding of tactile texture: Comparison of spatial and temporal mechanisms for roughness perception. Journal of Neuroscience, 12, 3414-3426.
DiCarlo, J. J., & Johnson, K. O. (1999). Velocity invariance of receptive field structure in somatosensory cortical area 3b of the alert monkey. Journal of Neuroscience, 19, 401-419.
DiCarlo, J. J., & Johnson, K. O. (2000). Spatial and temporal structure of receptive fields in primate somatosensory area 3b: Effects of stimulus scanning direction and orientation. Journal of Neuroscience, 20, 495-510.
DiCarlo, J. J., Johnson, K. O., & Hsiao, S. S. (1998). Structure of receptive fields in area 3b of primary somatosensory cortex in the alert monkey. Journal of Neuroscience, 18, 2626-2645.
Gamzu, E., & Ahissar, E. (2001). Importance of temporal cues for tactile spatial-frequency discrimination. Journal of Neuroscience, 21, 7416-7427.
Gescheider, G. A., Bolanowski, S. J., Greenfield, T. C., & Brunette, K. (2005). Perception of the tactile texture of raised dot patterns: A multidimensional analysis. Somatosensory and Motor Research, 22, 127-140.
Gescheider, G. A., & Wright, J. H. (1968). Effects of sensory adaptation on the form of the psychophysical magnitude function for cutaneous vibration. Journal of Experimental Psychology, 77, 308-313.
Hollins, M., Bensmaïa, S., Karlof, K., & Young, F. (2000). Individual differences in perceptual space for tactile textures: Evidence from multidimensional scaling. Perception & Psychophysics, 62, 1534-1544.
Hollins, M., Bensmaïa, S., & Risner, R. (1998). The duplex theory of tactile texture perception. In S. Grondin & Y. Lacouture (Eds.), Fechner Day 98. Proceedings of the Fourteenth Annual Meeting of the International Society for Psychophysics (pp. 115-120). Québec, QC: The International Society for Psychophysics.
Hollins, M., Bensmaïa, S. J., & Washburn, S. (2001). Vibrotactile adaptation impairs discrimination of fine, but not coarse, textures. Somatosensory & Motor Research, 18, 253-262.
Hollins, M., Delemos, K. A., & Goble, A. K. (1991). Vibrotactile adaptation on the face. Perception & Psychophysics, 44, 21-30.
Hollins, M., Faldowski, R., Rao, S., & Young, F. (1993). Perceptual dimensions of tactile surfaced texture: A multidimensional scaling analysis. Perception & Psychophysics, 54, 697-705.
Hollins, M., Fox, A., & Bishop, C. (2000). Imposed vibration influences perceived tactile smoothness. Perception, 29, 1455-1465.
Hollins, M., Lorenz, F., & Harper, D. (2006). Somatosensory coding of roughness: The effect of texture adaptation in direct and indirect touch. Journal of Neuroscience, 26, 5582-5588.
Hollins, M., & Risner, S. R. (2000). Evidence for the duplex theory of tactile texture perception. Perception & Psychophysics, 62, 695-705.
Johansson, R. S., & Westling, G. (1984). Roles of glabrous skin receptors and sensorimotor memory in automatic control of precision grip when lifting rougher or more slippery objects. Experimental Brain Research, 56, 550-564.
Johansson, R. S., & Westling, G. (1987). Signals in tactile afferents from the fingers eliciting adaptive motor responses during precision grip. Experimental Brain Research, 66, 141-154.
Katz, D. (1989). The world of touch (L. E. Krueger, Trans. & Ed.). Hillsdale, NJ: Erlbaum. [Original work published 1925.]
Klatzky, R. L., & Lederman, S. J. (1999). Tactile roughness perception with a rigid link interposed between skin and surface. Perception & Psychophysics, 61, 591-607.
Klatzky, R. L., Lederman, S. J., Hamilton, C., Grindley, M., & Swendsen, R. H. (2003). Feeling textures through a probe: Effects of probe and surface geometry and exploratory factors. Perception & Psychophysics, 65, 613-631.
Lamb, G. D. (1983). Tactile discrimination of textured surfaces: Psychophysical performance measurements in humans. Journal of Physiology, 338, 551-565.
LaMotte, R. H., & Whitehouse, J. M. (1986). Tactile detection of a dot on a smooth surface: Peripheral neural events. Journal of Neurophysiology, 56, 1109-1128.
Lederman, S. J. (1974). Tactile roughness of grooved surfaces: The touching process and the effects of macroand microsurface structure. Perception & Psychophysics, 16, 385-395.
Lederman, S. J., & Klatzky, R. L. (1987). Hand movements: A window into haptic object recognition. Cognitive Psychology, 19, 342-368.
Lederman, S. J., Loomis, J. M., & Williams, D. A. (1982). The role of vibration in the tactual perception of roughness. Perception & Psychophysics, 32, 109-116.
Lederman, S. J., & Taylor, M. M. (1972). Fingertip force, surface geometry, and the perception of roughness by active touch. Perception & Psychophysics, 12, 401-408.
Makous, J. C., Friedman, R. M., & Vierck, C. J., Jr. (1995). A critical band filter in touch. Journal of Neuroscience, 15, 2808-2818.
Meftah, E. M., Belingard, L., & Chapman, C. E. (2000). Relative effects of the spatial and temporal characteristics of scanned surfaces on human perception of tactile roughness using passive touch. Experimental Brain Research, 132, 351-361.
Morley, J. W., Goodwin, A. W., & Darian-Smith, I. (1983). Tactile discrimination of gratings. Experimental Brain Research, 49, 291-299.
O’Mara, S., Rowe, M. J., & Tarvin, R. P. C. (1988). Neural mechanisms in vibrotactile adaptation. Journal of Neurophysiology, 59, 607-622.
Phillips, J. R., & Johnson, K. O. (1981). Tactile spatial resolution. I. Two-point discrimination, gap detection, grating resolution, and letter recognition. Journal of Neurophysiology, 46, 1177-1792.
Pruett, J. R., Jr., Sinclair, R. J., & Burton, H. (2001). Neural correlates for roughness choice in monkey second somatosensory cortex (SII). Journal of Neurophysiology, 86, 2069-2080.
Randolph, M., & Semmes, J. (1974). Behavioral consequences of selective subtotal ablations in the postcentral gyrus of Macaca mulatta. Brain Research, 70, 55-70.
Roland, P. E., O’Sullivan, B., & Kawashima, R. (1998). Shape and roughness activate different somatosensory areas in the human brain. Proceedings of the National Academy of Sciences U.SA., 95, 3295-3300.
Sathian, K., Goodwin, A. W., John, K. T., & Darian-Smith, I. (1989). Perceived roughness of a grating: Correlation with responses of mechanoreceptive afferents innervating the monkeys fingerpad. Journal of Neuroscience, 9, 1273-1279.
Smith, A. M., Chapman, C. E., Deslandes, M., Langlais, J.-S., & Thibodeau, M.-P. (2002). Role of friction and tangential force variation in the subjective scaling of tactile roughness. Experimental Brain Research, 144, 211-223.
Srinivasan, M. A., Whitehouse, J. M., & LaMotte, R. H. (1990). Tactile detection of slip: Surface microgeometry and peripheral neural codes. Journal of Neurophysiology, 63, 1323-1332.
Sripati, A. P., Bensmaïa, S. J., & Johnson, K. O. (2006). A continuum mechanical model of mechanoreceptive afferent responses to indented spatial patterns. Journal of Neuropbysiology, 95, 3852-3864.
Sripati, A. P., Yoshioka, T., Denchev, P., Hsiao, S. S., & Johnson, K. O. (2006). Spatiotemporal receptive fields of peripheral afferents and cortical area 3b and 1 neurons in the primate somatosensory system. Journal of Neuroscience, 26, 2101-2114.
Taylor, M. M., & Lederman, S. J. (1975). Tactile roughness of grooved surfaces: A model and the effect of friction. Perception & Psycbopbysics, 17, 23-36.
Verrillo, R. T., & Ecker, A. D. (1977). Effects of root or nerve destruction on vibrotactile sensitivity in trigeminal neuralgia. Pain, 3, 239-255.
Yoshioka, T., Gibb, B., Dorsch, A. K., Hsiao, S. S., & Johnson, K. O. (2001). Neural coding mechanisms underlying perceived roughness of finely textured surfaces. Journal of Neuroscience, 21, 6905-6916.
Mark Hollins, University of North Carolina at Chapel Hill
Sliman J. Bensmaïa, Johns Hopkins University
Preparation of this article was supported by NIH grants NS045685, NS18787 and DC00023. Andrew Sparrow assisted in the experiment on texture discrimination on the face.
Correspondence concerning this article should be addressed to Mark Hollins, Department of Psychology, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 (E-mail: firstname.lastname@example.org).
Copyright Canadian Psychological Association Sep 2007
Provided by ProQuest Information and Learning Company. All rights Reserved