INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The question of SII's involvement in roughness discrimination motivated the present investigation. This study used passively applied stimuli to characterize the effects of tactile grating groove width, contact force, and scanning speed on responses of macaque SII neurons. Only two previous single-unit recording studies have investigated SII's role in roughness perception (Jiang et al. 1997; Sinclair and Burton 1993).

Several investigators have used single-unit recording techniques to explore involvement of macaque primary somatosensory cortex (SI) in roughness perception (Ageranioti-Bélanger and Chapman 1992; Chapman and Ageranioti-Bélanger 1991; Darian-Smith et al. 1982, 1984; Phillips et al. 1988; Sinclair and Burton 1988, 1991a,b; Sinclair et al. 1996; Tremblay et al. 1996). In a study where monkeys controlled their hand movements during surface examination, mean firing rates for many SI cells changed as grating groove width, force, and/or speed increased (Sinclair and Burton 1991a). Responses from many SI cells displayed positive functions of increasing groove width, but a small number of cells showed negative functions of groove width. Responses from most SI cells followed changes in two or three variables, groove width, force, and speed.

Sinclair and Burton described SII cells that coded roughness and/or movement parameters in their average firing rates (Sinclair and Burton 1993). The majority of these SII cells had response functions for these variables that resembled those found in SI (Sinclair and Burton 1991a) and the ventral posterior lateral nucleus of the thalamus (Sinclair et al. 1991). Factors that distinguished SII from SI cells appeared unrelated to the processing of roughness. These included SII cells with larger receptive fields, no passively definable receptive fields, responses to other stimulus events in the task but not to gratings, and a general lack of periodic entrainment to grating temporal periods.

Experiments involving passive touch removed the effects of hand movement by scanning surfaces across the immobilized finger pad (Jiang et al. 1997; Sinclair et al. 1996; Tremblay et al. 1996). Hand motion during voluntary movements extensively activates proprioceptive and cutaneous receptors (Hulliger et al. 1979), which may simultaneously affect activity of individual SII cells, obscuring the contribution of any single variable. The large number of connections between SII and motor cortices (for reviews: Burton and Sinclair 1996; Felleman and Van Essen 1991) creates a further complication from potential modulation contingent on motor execution and planning.

Jiang and colleagues employed a passive-touch paradigm to conduct a limited exploration of roughness processing in macaque SII (Jiang et al. 1997). They used embossed dots for roughness stimuli but did not test effects of force and speed. They reported that many SII cells responded in a nongraded fashion to differences in roughness: a step increase in firing rate above versus below some threshold dot spacing. Consequently, they argued that SII cell average firing rates only signal changes in roughness, providing no information about roughness magnitude. Using an active touch paradigm, however, Sinclair and Burton (1993) found graded changes in mean firing rates when grating groove width changed, indicating that SII cell response functions provide information about roughness magnitude. The present study with a passive-touch protocol also found significant linear correlations between SII cell mean firing rates and magnitudes of grating groove width, force, and speed.

In this study, monkeys dichotomously classified the roughness of passively applied gratings with trial-wise random levels of applied force or speed. Recordings characterized the effects of force and speed on SII cell groove width response functions. Findings strongly implicate, in subsets of SII neurons, mean firing rate codes that follow the parameters of roughness stimuli, bolstering the contention that SII critically serves roughness perception.¹

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

The experimental subjects were two male rhesus monkeys that weighed 9 kg each and were 12 (MM23) and 7 (MM26) years old. During training and recording, they acquired their daily fluid (150-375 ml) by working for a fruit-flavored drink. The monkeys received bits of fruit for cooperation during receptive field mapping and liberal amounts of water and fruit on days off. Access to dry food was unrestricted.

Stimuli

Tactual grating stimuli consisted of a 22 × 70 mm, smooth, raised start pad (~0.5 mm high) that was followed by a 22 mm wide by 50 mm long section of raised ridges (~0.5 mm high) (see Fig. 1, H and I). The start pad and ridges were made of plastic (polyamide) on a metal backing (Nyloprint; Process Color Plate, Chicago, IL). The start pad minimized physical and neural transients that would occur when a moving grating edge strikes the finger pad. The ridge land width was nominally 0.25 mm. The groove width between these ridges was the intended variable for roughness. For gratings with fixed ridge width, the human magnitude estimate of roughness is a positive function of increasing groove width, up to ~3.0 mm (Lederman and Taylor 1972; Sathian et al. 1989; Sinclair et al. 1995). Four different groove width gratings, 1.07, 1.42, 1.90, and 2.53 mm, were created for neural recordings, and these 1.07- to 2.53-mm groove widths, therefore, span the major ascending portion of the human roughness magnitude estimation function (Connor et al. 1990; Sinclair and Burton 1991b; Sinclair et al. 1995). Stimuli were inspected by microscope. We used gratings that were free of imperfections and that had ridge and groove widths within ~0.03 mm of intended values.

View larger version (57K): [in this window] [in a new window]   Fig. 1. Top: the equipment for passive touch. The robotic stimulator delivered gratings at controlled and measured force and speed levels, while the monkey responded to graphical cues on the monitor. A and inset: each finger of the hand contralateral to the recorded hemisphere closed a micro-switch in B, the hand-holder. The monkey manipulated a 3-position foot switch, C, with the foot ipsilateral to the recorded hemisphere. D: a strain gauge in the platter measured force. E: stepper motors repositioned the stimuli between trials. F: hydraulic pistons pulled the platter holding the gratings in a way that minimized vibration during stimulation. G: a linear potentiometer transduced platter position to measure speed. H: grating start pad. I: grating. Bottom: task schematic. The relative (not to scale) timing sequence of important events during a trial. Time moves from left to right. A trial begins at 2 with the white ready box and ends with the yellow reward or red error screen between 10 on 1 trial and 1 on the next trial, following the monkey's foot pedal response between 10 and 11. The interval 5-9 is variable and depends on the speed of motion for that trial. 9-10, The delay period is either 700 or 1,500 ms. Breaking finger switch closure or foot pedal neutral between 3 and 10 would trigger a trial abort. The unfilled portion of the box on the grating line before time 6 reflects the start pad. The filled portion of this box between 6 and 8 reflects the grating ridges and grooves. The data window, 7-8, reflects a period of time when the compressed area of the finger pad is completely and only in contact with the grating, which is now moving at constant speed. This interval was used to analyze the effects of groove width, force, and speed on cell average firing rate (AFR), and the procedure for setting this window is described in the text.

Stimulus control

As described previously (Sinclair et al. 1996), a robotic device positioned a platter with different grating surfaces underneath a single finger pad. Stepper motors (e.g., Fig. 1E) positioned stimuli vertically and horizontally during each inter-trial interval. Before recording neural activity, an iterative routine set platter heights to levels that gave desired force readings. These could be modified interactively by the experimenter during recording. During inter-trial intervals, the platter was shuffled, thereby eliminating potential auditory or vibratory cues about the platter location of the next grating. All parts of the stimulator and stimulus surfaces were obscured from view of the monkeys. The monkeys performed at chance, quit, or adopted a response bias (e.g., repeatedly pulling the foot pedal up) if the grating contact force approached 0 g, providing convincing evidence they found no alternative strategies for performing the task. An audio speaker, 2 m behind the monkey, delivered pink noise to mask room noises during the experiment. Neural signals were monitored with headphones. Oscilloscope displays could not be seen by the monkey.

Another motor (not shown) pulled hydraulic pistons (e.g., Fig. 1F), which mounted to the platter, moving the gratings across the finger pad. This arrangement smoothed the platter motion and damped potential vibrations produced by the motor, thereby eliminating unwanted vibratory stimuli. One personal computer (PC) controlled trial sequences, the stimulator, performance, and neural data acquisition; a second PC displayed behavioral control graphics on a computer monitor seen by the monkey.

A linear potentiometer (Fig. 1G, partly hidden by the platter) sampled position during the platter stroke with a resolution of 0.1 mm to calculate speed. Position was sampled at 500 Hz. Speeds were always within ~10% of the target values. Therefore for simplicity, some statistical analyses of the data used speed categories based on the intended speed values.

A strain gauge in the platter (Fig. 1D) measured average force with a resolution of ~1 gm and instantaneous force with a resolution of ~7 g. The force signal was low-pass filtered with a seventh-order Cauer elliptic filter (670 series, stop-band edge 147 Hz; Frequency Devices). Force was sampled at 500 Hz to resolve any fluctuations caused by the grating ridges during stimulation.

Behavioral control for passive touch

A hand and finger restraint was constructed for each monkey's left hand using elastomer putty (A95B-2; Smith and Nephew Rolyan) that was rigidly supported by fiber-glass, aluminum, and thermoplastic material (Hexcelite; Hexcel Medical Products). Prior to and during stimulation, the monkey had to insert its hand in this apparatus (Fig. 1B). The stimulated finger passed through an opening that provided access to the finger pad from below. The fingernail of the stimulated finger engaged a plunger that depressed a tube-mounted micro-switch (Fig. 1, inset). The four nonstimulated fingers depressed paddles made of thermoplastic, each of which closed a micro-switch (e.g., Fig. 1A). Minor finger movements would break switch contact. Maintained closure of these switches was necessary for successful completion of a trial. The monkeys could remove their hands at will. Generally, they maintained hand position in this device for 40-60 min. Video of the hand and finger during stimulation in training sessions verified that monkeys kept the hand and fingers essentially immobile during stimulation. We stimulated the left hand, only, and recorded cells in the right (contralateral) hemisphere. We trained the monkeys to use the right (ipsilateral) foot for the operant response, to confine any potential neural correlate for the subsequent foot movement to part of the body representation farthest from the finger representation in SII on the recorded side. In addition to closing the five finger switches, the monkey had to toggle a three-position foot switch (Fig. 1C) to the middle (neutral) position with the right foot to initiate a trial. Maintenance of this neutral foot pedal position was also required for trial completion.

Trial sequence

A computer monitor, placed directly in front of the eyes, at a distance of ~1 m presented task related cues to the monkey. Figure 1 summarizes the events of a trial. A gray, inter-trial screen (between times 1 and 2, Fig. 1) followed platter repositioning for the next grating (before 1, Fig. 1). The monkey stopped moving the foot pedal and waited a random 1-4 s for a white, ready box. Moving the foot pedal during presentation of a gray screen activated an alarm tone and set the inter-trial interval to 4 s from the time of each offense. This feedback discouraged continued movements following the reward period from the previous trial. After a white, ready box appeared (between 2 and 3, Fig. 1), the monkey had to keep all five finger switches closed and toggle the foot pedal to neutral. When the fingers and foot were in place, a 200ms pause (between 3 and 4, Fig. 1) followed the appearance of a green, neutral box, which remained lit. The robotic stimulator, then, began to compress the finger with the start pad of the grating (4, Fig. 1), by raising the platter (Fig. 1D) to a predetermined height that gave the desired level of contact force during grating stimulation. Compression of the finger pad reached a maximum force level when the platter stopped at the specified height for this trial (between 4 and 5, Fig. 1). A 450-ms pause allowed peak static contact force to drop to a steady-state level as the skin of the finger pad deformed with compression. This wait period minimized potential neural effects that accompany the changing conformation of the elastic finger-pad tissue. Then the stimulator accelerated the grating start pad (5, Fig. 1) in a direction proximal to distal along the finger pad, reaching constant speed before the ridges contacted the finger (6, Fig. 1). Deceleration from constant speed (9, Fig. 1) did not begin until after the last ridge of the grating had moved out of contact with the finger. The grating therefore always moved at constant speed when in contact with the monkey's skin. A 700- or a 1,500-ms delay period (9-10, Fig. 1), assigned randomly, followed the cessation of platter motion. The monkey had to maintain finger switch contact and neutral foot position through this delay. These requirements provided additional assurance that neural activity seen during grating stimulation was not confounded by neural activity associated with postural adjustments in preparation for the motor response. Display of a white screen replaced the green, neutral box after this delay (10, Fig. 1). The monkey, then, had 1.5 s to make a behavioral response with the foot pedal. In this period, one monkey pulled the pedal up for rough and pushed down for smooth. For the other monkey, the pedal directions had the opposite meanings. Display of a yellow screen, a relay click, and ~1 ml of fruit flavored drink followed a correct roughness classification. A red error screen, an alarm tone, and a 10-s inter-trial penalty time followed an error. The monkey could move freely at operant response time and for ~4 s while the platter was repositioned for the next trial. If the monkey broke micro-switch contact or moved the foot pedal from neutral before the white response screen, that trial was aborted. The error tone sounded, the red abort box appeared, the platter dropped, and a 2-s time penalty was added to the inter-trial interval. If the monkey exceeded the 1.5 s allowed for the operant response, the error tone sounded, the red screen appeared, the platter dropped, and a 2-s time penalty was added to the inter-trial interval.

Experimental design

The monkeys learned to classify tactual gratings as rough or smooth (wider or narrower groove width) regardless of applied force or scanning speed. Short blocks of 24-36 trials with different grating pairs alternated while recording isolation persisted to create groove width response functions for each cell. One block of trials used the 1.07 mm (smooth) and 1.90 mm (rough) gratings, and the other used the 1.42 mm (smooth) and 2.53 mm (rough) gratings. The grating pairs in each block were equally discriminable; the smooth grating had 44% narrower grooves than the rough grating (defined as the difference in groove widths divided by the maximum), and over this range of groove widths, equal percent differences have closely perceived differences in roughness for both humans and monkeys (Sinclair and Burton 1991b). In each block of trials, gratings were presented, one at a time, with the random scheduling of rough and smooth grating presentations constrained so that the same operant response was never required more than four trials in a row, which discouraged adoption of response bias (i.e., predilection to choose a particular pedal response).

Two different experiments were run, and subsequent statistical analyses treated both as four by three factorial designs, with an additional analysis model that included force as a continuous variable when speed and groove width were manipulated (see following text). In the force/groove-width experiment, each of the four gratings was delivered at three target force levels, 30, 60, or 90 g, which were trial-wise randomized, while speed was held constant at 80 mm/s. In the speed/groove-width experiment, three levels of grating scanning speed, 40, 80, and 120 mm/s, (proximal to distal on the finger pad), were trial-wise randomized, with a nominal force of 60 g. Force and speed levels were chosen from the range values previously used by humans (Lederman 1974) and monkeys in active touch studies (Sinclair and Burton 1991a,b, 1993). Engineering constraints placed an upper limit on the range of speed used. In each block the 12 trial types (a sub-block), consisting of one rough and one smooth grating at three forces or speeds and at two response delays (700 or 1,500 ms), repeated two to three times. Typically, recording isolation stability only allowed enough time for either the force/groove-width or speed/groove-width experiment to be run on any one cell. Mechanical repositioning of gratings between trials dictated a maximum stimulation rate of two to three trials per minute.

Both monkeys typically performed 80% correct in both experiments. If the monkey caused a trial abort or did not make a foot pedal choice before the response period elapsed, the stimulus combination used on that trial was repeated at the end of that sub-block. The data from aborted trials were excluded from statistical analyses. These measures forced the monkeys to classify groove width for all stimulus combinations and helped ensure balanced data sets. The monkeys usually aborted fewer than two trials in any recording session.

Single-neuron recording procedures

Surgical and postoperative care procedures for placing a recording chamber over lateral parietal cortex were similar to those described previously (Burton et al. 1997; Sinclair and Burton 1993). A grid pattern of burr marks was placed on the bone surface in the recording chamber. The centers of the recording chambers were approximately: AP +11 mm, ML +27 mm at 60° tilted away from sagittal. All surgical, preoperative, and postoperative procedures were in accord with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

A stepper-motor hydraulic microdrive (Model 650; David Kopf) advanced locally produced, glass-coated, platinum-iridium micro-electrodes through ~2-mm-diam craniotomy holes, that were opened as the experiment proceeded. Electrodes measured 1.2-4.5 M at 1 kHz (Model IMP-1; impedance meter, BAK, Rockville, MD).

Neural action potentials were amplified and filtered between 300 and 3,000 Hz. A commercial hardware and software system (DataWave, Longmont, CO) simultaneously recorded on separate channels the filtered analog neural signal, acceptance pulses from a dual amplitude and time window discriminator (BAK), video sync from the behavioral control graphics monitor, force on the platter, and position of the gratings. The first three channels were sampled at 20 kHz. This commercial system applied time stamps to recorded events with 100-µs resolution. Off-line waveform processing of the analog neural signal provided a means to verify the window discriminator acceptance pulses. Rarely, a second, distinct waveform could be isolated in this manner. Extremely strict criteria for single-unit isolation were used because of the quantitative nature of the analyses.

Selection criteria and receptive field assessments

Repeated penetrations were made at the same locations over successive days. Sampling proceeded from superficial to deep locations to maximize data yield from shallow depths before penetrations were driven through for deeper recording loci. Routinely cells were isolated for study while the animals worked. We selected cells that appeared to be engaged by the task. Bias, if any, was toward cells with high firing rates during application of the gratings. After collecting data during stimulation with gratings, we determined the spatial extent of receptive fields by manual palpation and examination with probes and brushes. Adaptation to manually applied punctate indentation was used to classify cells as slowly, intermediately, or rapidly adapting. Cells were classified as Pacinian or non-Pacinian based on responses to tuning forks. We assigned lateralities (contralateral, contralateral best, or bilateral) and submodalities (cutaneous, joint, hair, mixed) where possible. Force thresholds were determined with calibrated nylon filaments (Stoelting) for a subset of the recorded cells.

Data analyses

A custom program sorted the data and created rasters and histograms for visual inspection. From these we manually selected portions of response trains from each trial for statistical analyses in the following way. First, the end of the analysis interval (window) was defined as the time in each trial when the force first began to drop because the trailing edge of the grating was moving past the compressed finger pad (8, Fig. 1). Force then fell to 0 g as the grating moved completely past the finger (between 8 and 9, Fig. 1). Next, we set the start of this same analysis window to an earlier time that corresponded to a position on the grating 40 mm closer to the beginning of the ridges (7, Fig. 1). The proximal-distal axis of the monkey's compressed finger pad was ~7-9 mm. Therefore this last 40 mm of a 50-mm grating was chosen for statistical analysis. We chose these start and stop points because for part of the first 10 mm of grating (6-7, Fig. 1), the finger was in transition from the start pad (unfilled portion of grating box schematic before 6, Fig. 1) to the ridged grating surface (solid portion of grating box, 6-8, Fig. 1). The compressed part of the finger pad was, therefore wholly and only in contact with a grating during this analysis period (7-8, Fig. 1). As a final step, we inspected the responses during the defined window and adjusted its limits to ensure that they excluded transients, such as phasic bursts at the beginning and end of the gratings. A 40-mm grating spatial extent corresponded to 500 ms of stimulation at 80 mm/s and 1,000 ms and 333 ms at 40 and 120 mm/s, respectively. Windows shorter than 200 ms were not used for statistical analyses. Average force, speed, and neural firing rates were computed during the selected analysis intervals.

Further statistical analyses were carried out using SAS (SAS v6.03 and v6.12, SAS Institute). In the force/groove-width experiments, GLM ANOVA models (robust to unbalanced designs) examined effects of groove width and force and first-order interaction effects of groove width and force on average firing rate (AFR). Force varied from the intended 30, 60, and 90 g. Therefore each trial was assigned to one of five force bins, 0-25, 26-50, 51-75, 76-100, and 100+ g. These bins were defined for graphs in the figures in RESULTS by their centers: 12.5, 37.5, 62.5, 87.5, and 112.5 g. The force bins became levels of a class variable for force effects in the ANOVAs.

The speed/groove-width experiments were analyzed with both class variable GLM ANOVA models (different groove widths and speeds were the levels of class variables) and class and continuous variable GLM ANOVA models (groove width and speed were classes, and force was run as a continuous variable). The models tested for main effects of groove width, speed, and force and first-order interactions between these independent variables with AFR as the dependent variable. In the speed/groove-width experiments, some variation around the intended 60 g level allowed the assessment of force effects. In these experiments, the model with force as a continuous variable tested for global effects of force and for homogeneity of the slopes of the functions of AFR versus force (continuous variable) across all combinations of the two classification variables, groove width and speed. For both the force/groove-width and the speed/groove-width experiments, ANOVA models containing a block-replicate term (when sets of both pairs of gratings had been repeated 2 times) were considered as needed.

Partial Pearson correlation coefficients (SAS), which measure the strength of the linear relationship between two variables with effects of a third variable removed (McNemar 1969), determined the strengths of the linear relationships between AFR and groove width, AFR and force, and AFR and speed. Partial correlation analyses included groove width, force, and average firing rate terms for the force/groove-width experiments and groove width, force, speed, and average firing rate terms for the speed/groove-width experiments. These analyses assessed linear components in the SII cell response functions. A significant partial Pearson correlation coefficient does not exclude the presence of significant nonlinear components in the response functions. However, with only four groove widths, the precise characterization of the nonlinear components in the SII cell groove width response functions was not meaningful. The partial Pearson correlation coefficients, therefore approximated the degrees of response function linearity. Cells were then assigned as having positive or negative response functions for groove width, force, or speed, if the class-variable GLM ANOVA effect of groove width, force, or speed was significant (P < 0.05) and the partial Pearson correlation coefficient between AFR and the variable of interest was also significant (P < 0.05). The class-variable ANOVAs simply tested for differences in AFR, and the partial correlation analyses tested for the presence, strength, and directionality of linear components in the cell response functions; i.e., a cell could have a significant ANOVA effect but no significant partial correlation results if its AFR function was peaked. The sign of the AFR response function designation, positive or negative, was determined by the sign of the partial correlation coefficient. A simple correlation analysis looked for correlations between these partial Pearson correlation coefficients, themselves, in the sample cell population with significant ANOVA effects (i.e., the sample that was sensitive to groove width, force, and/or speed).

Assessment of SII cells' abilities to distinguish between changes in the magnitude of groove width within the categories of smooth (1.07 and 1.42 mm) and rough (1.90 and 2.53 mm) was achieved as follows. t-tests examined, post hoc, the significance of mean firing rate differences between adjacent pairs of gratings, 1.07 and 1.42 mm, 1.42 and 1.90 mm, and 1.90 and 2.53 mm, (1-tailed) Bonferroni corrected for the three grating comparisons. Means were compared in this way at or across force and/or speed levels. If a cell distinguished between 1.07 and 1.42 mm and/or 1.90 and 2.53 mm, regardless of its ability to distinguish between 1.42 and 1.90 mm, then that cell was classified as providing information about groove-width magnitude. If, alternatively, the cell AFR function only distinguished the 1.42 and 1.90 mm groove widths, that cell could only strictly be classified as discriminating rough and smooth categories, a possibly behaviorally related effect, and not groove width magnitude, a more stimulus-level effect. These tests proved to impose extremely strict criteria for discrimination of groove-width magnitude, because of the low numbers of trials (hence the high AFR variance) at the individual groove width and force or groove width and speed combinations.

Histology

Monkeys were killed with an intravenous lethal overdose of pentobarbital sodium after sedation with ketamine hydrochloride (5-10 mg/kg). Transcardiac perfusion with buffers and aldehydes fixed the brains. The blocked brains were frozen and sectioned at 50 µm. Serial sections through the recorded area were stained with thionin and examined for evidence of electrode tracks and electrolytic lesions. The Animal Care and Use Committee at Washington University approved the methods used to kill the monkeys, methods that are consistent with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Overview

We determined receptive field properties for 300 cells recorded near the upper bank of the lateral sulcus. Histological recovery of electrode tracks, electrolytic lesions, ink, and dye marks allowed 216 of these to be assigned to SII, 17 to area 2, 47 to area 7b, 7 to the lower bank of the lateral sulcus. The other 13 cells with defined receptive fields could not be assigned to a cortical area with any certainty. No electrode tracks crossed into the insula. Figure 2 shows representative sections where electrode tracks passed through relative anterior, B and D, and posterior, A and C, parts of SII in both monkeys.

View larger version (155K): [in this window] [in a new window]   Fig. 2. Second somatosensory cortext (SII) histology. These 4 photomicrographs contain portions of relative anterior and relative posterior regions of SII from the right hemispheres of both monkeys. The lateral sulcus runs from top right to bottom center in each. A: relatively posterior section from monkey MM26. The presumed posterior sub-region of SII lies close to crown of the upper bank of the lateral sulcus, where the electrode tracks and the lesion are visible. B: relatively anterior section from MM26. Note the well-developed insula at bottom and the more prominent layer 4 compared with A. Two lesions are visible in layer 6 of SII. C: section from relatively posterior SII in monkey MM23. There is little development of the insula. D: multiple tracks targeted the hand and digit representation in SII. As with B, there is a well-developed insula, and there is a more prominent layer 4. Lines at the surface point to evidence of electrode tracks. Lines with asterisks denote tracks ending in electrolytic lesions. Scale bars are 2 mm.

One hundred and sixty-two cells were studied with 20 trials in either a force/groove-width or a speed/groove-width experiment. Recording isolation persisted for 40 trials in 102/162 (63.0%) (well-characterized) cells. The histologies placed 86 of these in SII, 8 in 7b, and 8 could not be assigned. We could not separate the eight 7b and eight unassigned cells from the SII group based on responses to groove width, force, and speed. They do not constitute statistically reliable samples and will not be considered further. The well-characterized sample of 86 SII cells is the subject of this discussion because multiple force or speed levels were used for all four gratings. Effects seen in experiments with fewer than 40 trials were often qualitatively similar to those seen for cells with 40 trials. The force/groove-width experiment was run on 33/86 (38.4%) cells, the speed/groove-width experiment was run on 49/86 (57.0%) cells, and 40 trials of both experiments were run on 4/86 (4.7%) cells (i.e., 37 force/groove-width and 53 speed/groove-width experiments on 86 cells; 50 cells were from monkey MM23, and 36 cells were from monkey MM26). We observed significant ANOVA effects of groove width, force, and/or speed on AFR in 72/86 (83.7%) of these SII cells, and the results from the two monkeys were similar.

Most of these SII neurons responded to cutaneous stimulation of the glabrous skin on the hand contralateral to the recorded hemisphere. We normally encountered either small receptive fields, covering from part of one finger to all of one or more finger(s), or larger receptive fields that included the entire hand and digits. Most cells were rapidly adapting, and many responded, at least weakly, to manual stimulation of the mirror-symmetric site, ipsilateral to the recorded hemisphere. Specifically, 22/86 (25.6%) cells responded exclusively to contralateral stimulation; 24/86 (27.9%) responded more strongly to contralateral than to ipsilateral stimulation; 14/86 (16.3%) responded equally to contralateral and to ipsilateral stimulation; and none of the cells responded exclusively to or better to ipsilateral stimulation. In 21/86 (24.4%) cells, recording isolation was lost before laterality could be carefully evaluated, and in 5/86 (5.8%), laterality could not be determined. In the sample, 25/86 (29.1%) cells responded exclusively to cutaneous stimulation, 7/86 (8.1%) responded primarily to joint stimulation, 15/86 (17.4%) responded to mixed modal stimulation, 13/86 (15.1%) had a better response when the monkey actively engaged a stimulus and were not well engaged by passively applied stimuli, and 1/86 (1.2%) had an inhibitory response to manually applied stimuli (in 20/86 or 23.3% of cells, recording isolation was lost before submodality could be carefully evaluated, and in 5/86 or 5.8%, submodality could not be determined). Consistent with previous findings (Sinclair and Burton 1993), periodic entrainment to grating temporal periods, measured by inspection of autocorrelation histograms, was essentially absent in SII. For 19 SII cells, isolation persisted through force threshold determination with hand-held, calibrated nylon filaments. Force threshold mean and standard error of the mean (SE) were 422 ± 151 mg: minimum = 20 mg, maximum = 2,950 mg, median = 258 mg. Sampling procedures were biased toward cells responding to gratings. Therefore this sample may not reflect the general characteristics of SII. The sample was heterogeneous, but statistical analyses found no significant correlations between receptive field characteristics and the effects of groove width, force, and speed.

Positive and negative response functions for increases in groove width

Sixty-three of 86 SII cells had significant differences in AFRs when groove width changed (ANOVA P < 0.05). In 31/63 cells, these effects were additive or isolated (ANOVA main effects); 25 of these had effects of groove width and at least one other variable, and 6 had isolated effects of groove width. In 32/63 cells, these effects were nonadditive with effects of force and/or speed (ANOVA interactions). The cells with additive or isolated effects of groove width will be considered first (followed by those of force and speed), and the nonadditive effects will be explained and described later. Of the 31 SII cells with additive or isolated effects of groove width, 27 cells had response functions that could be described as either positive or negative (see METHODS: significant ANOVA and significant positive or negative partial Pearson correlation coefficient between AFR and groove width, r_AFR/GW). Four of 31 cells showed significant differences in AFR as groove width changed, but their r_AFR/GW's were nonsignificant. There were roughly equal numbers of positive and negative cells in the sample: 14 of the 27 assigned cells had positive AFR response functions for groove width, r_AFR/GW (mean ± SE) = 0.49 ± 0.03, and the other 13 cells had negative AFR response functions, r_AFR/GW = 0.55 ± 0.05.

Figure 3 shows an example of a cell with a positive response function for groove width. Its AFR rose from 55.1 ± 4.4 to 100.8 ± 3.5 (SE) spikes/s (averaged across all forces) as groove width increased from 1.07 to 2.53 mm [F (3, 51) = 24.67, P = 0.0001, r_AFR/GW = 0.70, P < 0.0001]. On average,¹ as can be inferred from the graph (Fig. 3A), this cell's AFR differences could have provided the monkey with sufficient information to classify correctly the groove widths of the comparison gratings (1.07 vs. 1.90 and 1.42 vs. 2.53 mm). Examination of AFR differences for adjacent gratings, 1.07 versus 1.42, 1.42 versus 1.90, and 1.90 versus 2.53 mm groove widths, (see METHODS) showed that this cell additionally provided statistically reliable information about changes in groove width magnitude within the smooth and/or rough categories (smooth = 1.07 and 1.42 mm, rough = 1.90 and 2.53 mm) at two force levels. Specifically, it showed significant (P < 0.05, Bonferroni corrected for the 3 grating comparisons) changes in AFR from 1.07 to 1.42 mm, 42.8 ± 6.5 to 68.7 ± 2.5 spikes/s [t = 3.30, df = 12, P < 0.01 (all Ps in this format are Bonferroni corrected for the 3 grating comparisons)] in the 37.5-g force bin and from 1.90 to 2.53 mm, 59.5 ± 7.9 to 86.7 ± 3.3 spikes/s (t = 2.82, df = 12, P < 0.03) in the 37.5-g force bin and 87.7 ± 5.3 to 115.3 ± 2.8 spikes/s (t = 4.63, df = 10, P < 0.002) in the 87.5-g force bin.

View larger version (51K): [in this window] [in a new window]   Fig. 3. Positive groove width effect. A: this SII cell (gc08cf  bak) showed increased AFRs as groove width increased from 1.07 to 2.53 mm for trials that fell into 3 different force bins, 37.5, 62.5, 87.5 g (see text). Error bars are 1 SE. Force bins with fewer than 3 trials have been omitted for clarity. B: raster dot density increases with groove width. The rasters and histograms have been sorted for each grating groove width, 1.07, 1.42, 1.90, 2.53 mm, but otherwise appear in chronological order, top to bottom, regardless of force level. In each raster line a tic marks the time of occurrence of an action potential. The vertical bars represent the interval in each trial selected for statistical analysis (see METHODS). The histograms are aligned in time with the rasters above, and the time axis applies to both histograms and rasters. Bin size is 5 ms for the histograms. The filled bins represent firing rate in spikes per second (s/s) on the left histogram axis. The line traces show average force in grams relative to the right histogram axis. The histograms sum across all force levels for each of the 4 gratings. Note the nearly equal average force levels for the 4 gratings. C: a photograph of the oscilloscope screen demonstrates the high quality of the isolation. Window discriminator acceptance pulses are visible below each spike. D: the box indicates the stimulated finger on the left (L) hand. Darker shading indicates more vigorous responses to manual stimulation in the receptive field. The cell responded to stimulation of skin and hair on the entire hand, bilaterally. Adaptation to punctate probe indentation was intermediate.

The example depicted in Fig. 3 had an effect of groove width on AFR that was additive with an effect of force on AFR, which is visible in the graph as the upward shift of the AFR functions for groove width at increasing force bins. Its AFR rose from 62.5 ± 4.2 to 89.2 ± 4.2 spikes/s (averaged across all groove widths) as force increased from the 37.5 to the 87.5-g bin [F (4, 51) = 14.92, P = 0.0001].² Its partial Pearson correlation coefficient for AFR and force (r_AFR/FORCE) was 0.64, P < 0.0001. The cell's groove width response functions have approximately the same shapes and slopes, illustrating the additive nature of this cell's ANOVA main effects. Also note for this cell that AFR changed 45 spikes/s for a 2.4-fold change in groove width, but it only changed by ~27 spikes/s for a 2.3-fold change in force.

Figure 4 shows an example of a cell with a negative response function for groove width. Its AFR dropped from 46.7 ± 2.4 to 14.0 ± 1.3 spikes/s as groove width increased from 1.07 to 2.53 mm [F (3, 51) = 84.14, P = 0.0001, r_AFR/GW = 0.81, P < 0.0001]. This cell's AFR functions provided information about groove width magnitude for the 1.07-, 1.42-, and 1.90-mm groove-width gratings, but its AFR functions did not discriminate between 1.90 and 2.53 mm at any force level. It had significant changes in AFR from 1.07 to 1.42 mm, 48.8 ± 3.8 to 35.6 ± 1.9 spikes/s (t = 3.13, df = 6, P < 0.05) in the 62.5-g force bin, 44.3 ± 1.6 to 31.1 ± 1.1 spikes/s (t = 6.83, df = 12, P < 0.0001) in the 87.5-g force bin, and from 1.42 to 1.90 mm, 39.0 ± 5.0 to 15.0 ± 2.9 spikes/s (t = 3.41, df = 6, P < 0.03) in the 37.5-g force bin, 35.6 ± 1.9 to 11.9 ± 2.6 spikes/s (t = 7.45, df = 6, P < 0.0005) in the 62.5-g force bin, and 31.1 ± 1.1 to 12.5 ± 3.4 spikes/s (t = 5.93, df = 10, P < 0.0002) in the 87.5-g force bin. As with the example in Fig. 3, the functions of AFR versus groove width at each force level in Fig. 4A have similar shapes and slopes. In contrast, many other cells in the sample (see following text) showed changes in the slopes and/or shapes of the firing rate functions for groove width as force or speed changed.

View larger version (40K): [in this window] [in a new window]   Fig. 4. Negative groove width effect. A: this SII cell (pb74cf  cl1) showed decreasing AFRs with increasing groove width for trials that fell into 4 different force bins, 37.5, 62.5, 87.5, 112.5 g. B: raster dot density decreases with groove width. Note that average force levels are equal for the 4 gratings. C: the cell responded to cutaneous stimulation of digits 2 and 3. Force threshold could not be determined with calibrated filaments because motion in the receptive field appeared to be required for a response to passive stimulation with hand-held stimuli. Format as in Fig. 3.

Figure 5 shows two examples of the minority of recorded SII cells that only showed significant effects of groove width (i.e., no effects of force and/or speed). The cell in Figs. 5A had a positive AFR response function for groove width [F (3, 67) = 12.96, P = 0.0001, r_AFR/GW = 0.45, P < 0.0001], and the cell in Fig. 5B had a negative response function [F (3, 72) = 34.43, P = 0.0001, r_AFR/GW = 0.69, P < 0.0001]. Each curve on these graphs represents the AFR versus groove width function for trials run at a different speed. There were no significant differences in AFR for the speed levels used, and the shapes and slopes of the AFR versus groove width functions are equivalent for all three speeds. The AFR functions depicted for these cells provide other examples whose AFRs discriminated between changes of groove width within (and/or across) the behavioral response categories. The cell in Fig. 5A had a significant change in AFR from 1.07 to 1.42 mm, 26.6 ± 2.9 to 38.9 ± 4.0 spikes/s (t = 2.55, df = 42, P < 0.03) for trials at all three speeds. The cell in Fig. 5B showed significant changes in AFR from 1.07 to 1.42 mm, 27.9 ± 2.0 to 19.1 ± 1.3 spikes/s (t = 3.69, df = 43, P < 0.001) for trials at all three speeds and from 1.42 to 1.90 mm, 19.1 ± 1.3 to 12.7 ± 0.9 spikes/s (t = 4.12, df = 45, P = 0.0003) for trials at all three speeds.

View larger version (22K): [in this window] [in a new window]   Fig. 5. Groove width response independence. A: this cell (go04cv  bak) had a positive AFR response function for increasing groove width that was not affected by a 3-fold change in speed in a statistically significant way. B: this cell (gg05cv  bak) had a negative response function that was unaffected by speed.

Figure 6, top, plots the 27 cells with positive or negative response functions of increasing groove width. Note the similar range of slopes for the positive and negative functions. Figure 6, middle and bottom, present data for force and speed in the same manner. This data will be discussed in the following text.

View larger version (35K): [in this window] [in a new window]   Fig. 6. Positive and negative effects of groove width, force, and speed. Cells with positive (left) and negative (right) AFR response functions for groove width (top), force (middle), and speed (bottom) are plotted together for comparison. There were approximately equal numbers of cells with positive and negative response functions for groove width and force. There were 4 SII cells with negative functions of increasing speed. Note the comparable range of slopes for the positive and negative functions for each independent variable. Each line represents the response function of a different cell. Different line types (solid, long dash, and short dash) were arbitrarily assigned to aid visual resolution of the individual curves. The vertical axis represents scaled average firing rate in spikes per second. AFR has been scaled by subtracting each cell's lowest AFR from each of its data points (spikes/s  minimum). The AFRs for each groove width were averaged across all levels of force and/or speed. The AFRs for each force or for each speed were averaged across all levels of groove width or groove width and force, respectively. Error bars are SE. Conditions with fewer than 2 trials have been omitted for clarity.

A majority of the 63 cells that showed significant effects of groove width had AFR functions that discriminated between either the groove widths of the two gratings designated smooth (1.07 and 1.42 mm) or those designated rough (1.90 and 2.53), with or without discriminating between adjacent groove widths of the gratings designated as smooth and rough (1.42 vs. 1.90 mm).¹ Specifically, 58/63 cells showed significant effects of changing groove width from 1.07 to 1.42, 1.42 to 1.90, or 1.90 to 2.53 mm for at least one force or speed level (SAS v6.12: Proc T Test). The other 5/63 cells failed to discriminate any adjacent grating pairs. Forty-one of 58 cells (70.7%) that discriminated adjacent grating groove width levels did so for either changes from 1.07 to 1.42 mm or from 1.90 to 2.53 mm, i.e., within (and, in 21/41 cases, also across) the categories of smooth or rough. Seventeen of 58 cells (29.3%) had AFR functions that by these criteria could only distinguish adjacent groove widths across smooth and rough categories (1.42 vs. 1.90 mm). Visual inspection of response trends in the data, however, allowed 60/63 cells to be classified as discriminating withinand/or acrossbehavioral response categories (3/63 cells had effects of multiple variables that were difficult to categorize). Forty-eight of these 60 cells (80%) showed AFR changes as groove width changed from 1.07 to 1.42 mm or from 1.90 to 2.53 mm for at least one force or speed level, and 12/60 (20%) of these cells showed AFR changes as groove width increased from 1.42 to 1.90 mm but not from 1.07 to 1.42 mm or from 1.90 to 2.53 mm.

Positive and negative response functions for increases in contact force

Twenty-nine of the 37 cells that were studied with systematic variation of force and groove width had significant (class variable ANOVA P < 0.05) effects of force. In 16/29 cells, these effects were additive with effects of groove width (ANOVA main effects of force and groove width), and 1 cell had an isolated effect of force (total n = 17). The remaining 12/29 cells had interactive effects and will be discussed later. Fifteen of the 17 SII cells with ANOVA main effects of force had response functions that we designated either positive or negative (significant ANOVA and significant r_AFR/FORCE). Two of the 17 cells showed significant differences in AFR for changing force, but their r_AFR/FORCE's were nonsignificant. As with groove width, there were roughly equal numbers of positive and negative cells in the force sample. Seven of the 15 assigned cells had positive AFR response functions for force, r_AFR/FORCE (mean ± SE) = 0.47 ± 0.07, and the other 8 cells had negative AFR response functions, r_AFR/FORCE = 0.42 ± 0.05.

Fourteen of 49 (53 speed/groove-width experiments minus the 4 cells studied with systematic variation of force, speed, and groove width) cells studied with systematic variation of speed and groove width had significant effects related to chance variation around the intended 60 g of target force. For these cells, it was not possible to construct 25-g force bins with equal numbers of samples in each. However, the effects were robust and clearly visible over a limited range of force. For these reasons, the class and continuous variable ANOVA model described in METHODS was employed to analyze these cells, where force was a continuous rather than a class variable. For these 14 cells, means ± SE were for force ranges 43.0 ± 2.0 g, for force minima 38.5 ± 2.2 g, for force maxima 81.5 ± 1.8 g, for force means 58.6 ± 1.2 g, and for force standard deviations 8.9 ± 0.3 g. Six of these 14 cells showed nonadditive effects of force with at least one other variable. Two of the remaining eight cells showed significant differences in AFR for changing force, but their r_AFR/FORCE's were nonsignificant. Three of the remaining six cells had positive AFR response functions for force, r_AFR/FORCE (mean ± SE) = 0.26 ± 0.01, and the other three cells had negative AFR response functions, r_AFR/FORCE = 0.39 ± 0.09. These 14 cells corroborated the findings from the 29 cells with force effects in the force/groove-width experiments; in total, 43 SII cells had significant differences in AFRs when force changed.

Figure 7 shows data from a cell that had a positive AFR response function for increasing force. It increased its AFR from 55.5 ± 3.4 to 80.5 ± 6.6 spikes/s (averaged across all groove widths) as force increased from the 37.5- to the 112.5-g bin [F (4, 81) = 22.85, P = 0.0001, r_AFR/FORCE = 0.70, P < 0.0001].² The points on the force axis in Fig. 7 represent the centers of 25-g force bins. The graph shows that this cell's force response functions have approximately the same shapes and slopes for each level of groove width. This cell's force effect was additive with a significant effect of groove width. It changed its AFR from 48.3 ± 2.5 to 71.1 ± 2.6 spikes/second (averaged across all forces) as groove width increased from 1.07 to 2.53 mm [F (3, 81) = 30.12, P = 0.0001, r_AFR/GW = 0.64, P < 0.0001]. The upward shift of the curves for trials on the 1.90- and 2.53-mm groove-width gratings relative to those on the 1.07- and 1.42-mm gratings illustrates this positive effect. This cell is representative of the minority in the sample that changed AFR for groove width in a way that only distinguished between the smooth and rough groove-width categories with no differentiation within those groups. Its AFRs distinguished the 1.42- and 1.90-mm groove widths at 35.3 ± 4.4 and 70.4 ± 3.8 spikes/s (t = 5.90, df = 6, P < 0.002) in the 37.5-g force bin, at 54.3 ± 3.1 and 80.2 ± 5.6 spikes/s (t = 4.30, df = 19, P < 0.001) in the 62.5-g force bin, and at 59.2 ± 5.6 and 80.9 ± 6.0 spikes/s (t = 2.54, df = 10, P < 0.05) in the 87.5-g force bin.

View larger version (46K): [in this window] [in a new window]   Fig. 7. Positive force effect. A: this SII cell (pk03cf  cl1) showed increased AFRs as force increased from 37.5 to 112.5 g (centers of force bins) for all 4 gratings, 1.07, 1.42, 1.90, 2.53 mm. Data points with fewer than 2 trials have been omitted for clarity. B: the rasters and histograms have been sorted by force bin, and trials in the rasters appear from top to bottom in order of increasing force. In the rasters and histograms, force means ± SEs for trials grouped by force bins 37.5, 62.5, 87.5, and 112.5 g were 41 ± 2, 61 ± 1, 88 ± 2, and 115 ± 3 g, respectively. The histograms show average responses across all 4 gratings. Note that discharge level increased with force and that average force levels increased as intended during the experiment. C: a photograph of the oscilloscope screen shows the isolation. D: the cell responded more strongly to passive, manual stimulation of the digits of the left (studied) hand than the right hand. Its best responses occurred when the monkey actively examined objects. Adaptation to punctate probe indentation was rapid. Format similar to Fig. 3.

An example of a cell that had a negative response function for increasing force appears in Fig. 8. Its AFR fell from 66.9 ± 4.5 to 40.4 ± 2.9 spikes/s as force increased from the 12.5to the 87.5-g force bin [F (4, 69) = 19.46, P = 0.0001, r_AFR/FORCE = 0.47, P < 0.0001].² The similarly shaped, decreasing functions of force for each grating in Fig. 8A illustrate this effect. This cell was unusual (as will be seen below) because it had a negative force effect and a positive groove width effect, which is visible as the separation of force curves for the gratings in the graph, especially obvious at a force of 87.5 g (interactive effects of multiple variables will be considered in the following text; the interaction between force and groove width for this cell failed to reach significance). Its AFR increased from 37.0 ± 3.2 to 53.7 ± 2.5 spikes/s as groove width increased from 1.07 to 2.53 mm [F (3, 69) = 13.59, P = 0.0001, r_AFR/GW = 0.38, P < 0.0002].

View larger version (40K): [in this window] [in a new window]   Fig. 8. Negative force effect. A: this SII cell (pj06cv  cl1) showed decreased AFRs as force increased from 12.5 to 87.5 g (centers of force bins) for all 4 gratings. Data points with fewer than 3 trials per condition have been omitted for clarity. B: raster density decreases as force increases. In the rasters and histograms, force means ± SEs for trials grouped by force bins, 12.5, 37.5, 62.5, and 87.5 g, were 15 ± 1, 39 ± 1, 59 ± 1, and 87 ± 1 g, respectively. The line traces show average force levels increased as intended during the experiment. C: the cell gave a crisp, rapidly adapting response to manual stimulation of the glabrous skin and dorsal hair of the hand and digits, bilaterally. Format as in Fig. 7.

The 15 cells with positive and negative effects of force that were studied with systematic variation of force and groove width are plotted in Fig. 6 (middle). AFRs were scaled for force as they were for groove-width (Fig. 6, top). These scaled average rates were plotted as functions of the centers of bins used to categorize force for the ANOVA models, positive functions on the left, negative on the right. As with groove width, a similar range of slopes was observed for positive and negative responses to force.

Positive and negative response functions for increasing scanning speed

Thirty-four of 53 SII cells studied in the speed/groove-width experiments had significant differences in AFRs when speed changed. Fourteen of 34 of these effects were additive or isolated; 11 of these had effects of speed and at least one other variable, and 3 had isolated effects of speed. In 20/34 cells these effects were nonadditive with effects of groove width and/or force (ANOVA interactionsto be discussed later). Of the 14 cells with additive or isolated effects of speed, 12 cells had positive or negative response functions (significant ANOVA and significant partial Pearson correlation coefficient between AFR and speed, r_AFR/SPEED). Two of 14 cells showed significant differences in AFR as speed changed without significant r_AFR/SPEED's. Eight of the 14 cells had positive AFR response functions for speed, r_AFR/SPEED (mean ± SE) = 0.57 ± 0.06, and the other four cells had negative AFR response functions, r_AFR/SPEED = 0.37 ± 0.10.

Figure 9 shows data from a cell that had a positive response function for speed. Its AFR increased from 44.9 ± 3.6 to 70.6 ± 5.9 spikes/s (averaged across all groove widths) as speed increased from 40 to 120 mm/s [F (2, 34) = 11.53, P = 0.0002, r_AFR/SPEED = 0.44, P < 0.005]. The curves plotted in Fig. 9A also show an upward shift with increasing groove width, demonstrating that groove width also had a significant positive incremental effect on the responses of this cell. AFR rose from 39.7 ± 5.5 to 67.1 ± 4.5 spikes/s (averaged across all speeds) as groove width increased from 1.07 to 2.53 mm [F (3, 34) = 14.79, P = 0.0001, r_AFR/GW = 0.62, P < 0.0001]. Even though the curves in Fig. 9A are not quite parallel, the speed-by-groove width interaction term in the ANOVA model was not significant, meaning there was no statistically reliable change for the slopes of this cell's AFR versus speed functions as groove width changed.

View larger version (47K): [in this window] [in a new window]   Fig. 9. Positive speed effect. A: This SII cell (gc23cv  bak) showed increased AFRs as speed increased from 40 to 120 mm/s for all 4 gratings, 1.07, 1.42, 1.90, and 2.53 mm. B: raster density increases with speed. The rasters and histograms have been sorted by target speeds. Actual speed means ± SEs for trials grouped by these target levels, 40, 80, and 120 mm/s, were 39 ± 0, 79 ± 0, and 118 ± 0 mm/s, respectively. From top to bottom in the rasters, trials with different gratings occur in the time order that the monkey touched them. The interval chosen for statistical analysis shortens with increasing speed because it corresponds to a fixed length of grating. The line traces in the histograms show that the average force was controlled at a constant level while different grating and speed combinations were delivered to the monkey's finger pad in random sequence. C: the cell gave a crisp, rapidly adapting response to stimulation of the glabrous digits and dorsal hairs, bilaterally, and a weak response to the hands. Force threshold was 85 mg on the stimulated finger pad. Format similar to Fig. 3.

The example cell shown in Fig. 10 had a negative response function for increasing speed. This cell decreased its AFR from 62.1 ± 2.6 to 39.1 ± 3.5 spikes/s (averaged across all groove widths) as speed increased from 40 to 120 mm/s [F (2, 77) = 30.07, P = 0.0001, r_AFR/SPEED = 0.66, P < 0.0001]. This effect was consistent on all four gratings (Fig. 10A). A negative effect of groove width can be appreciated as a downward shift in the AFR versus speed functions for gratings with wider grooves. AFR fell from 70.9 ± 2.6 to 33.7 ± 2.9 spikes/s (averaged across all speeds) as groove width increased from 1.07 to 2.53 mm [F (3, 77) = 11.03, P = 0.0001, r_AFR/GW = 0.80, P < 0.0001]. There was no statistically significant change for the slopes of this cell's AFR versus speed functions as groove width changed (i.e., no speed-by-groove width ANOVA interaction). The negative speed effect is clearly evident in the decreasing raster density and histogram bin heights, without additional sorting by groove width (Fig. 10B).

View larger version (54K): [in this window] [in a new window]   Fig. 10. Negative speed effect. A: this SII cell (pj17cv  cl1) showed decreasing AFRs with increasing speed from 40 to 120 mm/s for all 4 gratings, 1.07, 1.42, 1.90, and 2.53 mm. B: actual speed means ± SEs for trials grouped by target speed levels, 40, 80, and 120 mm/s, were 40 ± 0, 80 ± 0, and 120 ± 0 mm/s, respectively. In addition to an obvious decrease in raster density from 40 to 120 mm/s, note that average force levels are comparable in spite of trial-wise randomization of speed and grating groove width. C: low- and high-speed oscilloscope traces show that this cell sometimes gave burst discharges consisting of pairs of action potentials. There was no obvious relationship between the time of occurrence of these bursts and any stimulus feature. Isolation was lost before receptive field assessment could be made. Format similar to Fig. 9.

Figure 6, bottom 2 graphs, shows response functions for the 12 cells with positive and negative response functions for increasing scanning speed. These plots of scaled AFR versus speed illustrate the comparable range of slopes for positive and negative functions.

Multiple and nonadditive effects

Most SII cells showed effects of groove width and force, groove width and speed, or all three factors. The distributions of multiple effects for the two kinds of experiments were as follows. In the 37 force/groove-width experiments, 28/32 (87.5%) cells, with significant ANOVA effects of groove width or force, had significant effects for both variables. Three cells had isolated effects of groove width, one cell had an isolated effect of force. The five other studied cells were not sensitive to either groove width or force. In the 53 speed/groove-width experiments, 34/43 (79.1%) cells, with significant effects of groove width, force, or speed, had significant effects for combinations of two or three variables. Ten of these cells had significant effects of groove width, force, and speed. One cell had effects of speed and force, 20 cells had effects of groove width and speed, and three cells had effects of groove width and force. Three cells had isolated effects of speed, three cells had isolated effects of groove width, and three cells had isolated effects of force. The remaining 10 cells studied with speed and groove width had no statistically significant effects of any variable. All four cells that were well characterized for both force and groove width and speed and groove width responded to multiple stimulus components. One of these four cells had significant ANOVA effects of groove width and force, and three of these cells had significant effects of groove width, force, and speed. These four cells contributed to the frequencies reported in the preceding text. Only 6/63 (9.5%) cells that had a significant ANOVA effect of groove width responded to groove width in isolation; i.e., to groove width but not to force or speed. This number is probably an overestimate of the true number of groove width-independent responses in this sample population, because effects of speed could not be evaluated in the force/groove-width experiments, and effects of force could only be evaluated in a subset of the speed/groove-width experiments (see METHODS).

We found roughly equal numbers of additive or nonadditive combined effects: 31 additive versus 32 nonadditive effects for groove width, 25 versus 18 for force, and 14 versus 20 for speed. That is, SII cells in the sample typically had either ANOVA main effects of two or three variables (additive effects, where changing 1 variable shifted the AFR function of another variable up or down) or ANOVA interactions (nonadditive effects, where changing 1 variable changed the slope of the AFR function of the other variable, with or without shifting the absolute level of that function). The dual effects that were graphed in the previous figures were additive.

Figure 11 provides an example of a cell with nonadditive effects. This cell was studied in a speed/groove-width experiment. In Fig. 11A the slope of the AFR versus speed function increases with progressively rougher gratings. In Fig. 11B, there was essentially no change in raster dot density at different speeds on the 1.07 mm grating, but there was an increasingly steeper rise in raster density with greater speeds on rougher gratings (1.42-2.53 mm). This cell had a significant groove-width-by-speed interaction [F (6, 81) = 9.49, P = 0.0001, r_AFR/GW = 0.72, P < 0.0001, r_AFR/SPEED = 0.46, P < 0.0001]. The AFR response function depicted for this cell in Fig. 11A provides an example whose AFR discriminated groove widths between categories of rough and smooth at 80 and 120 mm/s but also within the rough category at 120 mm/s. It showed significant changes in AFR from 1.42 to 1.90 mm, 33.5 ± 3.1 to 57.3 ± 4.3 spikes/s (t = 4.57, df = 15, P < 0.001), for trials at 80 mm/s, and 40.4 ± 3.6 to 63.8 ± 3.2 spikes/s (t = 4.86, df = 16, P < 0.0005), for trials at 120 mm/s. It discriminated between 1.90 and 2.53 mm at 120 mm/s, with AFRs of 63.8 ± 3.2 to 86.5 ± 8.3 spikes/s (t = 2.67, df = 15, P < 0.03); a trend for discriminating all the gratings is clearly visible by the separation of the points at 120 mm/s.

View larger version (72K): [in this window] [in a new window]   Fig. 11. Speed by groove width interaction. A: this SII cell (pj14cv  bak) showed more rapidly increasing AFRs with increasing speed from 40 to 120 mm/s, as groove width increased from 1.07-2.53 mm. B: rasters have been sorted by speed, 40, 80, and 120 mm/s and groove width, 1.07, 1.42, 1.90, and 2.53 mm and show more rapid increases in density with speed as groove width increases. C: a photograph of the oscilloscope screen demonstrates the high quality of the isolation. D: the cell gave a rapidly adapting response to cutaneous stimulation of glabrous and dorsal digit 3. Format similar to Fig. 3.

The distribution of positive and negative AFR functions for cells showing interaction effects resembled that of the group with purely ANOVA main effects. For the cell in Fig. 12A, curves representing the AFR versus groove width functions shift upward with increasing force. Note that the slope of the curve for the 87.5 g force bin was steeper than the slopes for the other two force bins. Responses of this cell increased significantly with groove width [F (3, 130) = 18.76, P = 0.0001, r_AFR/GW = 0.50, P < 0.0001] and force [F (3, 130) = 12.74, P = 0.0001, r_AFR/FORCE = 0.48, P < 0.0001], and this cell had a significant groove-width-by-force interaction [F (7, 130) = 2.90, P < 0.01]. The example cell in Fig. 12B showed less activity on rougher surfaces and at higher force levels, and it showed more steeply sloped negative effects of groove width at the lightest applied forces. It had negative groove width [F (3, 123) = 8.37, P = 0.0001, r_AFR/GW = 0.24, P < 0.005] and force [F (4, 123) = 49.96, P = 0.0001, r_AFR/FORCE = 0.63, P < 0.0001] main effects and a groove-width-by-force interaction [F (11, 123) = 4.72, P = 0.0001].

View larger version (26K): [in this window] [in a new window]