Abstract
Dexterous object manipulation depends critically on information about forces normal and tangential to the fingerpads, and also on torque associated with object orientation at grip surfaces. We investigated how torque information is encoded by human tactile afferents in the fingerpads and compared them to 97 afferents recorded in monkeys (n = 3; 2 females) in our previous study. Human data included slowly-adapting Type-II (SA-II) afferents, which are absent in the glabrous skin of monkeys. Torques of different magnitudes (3.5–7.5 mNm) were applied in clockwise and anticlockwise directions to a standard central site on the fingerpads of 34 human subjects (19 females). Torques were superimposed on a 2, 3, or 4 N background normal force. Unitary recordings were made from fast-adapting Type-I (FA-I, n = 39), and slowly-adapting Type-I (SA-I, n = 31) and Type-II (SA-II, n = 13) afferents supplying the fingerpads via microelectrodes inserted into the median nerve. All three afferent types encoded torque magnitude and direction, with torque sensitivity being higher with smaller normal forces. SA-I afferent responses to static torque were inferior to dynamic stimuli in humans, while in monkeys the opposite was true. In humans this might be compensated by the addition of sustained SA-II afferent input, and their capacity to increase or decrease firing rates with direction of rotation. We conclude that the discrimination capacity of individual afferents of each type was inferior in humans than monkeys which could be because of differences in fingertip tissue compliance and skin friction.
SIGNIFICANCE STATEMENT We investigated how individual human tactile nerve fibers encode rotational forces (torques) and compared them to their monkey counterparts. Human hands, but not monkey hands, are innervated by a tactile neuron type (SA-II afferents) specialized to encode directional skin strain yet, so far, torque encoding has only been studied in monkeys. We find that human SA-I afferents were generally less sensitive and less able to discriminate torque magnitude and direction than their monkey counterparts, especially during the static phase of torque loading. However, this shortfall in humans could be compensated by SA-II afferent input. This indicates that variation in afferent types might complement each other signaling different stimulus features possibly providing computational advantage to discriminate stimuli.
Introduction
When manipulating held objects, rotational forces (torques) commonly develop at the thumb and fingers (Kinoshita et al., 1997). Sensing rotational forces at the digits is important to ensure grip safety and for controlling an object's orientation in the hand. For example, when cutting with a knife, rotational slips should be avoided while also maintaining a desired torque magnitude to counteract variations in reactive forces on cutting. One can reorient an object by controlling the grip force magnitude that is sufficient to prevent tangential slips, but weak enough to allow controlled rotational slips. For example, we may allow a glass filled with water to rotate between our fingers to ensure that it aligns with gravity without inducing net vertical translation (tangential slip) resulting in loss of the object (Johansson and Westling, 1984; Kinoshita et al., 1997). Therefore, realization of action goals requires precise grip-force adjustments to both tangential and rotational forces, achieving a desired object orientation by applying torque, or purposefully reducing it by controlling rotational slips.
Skin mechanoreceptors, and their corresponding afferents, contribute to the control of dexterous object manipulation by signaling gripped object features, fingertip forces, and grip safety (Johansson and Westling, 1987; Goodwin et al., 1998; Nowak et al., 2003; Johansson and Flanagan, 2009; Crevecoeur et al., 2011; Khamis et al., 2014; Delhaye et al., 2021; Schiltz et al., 2022). Sensory inputs can rapidly update internal representations of hand-object manipulations to counteract unexpected events and build predictive strategies (Johansson and Westling, 1988; Johansson and Cole, 1992; Miall and Wolpert, 1996; Birznieks et al., 1998; Witney et al., 2004; Flanagan et al., 2008), which might also include planned slips (Birznieks et al., 1998).
Tactile afferent responses to fingerpad torques have previously been characterized only in macaques. Most fast-adapting Type-I (FA-I) and slowly-adapting Type-I (SA-I) low-threshold mechanoreceptive afferents innervating the fingertip had response impulse rates that were scaled by torque magnitude (Birznieks et al., 2010). SA-I afferents showed a torque direction preference, whereas FA-I afferents responded to changes in torque magnitude regardless of direction. Models trained to decode afferent input demonstrated that several combinations of normal force, torque magnitude, and direction could be accurately classified (Birznieks et al., 2010; Redmond et al., 2010a, b; Fu et al., 2012), and instantaneous grip force and torque parameters concurrently extracted from a small number of tactile afferents' responses, in a real-time fashion (Khamis et al., 2015).
Until now, no human afferent torque responses have been available for such modeling. Unlike monkeys, human glabrous skin possesses slowly-adapting Type-II (SA-II) afferents showing exquisite sensitivity to the magnitude and direction of strain induced by tangential skin stretch (Knibestöl and Vallbo, 1970; Knibestöl, 1975; Johansson, 1978; Olausson et al., 2000; Birznieks et al., 2001, 2009). Moreover, they can elicit sustained percepts of diffuse pressure, squeezing, or strain (Macefield et al., 1990; Watkins et al., 2022), thus making them ideal torque sensing candidates. Fingertip biomechanical differences between species may also influence receptor response properties. Human fingertips are larger and flatter than most nonhuman primate species, and differ in viscoelastic and frictional skin properties (Dandekar et al., 2003), which could lead to torque-encoding differences.
Here, we quantify human single FA-I, SA-I, and SA-II afferent torque responses, recorded via microelectrodes inserted percutaneously into the median nerve at the wrist; the stimuli were too slow to activate FA-II (Pacinian) afferents. We were particularly interested in the potential signaling advantages offered by SA-II afferents, which are present in human, but not monkey hand glabrous skin (Johnson, 2001). We applied torques of three different magnitudes and two directions (clockwise and anticlockwise), superimposed on three different background normal forces, to a central site on the glabrous skin of human fingertips. First, we illustrate and characterize human afferent torque response properties, then we compare the torque discriminative capacity we observed in human afferents, and compare them to a monkey afferent dataset, previously published by Birznieks et al. (2010).
Materials and Methods
Single-unit tactile afferent recordings
Single-unit tactile afferent recordings were acquired from 34 participants (median age 22, range 19–54; 19 females) using the standard microneurography procedures employed in our laboratory (Condon et al., 2014). Each participant provided informed written consent to the procedures in accordance with the Declaration of Helsinki, and all procedures were approved by the UNSW Human Research Ethics Committee. Participants reclined in a dental chair with their forearm supinated and Velcro straps around the wrist were used to secure the forearm to a vacuum cast to immobilize the arm. The dorsum of the hand was embedded in plasticine up to the mid-level with the fingers splayed but the distal end of the phalanges did not contact the plasticine, thereby allowing normal skin deformation mechanics during mechanical stimulation. To stabilize the digits, the nails of the index, middle, and ring fingers were glued to metal posts firmly sunk into the plasticine, and the thumb and little finger were secured by U-shaped aluminum clamps embedded in the plasticine.
Electrical impulses were recorded from individual low-threshold mechanoreceptive afferents with cutaneous receptive fields in the distal phalanx of the index, middle, or ring finger by percutaneously inserting tungsten needle electrodes (FHC Inc) into the median nerve at the wrist. Neural signals were amplified (20,000×) and filtered (300 Hz to 5 kHz) through an isolated amplifier (ISO-Dam 80, WPI). The microelectrode was manually guided into a cutaneous fascicle by delivering weak electrical pulses (0.2 ms, 1 Hz, 0.01–1.00 mA) via an isolated stimulator (Stimulus Isolator, ADInstruments); radiating paresthesia at 0.02 mA indicated that the tip had impaled a fascicle. Once the fascicle had been identified, the microelectrode was manipulated while the distal phalanx was mechanically stimulated and action potentials from a single sensory axon encountered.
Classifying the afferent population
During experiments the response threshold of each afferent was detected to mark the location of the receptive field center (RFC) and to establish the receptive field size. Calibrated nylon filaments (Semmes-Weinstein aesthesiometers, Stoelting) were used to establish afferent response thresholds determined as the lowest force that evoked responses to at least three out of four (75%) indentations. If an afferent was spontaneously active, the smallest force that could be seen to modulate its discharge rate (typically by ∼10%) was defined as the threshold force. Afferents were classified according to their distinct response properties as described previously by Vallbo and Johansson (1984). Afferents could be differentiated into fast adapting (FA) or slowly adapting (SA) by their responses to sustained suprathreshold force indentation with a nylon filament, whereby FA afferents were differentiated from SA afferents by their rapid adaptation to sustained stimuli, and characteristic firing as the nylon filament was retracted, which was not seen in any SA afferents. The receptive field borders of isolated afferents were defined as the region of skin from which a response could be elicited by a force four times the threshold force at the most sensitive zone in the receptive field. Each afferent was characterized as Type I if they had small discrete receptive fields, typically <15 mm, and as Type II if they had large receptive fields with poorly defined borders, typically >50 mm (Johansson, 1978). To qualitatively differentiate SA-I and SA-II afferent responses during experiments, we also determined whether afferents fired spontaneously, the sharpness of the discharge rate increase at stimulus onset, and the variability of the discharge rates during the plateau phases of force and torque stimuli. SA-II afferents also exhibit characteristic sensitivity to the skin stretch in a particular direction.
To objectively categorize slowly-adapting afferents that responded to torque, we used factor analysis of mixed data (FAMD, famd function in the FactoMineR R package; Pagès, 2004; Lê et al., 2008), which is a method of principal component analysis that accounts for both quantitative and qualitative variables. Inputs to the FAMD included whether an afferent responded to skin stretch or not, the receptive field size in mm2, and an Irregularity Index; a measure of the regularity of an afferent's firing rate in response to sustained indentation. The irregularity index was calculated as the mean of the absolute discharge rate differences measured between every two pairs of consecutive spikes divided by the mean discharge rate during the normal force plateau phase for each individual trial (Birznieks et al., 2008). Following FAMD, observations were clustered in factorial space using K-means clustering (kmeans function in R) on the first two principal components and the silhouette values computed. This resulted in two easily separable clusters of 24 SA-I and 13 SA-II afferents, with a mean silhouette value of 0.55, which confirmed our initial classification as SA-II afferents respond well to directional skin stretch, have very large receptive fields, and more regular firing rates than SA-I afferents.
We recorded from 83 low-threshold mechanoreceptive afferents with receptive field centers (RFCs) on the fingertip glabrous skin of digits 2, 3, and 4 of the left hand. Thirty-nine afferents were classified as fast-adapting Type-I (FA-I), 31 afferents were classified as slowly-adapting Type-I (SA-I), and 13 as slowly-adapting Type-II (SA-II). We excluded recordings from FA-II afferents because they did not respond reliably to our stimuli (Birznieks et al., 2001, 2010).
Previously published neural spike data acquired from 25 FA-I and 58 SA-I afferents in three anesthetized Macaca nemestria monkeys (n = 3; 2 females), as described previously by Birznieks et al. (2010), were also used to compare human and monkey afferent torque responses. The stimulation protocols and equipment were identical to those used in the monkey study: the magnitudes of the background normal forces were 1.8, 2.2, and 2.5 N, torque magnitudes of 2 and 3.5 mNm were applied at each normal force, and depending on the normal force 4, 4.5, or 5.5 mNm torque was applied at 1.8, 2.2, and 2.5 N normal force, respectively.
Stimulus
A custom-made stimulator was used to deliver normal forces and torques (Birznieks et al., 2010). The stimulus applicator was a disk (24 mm in diameter) with a flat stimulation surface. The disk was positioned so that the rotational torque axis was aligned over the center of the flat portion of the volar surface of the fingertip and perpendicular to the skin surface (Fig. 1A). Because of the rounded shape and viscoelastic properties of the fingertip skin and underlying tissues, afferents with receptive field centers even at the edges of the distal fingertip have been shown to respond vigorously to a standard stimulus site in the center of the fingerpad, and encode features like the force, stretch, and curvature of the stimulus (Bisley et al., 2000; Birznieks et al., 2001, 2009; Jenmalm et al., 2003). Thus, the center of the fingertip typically contacted by gripped objects (Christel et al., 1998) was chosen as the standard test site on the receptor-bearing finger, referred to as the rotational center, regardless of receptor location. The advantage of this approach was that it would provide a more realistic representation of the variations in spatiotemporal response patterns of a population of afferents, considering the finger's anatomic and histologic features.
Normal force and torque stimulus application phases. A, Schematic representation of the stimulus applicator during fingertip stimulation. The surface was oriented parallel to the flat portion of the fingertip and was held just above the skin surface. The center of the rotational axis for torque applications is indicated by black arrows. In each trial, the stimulator was advanced to compress the skin until the desired normal force was reached, then torque was applied, held at the plateau then it was rotated back to the starting point after which the applicator retracted from the skin surface to return to zero normal force. B, The phases of normal force and torque applications over time. Different magnitudes (only 7.5 mNm is shown here) of torques were tested in clockwise (blue) and anticlockwise (red) directions.
Torques were applied at 3.5, 5.5, and 7.5 mNm in clockwise and anticlockwise directions. Each of the torque magnitudes and directions were applied with three different normal forces: 2, 3, and 4 N, corresponding to grip forces generated during object manipulation. Forces and torques applied to the fingertip were measured by a six-axis, force-transducer (Nano F/T; ATI Industrial Automation) at 1000 Hz sampling rate with a force resolution of 0.0125 N and a torque resolution of 0.0625 mNm. The stimulus applicator was lowered to slightly above the skin and the damper setting adjusted to facilitate smooth contact with the skin surface. Each trial consisted of a dynamic phase of increasing normal force for 0.2 s, a static phase in which normal force was constant for 4.0 s, and a dynamic retraction phase lasting 0.2 s. Torques were superimposed during the constant normal force phase, commencing 1 s after the beginning of this phase (Fig. 1B). The torque loading phase lasted 0.5 s, was held constant for 1.5 s during a plateau phase, followed by an unloading phase which lasted 0.5 s. Normal force unloading commenced 0.5 s after the end of the torque unloading phase (Fig. 1B). Example traces from an SA-II afferent responding to each of the torque phases, and the neural spikes sorted for analysis are shown in Figure 1C. An example of the calculated instantaneous discharge rate is shown at the bottom.
Each stimulus set with a given background normal force consisted of presenting two normal-force only trials followed by three trials, each with one of the three torque magnitudes (3.5, 5.5, and 7.5 mNm) superimposed in ascending order. One experimental run comprised this stimulus set applied three times, which took ∼2 min. An experimental run was then repeated, resulting in six recordings of responses to stimuli of each torque magnitude and direction. Sometimes, when recording quality and time permitted, additional experimental runs were obtained, resulting in nine or 12 recordings for each torque magnitude and direction. The experimental runs were presented with 2 N normal force and in one torque direction and then the two runs were repeated for the same normal force but in the opposite torque direction. The full procedure was repeated with 3 and 4 N normal forces. In most cases torques superimposed on 2 N normal force were tested before 3 N and 4 N, so there was a larger pool of data with 2 N normal force. Together, there were 18 normal force and torque combinations: three torques presented in each direction (n = 6), each superimposed on three normal forces. We aimed to achieve six presentations of each normal force torque combination. However, because of limited recording time in human studies, as signal-to-noise ratio sometimes declined before all stimuli could be tested not every afferent could be tested with all 18 normal force and torque combinations and the desired number of repeats.
Standardized finger
To standardize data collected from different subjects and fingers, receptive field centers RFCs were referenced to a standardized fingertip on the right hand, by normalizing the coordinates of each fingertip and the RFC distances from the stimulus center to the measurements of the average fingertip size. Afferent RFCs on the left hand and torque directions applied to the left hand were treated as mirror images of the right hand and plotted accordingly on the standardized finger (Birznieks et al., 2010). Therefore, clockwise was defined as rotation of the stimulus applicator from ulnar to distal to radial aspects of the fingertip, regardless of the hand stimulated.
Signal analysis
Afferent neural activity was acquired using a PowerLab 16/35 data acquisition system and viewed in LabChart 5 (ADInstruments). Recorded neural spikes were sorted and selected for further analysis using custom-written code in Igor Pro 5 (Wavemetrics). Sorted spikes were analyzed using custom-written code in MATLAB (The MathWorks). In response to each stimulus, we calculated the mean discharge rate in the torque loading phase, plateau phase, and unloading phase. We also calculated the instantaneous peak discharge rate during the dynamic loading and unloading phases. For afferents without ongoing activity (generally FA-I and SA-I) we also quantified the latency of occurrence of the first spike after the beginning of the torque loading phase.
Data analysis
To determine whether afferent discharge rates were correlated with torque magnitudes, we calculated Spearman's rank correlation coefficients between the mean discharge rate and torque magnitude. Spearman's rank correlations are nonparametric tests, and thus do not assume a linear relationship between torque magnitude and discharge rate, enabling the capture of various types of accelerating or decelerating stimulus-response relationships, without the assumption of linearity. Correlation coefficients were calculated separately for each torque direction and each normal force magnitude. Responses to normal force only (zero torque) were not included in the correlation analysis. Afferents that responded to torque but did not show response scaling to the torque magnitudes tested in this study were analyzed to determine whether they had a nongraded response to torque by comparing responses of normal force only trials to trials with torque applications, regardless of magnitude, using the Mann–Whitney–Wilcoxon test (MATLAB, The MathWorks). To compare torque effects between afferents and between stimulus conditions, we calculated torque sensitivity over the whole torque range 0–7.5 mNm, ST. The number of spikes per second was plotted against the torque magnitude, and torque sensitivities were calculated as the gradient of the regression line. Thus,
d-prime (d') analyses were calculated for torque magnitude as:
Where
Experimental design and statistical analysis
Unless otherwise stated, statistical analyses were performed in R (version 4.0.2; R Core Team, 2020). To assess differences in torque sensitivity and discriminability we used one-way, two-way, or three-way ANOVA or repeated measures ANOVA, where necessary (anova_test function, rstatix package version 0.7.0; Kassambara, 2023). For one-way nonparametric comparisons the Kruskal-Wallis rank sum test was used (kruskal_test function, stats package; R Core Team, 2020). Effect sizes were reported where appropriate, as η squared (ŋ2), a measure of the proportion of variance accounted for by a variable, or F values. Post hoc pairwise comparisons were performed with the Tukey's adjustments for ANOVAs and Bonferroni correction method to adjust for repeated measures ANOVAs. Results are written as mean ± SEM, or median and range. The probability selected as significant was p < 0.05, for all tests.
Results
Characteristics of human afferent responses to torque
We examined the instantaneous discharge rates of individual afferents of each type in response to different torque magnitudes and directions (example torque traces in Fig. 2A). SA-II afferents have large receptive fields that sometimes encompassed the whole fingertip, and some afferents had an ongoing discharge without the presence of a stimulus (Knibestöl and Vallbo, 1970; Knibestöl, 1975). Typically, SA-II afferents showed increased firing rates at the onset of normal force application, which adapted to a steady level during the normal force plateau. Examples of several distinct SA-II response profiles to torque magnitudes applied in both directions are shown in Figure 2B–D. Torque application in one of the two directions often caused SA-II afferents to show pronounced discharge rate increases and their responses could be decreased by torques applied in the opposite direction. In some SA-II afferents instantaneous discharge rate peaked during the loading phase, after which discharge rate slightly decreased to a steady level during the plateau phase, followed by a change in discharge rate with the opposite sign during the unloading phase (Fig. 2B–D). Figure 2B,D show afferents whose discharge rates were positively changed by anticlockwise torques and negatively changed (suppressed) by clockwise torques during the loading and plateau phases. Figure 2C shows a less common example of an afferent with a decreased discharge rate in response to torque in one direction and was not significantly influenced by torque in the opposite direction. In summary, SA-II response behaviors included having an (1) increased or decreased discharge rate by one torque direction, (2) increased discharge rate by one torque direction and decreased discharge rate by the opposite direction, or (3) increased or decreased discharge rate by both torque directions.
Responses of individual afferents to torque with 2 N background normal force. A, Example torque stimuli from single trials of 7.5 mNm in clockwise (blue) and anticlockwise (red) directions and to normal force only (black trace). B–D, Responses from three individual SA-II afferents with various response profiles showing enhanced discharge rates in response to one torque direction and suppressed discharge rates to the opposite torque direction (B, C), or suppressed discharge rates to both torque directions (D). Instantaneous rate from an (E) SA-I afferent and (F) FA-I afferent that had an enhanced discharge rates in response to both torque directions. Three torque magnitudes of 3.5, 5.5, and 7.5 mNm were applied in each direction, but only responses to 7.5 mNm torque are included here, for clarity. Vertical gray dashed lines separate the torque phases: loading (L), plateau (P), and unloading (U). G, Histograms of the number of impulses per 100-ms bins in response to 7.5 mNm torque in either direction (example torque traces are shown above). Responses are represented as Z-scores of the difference in the number of impulses between torque trials and normal force only (0 mNm torque) trials. Each afferent tested in two directions is represented with anticlockwise and clockwise torque responses on the left and right sides respectively, one for each torque direction. Raster rows filled in with light gray indicate afferents in which one torque direction was not tested. L, loading phase; P, plateau phase; U, unloading phase.
SA-I afferents had small and discrete receptive fields, and increased discharge rates to normal force application, showing a rapidly increasing discharge rate at the onset, which either continuously declined or declined to a steady state during the normal force plateau. At torque onset, SA-I afferents most frequently showed positive discharge rate changes, but we also observed some SA-I afferents in which torque decreased their response to background normal force. Figure 2E shows an example response of a typical SA-I afferent. During the loading and unloading phases, this afferent had an increased discharge rate to both clockwise and anticlockwise torques. During the plateau phase, only the clockwise torques caused clear enhancement of the discharge rate (Fig. 2E).
As expected, FA-I afferents responded to normal force and torque magnitudes exclusively during the dynamic phases of stimulation (loading and unloading phases). An example of a torque response profile from a typical FA-I afferent is shown in Figure 2F. This afferent's discharge rate increased by torque magnitude applied in both directions during the loading and unloading phases, but it was silent during the torque plateau phase. As FA-I afferents are not activated by a steady background normal force, they cannot show negative changes in discharge rate by torque.
Next, we examined responses from individual afferents across the whole population (Fig. 2G). Z-scores were calculated and plotted in histograms for the number of impulses in 100-ms bins across six trials in response to 7.5 mNm torque of both torque directions, compared with normal force only (0 mNm torque) trials. Most afferents followed the typical response patterns of the example afferents in Figure 2B–F. Notable contrasts between the afferent types are the delayed rise to peak responses and sustained positive discharge rate change of SA-II afferents, compared with the more rapid rise to peak firing of SA-I afferents and their more rapid decline after peaking. Firing rates of SA-I afferents also tended to drop below background levels over the course of the plateau period (purple bins; Fig. 2G) even when the torque loading phase initially increased their discharge rate.
For SA-I and SA-II afferents, responses that were decreased during the torque loading and plateau phases often had increased discharge rates during the unloading phase. At the end of the unloading phase the stimulation surface was returned to its starting rotational position from which torque was applied. However, this did not result in torque returning to zero, there was always a torque overshoot (see Fig. 2A, unloading phase) in the opposite direction, which decreased over time while the surface remained fixed in position indicating that, at least in part, this effect was caused by viscoelastic properties of the skin tissue (de Dunilac et al., 2023). It is also possible that, during torque loading, localized “slip-and-stick” events occurred at the skin-contact area edge, particularly at higher torque magnitudes. When the surface was returned to its starting position, these skin regions, which because of “slip-and-stick” events did not move all the way with the probe, were now moved back by the full angular distance, thereby pushing those skin regions in the opposite direction, and generating a torque overshoot at the end of torque loading. The torque overshoot in the opposing direction could cause the discharge rate to peak (Fig. 2B,C,E) in slowly-adapting afferents which, unlike FA afferents, do not typically respond to stimulus-off events. Similarly, afferents with increased discharge rates in response to torque during the loading and plateau phases often returned to the baseline normal force discharge rate and might show negative peaks during the torque unloading overshoot, although this effect was less frequent. This mechanical phenomenon that causes a hysteresis effect and enhances afferent responses during unloading seems to be representative of natural movements during object manipulation. In such cases, torque increases and decreases during consecutive stages of manipulation when the object is tilted and then returned to its original orientation, like when cutting with a knife, or painting with a brush.
SA-II afferent responses to torque
Size of SA-II afferent population influenced by torque
The SA-II afferents had RFCs distributed across most of the distal phalanx and only one of these afferents' responses was not scaled by torque, suggesting that, regardless of the RFC location on the fingertip, SA-II afferents are capable of signaling changes in torque (Fig. 3A). A more granular analysis and generalization of the relationship between SA-II receptive field center locations and their response properties was not attempted because of the large size of their receptive fields. SA-II receptive fields often spanned large portions of the finger and could encompass the whole stimulus contact area. Receptive field's size and each afferent's stretch sensitivity in its own specific preferred direction, would make such analyses difficult to interpret.
Effect of torque on SA-II afferents in each torque phase. A, The receptive field center location (black dots) of each SA-II afferent and the direction in which its response was scaled by torque are plotted for each torque phase across a standard fingertip. Red and blue arrows indicate positive discharge rate change in anticlockwise and clockwise directions, respectively, and black arrows indicate negative change in discharge rate (suppression) in the respective direction. Black cross indicates torque stimulus rotational center and the gray circle indicates the approximate stimulus contact area. B, The proportion of SA-II afferents with scaled mean discharge rates in response to torque, with 2 N background normal force, is shown in darker columns. Red and blue bars represent afferents sensitive to anticlockwise and clockwise directions, respectively; gray bars represent afferents sensitive in at least one of the torque directions. The lighter colored stacked columns show the number of afferents that had nongraded responses to torque. Colored numbers above indicate the number of afferents with torque responses and in brackets the proportion of these that only showed nongraded responses.
Mean discharge rate
Using the mean discharge rate during each torque phase as the response measure we used Spearman's rank correlations to determine whether torque magnitude significantly influenced the mean discharge rate. Most SA-II afferents were significantly influenced by torque with 2 N background normal force and had directional differences. For all mean discharge rate comparisons, afferents that were only tested with one torque direction were excluded, so that the proportions of afferents with responses scaled by clockwise and anticlockwise torques could be compared. Moreover, only afferent responses to torques applied with 2 N background normal force were included, as 82 of 83 afferents were stimulated with this background normal force, but not necessarily with higher background normal forces. Finally, we also examined the torque effects on afferent peak discharge rates, to investigate dynamic response scaling, and we found the relationship to be the same as for mean discharge rate (data not shown).
Proportions of SA-II afferents with mean discharge rates scaled by the torque loading phase are shown in Figure 3B. In some trials, afferent responses to torque were significantly different from trials with normal force only (0 mNm torque) but did not show discernible response scaling to the torque magnitudes in each direction, within the magnitude range tested in this study. We refer to such afferents as having a nongraded response to torque in each direction, which were determined using the Mann–Whitney–Wilcoxon test. In total, 100% (13/13) of SA-II afferents were either scaled or had nongraded responses to torque in at least one direction. Including graded and nongraded responses, SA-II afferents had strong torque direction selectivity among the population, as five afferents' positive-negative changes in discharge rate switched between opposite directions, seven responded with positive or negative changes in discharge rate to torque, selectively in one of the directions, and only two out of 13 seemed to be indifferent and had positively changed discharge rates to torque in either direction. Torque scaling was similar during the plateau phase (Fig. 3B). When nongraded responses to torque were taken into account, then 92% (12/13) of SA-II afferents were either scaled or had nongraded responses to torque in at least one direction. Fewer SA-II afferents had responses scaled by torque during the unloading phase (Fig. 3B). When nongraded responses to torque were considered, 92% (12/13) of SA-II afferents were either scaled or had nongraded responses to torque in at least one unloading direction.
Effect of normal force on SA-II torque responses
Individual SA-II afferent responses
Responses to torque of many individual SA-II afferents were clearly influenced by the background normal force, despite the torque stimulus remaining the same (Fig. 4A). An example afferent #1 had high torque sensitivity with 2 N background normal force (Fig. 4A, left), but the torque responses were greatly reduced with higher background normal force 3 N (Fig. 4A, middle) and 4 N (Fig. 4A, right). Another interesting example of an SA-II afferent (#2) with responses that were influenced by background normal force is shown in Figure 4B. With 2 N background normal force, this afferent showed an increased discharge rate during the loading and plateau phases to torques in both directions but, when the background normal force was increased to 3 N (Fig. 4B, middle) and 4 N (Fig. 4B, right), the effect of anticlockwise torques switched from an increased to decreased discharge rate. These examples suggest that background normal force is important in determining the torque sensitivity of an afferent and increasing the normal force can either increase or decrease torque sensitivity of SA-II afferents. It is likely that fingertip geometry and mechanics, in combination with the receptive field location determines those complex interactions between the afferent torque and normal force responses.
The effect of background normal forces on SA-II afferent responses to torque. A, Responses to torque from the same single afferent are shown when the torques were applied with background normal forces increasing from 2 to 4 N (example torque at the top). The first SA-II afferent shows a typical response profile with decreasing torque sensitivity as background normal force increases. B, Torque responses from another afferent show sensitivity to clockwise torques that remained similar with increasing background normal forces but, in response to anticlockwise torques, responses transitioned from positive to negative change in discharge rate with increasing background normal force. See Figure 2 legend for detailed description.
SA-II afferent population effects
Across the whole SA-II afferent population there was a decreasing proportion of afferents with responses scaled by torque during the loading phase with increasing background normal force (Fig. 5A). During the plateau and unloading phases, the relationship was less clear, but in both cases the proportion of afferents with responses scaled by torque were greater with 2 N background normal force than with 4 N.
Effect of normal force and receptive field center location on SA-II afferent's torque responses. A, The proportion of SA-II afferents that showed responses that were scaled by torque in at least one direction as measured by mean discharge rate, for each normal force. B–D, SA-II torque sensitivities with 2, 3, and 4 N background normal forces, separated by afferents that had significantly (B) increasing or (C) decreasing torque sensitivity with increasing background normal force, or were unaffected by background normal force (D). Red and blue semi-circles indicate that anticlockwise and clockwise torque sensitivity, respectively, was significantly increasing or decreasing with increasing background normal force. Black semi-circles indicate no influence in the respective direction. Only afferents that were tested with torque at each of the three normal forces are included. Values from individual afferents are connected by lines. The fingertip contours to the right of panels indicate only approximate receptive field center locations of these afferents as SA-II afferents have large receptive fields with diffuse borders.
To determine whether individual afferent sensitivity measures to torque (ST) were scaled by the normal force on which torque stimuli were superimposed, we used Spearman's rank correlation analyses. Afferents that were tested with all three normal forces (n = 10) were used in these analyses. We investigated this relationship only during the torque loading phase, as this is when encoding torque changes is most important for object manipulation and lifting. In Figure 5B–D, we separated the afferents into three groups; in which normal force increased, decreased, or had no effect on torque sensitivity. We found that 60% (6/10) of SA-II afferents were significantly scaled by background normal force in at least one direction (median correlation coefficient = −0.37, range = −0.79–0.77), of which all six had decreasing torque sensitivity with increasing background normal force, but one of these afferents also had increasing torque sensitivity for torques in the opposite direction (example afferent from Fig. 4B). The distribution of SA-II afferent RFCs over the distal fingertip is illustrated for groups of afferents with increasing, decreasing, or no effect on torque sensitivity (see schematic fingertips to the right of Fig. 5B–D).
SA-I afferent responses to torque
Size of SA-I afferent population influenced by torque
Mean discharge rate
SA-I torque responsive afferents had receptive fields all over the distal phalanx, and in most the mean discharge rate was scaled by torque superimposed on 2 N normal force (Fig. 6A,B). Altogether, 65% (20/31) of SA-I afferents had mean discharge rates scaled by torque in at least one direction during the loading phase from which 12 afferents' responses were scaled by both torque directions (Fig. 6B). Two of the nonscaled afferents had nongraded responses to torque. Proportions of SA-I afferents with torque scaled responses were slightly reduced during the plateau phase (Fig. 6B). Altogether, 52% (16/31) of SA-I afferents had responses that were scaled by torque in at least one direction during the plateau phase from which only three were scaled in both directions (Fig. 6B) and six scaled only in one torque direction had nongraded responses to the opposite torque direction. In two afferents, only nongraded responses to torque in one direction could be detected (Fig. 6B). During the unloading phase, 55% (17/31) of SA-I afferents had responses that were scaled by torque in at least one direction (14 clockwise, 12 anticlockwise, and nine in both directions; Fig. 6B). Four of the nonscaled afferents had nongraded responses to torque (p < 0.05, Mann–Whitney test). Peak discharge rate during the dynamic loading and unloading phases showed slightly lower proportions of scaled afferents, but the relationships were similar to mean discharge rates, and as such data are not shown.
Effect of torque on SA-I afferents in each torque phase. A, The receptive field center location (black dots) of each SA-I afferent and the direction in which its response was scaled by torque are plotted for each torque phase across a standard fingertip. B, C, The proportion of SA-I afferents that showed responses that were scaled by torque in at least one direction are shown in darker columns, as measured by (B) spike count and (C) first spike latency, with 2 N background normal force. Stacked columns in a lighter color in B, C show the number of afferents that had nongraded responses to torque. Colored numbers above indicate the number of afferents with torque responses and in brackets the proportion of these that only showed nongraded responses. See Figure 3 legend for detailed description.
First spike latency
Encoding mechanisms such as the recruitment order of tactile afferents might be a fast way to signal torque parameters at the fingertips. To investigate whether first spike latencies depend on torque, we calculated Spearman's rank correlations between the latency to the first spike in the torque loading phase and torque magnitude. First spike latencies were only calculated for torques applied with 2 N background normal force. Loading phase first spike latencies were scaled in 13% (4/31) of SA-I afferents, in response to either clockwise or anticlockwise torques, and none of the SA-I afferents' first spike latencies were scaled by both torque directions (Fig. 7C). Altogether, 26% (8/31) of SA-I afferents had first spike latencies scaled by torque in at least one of directions (Fig. 6D). On average, SA-I afferent first spike latencies were 50.3% shorter in response to 7.5 mNm compared with 3.5 mNm torque. This analysis was not performed for the SA-II afferents given their spontaneous discharge.
Effect of normal force and receptive field center location on SA-I afferent's torque responses. A, B, The proportion of afferents that were scaled by torque in at least one direction as measured by (A) spike count and (B) first spike latency for each normal force. C–E, SA-I torque sensitivities with 2, 3, and 4 N background normal forces, separated by afferents that had significantly increasing (C) or decreasing (D) torque sensitivity with increasing background normal force, or were unaffected by background normal force (E). F, Sensitivity values from the torque loading phase are shown for SA-I afferents against their receptive field center distance from the stimulus rotational center. ST was calculated from the maximum torque sensitivity value between either clockwise or anticlockwise torque directions. SA-I afferents showed increasing torque sensitivity with distance from the stimulus rotational center, up to 12 mm, and the most pronounced effect was from 6 to 12 mm. SA-I afferents with receptive field center >15 mm from the stimulus rotational center were unresponsive to torque. Vertical dashed line indicates the edge of the probe. Inset to the right shows a radar chart of the torque sensitivity in impulses mNm−1 s−1 (sensitivity increments shown at 0.325 and 0.75 impulses mNm−1 s−1) of afferents with receptive field centers in each quadrant of the fingertip, for each of the three background normal forces. Quadrants are labeled D, distal, P, proximal, U, ulnar, and R, radial. See Figure 5 legend for detailed description.
Effect of normal force on SA-I afferent torque responses
The background normal force influenced the number of SA-I afferents that had mean discharge rates scaled by torque magnitude. When using mean discharge rate as the response measure, there were fewer afferents scaled by torque when normal force increased during the loading and plateau phases (Fig. 7A), and this was also apparent for first spike latencies (Fig. 7B).
To visualize the effect of force on individual afferents within the population, we separated afferents into three groups; those that showed increasing, decreasing, or no effect on torque sensitivity, respectively (Fig. 7C–E) with increasing background normal force. Only afferents in which the effect of torque was tested with all three background normal forces were included in this analysis (n = 17). We found that 53% (9/17) of SA-I afferents were significantly scaled by background normal force (median correlation coefficient = −0.21, range = −0.84–0.87), of which 78% (7/9) had decreasing torque responses with increasing background normal force, and 44% (4/9) had increasing torque responses with increasing background normal force. Interestingly, three of these SA-I afferents had increasing torque sensitivity in one torque direction and decreasing in the opposite direction, when background normal force increased. There was no difference in the distance of SA-I afferents' RFC from the rotational center whether they had increasing (mean distance = 7.0 mm, SD = 5.4 mm) or decreasing (mean distance = 7.2 mm, SD = 4.5 mm) torque sensitivity with background normal force, and no clear relationship was observed in the distribution of their RFC over the distal part of the fingertip (Fig. 7C,D).
SA-I relationship between receptive field location and torque sensitivity
Further away from the rotational center the ratio between the local normal force to local tangential torque decreases thus increasing the likelihood of torsional slips between the skin and the surface. Shear forces experienced by the skin are also greatest just outside the edge of the contact area, hence we expected to see torque sensitivity (ST) increase with SA-I RFC distances from the stimulus center. SA-II afferents were excluded for these analyses as visual inspection of the data did not reveal any trend as afferents in any location on the fingertip were sensitive to torque, possibly because of the large size of SA-II afferent receptive fields and difficulty locating their RFC. Figure 7F shows the torque loading phase sensitivity with three background normal forces. Each SA-I afferent was binned using the RFC distance from the stimulus center in 3-mm bins and the torque sensitivities plotted for each bin. SA-I afferents had lower torque sensitivity when their RFC was within a 6 mm radius of the stimulus center, compared with afferents with RFCs between 9–12 mm from the stimulus center (p < 0.05, F(7,150) = 8.0, two-way repeated measures ANOVA, Tukey's adjusted pairwise post hoc test), but there was no differences between forces (p > 0.05, F(2,150) = 1.1; Fig. 7F). There were no SA-I afferents sensitive to torque with RFCs >15 mm from the stimulus center. To demonstrate the afferent RFC locations influenced their torque sensitivity because of variation in the properties of the skin near the nail or the interphalangeal joint, we compared the sensitivity of afferents in each of four spatial quadrants (Fig. 7F, inset). We found no difference in torque sensitivity across the four spatial quadrants for either torque direction (p > 0.1, F(3,150) = 2.0, ŋ2 = 0.075, two-way repeated measures ANOVA); however, this might be because of the relatively small number of afferents relative to the range of their interindividual differences, which is insufficient to evenly represent torque sensitivity in each quadrant.
FA-I afferent responses to torque
Size of FA-I afferent population influenced by torque
Mean discharge rate
FA-afferents with receptive fields across the whole distal fingertip were responsive to torque, and most had mean discharge rates scaled by torque during the loading and unloading phases (Fig. 8A). During the loading phase, 56% (22/39) of FA-I afferents' responses were scaled by torque (15 in clockwise, 21 in anticlockwise and 14 in both directions; Fig. 8B). Five FA-I afferents that had torque-scaled responses to one torque direction, had nongraded responses to the opposite torque direction, and one that was not scaled by torque at all, had nongraded responses to anticlockwise torques (p < 0.05, Mann–Whitney–Wilcoxon test). During the unloading phase the pattern was similar, 59% (23/39) of FA-I afferents' responses were scaled by torque (15 in clockwise, 20 in anticlockwise and 15 in both directions; Fig. 8B). Five FA-I afferents that had torque-scaled responses to one torque direction, had nongraded responses to the opposite torque direction, and two that were not scaled by torque at all had nongraded responses (p < 0.05, Mann–Whitney–Wilcoxon test). Peak discharge rate showed slightly lower proportions of scaled afferents, but the relationships were similar to mean discharge rates, and as such data are not shown.
The effect of torque on FA-I afferent responses. A, The receptive field center location (black dots) of each FA-I afferent and the direction in which its response was scaled by torque are plotted for each torque phase across a standard fingertip. B, C, The proportion of FA-I afferents that showed responses scaled by torque are shown in darker colors based on (B) mean discharge rate and (C) first spike latency, with 2 N background normal force. Stacked columns in a lighter color in B, C show the number of afferents that had nongraded responses to torque. Colored numbers above indicate the number of afferents with torque responses and in brackets the proportion of these that only showed nongraded responses. See Figure 3 legend for detailed description.
First spike latency
First spike latencies were scaled by torques in 53% (21/39) of FA-I afferents (14 in clockwise, 15 in anticlockwise and eight in both directions; Fig. 8D). Thus, the number of afferents capable of encoding torque magnitude by means of spike timing is similar to that based on mean discharge rates. On average, FA-I afferent first spike latencies were 45.6% shorter in response to 7.5 mNm compared with 3.5 mNm torque.
Effect of normal force on FA-I afferent torque responses
The background normal force appeared to influence FA-I afferent sensitivity to torque. Figure 9A shows the proportion of afferents that had mean discharge rates influenced by torque in at least one direction with different levels of background normal force. During the loading phase, the highest proportion of afferents with torque-scaled mean discharge rates, and first spike latencies were found with 2 N background normal force (Fig. 9A–C). Next, we investigated the relationship between background normal force and torque sensitivity for each FA-I afferent that was tested with all three normal forces (n = 13). In most FA-I afferents the background normal force scaled torque sensitivity 62% (8/13; median correlation coefficient = −0.33, range = −0.99–0.25), and in all cases torque sensitivity decreased with increasing background normal force (Fig. 9D). Torque sensitivities and RFCs of FA-I afferents that were unaffected by changes in background normal force are shown in Figure 9E.
The effect of normal force and receptive field center location on FA-I afferent's torque responses. A, B, The proportion of afferents that were scaled by torque in at least one direction as measured by (A) mean discharge rate and (B) first spike latency for each normal force. All three metrics were calculated in the loading and unloading phases, as FA-I afferents were unresponsive to torques in the plateau phase. C, D, FA-I torque sensitivities with 2, 3, and 4 N background normal forces, separated by afferents that had significantly decreasing torque sensitivity with increasing background normal force (C), or were unaffected by background normal force (D). E, Sensitivity values from the torque loading phase are shown for FA-I afferents against their receptive field center distance from the stimulus rotational center. FA-I afferents were mostly unresponsive to torque within 3 mm of the stimulus rotational center. With 2 N background normal force, FA-I afferents showed the highest torque sensitivity when their receptive field centers were between 3 and 9 mm from the stimulus rotational center. Vertical dashed line indicates the edge of the probe. Inset above shows a radar chart of the torque sensitivity in impulses mNm−1 s−1 (sensitivity increments shown at 0.5 and 1.0 impulses mNm−1 s−1) of afferents with receptive field centers in each quadrant of the fingertip, for each of the three background normal forces. Quadrants are labeled D, distal, P, proximal, U, ulnar, and R, radial. See Figure 5 legend for detailed description.
Relationship between receptive field location and torque sensitivity
Like SA-I afferents, we expected to see torque sensitivity increase with FA-I RFC distances from the stimulus center. Figure 9E shows the torque loading phase sensitivity (ST) with three background normal forces. Generally, the relationship between torque sensitivity and distance from the stimulus center appeared to show an inverted U-shaped curve with minimal torque sensitivity at the stimulus center and beyond 18 mm from the stimulus center; this resembled that seen with SA-I afferents, but was less consistent. FA-I afferents were significantly less sensitive to torque within a 3-mm range of the stimulus center compared with ranges 3–6, 9–15 mm from the stimulus center (p < 0.05, F(12,132) = 4.5, two-way repeated measures ANOVA with Tukey's adjusted pairwise post hoc tests), and were significantly more sensitive with 2 N background normal force compared with 3 and 4 N (p < 0.05, F(2,132) = 8.9, two-way repeated measures ANOVA with Tukey's adjusted pairwise post hoc test; Fig. 9E). To determine whether FA-I RFC locations influenced their torque sensitivity because of variation in the properties of the skin near the nail or the interphalangeal joint, we compared the sensitivity of afferents in each of four spatial quadrants (ulnar, radial, distal, and proximal; Fig. 9E, inset). As with SA-I afferents we found no difference in torque sensitivity across the four spatial quadrants for either torque direction (p > 0.4, F(3,132) = 0.9, two-way repeated measures ANOVA).
Comparison of torque sensitivity and discriminability of human and monkey afferent types
It is well established that all tactile afferent types respond to most somatosensory stimuli of sufficient strength, but each afferent is well-adapted to encode some types of stimuli better than others. The stimuli used in the present study were too slow to excite FA-II afferents, but did reliably excite FA-I, SA-I, and SA-II afferents. Previously it was demonstrated that FA-I and SA-I afferents in monkeys could encode torque magnitude, while SA-I afferents were better for encoding torque direction (Birznieks et al., 2010; Redmond et al., 2010a, b; Fu et al., 2012; Khamis et al., 2015). Here, we sought to determine which human afferent types were better at discriminating torque magnitude and direction, and compare torque discriminability between human and monkey afferents, noting that SA-II afferents are absent from the fingers of monkeys. The monkey dataset which was previously published by Birznieks et al. (2010) was recorded from three anesthetized macaques using the same experimental conditions as in the present study.
Torque magnitude sensitivity
When comparing data obtained in the current study with Birznieks et al. (2010), it was generally observed that monkey SA-I afferents responded with greater sensitivity to torque magnitude changes than human SA-I afferents. Monkey SA-I mean ST in the torque loading phase was 1.20 ± 0.11 impulses mNm−1 s−1, which was nearly twice that of human afferents, 0.68 ± 0.08 impulses mNm−1 s−1 (p = 0.0007, W = 756; Mann–Whitney–Wilcoxon test; Fig. 10A). In the torque plateau phase, monkey SA-I afferents had a mean ST of 2.00 ± 0.22 impulses mNm−1 s−1 which was more than five times greater than human afferent ST of 0.37 ± 0.05 impulses mNm−1 s−1 (p = 3.4e-12, W = 234).
Comparison of human and monkey FA-I, SA-I, and SA-II torque magnitude and direction discriminability. A, Afferent torque sensitivity (ST) of each afferent type in humans (left) and monkeys (right). B, The torque magnitude discriminability of each human afferent type during the torque loading and plateau phases, with 2 N background normal force, in box and whisker plots. Each filled circle is the d' value computed from the mean and standard deviation of the spike count in response to six torque trials with 3.5 and 5.5 mNm torque, applied to a single afferent. d' values were calculated from between the torque magnitudes, applied in the same direction, and the figure includes d' values for both torque directions. Blue numbers above are the percentage of each afferent population that had d' > 1. Monkey (right) FA-I and SA-I afferent torque responses with 2.5 N background normal force. C, The torque direction discriminability during loading and plateau phase of human (left) SA-II, SA-I, and FA-I afferents. d' values were computed from the mean and standard deviation of spike counts in response to 5.5 mNm torque for clockwise and anticlockwise torques. Monkey (right) SA-I and FA-I afferents in response to 5.5 mNm torques, with 2.5 N background normal force. Blue percentages above bars indicate the proportion of each afferent with d' values >1. Red horizontal lines and red circles indicate medians and means, respectively, and box boundaries indicate the upper and lower quartiles. Upper whisker shows the largest observation less than or equal to the upper box hinge + (1.5 × interquartile range), and the lower whisker shows the smallest observation greater than or equal to the lower box hinge – (1.5 × interquartile range). Gold colored lines indicate the threshold d' value of 1.
Similarly, investigation of FA-I afferent torque magnitude sensitivity between the two species found that monkey afferent ST, 1.60 ± 0.18 impulses mNm−1 s−1, was significantly greater than human FA-I afferents, 0.77 ± 0.08 impulses mNm−1 s−1 in the torque loading phase (p = 8e-05, W = 247; Fig. 10A).
Comparatively, human SA-II torque sensitivity averages in the loading (0.48 ± 0.07 impulses mNm−1 s−1) and plateau (0.46 ± 0.08 impulses mNm−1 s−1) phases were similar to human SA-I and FA-I afferents but were much lower than monkey SA-I and FA-I afferents.
Torque magnitude discriminability
Human d-prime (d') values were computed between the number of spikes evoked in response to 3.5 and 5.5 mNm torque and compared with these in monkeys tested with 2.5 N background normal force. Two d' values were obtained for each afferent, one for each of the two torque directions. A d' value of 1 was set as a threshold, such that if d' was >1, the two torque magnitudes were deemed discriminable based on afferent's responses.
We compared human and monkey torque magnitude discriminability of SA-I afferents in the torque loading phase, as measured by d', and found no significant difference (p = 0.57, W = 2340), as 47% of monkey SA-I afferents (55/116 afferent responses with d' > 1, median d' = 0.94, data from both torque directions included) were able to discriminate torque magnitude in the loading phase, compared with 58% of human afferents (22/38 responses d' > 1, median d' = 1.26). However, in the plateau phase monkey SA-I afferent torque magnitude discriminability was significantly better than in human afferents (p = 0.0027, W = 1487), with 63% (73/116 afferent responses with d' > 1, median d' = 1.53) of monkey SA-I afferent responses able to discriminate plateau torque magnitude, in comparison to 45% in humans (17/38 responses d' > 1, median d' = 0.78; Fig. 10B). In humans the discrimination of torque magnitude is aided by SA-II afferents in which 38% (10/26 responses d' > 1; median d' = 0.63) could discriminate torque magnitude in the loading phase, and 35% (9/26 afferents with d' > 1; median d' = 0.75) in the plateau phase (Fig. 10B).
FA-I afferents' torque magnitude discriminability across species was only compared for the torque loading phase, as they did not respond in the plateau phase. Monkey FA-I afferents could discriminate torque magnitude significantly better than human FA-I afferents in the torque loading phase (p = 1.7e-05, W = 595), with 82% of monkey FA-I afferent responses able to discriminate torque magnitude (41/50 afferents with d' > 1) and a median d' of 3.2, which was more than double the human FA-I median d' of 1.5 and lower fraction 69% (33/48) responses with d' > 1 (Fig. 10B). In both species larger fraction of FA-I afferents could discriminate torque magnitude in comparison to SA afferents.
Torque direction discriminability
Next, we compared torque direction discriminability among human and monkey afferent types by computing d' from spike counts to both torque directions, of magnitude 5.5 mNm.
In the loading phase torque direction discriminability of FA-I and SA-I afferents, as measured by d', did not reach any statistical differences between species (FA-I: p = 0.15, W = 229; SA-I: p = 0.93, W = 544).
In the torque plateau phase direction discriminability of SA-I afferents, was better in monkey afferents than human (p = 0.0012, W = 276), as 91% of monkey SA-I afferents (53/58 afferent responses with d' > 1, median d' = 5.12) were able to discriminate torque direction, compared with 79% of human afferents (15/19 responses d' > 1, median d' = 1.83; Fig. 10C). Between afferent types SA-I showed largest number of afferents being able to discriminate torque direction leaving behind SA-II and FA-I afferents.
Discussion
This study is the first to demonstrate how human SA-II, SA-I, and FA-I afferents innervating the fingerpads respond to torque. Previous studies (Birznieks et al., 2010; Redmond et al., 2010a,b; Fu et al., 2012; Khamis et al., 2015) have used data obtained in monkeys, which despite possessing SA-II afferents in hairy skin, lack them in the glabrous skin of the hand (Johnson, 2001). We show that human SA-II afferents are highly sensitive to torque magnitude and direction and readily encode these torque features. SA-II, SA-I, and FA-I afferent responses could discriminate between different torque magnitudes and directions, but a population of the three afferent types are likely necessary to encode the full range of torque parameters. Finally, human SA-I and FA-I afferents generated an overall smaller number of spikes per unit of torque, as measured by torque sensitivity index (ST), resulting in a lower torque magnitude discrimination ability (d') in humans for FA-I afferents during the loading and SA-I afferents during the plateau phase. Discrimination ability of torque direction was similar in the loading phase regardless of afferent type; however stronger directional preferences were seen in monkey SA-I afferents during the plateau phase.
Single afferent responses to torque
Both the magnitude and direction of torque influenced FA-I, SA-I, and SA-II afferent activity in humans. During the loading phase, most (85%) SA-II afferents had mean discharge rates that were scaled by torque in at least one direction. Torques applied in either direction could positively or negatively change SA-II discharge rates, resulting in diverse response behaviors to different torque magnitudes and directions that may, in part, derive from an SA-II afferent's associated receptor (presumably the Ruffini ending) location relative to the rotational center, and its anchoring within tissue. SA-II afferents are commonly found at the fingernail border (such as the afferent in Fig. 4A) and might experience distinctive tangential force effects because of shear stress developing in the skin where the compliant tissue borders the stiff fingernail (Dandekar et al., 2003; Shimawaki and Sakai, 2007; Birznieks et al., 2009; Ho and Hirai, 2015; Wolterink et al., 2019).
Human torque parameter sensitivity and discriminability
Encoding torque changes during the loading phase is critical to ensure grasp stability, but some dexterous tasks may also require maintenance of an object's orientation; decreasing torque could indicate that an object is rotating and losing the desired orientation in the grip (Goodwin et al., 1998). SA-II afferents were more suitable to signal changes in torque magnitude during the plateau phase, corresponding to situations involving object orientation maintenance. By contrast, SA-I and FA-I afferents typically were more sensitive to dynamic changes in torque loading and unloading.
All three afferent types could discriminate torque direction during torque loading. d' analyses showed that all three afferent types had discriminable responses between torque directions and there did not appear to be an afferent type that was best for encoding torque direction. It is expected that the contribution of different individual afferents and afferent types signaling torque parameters changes depending on the task and stimulus parameters.
When fast signaling of torque magnitude is considered, FA-I afferents could deliver superior performance as first spike latencies were scaled by torque in (58%) FA-I afferents, which was also the case in macaques (Birznieks et al., 2010). Therefore, FA-I afferents can rapidly signal torque changes precisely at onset, providing an advantage in controlling object orientation during object manipulation. However, as noted above, they cannot differentiate the direction of rotational forces.
Human and monkey afferent torque sensitivity and discriminability
We compared torque sensitivity and used d' analysis to compare the discriminability of torque parameters by individual afferents of each afferent type in humans and monkeys. Remarkably, monkey afferents, particularly FA-Is during the loading phase and SA-I afferents during the plateau phase, were significantly more sensitive and could discriminate torque magnitude better than human afferents.
SA-II afferents may, therefore, become a valuable addition to improve information content about torque especially during longer periods of holding an object in the grip beyond what was tested in the current study. Sustained firing with little adaptation over time may be crucial to follow and make corrective motor responses when an object's orientation should be maintained without visual input (Johansson and Westling, 1987; Kinoshita et al., 1997). Because of similar sensitivity, one may wonder whether the nervous system would preferentially use one afferent type. After all, SA-II afferents are relatively poor at differentiating between object characteristics such as the curvature of an applied object (Goodwin et al., 1997) yet clearly can signal the direction and magnitude of rotational forces. We believe that in natural environments the challenge is to extract relevant information when multiple stimuli simultaneously interact, which may be unlike the changes in constrained stimulus parameters in most studies. In such situations, combining inputs from a diverse afferent population might represent a computational advantage to disentangle the stimulus feature of interest (Khamis et al., 2015). We intend to test this in follow-up studies by computational modeling of discrimination capacity of afferent populations when subjected to simultaneous effects of multiple stimulus parameters.
Normal force effects and the spatial pattern of afferent torque responses
Macaque fingertips are approximately half the size of human fingertips, and typically have more bulged fingertip pulp tissues creating softer volar surfaces. This results in larger tissue strain developing when exposed to the same stimuli and a larger proportion of the receptors of the distal finger segment being stimulated. Most human SA-I and FA-I afferents that were not influenced by torque had RFCs >12 mm from the torque rotational center, which suggests that torque stimulus effects did not reach their receptive fields.
During the loading phase, most afferents of each type had decreased torque sensitivity with larger background normal forces. Better response modulation to torque at lower normal forces might, in part, be caused by afferents responding to microslips arising close to the edge of the probe (Johansson and Westling, 1984; Westling and Johansson, 1987; de Dunilac et al., 2023). In contrast, for other afferents, especially those with RFCs located outside the contact area, a larger normal force may cause a more profound fingertip deformation and thus a torque-generated skin deformation pattern could reach fingertip skin further away from the rotational center. In the central “stuck” zone, which does not slip, FA-I afferents did not reliably respond to torque (André et al., 2011; Adams et al., 2013; Delhaye et al., 2014, 2021; Schiltz et al., 2022). Nearer the edge of the contact area, and just outside it, both FA-I and SA-I afferents were typically more sensitive to torque and most likely responded to changes in tangential or shear stress in the skin as well as slip events (de Dunilac et al., 2023), which may also be reflected in the instantaneous discharge rate profile irregularities of the of SA-II afferents we observed.
Although we did not systematically change the stimulus friction, it is expected that the contact area size would depend on frictional conditions (Willemet et al., 2021) and thus would influence afferent torque responses. This is supported by previous studies suggesting that torque information is encoded while superimposed on a range of normal forces (Birznieks et al., 2010; Khamis et al., 2015), in which SA-I and FA-I afferents are generally more sensitive to torque with lower normal force and peripheral slips are more likely to occur. We used a high friction surface which, in combination with higher background normal force, would limit partial slip probability under the stimulation probe. With more slippery surfaces, the slipping area with the same contact force will be larger, but rotational strain will be smaller. Thus, it is expected that the SA-II afferent input would decrease while FA-I and SA-I afferent input would increase within the slipping zone under the probe. Like submillimeter lateral slips (Afzal et al., 2022), we hypothesize that during natural manipulative tasks small rotational slips are a significant source providing frictional information, as they are permitted to occur without endangering grip safety and torques are almost always present during object manipulation. Therefore, future studies should investigate frictional effects and torque effects in combination with linear tangential force mimicking object manipulation.
In conclusion, for the first time, we have demonstrated that human SA-II afferents encode torque magnitude and direction with high sensitivity. Interestingly, monkey afferents were generally superior in their sensitivity and discrimination of torque parameters. SA-II afferents may increase the richness of information about torque stimuli that populations of human tactile afferents provide to the sensorimotor system during natural manipulations. This is particularly important as the torque sensitivity of human SA-I afferents appears to be inferior to that of monkeys, especially during the plateau phase. Finally, SA-II afferents could contribute unique inputs to the nervous system, providing computational advantages to disentangle and extract information about multiple concurrently changing tactile stimuli during natural object manipulation.
Footnotes
↵*H. E. Wheat is now retired.
This project was funded by the Australian Research Council (ARC). We thank Prof. Antony W Goodwin for contribution to the inception of this study as well designing and providing stimulators from his laboratory. We thank Dr. Rachael Brown for help with spike sorting procedures.
The authors declare no competing financial interests.
- Correspondence should be addressed to Alastair J. Loutit at a.loutit{at}neura.edu.au
