of the gripped surfaces varied among sandpaper, suede, and rayon, which represented increasingly slippery sur- faces. The grip force at the start of the load ...
Exp Brain Res (1993) 95:523-532
O Springer-Verlan 1993
Friction at the digit-object interface scales the sensorimotor transformation for grip responses to pulling loads Kelly J. Cole1, Roland S. Johansson2
' Department of Exercise Science, University of Iowa, Iowa City, IA 52242, USA Department of Physiology, University of Umei, Umei, Sweden Received: 29 October 1992 / Accepted: 31 March 1993
Abstract. When restraining a mechanically "active" object (one that exerts unpredictable changes in loading forces) with a precision grip of the digits, we maintain a stable grasp by modulating our grip force using somatosensory information related to the loading forces. The response to ramp load increases consists of an initial fast rise in grip force ("catch-up") followed by a secondary response that steadily increases the grip force in parallel with the load force ("tracking"). The sizes of these response components scale in proportion to the loading rate. However, maintaining a stable grasp without employing an exceedingly large grip force may require further scaling of this load-to-grip sensorimotor transformation based on two additional factors: (1) the friction at the digit-object interface and (2) the grip force present at the start of the load increase. The present experiments sought to determine whether such scaling occurs and to characterize its control. Subjects restrained a manipulandum held between the tips of the thumb and index finger. At unpredictable times a pulling force appeared, directed away from the subject's hand. Each pull had a trapezoidal load profile beginning and ending at 0 N with 4-N/s ramps; each ramp was 1 s in duration. The texture of the gripped surfaces varied among sandpaper, suede, and rayon, which represented increasingly slippery surfaces. The grip force at the start of the load ramp (intertria1 grip force), and the amplitudes of the catch-up and secondary grip responses scaled in proportion to the inverse friction. We interpret these results to indicate a uniform scaling of the transformations controlling the intertrial grip force, the catch-up response, and the secondary response. Initial-state information from tactile cues available upon object contact appeared to update the frictional scaling value. This conclusion is based on observations of immediate changes in the intertrial grip force upon contact with a new surface, and because differences in force-rate profiles appeared virtually by the onset of the catch-up response. Similarly, the intertrial grip force also constituted initial-state information. The size of the Correspondence to: K.J. Cole
catch-up and secondary grip force responses varied inversely with the size of the intertrial grip force. These scalings of the load-to-grip-force sensorimotor transformation for friction and intertrial grip force level appear to be functionally adaptive, because they contribute to a stable grasp (prevent object slips) while avoiding exceedingly large safety margins. Key words: Motor control - Precision grip - Hand - Grip force - Human
Introduction During tasks such as operating power tools or holding a dog's leash, our grasp may be subjected to unpredictable loads that may cause the object to slip. When we hold objects with a precision grip, unexpected increases in the forces oriented tangentially to the gripped surfaces ("load forces" in the present paper) trigger active adjustments of the force perpendicular to the gripped surfaces (grip force) that help stabilize the object (Cole and Abbs 1988; Johansson et al. 1992a-c; Johansson and Westling 1988; Winstein et al. 1991). These adjustments rapidly reestablish and maintain an adequately high ratio between the grip and load forces to protect against slips, and thus contribute to a stable grasp. A ramp increase in load force triggers an initial fast increase in grip force, termed the "catch-up" response (Johansson et al. 1992a,b).This response compensates for the response delay (minimum approximately 80 ms) by eventually matching the grip force to the load demands; that is, it catches up to the load. Apparently, tactile mechanoreceptors encode the load rate (Johansson et al. 1992c), which in turn scales the catch-up response amplitude and peak rate to meet the load demands (Johansson et al. 1992b). Continued ramp increases in load trigger further increases in grip force that appear to "track" the load change, maintaining an approximately constant ratio between the grip and load forces (Johansson et al. 1992a,b).
To ensure a stable grasp, the skin-object friction also should influence the grip responses to provide adequate safety margins against slips during various frictional conditions (see Westling and Johansson 1984).Whether these grip responses to rapid loading indeed scale according to friction is unknown, as are the control principles that underlie this possible scaling (e.g., Jones and Hunter 1992). Efficient and rapid scaling of sensorimotor transformations occurs during various forelimb movements (Ghez and Vicario 1978; Ojakangas and Ebner 1991; Vicario and Ghez 1984), posture (Nashner 1981), and eye movements (Deubel et al. 1986; McLaughlin 1967; Optican et al. 1985). Therefore, we hypothesize that the transformations governing the grip force responses likewise scale rapidly when meeting different skin-object frictions, as when subjects lift mechanically "passive" objects that differ in surface texture (Edin et al. 1992; Johansson and Westling 1984; Westling and Johansson 1984). This scaling may occur based on foreknowledge of an object's frictional characteristics, or from tactile information that is available upon contact with the object (see Johansson and Westling 1987). Anticipatory scaling is consistent with current views in which the brain, during control of volitional limb movement, uses internal (neural) models of our limbs and external objects based on previous experiences and/or sampling of information about initial conditions (Ghez et al. 1991;Johansson 1991; Johansson and Cole 1992; Lacquaniti et al. 1992). Alternatively, experience with the load or its consequences (e.g., frictional slips) may be necessary under a new frictional condition to scale the sensorimotor transformation, or to achieve sufficiently accurate scaling. We investigated the influences of surface friction on the grip responses employed when subjects restrained objects that pulled upon the hand at unpredictable points in time. In some experiments we also asked subjects to vary the force used to hold the object between load occurrences, to investigate whether this change in initial conditions also may affect the scaling of the sensorimotor transformations. Materials and methods Subjects and general procedure Thirteen healthy, right-handed individuals (7 women and 6 men) between the ages of 17 and 41 years (mean 25 years) participated in these experiments. We obtained informed consent from all subjects according to the Declaration of Helsinki. Subjects sat with their right arm abducted about 30°, the elbow joint flexed at an angle of about 10O0, and the forearm extended anteriorly in intermediate pronosupination. A vacuum cast supported the forearm up to the wrist. Subjects grasped a manipulandum using a precision grip with the thumb and index finger, while slightly extending their remaining three fingers. They did not view their hand and attempted to restrain the manipulandum against pulling forces directed distally (and tangentially to the grip surfaces).
Apparatus Except for certain modifications detailed below, the experiments used the same apparatus described previously (Johansson et al.
1992a). The manipulandum consisted of two parallel, vertically oriented circular surfaces for grip (30 mm diameter), spaced 31 mm apart. Sandpaper (320-grit), suede leather, or rayon-polyester (rayon) covered the circular surfaces. A rigid, immovable plate secured the surface for the thumb. The index finger surface connected to a servoregulated force motor that could generate pulling forces (&I0 N) in proximal and distal directions (&I5 Hz bandwidth, noise less than 0.05 N). A laboratory computer controlled the force motor. The manipulandum was servo-regulated to a constant position (stiffness 1.2 N/mm) when the finger did not touch the manipulandum and when we measured the coefficient of static friction present between the index finger and the various grip surfaces (see below). The two parallel grip surfaces were superimposed initially (i.e., a line joining their centers would be perpendicular to both surfaces). Thus, the digit pulps would fall into apposition if it were not for the manipulandum. Strain gauges transduced the grip force produced by the index finger (force perpendicular to the gripped surface) and the load force on this digit (force tangential to the gripped surface, in the direction of handle motion). The recorded forces (d.c.-120 Hz) showed less than 5% cross talk. The position of the index finger grip surface was transduced to a resolution of 0.05 mm in the pulling direction. The moving part of the manipulandum also carried an accelerometer (10-600 Hz) that helped to detect slips between the finger and the manipulandum.
Pulling loads Nine subjects each participated in four experiments with trials blocked according to surface texture; one experiment each with the sandpaper and rayon surfaces, and two experiments with the suede surface. Experiments were randomly ordered for each subject. Each experiment consisted of trials of distal pulling loads (Fig. 1). For 32 consecutive trials the load increased at a constant rate of 4 N/s for 1 s. A plateau phase followed (4 N constant force) that was maintained from 0.5 to 2.0 s (randomly selected interval) before the load decreased to zero at a rate of 4 N/s (not shown). The intertrial delay varied randomly from 2.0 to 5.5 s. The subjects attempted to prevent the manipulandum from moving during the loading trials. No penalty occurred if the manipulandum slipped; in this case the manipulandum returned within 2 s to its starting position and the trial was repeated. Subjects were not instructed explicitly to alter their grip force in response to the increased loads. However, during one experiment with the suede surface subjects attempted to maintain different static grip force levels between trials. The experimenter asked for greater or less force from trial-to-trial to obtain a wider range of background (intertrial) grip forces than were adopted naturally. In separate experiments on six subjects (including two from the previous series of experiments) we investigated the sequential effects of surface structure type on the grip responses. We varied the grip surfaces between sandpaper and rayon according to a prescribed order that subjects could not predict. There were 54 trials in each experiment. On at least 12 occasions a trial with sandpaper followed a trial with rayon, and vice versa. Likewise, a trial with sandpaper preceded at least nine other trials with sandpaper, and a rayon trial preceded at least nine other trials with rayon. After each load trial the subject released the object and awaited the command to replace their digits on the manipulandum. The load ramp began within 2-5 s (randomly selected) after placement of the digits. Except for the load rate, which was 2 N/s, the load trials were identical to those described above.
Frictional properties of the contact surfaces We measured the coefficients of static friction between the index finger skin and each surface type at the end of each experiment. The subject placed their index finger on the grip surface and increased their finger flexion force to at least 1 N, while the manipulandum
was servo-regulated to a constant position. The subject then dragged their finger proximally while maintaining the finger flexion force until one or more slips occurred. Slips appeared as sudden decreases in load force and as vibrations of the manipulandum (measured by the accelerometer). The ratio of the grip and load forces was measured at the onset of the slips (i.e., the slip ratio) and provided an estimate of the inverse coefficient of static friction. The difference between the employed grip/load force ratio and the slip ratio is the safety margin preventing slips. The safety margin is expressed throughout the results as a fraction of the employed force ratio. The chosen surface textures effectively varied the skin-object friction. The average inverse coefficient of static friction (f SD) was 0.92f 0.10, 1.38k0.38, and 2.13f 0.93 for sandpaper (least slippery), suede, and rayon (most slippery), respectively. There were large intersubject variations (Fig. 3B, top row).
Data collection and analysis Data were collected and analyzed with .a laboratory computer system (SC/ZOOM; Department of Physiology, University of Umel). The accelerometer (root-mean-square processed, rise and decay time constants of 1 ms and 3 ms) and force signals were sampled with 12-bit resolution at 400 samples/s, while the position signal was sampled at 100 samples/s. Event markers related to onsets and offsets of the various phases of each load trial were sampled as well (f0.1 ms resolution). Force rates were obtained using symmetrical numerical time differentiation (+ 5 points). The instantaneous ratio between the grip and load forces was computed numerically. Several measurements of the grip force response were taken from individual trials. The grip force present at the onset of the load ramp (marker a in Fig. 1A) was defined as the intertrial grip force. The initial response (catch-up) began when the grip force rate signal became greater than during the prior 0.5-s interval, provided that it continued to increase for at least 100 ms (marker b). The return of the grip force rate signal to a lower stable level (marker c) marked the end of the catch-up response. The amplitude of the catch-up response was the difference between the grip force at the end and start of the catch-up response. The peak grip force rate was the greatest force rate during the catch-up interval (marker f, in interval b-c).. Grip - force increases that continued after the catch-up- response (e.g., tracking responses) were designated as secondary responses. The amplitude of the secondarv response was the difference between the grip force measured at thk peak grip force (marker d) and the end of the catch-up response (marker c). The sum of the intertrial, catch-up, and secondary response amplitudes yields the peak grip force. The plateau grip force was the grip force 0.5 s after the beginning of the plateau force (marker e). All values are reported as the mean f 1 SD, either pooled across the average values of individual subjects, or pooled across data from all subjects, as indicated. Results for the dependent variables (each subject's mean value) were assessed for difference across surfaces (sandpaper, suede, rayon) using an analysis of variance (ANOVA) with repeated measures for subjects. A probability of 0.05 or less was required for a significant test result. Data are reported first from experiments in which surface textures were blocked across trials. Data from experiments with unanticipated changes in surface texture and with changes in the intertrial grip force are reported in separate subsections of the Results section, respectively. A
Results
Frictional condition scales the task-related grip force responses The friction between the contact surfaces and the skin scaled the grip force responses; more slippery surfaces yielded higher grip forces (Figs. 1, 2). Figure 3 reveals a
close relationship between the inverse friction and the peak grip force. Note that each force component changed, namely, the intertrial grip force, catch-up response, and secondary grip force response. Intertrial grip force. The intertrial grip forces for all surfaces typically were small in comparison with the grip forces finally employed during the load plateau phase (Fig. 3). Nevertheless, the intertrial grip force was important because it solely determined whether the handle would slip during the latent period before the catch-up response (see grip-load force ratio and slip ratio, Fig. 1A). Subjects displayed anticipatory control in this regard and increased their intertrial grip force for one or both of the more slippery surfaces (P 0.58, F2,,, = 0.719), while, as expected, differences due to response component were significant (P < 0.00001; F2,,,= 37.6). The similar relative sensitivities of the grip
.
lntertrial grip force
'Catch up' response
Secondary ('track') response
Surface structure Fig. 6. Plots of the fraction of the peak grip force occupied by the intertrial grip force, catch-up response, and secondary response for each surface and for each subject (circles). The bars represent the means across subjects, which are consistent across surfaces for each response component. Symbols joined by lines refer to the mean values of individual subjects. Note the within-subject variability evident across surfaces
response components to friction also are illustrated in Fig. 1B. After normalizing the grip force signals from Fig. 1A to the peak grip force for the rayon surface, the various response components maintained proportional relationships under the different frictional conditions.
0
Responses to unpredictable changes in frictional condition
We changed surface texture unpredictably from trial-totrial to investigate how the grip force response scaling developed (see Materials and methods). Changing the surface immediately influenced the intertrial grip force, catch-up amplitude, and the peak grip force (Fig. 7). In all instances the various force measures differed (P
