INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A current proposal is that that the cerebellum is the site of internal models of arm dynamics for learned movements (Kawato 1996; Wolpert et al. 1998). Internal models are hypothetical neural representations that enable predictive (anticipatory) commands to be generated which compensate for movement related loads. One reason for implicating the cerebellum in anticipatory control of arm dynamics is that patients with cerebellar lesions show disorders in control of force. However, relatively few studies have investigated force control in skilled movements in cerebellar patients, and for those that have, some inconsistencies exist. For example, for the case of the fingers, Holmes (1917) reported that acute cerebellar patients showed longer rise times to peak force and lower levels of maximum pinch forces. However, more recent studies on chronic cerebellar patients have shown that the maximum level of force produced in the precision (pinch) grip is unaffected by the cerebellar lesion (Mai et al. 1988; Müller and Dichgans 1994a,b). Furthermore, in a drawer-pulling task, cerebellar patients actually showed an increase in grip force compared with controls (Serrien and Wiesendanger 1999). These studies agree with Holmes that deficits occur when patients perform tasks in which there is a force change. For example, patients showed a slowness in making fast alternating changes between two levels of isometric force and deficits in isometric force tracking (Mai et al. 1988) and a reduced ability to increase the rate of isometric pinch force when adapting to heavy loads (Müller and Dichgans 1994a).

One task where precise and rapid change of finger force is required is gripping and releasing a ball in an overarm throw. We have previously shown that during the period of ball release in a throw, as the hand accelerates forward and downward, a back force occurs from the ball on the fingers (Hore et al. 1999). In the preceding paper (Hore et al. 2001), we show that skilled throwers keep finger amplitude relatively constant for throws with balls of different weights by estimating the magnitude of the back forces and generating fast force changes in finger muscles throughout the throw to oppose them. If the cerebellum is the site of internal models of hand dynamics for learned movements, then it would be expected that patients with lesions of the cerebellum would have disorders in their ability to correctly predict back forces from the ball on the fingers and in their ability to produce the appropriate force change.

The overall aim was to test the ability of cerebellar patients to predict and produce the appropriate levels and rate of change of finger forces in an overarm throw, which is an example of a skilled multijoint movement. However, it was considered that it was not practical in cerebellar patients to measure force on the fingers directly with force transducers taped to the fingers as was done with skilled throwers (Hore et al. 2001). This was because this technique requires that the ball should roll directly over the transducerany deviation to the side results in a decreased force being recorded. These transducers also result in the subjects having decreased tactile information, i.e., they cannot clearly feel the middle finger gripping the ball. While these were not problems for skilled throwers, it was felt that they could result in spurious results in the cerebellar patients, e.g., a decrease in measurement of force could result either because of a real decrease in force of grip by the middle finger or because the ball was not perfectly centered on the middle finger. An alternative way to gain information about finger force is to record finger position with respect to the hand. For example, once the ball has left the hand, finger position (i.e., the amplitude of the finger flexion flick that occurs immediately after ball release) is directly related to the magnitude of force recorded on the distal phalanx before ball release (Hore et al. 2001). Consequently, in the present study, we measured finger position with respect to the hand with the search-coil technique (at 1,000 Hz) that records angular position at high resolution and causes no impediment to gripping. The specific objective of the present experiments was to investigate the ability of cerebellar patients to control finger position throughout an overarm throw. Because not all patients were skilled throwers before the lesion, it was necessary to test controls who were unskilled to various degrees. In the process, it became clear that unskilled throwers showed similar deficits to the cerebellar patients. Therefore a second objective was to compare the ability of unskilled subjects to control finger position throughout a throw with that of skilled subjects and cerebellar patients.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Data were analyzed in detail from experiments performed on nine right-handed cerebellar patients, nine controls, and six unskilled throwers and from previously published experiments (Hore et al. 1999) on six skilled throwers. Additional experiments with force transducers taped to the fingers were performed on four unskilled throwers. The experiments were approved by the local ethics committee, and all subjects and patients gave informed consent. Patient information is given in Table 1. In brief, five patients had right-sided surgical lesions that affected the region of the cerebellar nuclei, one a right-sided cerebellar infarction (superior cerebellar artery) and three diffuse cerebellar cortical atrophy. In all cases, the extent of the lesion was confirmed by magnetic resonance imaging (MRI) or computerized tomography (CT) scans. All patients had a full neurological examination at the time of the experiments. These 9 cerebellar patients were selected from a total of 16 cerebellar patients who performed the experiment. Data from seven patients was excluded from the final analysis either because they showed significant signs of brain stem damage or polyneuropathy on clinical examination (4 patients) or because they showed no evidence that the cerebellar lesion had affected the fingers (3 patients). Disorder of the fingers was assessed in two ways. First, patients and controls were asked to make fast and accurate small amplitude finger extensions to a target ~10° away (cf. Hore et al. 1991). Patients were excluded if they did not show hypermetria compared with controls. On average, controls overshot the target slightly (mean amplitude of total movement was 15°), whereas the included patients showed hypermetria (mean amplitude, 39°) followed in some cases by a damped intention tremor. Second, finger disorder was assessed by the size of the timing window for ball release. As before (e.g., Hore et al. 1995; Timmann et al. 1999), we defined variability in timing of ball release (timing windows) in terms of the standard deviation (SD) about the mean time of ball release with respect to the moment in the throw when the hand was vertical in space (the timing window for 95% of the throws was calculated by multiplying the SD by 3.92). Patients were excluded if they had shorter timing windows for 40 throws with the tennis ball than their matched controls (Table 2). Previous results from our laboratory have shown that average timing windows for normal subjects are ~5 ms for very accurate male competitive ball players, 9-10 ms for male recreational ball players, 15 ms for female recreational ball players, 15-30 ms for unskilled female throwers, and 55 ms for cerebellar patients (cf. Hore et al. 1995, 1996a,b; Timmann et al. 1999).

Control subjects were chosen to match cerebellar patients with respect to the estimated skill of the patient at throwing before the cerebellar lesion, their sex, and their approximate age. Subjects and patients rated their prelesion throwing ability on a score from 1 to 5 (1, very bad; 2, bad; 3, reasonable; 4, good; 5, very good thrower). Six of the cerebellar patients were from Germany, and their self-assessed throwing ability before the lesion was verified by asking them how far they threw in the standardized test for athletic ability given every year to German school children. In addition, further information was obtained by asking all subjects how often in a series of throws they felt they would hit our 6 cm target which was located 3 m away (in the case of the patients, how accurate they felt they were before the lesion). Unskilled subjects were selected on the basis of being self-assessed "bad or very bad" throwers and having never played recreational sports involving throwing. Four of the six unskilled throwers were female and one (subject Lt) was a 70-yr-old male.

General procedures

All subjects threw from a sitting position because some patients could not stand for long periods. Subjects were instructed to throw tennis-sized balls with an overarm motion (using shoulder adduction) at a central target which was located at eye level on a vertical target grid of numbered 6-cm squares 3 m away. Each throw was scored for accuracy by an experimenter calling out the square that was hit. The time of final ball release from the fingertip was defined by a number of criteria including: 1) the time the ball rolled off a small, light-weight microswitch (distal trigger) attached to the distal phalanx of the middle finger, 2) the moment of a deflection or peak in finger extension, or horizontal finger motion all with respect to the hand (these motions result from reactive forces associated with the ball leaving the hand), and 3) the duration of finger opening to ball release (for throws of the same speed this time is fixed because there is a fixed time for the ball to roll along the finger). The small number of throws not satisfying at least two of these criteria were omitted. All patients and controls were instructed to throw 40 tennis balls (55 g weight, 65 mm diam) accurately at a medium speed. In addition, six of the cerebellar patients (Ol, Hv, Dv, Vn, Hn, and Pr), six controls, and six very unskilled throwers made an additional 40 throws with a light plastic ball that had a hard surface (14 g, 70 mm) and 40 throws with a heavy tennis ball filled with concrete (196 g, 65 mm). Balls were thrown in the order of 20 tennis balls, 20 light balls, and 20 heavy balls, and the sequence was repeated. In all cases, subjects were instructed to center the ball on the middle finger so that during release it rolled along that finger. This grip position was monitored throughout the experiment.

As previously, arm segment orientations were measured with the use of a modification of Robinson's (1963) magnetic-field search-coil technique (Tweed et al. 1990). Arm movements were sampled at 1,000 Hz by means of a pair of orthogonal search coils taped to four arm segments and the distal phalanx of the middle finger. A reference arm position was recorded where the upper arm was held horizontal and the forearm and fingers were held vertical. Calculations were performed to obtain rotations of the distal phalanx with respect to the hand.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Force on the fingers changes during an overarm throw

In the preceding paper (Hore et al. 2001), we show how force on the fingers changes throughout an overarm throw in skilled throwers. The essential findings are summarized in Fig. 1. Figure 1A shows a diagrammatic representation of the three middle finger phalanges and hand during the forward throw, which began ~150 ms before final ball release (time 0). A force transducer on the distal phalanx is represented (). Figure 1B shows average force records from the distal phalanx for 10 throws made with balls of 14, 55, 196, and 360 g weight; and Fig. 1C shows the average angular positions of the distal phalanx of the finger with respect to the hand (finger extension) for the same throws. Early in the forward throw (100 ms), the hand grips the ball with a force proportional to ball weight (Fig. 1B). Between 60 and 30 ms (Fig. 1A), the finger starts to extend (open), and there is a decrease in force applied to the ball (Fig. 1B); between 30 and 10 ms, the ball rolls up the finger and the force on the distal phalanx increases and then peaks, and between 10 and 0 ms, the ball rolls off the transducer and is released from the fingertip. These records show three findings that have been reported previously for skilled recreational ball players (Hore et al. 1999, 2001): for throws of similar speeds, the force on the finger throughout the throw is proportional to ball weight (Fig. 1B); the amplitude of finger opening (from the baseline to the moment of ball release) does not increase with balls of increasing weights and in some subjects may actually decrease (Fig. 1C); and the amplitude of the finger flexion (Fig. 1C) that begins at the moment of ball release (time 0) is proportional to the peak force on the distal phalanx (Fig. 1B). This latter finding presumably occurs because the sudden release of the ball from the fingertip leaves the finger flexor torque unopposed and the larger flexor torques with the heavier balls produce larger finger flexions. In summary, these records illustrate that skilled throwers adjust the level and rate of change of finger force throughout a throw and that this is associated with a finger extension amplitude that does not increase with balls of increasing weight.

View larger version (29K): [in this window] [in a new window]   Fig. 1. Finger forces and finger positions during the forward component of overarm throws made by a skilled thrower. A: diagrammatic representation of the angular positions of the 3 finger phalanges of the middle finger and the hand. Times are with respect to ball release (time 0). , force transducer on distal phalanx. B: mean forces recorded on the distal phalanx for 10 throws each with balls of 14, 55, 196, and 360 g weight. C: mean angular positions of distal phalanx of middle finger with respect to hand for the same throws; subject De.

However, not all cerebellar patients were skilled throwers before the lesion. To verify that the preceding relations between finger force and finger position also apply for unskilled subjects, we asked four unskilled throwers to throw the 14-, 55-, and 196-g balls when force transducers were taped to the distal and middle phalanges of the middle finger. Figure 2, A-C, shows averages of 30 throws aligned on the moment of ball release (time 0) for one unskilled subject. Although this subject threw at a slower speed and had a different pattern of force on the distal phalanx than previously found for skilled throwers (because she gripped the ball with only the middle phalanx), two results were the same. First, the force on the distal and middle phalanges increased in proportion to ball weight. Figure 2, A and B, shows that mean forces with the heavy ball (196 g, thick lines) were greater than those for the tennis ball (55 g, medium lines), which in turn, were greater than those with the light ball (14 g, thin lines). Second, on average, the amplitude of the finger flexion 10 ms after the moment of ball release (dashed line; Fig. 2C) was proportional to the peak force on the distal phalanx prior to ball release (Fig. 2A). Figure 2D shows this relation in plots of the means ± SD for each parameter for each of the three balls of different weights. A similar increase in force on the distal and middle phalanges and increase in finger flexion for the heavier balls was found in one other unskilled subject, but in the other two unskilled subjects, the force on the distal phalanx was highly variable from throw to throw, presumably because the ball did not consistently roll over the force transducer on the distal phalanx. Nevertheless, the fact that in these two latter subjects the middle finger was not pushed back into extreme extension with the heavy ball (i.e., their finger extensions were similar to those in Fig. 2C) indicates that they were also scaling finger force to ball weight. This point will be taken up in a later section.

View larger version (16K): [in this window] [in a new window]   Fig. 2. Finger forces and finger positions for throws made by an unskilled thrower. A: mean force recorded on distal phalange for 30 throws made with each ball of 14-, 55-, and 196-g weight. B: mean force on middle phalanx. C: mean angular position of distal phalanx with respect to the hand for same throws. Traces have been aligned vertically (as well as horizontally) to be at the same finger position at ball release. D: relation for each ball for same throws between mean amplitude and SD of finger flexion 10 ms after ball release from the fingertip and mean peak force and SD on distal phalanx before ball release; subject Sh.

Cerebellar patients show greater variability in amplitude of finger opening when throwing

We now consider whether cerebellar patients can control these levels and changes in finger force throughout an overarm throw to produce a consistent finger grip and consistent amplitude of finger opening. As a starting point, we examined the ability of patients to open the fingers when no backforce due to arm movement was applied to the fingers. This was achieved by instructing the controls and patients to hold the hand stationary at about the position in a throw where the ball was released, to grip the ball as if throwing, and to open the hand fully and drop the ball. Figure 3A shows for 10 trials that both control (Mi) and cerebellar patient (Ol) opened the fingers smoothly and consistently from trial to trial. No cerebellar patient showed any abnormality such as hypometria or tremor of the fingers in this task. The likely reason that finger extension was constant in amplitude from trial to trial in both control and cerebellar subjects was that the movement started from one fixed position (finger grip on the ball) and finished at another (fingers fully open) and did not require active braking because of the effect of passive visco-elastic forces.

View larger version (33K): [in this window] [in a new window]   Fig. 3. Variability of finger opening in a cerebellar patient when throwing but not when opening the hand to drop a ball. In all cases, traces are 10 superimposed records of angular position of the distal phalanx of the middle finger with respect to the hand. A: finger opening aligned on end of movement when a control subject (Mi) and cerebellar patient (Ol) opened the hand and dropped a tennis ball. For all 20 ball drops, mean finger extension amplitudes were 94 ± 8.3° (mean ± SD) for the control and 136 ± 6.3° for the cerebellar patient. The difference in amplitude occurred because the cerebellar patient's finger grip position on the ball was more flexed. B: finger opening for same subjects aligned on ball release (time 0) for throws made at a medium speed with a tennis ball. C: finger opening for same throws aligned to be at the same finger position at ball release.

In contrast, when both the control and cerebellar patient threw a 55-g tennis ball, which is a situation where back force from the ball has to be controlled (Figs. 1 and 2), major differences were found in finger opening. Figure 3B shows that for 10 throws aligned on the time of ball release, angular finger position with respect to the hand in the control subject was similar from throw to throw but was highly variable in the cerebellar patient. One obvious difference was the greater variability in amplitude of finger opening in the cerebellar patient (from onset of finger extension to ball release). To examine finger opening in detail, the mean amplitude of finger extension and its SD was measured for 40 throws made with a tennis ball in all cerebellar patients and their controls. The amplitude of finger extension was measured from when finger extension velocity crossed a low threshold (200°/s) to the moment of ball release (time 0; Fig. 4). This threshold was chosen because it gave a reliable measure of the onset of finger opening. Mean finger extension amplitudes and SDs for the patients and matched controls, which are in equivalent positions across the page, are shown in Fig. 5. Although there was a tendency for mean finger amplitude to be larger in the cerebellar group (i.e., in 7 of the 9 pairs finger amplitude was larger for the cerebellar patients), this did not reach statistical significance (P = 0.103; unpaired t-test). However, a more striking result can be seen when considering variability of finger amplitude as measured by the SD: for each pair the SD was larger for the cerebellar patient. In agreement, unpaired t-test revealed a significant group difference (P = 0.023). Thus cerebellar patients show a greater variability in the amplitude of finger opening.

View larger version (19K): [in this window] [in a new window]   Fig. 4. Diagram illustrating how 3 parameters of finger position were measured during a throw. Trace is angular position of the distal phalanx of the middle finger with respect to the hand before and after ball release (time 0); control subject Mi.

View larger version (29K): [in this window] [in a new window]   Fig. 5. Means and standard deviations of the amplitude of finger extension for 40 throws made with a tennis ball. Data from controls and their matched cerebellar patients are in equivalent positions going from left to right.

Could this variability in finger amplitude be due to more trial-to-trial variability in throwing speed? We used hand angular velocity at ball release as a measure of throwing speed because many throws made by cerebellar patients missed the target and ball speed could not be calculated (from flight time to target impact). The mean throwing speed for patients (1,586°/s) was lower than that for controls (2,440°/s; P = 0.016; unpaired t-test). However, the variability of throwing speed, although larger for the cerebellar group (mean 204°/s) than the control group (mean 144°/s), was not significantly different (P = 0.203). In keeping with this, only one cerebellar patient (Sc) showed a relation between finger amplitude and throwing speed (R² = 0.24) that could have resulted from increased variability in throwing speed compared with their matched control. In short, considering all patients, the increased variability in finger amplitude was not due to more trial-to-trial variation in throwing speed.

Cerebellar patients have more variable amplitudes of finger movement before and after finger opening

Inspection of Fig. 3B reveals further differences between the finger position records from the control subject and cerebellar patient. A second difference is that there is an increase in variability from throw to throw in the change in finger position when gripping in the cerebellar patient. That is, for each throw before the onset of finger opening (i.e., the period between 200 and 100 ms in Fig. 3B), the control shows a fairly constant finger position for each throw as the hand grips the ball, whereas the cerebellar subject shows fluctuations in finger position. This often consisted of a drift toward extension, but in some cases, there was extension followed by flexion. Inspection of traces over a long time course showed that this flexion movement was not a rhythmic tremor. No patient showed cerebellar tremor of the fingers when gripping the ball. The amplitude of the change in finger grip position for each throw was measured as shown in Fig. 4. A fixed finger position was taken 200 ms before ball release (200 ms), and the amplitude of finger position with respect to this position (i.e., change in grip position) was measured 100 ms later (100 ms). For this single throw from control subject Mi, there was a slight flexion of the fingers at the 100 ms point. Figure 6A (top) shows that 100 ms after the fixed 200-ms point, cerebellar patients on average had finger grip positions slightly more toward extension than the controls. However, a more striking difference can be seen in Fig. 6A (bottom), i.e., variability in the amplitude of the change in finger grip position (as measured by the SD) was larger in the cerebellar group (P = 0.014; unpaired t-test). Only one patient (Hs) showed a relation between the change in finger grip position and throwing speed (R² = 0.21), but he had a lower variability in throwing speed than the matched control.

View larger version (20K): [in this window] [in a new window]   Fig. 6. Means and SD of the amplitude of the change in finger grip position (A) and amplitude of finger movement 10 ms after ball release (B). Amplitudes measured as shown in Fig. 4 for 40 throws made with a tennis ball in cerebellar patients and their matched controls.

A third difference between the cerebellar patients and controls occurred in the finger positions after release of the ball from the fingertip. Figure 3B shows that following ball release (0 time) for all throws in the control, the finger flexed slightly then extended. In contrast, in the cerebellar patient, after the moment of ball release, the finger flexed in some throws and extended in others. This difference is seen more clearly in Fig. 3C in which the finger position records have been aligned vertically (as well as horizontally) to be at the same finger position at the time of ball release. At any time after ball release, finger position is more variable in the cerebellar patient. This was determined quantitatively as shown in Fig. 4 by measuring for each throw the amplitude of finger movement after ball release, i.e., the amplitude difference of finger position 10 ms after ball release with respect to finger position at ball release. A value of 10 ms was chosen because in control subjects (e.g., Figs. 4 and 7B), it was the optimal time for measuring finger flexion before the rebound of the finger toward extension. Figure 6B (bottom) shows that the variability (SD) of the amplitude of finger movement 10 ms after ball release was larger in the cerebellar group (P = 0.024; unpaired t-test). Again, this could not be explained by increased variability in throwing speed in the cerebellar patients.

Evidence that cerebellar patients can scale finger force for balls of different weight

The preceding results demonstrate that cerebellar patients have more variable changes in finger positions throughout a throw than controls. The likely explanation is that this resulted from disordered control of finger force. Considering that throws with balls of different weights require development of different finger force (Figs. 1B, and 2, A and B), the question arose whether cerebellar patients can scale the appropriate level of finger force for throws with the different balls. To answer this question, we instructed six cerebellar patients and their controls to make 40 throws with each of the 14-, 55-, and 196-g balls. We determined their ability to scale finger force by means of an indirect approach based on previous results (Hore et al. 2001), which showed that the amplitude of the finger flexion that occurred after ball release is proportional to the force on the fingers before ball release, e.g., Figs. 1, B and C, and 2D. Records of finger position for 20 throws made with each of the three balls are shown in Fig. 7A aligned on the time of ball release for a cerebellar patient (Hv) and her control (Rn). For each of the three balls, the patient, as before, showed more variable changes in grip position for each throw, more variable amplitudes of finger opening to ball release, and more variable changes in finger position after ball release. For the first throw with a different ball (e.g., heavy after light), the amplitude of finger extension was no different from that in subsequent throws with the same ball. Figure 7B shows averages of finger position for all 40 throws for each ball aligned on ball release and aligned vertically so that finger positions were at the same point at ball release. For both the control and cerebellar patient, there was, on average, a flexion proportional to ball weight with the heavy ball having the largest flexion. This was determined quantitatively (as shown in Fig. 4) by measuring finger position 10 ms after ball release (Fig. 7B, - - -). The height of the bars in Fig. 8A represents mean finger position at this 10-ms point with respect to finger position at ball release (0°). When going from the light ball to the tennis ball to the heavy ball ( to  to ), although there is variability in absolute finger position, all controls and all cerebellar patients show a progression to greater flexion (or less extension) for the heavy ball. Accordingly, a repeated-measures ANOVA (with ball weight as repeated measure) revealed a significant effect of ball weight (P < 0.001), but no significant difference between patient and control groups (P = 0.77) and no significant ball weight × group interaction (P = 0.35). In view of the finding that the amplitude of finger flexion 10 ms after ball release is proportional to the force on the distal phalanx before ball release in skilled subjects (Hore et al. 2001) and in unskilled subjects (Fig. 2), we take this as evidence that, on average, both control subjects and cerebellar patients scaled finger force in proportion to ball weight.

View larger version (29K): [in this window] [in a new window]   Fig. 7. Angular finger position during throws made with balls of 3 different weights. A: superimposed traces of angular position of the distal phalanx of the middle finger with respect to the hand for 20 throws made with the light balls (14 g), tennis balls (55 g), and heavy balls (196 g). Traces aligned on ball release; control (Rn); cerebellar patient (Hv). B: average of 40 throws made with each ball aligned on the same angular finger position at ball release. - - -, 10 ms after ball release.

View larger version (33K): [in this window] [in a new window]   Fig. 8. Mean finger positions 10 ms after ball release are flexed in proportion to ball weight for both controls and cerebellar patients. A: mean finger angular positions with respect to the hand 10 ms after ball release for 40 throws plotted with respect to a baseline of 20° flexion from finger position at ball release (0°). B: SDs for amplitudes of finger positions 10 ms after ball release with respect to finger positions at ball release (measured as shown in Fig. 4).

Variability of finger position in unskilled throwers

Because cerebellar patients were of variable throwing ability before the lesion and because controls were matched in part on the basis of this prelesion throwing ability, it resulted that controls were also of variable throwing ability. We noticed that for some finger position parameters, variability (as measured by the SD), was higher for the unskilled controls compared with the skilled controls. For example, in Fig. 5B, the SDs of the amplitude of finger extension were larger for Mi, Rn, and Tm, who were relatively unskilled throwers, than they were for Kp, Bs, and Jn, who were relatively skilled. This raised the question of whether unskilled throwers, who did not have a cerebellar lesion, would show variability in finger position parameters equivalent to that observed for the cerebellar patients. To answer this question, the same experiment of throwing balls of three different weights was performed by six further subjects who were selected on the basis of their self assessment that they were unskilled (bad or very bad) throwers. Four subjects were 20- to 35-yr-old females (Kt, Ab, Tn, and Ss), one was a 22-yr-old male (Ml), and one was a 70-yr-old male (Lt). Of particular interest was subject Ss, a 24-yr-old female dental student who had passed the dental aptitude test for finger dexterity and who claimed to have almost no previous throwing experience.

In the very unskilled subject Ss, finger position varied in a similar way from throw to throw as found for the cerebellar patients. Figure 9 shows angular position of the distal phalanx with respect to the hand for 20 throws made with the heavy ball by a control subject (Mi), the very unskilled subject (Ss), and a cerebellar patient (Ol). Compared to the control, both the very unskilled subject and the cerebellar patient show more variable finger grip positions during each throw, more variable amplitudes of finger extension, and more variable finger movements after ball release than the control.

View larger version (24K): [in this window] [in a new window]   Fig. 9. Very unskilled throwers show a similar increase in variability in finger angular position throughout the throw as found in cerebellar patients. Traces are superimposed records of angular position of the distal phalanx of the middle finger with respect to the hand aligned on ball release for 20 throws made with the heavy ball by a control subject (Mi), a very unskilled thrower (Ss), and a cerebellar patient (Ol).

The mean amplitudes and SDs of finger opening (extension) for 40 throws made with the balls of different weight are shown in Fig. 10 for the controls, cerebellar patients and unskilled throwers (light ball, ; tennis ball, ; heavy tennis ball, ). One similarity between unskilled subjects and cerebellar patients was that both showed an increase in finger extension amplitude with an increase in ball weight. Regression analysis showed that such an increase was found in five of the six cerebellar patients (not in Hn) and in six of the six unskilled throwers. In contrast, most of the controls did not show a statistically significant increase (Mi, An, and Wt), and Jn showed a decrease. This fits with previous studies on skilled subjects that have shown no change or a decrease in the amplitude of finger extension with balls of increasing weight (Hore et al. 1999, 2001). Although control subjects Rn and Sh both showed an increase, this could have been because they were unskilled subjects. Like controls and cerebellar patients, all unskilled subjects also showed a finger flexion flick after ball release that was proportional to ball weight (e.g., Fig. 2, C and D).

View larger version (29K): [in this window] [in a new window]   Fig. 10. Means and standard deviations of finger extension amplitude for 40 throws made with balls of 3 different weights by control, cerebellar, and unskilled subjects.

As far as variability of finger extension amplitude is concerned, Fig. 10 shows that unskilled subjects Ss and Ab had larger SDs than any of the controls. Considering groups, SDs were larger for the cerebellar patient group than the control group for all three balls (P = 0.008, repeated-measures ANOVA) and were larger for the unskilled than the control group. However, in the latter case this did not reach statistical significance. The likely reason was that the control group also contained a number of relatively unskilled throwers. In fact, only two throwers in the control group were recreational ball players (Jn and Wt).

To investigate this further, we examined the variability in finger extension amplitude for each of the balls of different weights for four groups: the same three as before (Fig. 10) and a new group of six skilled male recreational ball players from a previous study (Hore et al. 1999). The bar heights in Fig. 11, which give the mean value of the SD of finger extension for throws with a particular ball made by the six subjects in each group, show that there was a progressive increase in variability of finger extension amplitude for all three balls in the order of skilled, control, unskilled, and cerebellar. In keeping with this, repeated-measures ANOVA showed a group difference between cerebellar and skilled (P = 0.001) and between unskilled and skilled (P = 0.043). In summary, unskilled throwers to various extents showed the same increased variability in finger position throughout and after the throw and the same increased amplitudes of finger extension for balls of increasing weights as found for the cerebellar patients.

View larger version (22K): [in this window] [in a new window]   Fig. 11. Mean variability of finger extension amplitude for different groups of throwers. Each bar was obtained by determining the mean and SD of the amplitude of finger extension for 40 throws made by each of 6 subjects then determining the mean of the SDs.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Indirect evidence for deficits in force production in cerebellar patients

Cerebellar patients perform different finger opening tasks with different degrees of disorder. For example, when asked to make small-amplitude finger flexions fast and accurately, cerebellar patients showed marked hypermetria followed by intention tremor (Hore et al. 1991). This was confirmed for small amplitude finger extensions in the present study (METHODS). However, the same patients were able to open the stationary hand to drop the ball with no obvious disorder (Fig. 3A). And when throwing, rather than marked hypermetria, patients showed increased variability in the amplitude of finger opening (Fig. 5). The likely explanation for the differences in performance in the different tasks is that the control of force and braking are different. Whereas fast small movements require early onset of antagonist muscle activity to brake the movement, which cannot be generated by cerebellar patients (Hore et al. 1991), hand opening in the ball drop task is presumably braked in large part by visco-elastic forces. In contrast, in throwing, accurate finger opening requires prediction of the back force from the ball on the finger based on motions at other joints, generation of the appropriate magnitude and rate of change of net finger flexor torque (grip force) to oppose the changing back force, and extension of the fingers while still applying the net finger flexor torque.

Indirect evidence indicates that the increased variability in finger position throughout the throw in cerebellar patients was caused by increased variability in finger forces. One piece of evidence comes from the finding that finger position shortly after ball release from the fingertip was more variable in the cerebellar patients (Figs. 3C, 6B, 7, 8B, and 9). We previously found in skilled throwers that the peak amplitude of the finger flexion (flick) that occurs immediately after ball release is directly proportional to the peak force on the distal phalanx prior to ball release (Hore et al. 2001). The same relation was found in some unskilled subjects in averages from throws made with the 14-, 55-, and 196-g balls (Fig. 2). It is therefore likely that the finger flexion that occurs after ball release in cerebellar patients and that on average is proportional to ball weight (Fig. 8A), also reflects the magnitude of net finger flexor torque prior to ball release. If this is the case, then the increased variability in finger position after ball release in cerebellar patients reflects increased variability in finger flexor force prior to ball release. Furthermore, because the amplitude of finger extension (to ball release) reflects the sum of all the forces acting on the finger, the increased variability in the amplitude of finger extension in cerebellar patients (Fig. 5) presumably reflects the same increased variability in these forces. Overall this finding in throwing of increased variability in finger position fits with the view of Thach et al. (1992) that cerebellar disorders are larger in multijoint than in single-joint movements because control of force is more complicated.

Could some other factor explain the increased variability in finger position throughout the throw? Although cerebellar patients were instructed to center the ball on the middle finger, it is possible that they had more variable middle finger configurations when gripping the ball. This could result in variable finger positions for the same overall grip force during each throw, which could have contributed to the larger variability in baseline finger grip positions (Figs. 3B, 7A, and 9). However, rather than measuring the differences in baseline levels between throws, which would have determined differences in initial (static) finger configurations, we measured the amplitude of the dynamic change in finger position when gripping during each throw and averaged this across throws. This measure of dynamic change should not be greatly affected by different initial finger configurations. Furthermore, variability in the middle finger configuration would not affect the ball rolling along the finger (Fig. 1A) and therefore would not cause the increased variability in finger amplitude or in finger flexion after ball release.

Although as a group cerebellar patients showed evidence of disordered control of the fingers, the degree of this disorder varied from patient to patient. In general, it appeared that the degenerative patients (Vn, Hn, and Pr) showed less disorder than patients with surgical or ischemic lesions, which affected the cerebellar nuclei. It was somewhat surprising that the two male degenerative patients (Hn and Pr), who had severe disorders in their posture and gait and who had abnormalities in braking fast small-amplitude finger movements (METHODS), showed a relatively small increase in variability in finger extension amplitude when throwing balls of different weights (Fig. 10). This may reflect a relatively small affect of the lesion on the control of the fingers (which would fit to some extent with the relatively small timing windows of 39 and 23 ms shown in Table 2) or it may have something to do with the fact that they were both good throwers before the lesion.

The question arises whether these findings of disordered finger control in cerebellar patients are related to the classic cerebellar signs of asthenia (decrease in force in voluntary movements) and hypotonia (decreased resistance to passive muscle stretch) (Holmes 1917, 1922; Luciani 1915). Although we did not directly measure muscle strength and muscle tone and therefore cannot make definitive conclusions, the evidence does not support this possibility. First, these classic disorders are only prominent in acute cerebellar patients and we studied chronic patients. In chronic patients, Mai et al. (1988) found no decrease in maximum force developed by the fingers. Second, cerebellar patients do not have to be hypotonic to show disorders in movement. Although hypotonia was originally believed to be a cause of cerebellar movement disorders (Holmes 1922), more recent work has failed to find any relation between hypotonia and arm ataxia and tremor (Growdon et al. 1967), gait ataxia in cats (Gorassini et al. 1993), and arm dysmetria in monkeys (Flament and Hore 1986). And third, unskilled subjects who were not weak or hypotonic showed the same finding as the patients of variable finger positions and a small increase in the amplitude of finger opening with balls of increasing weight. In summary, the evidence suggests that the increased variability in finger position throughout an overarm throw in cerebellar patients is due to variable control of finger force and not to weakness or hypotonia. This finding complements previous results where we have shown that cerebellar patients have increased variability in the timing of finger opening in this same task (Timmann et al. 1999).

Cerebellar involvement in prediction and production of force

A likely possibility is that increased variability in the production of finger force and in the timing of finger opening in cerebellar patients arises from disorder in the central commands to the fingers during a throw. Although most studies have interpreted the disorder in hand and arm movements in cerebellar patients as loss of predictive programming (Brooks and Thach 1981), the precise cause of the central deficit is still unclear. This is because loss of predictive programming could result from loss of the ability to predict movement dynamics per se or from loss of the ability to generate precise movement commands for the production of phasic force. As yet there is no consensus as to the relative importance of each of these mechanisms. One situation that has been studied in detail is control of interaction torques in fast multijoint reaching movements. There is agreement that cerebellar patients show disorder in control of these torques, which then contributes to cerebellar arm ataxia (Bastian et al. 1996; Topka et al. 1998). However, there is controversy as to whether this results from an inability to predict and compensate for interaction torques (Bastian et al. 2000) or from an inability to generate sufficient levels of phasic muscle torque (Boose et al. 1999; Topka et al. 1998).

For throwing, some results can be interpreted as being consistent with the proposal that cerebellar patients have lost the ability to generate precise movement commands for the production of phasic force. For example, the finding that five of the six cerebellar patients showed a tendency for larger finger amplitudes with balls of increasing weight could be explained as being due to loss of the ability of the fingers to generate a large-amplitude, fast-changing flexor force. This would fit with previous observations of deficits in finger control in cerebellar patients. It has previously been shown that patients have slowed speed in a repetitive isometric task involving the fingers (Mai et al. 1988), slowed development of finger grip force (Müller and Dichgans 1994b), and a reduced ability to adapt to heavy loads by increasing the peak pinch force rate (Müller and Dichgans 1994a). According to this scheme, the same finding in unskilled subjects of increased finger amplitudes with heavier balls would be due to the subjects failure to have learned the skill of producing large, fast-changing finger flexor torques. Alternatively, the increase in finger amplitude in patients and unskilled subjects could simply have been caused by a small passive movement at the distal interphalangeal joint with the heavier balls that was due to a failure to stiffen this joint properly as the ball rolled along the finger.

The present results indicate that patients had neither lost all of their ability to predict force nor to produce it. Two pieces of evidence indicate that patients, at least to some extent, were able to scale up finger flexor forces for the heavier balls. First, although back forces were 14 times larger for the heavy 196-g ball compared with the light 14-g ball (for the same hand acceleration), the fingers were not pushed back into extreme extension in throws with the heavy ball. Second, on average, cerebellar patients (like controls) had larger finger flexions after ball release for balls of heavier weights (Fig. 8A). Given that the amplitude of finger flexion 10 ms after ball release is proportional to peak force on the finger before ball release in skilled subjects (Hore et al. 2001) and in unskilled subjects (Fig. 2), this result suggests that in cerebellar patients finger force was on average scaled up for the heavy balls.

This ability of cerebellar patients to scale up finger force to some extent to the expected load agrees with other findings. For example, Müller and Dichgans (1994a) found that cerebellar patients were able to scale their pinch force levels to different loads in the 1- to 8-N range. Furthermore, in single-joint movements, cerebellar patients increased agonist electromyographic activity in response to both an increase in mass on the limb (Manto et al. 1994) and to an increase in movement amplitude (Hore et al. 1991). These findings bear some resemblance to those of Timmann and Horak (1997), who studied postural responses to perturbations of stance. They found that cerebellar patients could predict perturbation amplitudes based on prior experience, but they could not use this prediction to modify precisely the gain of responses. Taken together, these findings indicate that cerebellar patients retain considerable ability to scale force to the expected load, i.e., to predict and generate appropriate force levels, though this ability is disordered. Whether this residual ability results from function of intact areas of the cerebellum or from other extra-cerebellar motor areas remains unclear.

Increased variability is associated with lack of skill and with cerebellar lesions

The most striking result in the present study was increased variability in the changes in finger position throughout the throw in both the unskilled throwers and the cerebellar patients. For the unskilled throwers, this fits with previous studies that have shown that an increase in skill is associated with a decrease in variability of hand trajectories. For example, an increase in skill as a result of practice produced more consistent trajectories in a dart throwing task (Higgins and Spaeth 1972), in elbow movements in humans (Darling and Cooke 1987), and in an arm aiming task in monkeys (Georgopolous et al. 1981). As far as timing is concerned, we have previously found that lack of skill results in an increase in timing windows for finger opening and ball release. For example, male recreational ball players had timing windows for ball release of 9 ms when throwing with their skilled (dominant) arm and 22 ms when throwing with their unskilled (nondominant) arm (Hore et al. 1996b). Similarly, in the present study mean timing windows were 7 ms for skilled subjects and 27 ms for unskilled subjects (Table 2).

For cerebellar patients and monkeys with lesions of the cerebellum, increased trial-to-trial variability in movement appears to be a cardinal sign of their disorder. Such variability has been observed in saccades where the trial-to-trial variability affects both amplitude and direction (e.g., Robinson 1995; Robinson et al. 1993; Takagi et al. 1998). Similarly, for arm movements, Holmes (1922) described disorders and variability in their direction, range and rate. In agreement, recent studies of multijoint arm movements in cerebellar patients have found increased trial-to-trial variability in hand trajectories in throwing (Becker et al. 1990; Timmann et al. 1999, 2000) in reaching (Becker et al. 1991; Day et al. 1998), and in a pointing task (Bastian et al. 1996) and increased variability in the timing of finger opening in throwing (Timmann et al. 1999, 2000).

At first sight the results shown in Figs. 9-11 might be interpreted to suggest that in throwing, finger control in unskilled subjects and patients with a cerebellar lesion is equivalent. But this view would not be correct because cerebellar patients had larger timing windows (Table 2) and were considerably less accurate on the target. Furthermore, it is important to emphasize that we did not study patients with a nonfunctioning cerebellum. Rather the present patients had mild to medium cerebellar disorders (Table 1), which could be associated with the retention of considerable cerebellar function. Similarly, for the unskilled subjects, the cerebellum presumably contributed significantly to the control of their motions even if they were less accurate than those of skilled subjects. If one assumes that precise control of the fingers in throwing depends on the development of skill, which in turn depends on the cerebellum, it is not surprising that the performance of patients with various sized cerebellar lesions overlaps that of unskilled subjects with various degrees of skill.

Models of cerebellar function

Can these results be explained by current theories of cerebellar function? A recent extension of the concept that the cerebellum operates in part by predictive programming is that the cerebellum is the site of internal models of the motor apparatus and load (Kawato 1996; Wolpert et al. 1998). Considerable experimental evidence has been found to support the idea that skilled hand movements are controlled by such internal models (e.g., Flanagan and Wing 1997; Hore et al. 1999; Johansson and Westling 1988a,b; Laquaniti and Maioli 1989a,b). The cerebellum is implicated because of its adaptive capability (synaptic plasticity), its broad range of sensory inputs, and the properties of its circuitry, which it is argued, possess a capacity to approximate complex dynamics (Kawato 1996). The present results are consistent with this proposal in so far as they show loss of accurate and precise feedforward control. Beyond this, the proposal has not yet been developed so that it accounts for the retention in cerebellar patients of some predictive ability and the increased variability in their movement kinematics.

In contrast, a recent model of the function of the cerebellum in the control of saccades has been proposed which accounts for increased variability with cerebellar lesions (Quaia et al. 1999). In this model, the cerebellum accounts for both the accuracy and consistency (lack of variability) of saccades. This cerebellar contribution occurs during each saccade to compensate for both the characteristics of the occulomotor plant and the variability present in the rest of the saccadic system during the preparation and execution of the movement. Presumably the cerebellum is playing a similar, but more complex, role in arm movements.

In conclusion, the present study has demonstrated that, in a multijoint arm movement where generation of the appropriate finger force is based on prediction of the back force from the ball and production of the appropriate finger flexor torque, both cerebellar patients and unskilled subjects show the same deficits to various degrees: both can predict and scale force to some extent to the expected load, but both have an abnormally large variability from throw to throw in the amplitude of the finger position changes that occur throughout each throw. This presumably reflects increased variability in the control of finger force that depends in turn on the degree of cerebellar dysfunction and the degree of lack of skill. This result is consistent with the hypothesis that an essential role of the cerebellum is to achieve skill in movements by reducing variability in both timing and force of muscle contractions.