Fulltext for PDF in Flash Viewer:

Exp Brain Res (2004) 156: 149163 DOI 10.1007/s00221-003-1776-y RESEARCH ARTICLES Susanne M. Morton . Goran S. Dordevic . Amy J. Bastian Cerebellar damage produces context-dependent deficits in control of leg dynamics during obstacle avoidance Received: 29 July 2003 / Accepted: 7 November 2003 / Published online: 31 January 2004 # Springer-Verlag 2004 Abstract It has been suggested that the cerebellum is an important contributor to CNS prediction and control of intersegmental dynamics during voluntary multijoint reaching movements. Leg movements subserve different behavioral goals, e.g., locomotion versus voluntary stepping, which may or may not be under similar dynamic control. The objective was to determine whether cerebellar leg hypermetria (excessive foot elevation) during obstacle avoidance in locomotion and voluntary stepping could be attributed to a particular deficit in appropriately controlling intersegmental dynamics. We compared the performance of eight individuals with cerebellar damage to eight healthy controls as they walked or voluntarily stepped in S. M. Morton . A. J. Bastian Kennedy Krieger Institute, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA S. M. Morton . A. J. Bastian Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA e-mail: bastian@kennedykrieger.org S. M. Morton Interdisciplinary Program in Movement Science, Program in Physical Therapy, Washington University School of Medicine, St. Louis, MO 63108, USA G. S. Dordevic Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA G. S. Dordevic Faculty of Electronic Engineering, University of Nis, Serbia and Montenegro A. J. Bastian (*) Motion Analysis Laboratory, Room G-05, Kennedy Krieger Institute, 707 N. Broadway, Baltimore, MD 21205, USA e-mail: bastian@kennedykrieger.org Tel.: +1-443-9232718 Fax: +1-443-9232715 place over a small obstacle. Joint kinematics and dynamics were calculated during swing phase for both movement contexts. The kinematic analysis showed that hypermetria occurred during both walking and stepping and was associated with excessive knee flexion. When present, the amplitude of hypermetria was greater during stepping compared to walking. During stepping, subjects with cerebellar damage produced excessive knee flexor muscle torques and consequently overcompensated for interaction and gravitational torques normally used to decelerate the limb. During walking, the torque pattern was very similar to that of control subjects walking over a taller obstacle, and therefore might be a voluntary compensatory strategy to avoid tripping. Our results show that the extent of kinematic and dynamic abnormalities associated with cerebellar leg hypermetria is context-specific, with more fundamental abnormalities of leg dynamics being apparent during stepping as opposed to walking. Keywords Cerebellum . Dysmetria . Walking . Stepping . Human Introduction Most limb movements require simultaneous rotations about several joints. In order to produce movements that are smooth and accurate, the central nervous system (CNS) must therefore generate motor commands that guide rotations at multiple joints concurrently. This task is complicated by the fact that rotation at any one joint is caused not only by the muscular forces acting about that joint; gravity and other forces also play a role. For example, mechanical interaction torques, those caused by motions at other linked joints, can accelerate otherwise nonmoving joints and powerfully affect limb trajectory (Hollerbach and Flash 1982). Therefore control of interaction torques is a vital component of the coordination of multijoint limb movements. The cerebellum has been implicated as an essential contributor to CNS control of multijoint dynamics during 150 voluntary arm movements. Cerebellar damage affects multijoint arm movements to a much greater extent than single joint motions (Thach et al. 1992; Goodkin et al. 1993; Bastian et al. 2000; Goodkin and Thach 2003), suggesting that the cerebellum plays a critical role in coordinating movements across multiple joints. Hypermetria, or overshooting of targets, and curved hand paths are hallmarks of cerebellar dysfunction (Holmes 1939; Flament and Hore 1986; Hallett and Massaquoi 1993; Diener et al. 1993; Topka et al. 1998a; Bastian et al. 1996, 2000) and have been shown to be related to impaired control of interaction torques during reaching (Topka et al. 1998b; Bastian et al. 1996, 2000; Cooper et al. 2000). Mechanically limiting the number of moving joints (thus reducing the extent to which interaction torques affect a movement) markedly improves cerebellar hypermetria during reaching (Bastian et al. 2000). Despite good evidence to support a role for the cerebellum in the dynamic control of reaching movements, no study has addressed cerebellar control of intersegmental dynamics during movements of the legs, such as stepping or walking. Leg movements differ from arm movements in terms of the behavioral goal: most arm movements are intended for manipulating objects in the environment, while most leg movements are intended for moving the body from one location in space to another. Leg movements also differ from arm movements in that they probably rely heavily on specialized spinal cord control, especially during locomotion (Brown 1911; Forssberg and Grillner 1973; Grillner 1975). Still, people with cerebellar ataxia show hypermetria of the leg that appears similar to that seen in the arm during reaching. Stepping movements in response to a postural perturbation (Timmann and Horak 1998) and foot placements during walking on a treadmill (Stolze et al. 2002) are often hypermetric in people with cerebellar damage. Other studies have reported hypermetric leg movements during stepping and walking in cats and monkeys with cerebellar lesions (Botterell and Fulton 1938b; Chambers and Sprague 1955b; Udo et al. 1980; Yu and Eidelberg 1983). Yet it remains unknown whether, like the hypermetric arm movements, any of the leg forms of hypermetria are related to impaired control of interaction torques. If a general role of the cerebellum is to predict and adjust for the mechanical interactions between linked segments as has been suggested (Bastian et al. 1996), one might hypothesize that cerebellar damage would result in impaired control of interaction torques during all types of movements, including stepping and walking. On the other hand, the cerebellum is known to contain relatively discreet functional zones (Jansen and Brodal 1940; Botterell and Fulton 1938a, 1938b; Chambers and Sprague 1955a, 1955b; Yu and Eidelberg 1983; Thach et al. 1992), such that different longitudinal regions of the cerebellum are involved in movement control to varying extents depending on the specific type of movement being made. For instance, medial cerebellar regions are thought to be heavily involved in postural control and locomotion, intermediate cerebellar regions involved in controlling agonist-antagonist muscle pairs during single limb movements, and lateral cerebellar regions especially involved in planning and generating complex multijoint limb movements and those guided by vision (Thach et al. 1992; Ito 1984; Dichgans and Diener 1985). Therefore it is possible that the cerebellar regions utilized for the control of interaction torques during voluntary arm movements might not be utilized during leg movements. Thus, hypermetria during stepping or walking might not be attributable to abnormal control of multijoint dynamics. In the present study, we examined the kinematic and dynamic abnormalities associated with two types of leg movements that were similar kinematically but different in context, in people with cerebellar damage. Specifically, we examined the extent to which subjects with cerebellar damage demonstrated hypermetric leg movements, e.g., excessive foot elevation during swing, during obstacle negotiation in locomotor versus voluntary stepping movements. We chose to have subjects perform an obstacle avoidance task because the dynamics for this are well known (Patla and Prentice 1995; McFadyen and Carnahan 1997; Eng et al. 1997; McFadyen et al. 1999) and because cerebellar subjects would be less able to use compensatory strategies such as decomposition to avoid dealing with interaction torques, as had been described during level walking (Earhart and Bastian 2001; Morton and Bastian 2003). We then determined whether the hypermetria, when present, was related to a particular deficit in producing appropriate multijoint leg dynamics. Preliminary findings from portions of this study have been reported in abstract form (Morton and Bastian 2002). Materials and methods Subjects Eight subjects with cerebellar damage (three females and five males; 52.25��.10 years, mean age ��1 SE) and eight age- and gendermatched healthy control subjects (50.88��.70 years) participated in the study. All subjects gave their informed consent prior to participating and a University Human Studies Committee approved the study. Prior to testing, all cerebellar subjects underwent a thorough motor neurological examination. Cerebellar damage was confirmed by MRI or CT scan. Subjects with radiological evidence of damage beyond the region of the cerebellum and/or clinical evidence of involvement of other brain structures (e.g., motor weakness, sensory loss, hyperreflexia, bradykinesia, rigidity, dystonia) were excluded from the study. See Table 1 for cerebellar subject information. Further detail regarding diagnoses for these subjects is provided below. Of the eight individuals with cerebellar damage, seven had cerebellar atrophy. In these subjects, damage was diffuse and evident throughout the cerebellum. Three of the subjects with atrophy (CBL-1, -2 and -8) tested positive for a specific spinocerebellar ataxia, SCA6. Subject CBL-2 also tested positive for SCA8. The other four subjects with cerebellar degeneration (CBL-4, -5, -6, and -7) were diagnosed with idiopathic pancerebellar atrophy. Upon neurological examination, none of these subjects had evidence of involvement of any brain region beyond the cerebellum, none had any signs of autonomic system dysfunction, and none reported a history of family members with similar symptoms. One subject, 151 speed for each subject averaged over all trials; Overshoot freq. column indicates how frequently each cerebellar subject showed overshooting of the foot height (e.g., the ratio of the number of overshoot trials for each subject to the total number of trials performed by that subject). See ��aterials and methods��for details regarding the criteria for categorizing trials as overshoot Overshoot freq. (%) CBL-3, had a focal lesion. This subject had an infarct of the left posterior inferior cerebellar artery. Step 40 0 0 60 0 0 67 100 - Paradigm All subjects participated in two leg movement conditions: walking over a small obstacle and stepping over a small obstacle from a standing position. Control subjects repeated both movement conditions with a second, taller obstacle. Detailed information regarding the two movement conditions is given below. In the walking condition, subjects walked across a level walkway, approximately 8 m in length, that had a small obstacle (a box, 5 cm tall ��10 cm wide) placed across either the right or left (but always the same) side of the walkway, such that subjects were forced to walk over the obstacle with one leg. The obstacle was placed in the walkway at a location where subjects were easily able to approach it during the swing phase of their gait cycle. We wanted subjects to be able to incorporate clearance of the obstacle into their normal walking pattern as effortlessly as possible. Subjects were instructed to approach the obstacle, walk over it, and continue moving past the obstacle without pausing. In the stepping condition, subjects did not walk, but rather stood in place and made single steps (one step per trial) over the same small obstacle. A step was considered to consist of the movement of the foot over the obstacle and did not include the return of the foot to the starting position. As in the previous task, the obstacle was placed in front of either the right or left (but always the same) foot. Subjects��feet were positioned approximately shoulder-width apart with the toe of the stepping foot parallel to the heel of the stationary foot and the stationary foot parallel to the obstacle. This position was intended to closely resemble the limb configuration during walking over the obstacle. All subjects performed the stepping condition while holding onto a fixed bar with the hand contralateral to the stepping foot, thus reducing the effects that a balance deficit may have had on performance. A second box (20 cm tall ��10 cm wide) was used for the trials where control subjects repeated the two conditions with a taller obstacle. The purpose of this secondary analysis was to determine whether any dynamic differences seen between control and cerebellar groups, if they existed, could be explained by cerebellar subjects using a voluntary ��afety��strategy. That is, if cerebellar subjects took high (hypermetric) steps and also demonstrated dynamic differences, it could be because they were simply utilizing a safety strategy to avoid hitting the obstacle and possibly falling. By testing control subjects navigating a taller obstacle, we approximated the hypermetric movements of cerebellar subjects navigating the small obstacle. If no dynamic differences remained when comparing cerebellar subjects navigating the small obstacle to control subjects navigating the taller obstacles (thus controlling for kinematic differences), we would conclude that the dynamic differences could be largely attributed to a voluntary strategy. In both conditions, cerebellar subjects were instructed to walk or step at a fast speed. Cerebellar subjects always moved slower than control subjects. Because of this, controls were asked to pace their movements to a metronome, which was set to the cadence of the cerebellar subjects. We chose to match cadences so that speed differences between the two groups would be minimized and we could therefore more fairly compare torque magnitudes, which are dependent upon velocity and acceleration. All subjects received practice with both movement conditions prior to recording data. Subjects performed three to five trials of each condition. An examiner guarded subjects during all tasks and all subjects received frequent rest breaks throughout the testing session in order to minimize fatigue. One subject from the cerebellar group (CBL-7) was unable to walk alone safely, and was given handhold assistance by an examiner during the walking condition. Subjects with cerebellar damage were instructed to clear the obstacle with their more involved leg. If, upon neurological examination, neither leg appeared to be more involved, we tested their preferred leg (see Table 1). Walk Step Walk 100 100 40 20 33 60 67 33 CBL-1 CBL-2 CBL-3 CBL-4 CBL-5 CBL-6 CBL-7 CBL-8 Mean (SE) 33 48 41 68 57 51 64 56 52.25 (4.10) F M M M M F F M R R R R R R R R L R L R R R R R SCA6 SCA6, SCA8 L PICA stroke Pancerebellar atrophy Pancerebellar atrophy Pancerebellar atrophy Pancerebellar atrophy SCA6 2y 5y 6d 7y 3y 8y 8y 3y 4.50 y (1.05) 36 43 17 13 36 31 74 19 33.62 (6.88) 0.89 0.77 0.92 1.65 1.26 1.16 0.14 1.11 0.99 m/s (0.15) 0.90 0.70 0.81 1.00 1.98 0.93 0.37 0.62 0.81 m/s (0.20) Table 1 Cerebellar subject information [LOI length of time since onset, ICARS score on International Cooperative Ataxia Rating Scale (Trouillas et al. 1997; range of possible scores is 0100, with 0 indicating no ataxia and 100 indicating severe ataxia), SCA spinocerebellar ataxia, PICA posterior inferior cerebellar artery]. Preferred leg column indicates the leg with which subjects preferred to kick a ball; Movement speed column indicates the movement Subject Age (years) Gender Preferred leg Tested leg Diagnosis LOI ICARS Movement speed 152 Data collection Joint positions were recorded in three dimensions using the OPTOTRAK System (Northern Digital, Waterloo, ON). Two OPTOTRAK sensors were placed on opposite sides of the laboratory space so that recordings could be made simultaneously from both sides of the body. Twelve infrared light emitting diodes (IREDs) marked the positions of the shoulders (head of humerus), pelvis (superior iliac crest), hips (greater trochanter), knees (lateral knee joint space), ankles (lateral malleolus), and feet (head of fifth metatarsal). Four additional IREDs were placed on four corners of the obstacle, marking its location and dimensions. We defined the coordinates of our laboratory space such that forward-backward movements occurred in the x-direction, vertical movements in the ydirection, and lateral movements in the z-direction. Data were collected at 100 Hz. Data analysis Joint position data were low-pass filtered at 6 Hz. OPTOTRAK software was used to calculate marker positions and joint angles, and to generate animated stick figures that were used to identify the times of foot contact with the ground. The joint angles were studied only in the sagittal plane. Custom software from MATLAB (Mathworks, Inc.) was used to calculate angular velocities and accelerations from the angle data, and the MATLAB Robotics Toolbox (Corke 1996) was used for calculating the joint dynamics. Kinematics For each trial, we used the animated stick figures to identify the times of initial contact and lift off of the foot from the floor. During the walking condition, a full gait cycle was defined as the period from initial contact to initial contact, which encompassed stance (initial contact to lift off) and swing (lift off to initial contact) phases. Fig. 1AC Experimental setup. A Knee joint kinematics (top three traces) during a full gait cycle (initial contact to initial contact) and knee joint dynamics (bottom trace) during the period of flexor deceleration from a typical control subject walking over the small obstacle. Vertical dashed line indicates peak knee flexor velocity after lift off; vertical dotted line indicates the first zero crossing of the knee velocity trace. Gray shaded area indicates the period of flexor deceleration and the period for which the knee dynamics are drawn. B Example of limb configuration during swing phase (lift off to initial contact) for the same control subject as in A. Knee flexor deceleration phase is indicated in red. Flexor deceleration ranged from peak flexor velocity (which typically occurred near lift off) to the first zero crossing of the velocity trace (which always coincided with peak joint flexion). C Schematic of the leg segments, joint angles, and degrees of freedom considered in the dynamic analysis. Segments were the foot, shank, thigh, and pelvis. DOF included sagittal-plane translations in the x-dimension (t1) and y-dimension (t2) and sagittal-plane rotations about the trunk (r3), hip (r4), knee (r5), and ankle (r6). The linear movements (t1 and t2) were assigned to be occurring only at the trunk, and were thus incorporated into the net torque term for the trunk and the interaction torque term for all other joints. Direction of arrows indicates the direction of positive torques. Direction of rotation at the knee was reversed so that flexor rotations at all joints would always be given a positive sign 153 The stepping condition contained only a swing phase (lift off to initial contact). Kinematic measurements included: (1) movement speed, to confirm that pacing control subjects��movements to the metronome was successful in controlling for speed differences between the two groups, (2) foot positions relative to the obstacle, to determine whether swing phase starting positions differed between control and cerebellar groups, (3) the maximal vertical height of the foot, e.g., hypermetria, (4) peak flexion angles at the ankle, knee, and hip joints, to determine which joint(s) were responsible for the hypermetria when it occurred, and (5) acceleration and deceleration phase durations at the ankle, knee, and hip joints, because hypermetria during reaching movements can be associated with a prolongation of deceleration (Brown et al. 1990; cf. Hore et al. 1991; Flament and Hore 1986). We determined the extent of dysmetria on a trial-by-trial basis for each of the cerebellar subjects. Trials were categorized as either ��vershoot��(OS), ��o error��(NE), or ��ndershoot��(US) trials, e.g., either hypermetric, within normal range, or hypometric. The categorization was based upon whether the peak vertical foot height was greater than, within, or less than two standard deviations of the mean peak vertical foot height from the control group (Bastian et al. 2000). We chose to categorize trials in this manner because it allowed for a direct group comparison of kinematics and dynamics between the occasions when cerebellar subjects were able to perform the leg movements without hypermetria and the occasions when they demonstrated the abnormal hyper- or hypometria. We also identified a ��lexor deceleration phase��for each joint, which was defined as the time from peak flexor velocity after lift off to the subsequent zero velocity crossing. This is the time period when the joint movement was slowing just before reaching the maximum flexion angle. Figure 1A illustrates the method for calculating the flexor deceleration phase at the knee joint. Joint dynamics were calculated over this time period. The flexor deceleration phase was selected because it encompassed nearly the entire period from lift off to peak joint flexion. Our dynamic model did not account for ground reaction forces and was thus invalid whenever the leg was in contact with the floor. Figure 1B shows an example of the typical limb configuration during flexor deceleration at the knee joint. refers to torques produced by active and passive tissue forces about the joint. The interaction torque was defined as the torque produced at a segment that was caused by movement at other linked segment (s). The interaction torque is composed of acceleration-dependent (inertial) and velocity-dependent (Coriolis and centrifugal) terms. The gravitational torque was defined as the torque produced by the weight of the segment in question (assigned to be acting at the segment center of mass) and all segments distal to it. The muscle torque is equivalent to the ��esidual torque��of Cooper et al. (2000), the ��eneralized muscle torque��of Hoy et al. (1985), and the ��et torque��of Hollerbach and Flash (1982) and Soechting and Lacquaniti (1981). The muscle torque reflects active muscle contractions about a given segment as well as other passive, noncontractile forces, such as muscle stiffness, viscous, and joint friction forces. The net, interaction, and gravitational torques are computed using the recorded data; the muscle torque cannot be directly measured and is instead calculated as the difference between the net torque and the sum of the interaction and gravitational torques. Torques were normalized to body mass so that comparisons of magnitude could be made between subjects. We compared net, interaction, gravitational, and muscle torque magnitudes averaged over the flexor deceleration phase (see Fig. 1A). We used factorial ANOVA to compare kinematic variables between control and cerebellar groups across the walking and stepping conditions. One-way ANOVA was used to compare dynamic measures between control and cerebellar subgroups (NE and OS trials) in each condition. When the ANOVA yielded a significant effect, post hoc analyses were completed using Tukey�� honest significant difference test. Statistica (StatSoft, Tulsa, OK) software was used for all statistical analyses. Results Kinematics Prior to the main analysis, we verified that pacing control subjects��movements to those of the cerebellar subjects was effective in matching speeds across the two groups. Accordingly, we compared movement speeds between the groups and across the testing conditions. Factorial ANOVA showed no significant group (1.08��.04 m/s control vs. 1.01��.05 m/s cerebellar, p=0.311), condition (1.09��.03 m/s walking vs. 1.00��.05 m/s stepping, p=0.143), or group ��condition interaction (p=0.474) effects. Table 1 provides the average movement speeds for each of the cerebellar subjects. Given the lack of a significant difference between groups, we concluded that it would be legitimate to compare torque magnitudes between groups. We also compared the (x-dimension) distance between the obstacle and the positions of the swing and stance feet at the time of lift off. For the swing foot, there was no significant group (47.94��.75 cm control vs. 47.45 ��.78 cm cerebellar, p=0.960), condition (49.51��.73 cm walking vs. 47.23��.32 cm stepping; p=0.211), or group ��condition interaction (p=0.541) effects. Results were similar for the stance foot: there was no difference between control and cerebellar groups (11.90��.57 cm control vs. 12.61��.69 cm cerebellar, p=0.538), between walking and stepping (17.16��.30 cm walking vs. 15.39 ��.57 cm stepping, p=0.183), and no interaction effect (p=0.608). Therefore, any difference in torque patterns Dynamics Using the recorded kinematic data and anthropometric estimates of segment masses, lengths, and moments of inertia (Winter 1990), we employed inverse dynamics equations of motion to calculate torques in joint-centered coordinates for a four-segment, 6-degree-offreedom (DOF) sagittal-plane model of the leg during swing phase. The equations of motion were generated using the MATLAB Robotics Toolbox (Corke 1996) and have been verified against published equations of motion written for a four-segment, 6-DOF model of the cat forepaw during reaching (Cooper et al. 2000). Figure 1C shows our leg model, the limb segments, and the DOF. The four segments were: the foot, the shank, the thigh, and the pelvis. The DOF included: sagittal-plane (flexion-extension) rotations about the ankle, knee, hip, and trunk, as well as translations in the forward-backward (x) and vertical (y) dimensions to account for the linear movements of the leg. The linear movements were assigned to be occurring at the trunk. Joint torques were computed and sorted into four types: net, interaction, gravitational, and muscle. We defined these terms in the same manner as previously described for the arm (Bastian et al. 1996, 2000). The net torque was defined as the sum of all of the torques acting about a joint. It is proportional to the acceleration of the segment in question and was estimated by the product of the angular acceleration of the segment and the moment of inertia of that segment and all distal segments. The net torque is equivalent in magnitude to the ��elf torque��reported by Cooper et al. (2000) and should not be confused with the net torque described by others (Hollerbach and Flash 1982; Soechting and Lacquaniti 1981), which 154 between groups or conditions would not have been caused by differences in initial foot positions. We then identified if and when individuals with cerebellar damage produced hypermetric leg movements, and determined which joint(s) were responsible for the increases in foot height. Swing phase foot path traces and joint angles from a control and a cerebellar subject are shown in Fig. 2. The traces show single trials from the obstacle walking and stepping conditions. Stick figures depicting the limb configurations and foot paths are shown above (Fig. 2A), with the accompanying joint angles below (Fig. 2BD). During obstacle walking, the cerebellar subject showed a somewhat elevated foot height compared to the control subject; during obstacle stepping, the cerebellar subject showed an even greater increase in foot height. The joint angle traces indicated that much of this subject�� hypermetria was caused by increases in flexion at the knee joint (note arrows). Ankle angles Fig. 2AD Individual examples of swing phase kinematics from single trials from a cerebellar and a control subject during the obstacle walking (left column) and obstacle stepping (right) conditions. A Stick figures showing limb configurations and foot path traces for the cerebellar (top row, black traces) and control (bottom row, gray traces) subjects. Stick figures are plotted every 50 ms for all of swing phase. Red arrows indicate a moderately increased foot height during obstacle walking and a markedly increased foot height during obstacle stepping by the cerebellar subject (S shoulder, T trunk, H hip, K knee, A ankle, MT fifth metatarsal). BD Swing phase joint angle traces for the same trials as in A; B ankle angles; C knee angles; D hip angles. Red arrows indicate increased peak knee flexion by the cerebellar subject indicated primarily reduced plantarflexion by the cerebellar subject and hip angles revealed little difference between the two subjects. Table 2 shows the results from the categorization used to identify the hypermetric trials (foot height more than 2 SD above controls). Hypermetria was common during both walking and stepping over the obstacle, but was somewhat more frequent during obstacle walking (54%) than obstacle stepping (30%). All eight cerebellar subjects had at least one overshoot trial during the obstacle walking condition; four had at least one overshoot trial during the obstacle stepping condition (see Table 1). No subjects exhibited hypometria in either the walking or the stepping condition. Based on the categorization, we divided the cerebellar group and then compared joint kinematics between control, cerebellar-NE (no error), and cerebellar-OS (overshoot) groups during the walking and stepping 155 Table 2 Categorization of cerebellar trials. Values indicate the percentages of trials from the cerebellar group that fell into each category for walking versus stepping over the obstacle Walking (%) Overshoot No error Undershoot 54 46 0 Stepping (%) 30 70 0 conditions. Figure 3 shows average peak foot heights (Fig. 3A) and peak flexion angles at the hip (Fig. 3B), knee (Fig. 3C), and ankle (Fig. 3D) joints for the three groups. Notice that while peak foot heights from the cerebellar-NE trials appear greater than those from the control trials, they are still within the 2 SD criteria we used for categorization (error bars in the figure represent SEs; shaded region represents the �� SD criterion for categorization). Thus, although not elevated enough to meet our criteria, the foot heights from the cerebellar-NE trials were still slightly higher than control subjects. The statistical comparison of peak foot heights revealed main effects of group (p<0.001), condition (p<0.001), and a group ��condition interaction effect (p=0.033). The interaction shows that when cerebellar subjects overshot, the hypermetria was of greater magnitude during stepping than walking, compared to controls. To determine which joint(s) were responsible for the increase in peak foot heights, we looked to the measures of Fig. 3 A Peak foot heights from control, cerebellar-NE, and cerebellar-OS groups during the obstacle walking and stepping conditions. There were significant group, condition, and group ��condition interaction effects. Gray shaded rectangles indicate the �� SD criteria used to categorize cerebellar trials. B Peak hip flexion angles. C Peak knee flexion angles. D Peak ankle dorsiflexion angles. There was a trend toward a significant group ��condition interaction effect only at the knee joint. Error bars represent �� SE peak flexion at the hip, knee, and ankle joints. At the hip (Fig. 3B), all groups increased peak flexion during stepping compared to walking, but there was no difference between groups (condition effect, p<0.001; group, group ��condition interaction effects, ns). Therefore, excessive flexion at the hip did not appear to be responsible for the increased peak foot height in the cerebellar-OS trials. The knee joint angles (Fig. 3C) showed a similar pattern to that observed for the measure of peak foot height. All groups showed an increase in peak knee flexion in the stepping condition compared to walking (condition effect, p<0.001). Additionally, peak knee flexion was increased overall in the cerebellar-OS trials compared to the other groups (group effect, p<0.001). Finally, the extent of excessive knee flexion appeared even greater in the cerebellar-OS group when stepping over obstacles compared to walking over obstacles (trend toward significant group ��condition interaction, p=0.086). Thus, the knee was an important contributor to the patterns of leg hypermetria that we observed in the cerebellar-OS trials. At the ankle (Fig. 3D), both the cerebellar-NE and cerebellar-OS groups had increased peak dorsiflexion compared to controls, but no group showed a difference between walking and stepping (group effect, p<0.001; condition, group ��condition interaction effects, ns). Excessive dorsiflexion might have explained some of the leg hypermetria, but it was unlikely to have explained a considerable deal of it for at least two reasons. First, there was no relative increase in dorsiflexion in the cerebellar- 156 OS group during stepping compared to walking. So increases in dorsiflexion could not explain why leg hypermetria was worse during stepping. Second, although both the cerebellar-OS and cerebellar-NE groups showed nearly identical increases in peak dorsiflexion, only cerebellar-OS trials were significantly more hypermetric. The foot has a shorter segment length than the shank or thigh, so the extent to which ankle movements contribute to the overall elevation of the leg is minimal compared to knee or hip movements (e.g., increasing knee flexion by 5��esults in more foot elevation than increasing dorsiflexion by 5�� Therefore, the joint most likely to be responsible for the leg hypermetria appeared to be the knee. To verify that there was a relationship between the extent of hypermetria and the extent of excessive knee flexion, we measured the correlation between peak vertical foot height and peak knee flexion for all trials (both NE and OS trials) from the cerebellar group during the walking and stepping over obstacles conditions. There was a highly significant positive correlation between the two variables (r=0.747, p<0.001), indicating that cerebellar leg hypermetria was related to excessive knee flexion. Hence, in this report we focused much of the subsequent kinematic and dynamic analyses on the knee joint. We next examined acceleration and deceleration durations. Excessive flexion of a joint can be associated with a prolonged deceleration phase and prolonged deceleration has previously been associated with hypermetric arm movements in cerebellar subjects (Brown et al. 1990). The average flexion acceleration phase durations at the knee joint were slightly increased in the cerebellar-OS group during stepping, but this increase was not significant (no significant group, condition, or group ��condition interaction effects). However, deceleration durations were significantly prolonged. Figure 4 shows the average flexion deceleration phase durations at the knee joint for control, cerebellar-NE, and cerebellar-OS groups during the walking and stepping conditions. All groups increased the time spent in deceleration during stepping, but the cerebellar-OS group did so to a greater extent. The ANOVA revealed a condition (p<0.001), a trend toward a group (p=0.085), and a group ��condition interaction (p=0.037) effect. The statistics indicated that cerebellar overshooting was associated with a prolonged knee flexion deceleration phase during stepping over the obstacle, but not during walking over the obstacle. Dynamics: knee The kinematic findings suggested that most of the leg hypermetria was caused by excessive flexion at the knee. During stepping, hypermetria was greater and was associated with a prolongation of knee deceleration. Therefore, we investigated knee torque magnitudes during the knee flexor deceleration period. Prior to the torque analysis, we checked the planarity of the movements (e.g., we checked to see if movement was constrained to the Fig. 4 Duration of the deceleration phase at the knee joint from control, cerebellar-NE, and cerebellar-OS groups during the obstacle walking and stepping conditions. There was a significant condition effect, a trend toward a group effect, and a significant group ��condition interaction effect. Error bars represent �� SE parasagittal plane) by calculating the apparent limb lengths of the thigh and shank segments in the parasagittal plane. Rotations out of this plane will cause apparent changes in limb length (Hoy et al. 1985). We found that the apparent length of the thigh and shank changed less than 10% for all trials for control and cerebellar subjects, which has been previously considered acceptable for this type of analysis (Hoy et al. 1985). Figure 5 shows individual traces of the knee net, interaction, gravitational, and muscle torques from control and cerebellar subjects during the knee flexor deceleration phase for the walking and stepping conditions. In both conditions, knee flexor deceleration was caused primarily by the extensor interaction and gravitational torques. In the walking condition, the muscle torque produced by the control subject (Fig. 5A) started in the extensor direction, assisting with deceleration of the knee, and barely crossed into flexion during the second half of flexor deceleration. In contrast, the cerebellar subject (Fig. 5B) exerted a flexor muscle torque almost immediately, resisting knee deceleration, and the muscle torque reached a greater flexor magnitude than that produced by the control subject. Consequently, the average muscle torque for the trial from the cerebellar subject was 0.005 Nm/kg (an average flexor muscle torque), while the average muscle torque for the trial from the control subject was ��.033 Nm/kg (an average extensor muscle torque). Recall that the knee was decelerating during this period (e.g., the net torque was extensor), and the increased flexor muscle torque therefore resisted slowing knee flexion. In the stepping condition, the muscle torque produced by the control subject (Fig. 5D) started in the flexor direction and moved into the extensor direction after approximately two-thirds of the flexor deceleration duration. The greater flexor muscle torque during stepping compared to walking was required because the interaction torque contributed more to knee deceleration, and there- 157 Fig. 5AF Individual examples of knee flexor deceleration torques from single trials from four control and two cerebellar subjects. Obstacle walking (top row) and stepping (bottom) conditions are shown. A Control subject walking over the small obstacle. B Hypermetric cerebellar subject walking over the small obstacle. C Control subject walking over the taller obstacle. D Control subject stepping over the small obstacle. E Hypermetric cerebellar subject stepping over the small obstacle. F Control subject stepping over the taller obstacle. Knee muscle torques, averaged over the flexor deceleration phase, are indicated in the bottom right corner of each graph fore some active countering of the interaction torque by the muscle torque was needed. The cerebellar subject (Fig. 5E) also exerted flexor muscle torque immediately, but maintained it throughout the entire flexor deceleration phase. Consequently, the average muscle torque for the trial from the cerebellar subject was 0.082 Nm/kg, while the average muscle torque for the trial from the control subject was 0.032 Nm/kg. Both subjects produced average flexor muscle torques, but the magnitude of the torque was more than doubled for the cerebellar subject. As with walking, the knee was decelerating during this period and the increased flexor muscle torque acted to resist slowing knee flexion and could have contributed to excessive knee Fig. 6 Mean knee flexor deceleration torques from control, cerebellar-NE, and cerebellarOS groups during the obstacle walking (A) and stepping (B) conditions. Significant or trending toward significant group (muscle torque during obstacle walking) or post hoc (all others) statistics, comparing the control group to the cerebellar-OS group, are indicated at the bottom of each graph. Error bars represent �� SE flexion. Also notice that during stepping, the cerebellar subject required increased time to complete flexor deceleration. This corresponds to the prolonged knee deceleration previously described (see Fig. 4). Group data depicting average knee flexor deceleration torque magnitudes are shown in Fig. 6 for control, cerebellar-NE, and cerebellar-OS groups during the obstacle walking and stepping conditions. Average net, interaction, gravity, and muscle torques are shown. Since the net torque is effectively the sum of all other torque components, it is easy to appreciate the relative contributions that the interaction, gravity, and muscle torque each make to the net torque. The average knee muscle torque 158 produced by control and cerebellar-NE groups during walking (Fig. 6A) was extensor, indicating that it assisted with deceleration (e.g., assisted in slowing knee flexion). The average knee muscle torque produced by the cerebellar-OS group however was, on average, flexor, indicating that it resisted deceleration (e.g., resisted slowing knee flexion). The ANOVA revealed that differences in average muscle torques across the three groups did not reach statistical significance (p=0.105), likely because of the somewhat large within-group variability. Average interaction torques were different across the three groups (p<0.001). Post hoc analysis indicated that the interaction torque from the cerebellar-OS group was larger than control subjects (p<0.001). Neither average net torques nor average gravitational torques were significantly different among the three groups. In the stepping condition (Fig. 6B), the average knee muscle torque produced by all three groups was flexor, indicating that it resisted deceleration. However, the ANOVA showed that the magnitude of the flexor muscle torque was different among the three groups (p=0.010). Post hoc analysis revealed that the cerebellar-OS group produced a flexor muscle torque that was greater than the control group (p=0.007) and nearly greater than the cerebellar-NE group (p=0.055), but there was no significant difference between the control and cerebellar-NE groups. The average interaction and net torques during stepping appeared reduced in the cerebellar-OS group; however, these differences were not quite significant (p=0.251 and p=0.152, respectively). Average gravitational torques were significantly different among the three groups (p=0.001), but the post hoc analysis showed no difference between control and cerebellar-OS groups (p=0.453). The average knee torque comparisons indicated that excessive flexor muscle torque production might be responsible for the hypermetric leg movements during walking and stepping. We examined this relationship further by measuring the correlation between peak foot Fig. 7 Mean knee flexor deceleration torques from the control group navigating the taller obstacle and the cerebellar-OS group navigating the small obstacle are shown during the walking (A) and stepping (B) conditions. Note that the cerebellar-OS group data are identical to that previously shown in Fig. 6. Significant statistics are indicated at the bottom of each graph. Error bars represent �� SE height and average knee muscle torque. In control subjects, we found no correlation between these two variables. Cerebellar subjects showed no significant relationship between foot height and knee muscle torque during walking, but a moderate and significant correlation (r=0.34; p=0.028) between these two variables during stepping. The positive correlation indicated that the greater (e.g., more flexor) the average knee muscle torque, the greater the peak foot height. The finding of increased flexor muscle torque at the knee and resulting elevation of foot heights by cerebellar subjects could be caused by different processes. One possibility is that cerebellar subjects were unable to appropriately scale their muscle torque to account for interaction torques occurring during the movement. Another possibility is that cerebellar subjects simply used a voluntary strategy of increasing knee flexor muscle activity as a ��afety margin��in order to insure clearance of the obstacle. Cerebellar subjects, knowing they have difficulty controlling the leg, might opt to purposely ��verflex��at the knee to make sure that they do not hit the obstacle and risk tripping or falling. To address this possibility, in a second analysis we compared the data from cerebellar subjects to that of control subjects walking and stepping over a taller obstacle. In this comparison, the movement kinematics were more equivalent between groups (e.g., peak foot heights and knee flexion angles were not significantly different between control and cerebellar groups). If cerebellar subjects had no dynamic abnormalities and were merely using a voluntary strategy to lift the leg higher, their patterns of torque production navigating the small obstacle should closely resemble those produced by control subjects navigating the taller obstacle. Figure 5C and F show individual traces of the knee net, interaction, gravitational, and muscle torques from two control subjects navigating the taller obstacle. These traces can be compared to the two cerebellar subjects who 159 overshot navigating the small obstacle (Fig. 5B, E). During obstacle walking, the torque patterns were very similar between the two subjects. During obstacle stepping, the average muscle torque produced by the cerebellar subject was still excessive compared to the control subject and the net torque was therefore still relatively reduced. Group knee torque data from controls navigating the taller obstacle and the cerebellar-OS group are shown in Fig. 7 (note that the cerebellar-OS group data are the same as that presented in Fig. 6). In the obstacle walking condition (Fig. 7A), the weak difference previously noted between control and cerebellar-OS groups in the average knee muscle torque magnitudes no longer remained (p=0.471). Only interaction torques were statistically different (p<0.001), being somewhat increased in the cerebellar-OS group. In the obstacle stepping condition (Fig. 7B), the significant difference previously noted between control and cerebellar-OS groups in the average knee muscle torque remained. Here, knee muscle torque magnitudes from the cerebellar-OS group were still approximately double that produced by the controls navigating the taller obstacle (p=0.006). Average net torques were also significantly reduced in the cerebellarOS group compared to controls (p=0.014). Interaction torques were not different between the groups, but gravitational torques were (p=0.022), likely because of slight differences in limb configuration. Thus, even when accounting for the limb kinematics, trials where cerebellar subjects overshot still had excessive flexor muscle torque production during obstacle stepping. This was not true during obstacle walking. Therefore, a simple voluntary strategy alone might be responsible for the dynamic abnormalities noted during walking, but could not explain the dynamic abnormalities noted during stepping. Dynamics: hip and ankle Though the knee appeared most responsible for the leg hypermetria, we also assessed hip and ankle joint dynamics. Since the dynamic differences during walking appeared to be largely due to a voluntary strategy, we show here only the data for the hip and ankle joints during stepping. Figure 8 shows group data for the average torque magnitudes at the hip and ankle during the same deceleration phase of the stepping condition. At the hip (Fig. 8A), there was a significant group effect for all torque components (all p<0.05). The cerebellar-NE group produced torques similar to controls, but the cerebellar-OS group had many differences. Post hoc analysis showed the cerebellar-OS group had a greater extensor gravitational torque (p<0.001), a smaller extensor interaction torque (p=0.028), and a slightly larger flexor muscle torque (p=0.058). This resulted in a trend towards a slightly reduced extensor net torque to decelerate hip flexion (p=0.143). The smaller interaction torque was likely due to the smaller net torque at the knee (knee acceleration is the major contributor to the hip interaction torque). The larger gravitational torque was due to the difference in limb configuration (greater knee and hip flexion results in a larger hip gravitational torque). An appropriate response to the reduced extensor interaction torque would be to reduce flexor muscle torque. However, cerebellar subjects who were hypermetric instead produced flexor muscle torque that was slightly excessive. Despite these dynamic differences, cerebellar-OS group hip kinematics were not, on average, different than controls. This is not surprising, given that the hip net torque was not significantly different from controls. Unlike the knee and hip, the ankle motion during this phase was accelerating into dorsiflexion (e.g., the net torque was flexor). Ankle dynamics (Fig. 8B) tended to look more similar among the three groups. There were, Fig. 8 Mean hip (A) and ankle (B) torques from control, cerebellar-NE, and cerebellar-OS groups during the knee flexor deceleration phase of the obstacle stepping condition. Significant or trending toward significant post hoc statistics, comparing the control group to the cerebellar-OS group, are indicated at the bottom of each graph. Error bars represent �� SE 160 however, group differences for interaction and gravitational torques (p<0.01 and p<0.001, respectively). Post hoc analysis showed that, compared to controls, the cerebellar-OS group had a slightly larger extensor (plantarflexor) gravitational torque (p=0.022) and a smaller extensor interaction torque (p=0.007). Presumably, the smaller interaction torque is due to the smaller net torque at the knee (knee acceleration is a major contributor to the ankle interaction torque). The larger gravitational torque is related to the orientation of the foot at this time (slightly more dorsiflexed). Unlike the hip, the ankle muscle torque appeared to show an appropriate compensation for the reduced interaction torque by decreasing (though not significantly) the flexor torque production. Overall, the dynamic differences at the hip and ankle were less marked than those at the knee. Interaction torque differences at hip and ankle could largely be explained by the decreased knee net torque. Context specificity An important result from this study is that the extent of kinematic and dynamic abnormalities varied depending upon the context in which the leg motion was performed. Stepping over obstacles elicited hypermetric foot elevation to a greater extent than walking over obstacles, and appeared to result from a fundamental deficit in producing the appropriate torque pattern. Hypermetria during walking over obstacles appeared to be largely due to a voluntary safety strategy. Our previous work has shown that cerebellar subjects can over- or undercompensate for interaction torques during reaching, depending upon the demands of the task. During fast reaching movements, the deficit observed was that of undercompensation for interaction torques (Bastian et al. 1996). However, during arm motion requiring shoulder stabilization, cerebellar subjects overcompensated for interaction torques at the shoulder and produced excessive, unwanted shoulder flexion (Bastian et al. 2000). This pattern of overcompensation is similar to what we observed at the knee during our obstacle stepping task. The differences in performance by cerebellar subjects suggest some elementary differences in the role of the cerebellum during different types of tasks, a concept that is not new to the cerebellar literature. In the monkey, Thach et al. (1992) showed that focal inactivations to the three pairs of deep cerebellar nuclei produced three distinct patterns of motor deficits: fastigial inactivation produced primarily deficits of balance and locomotion, interpositus inactivation produced action tremor of the arm and impaired contact placing of the foot during locomotion, and dentate inactivation produced mainly deficits of voluntary, multijoint arm and hand movements. Thus, different cerebellar regions are more or less heavily recruited, depending upon the particular motor behavior required. Locomotor studies in cats have shown somewhat similar results. Neurons in the cat lateral cerebellum fire much more so during visually guided walking tasks (Armstrong and Marple-Horvat 1996; Armstrong et al. 1997; Marple-Horvat et al. 1998) or during perturbed walking (Schwartz et al. 1987) compared to uninterrupted walking over a level surface. Other studies have suggested that the intermediate region of the cerebellum may be particularly important for the control of limb dynamics and the relative timing of flexor and extensor muscle activity during certain tasks. Interpositus inactivation impairs intersegmental dynamics in the cat forepaw during reaching (Cooper et al. 2000). Extracellular recordings from the cat have indicated a specific role for the interpositus in facilitating certain portions of the gait cycle during treadmill locomotion (Armstrong and Edgley 1984). Likewise, decerebrate cats showed hyperflexed forelimb movements during swing phase of the step cycle after cooling of lobule V of the intermediate cerebellar cortex (Udo et al. 1980; cf. Thach et al. 1992). Their finding of hyperflexed forelimbs during swing resembles our findings of leg hypermetria and excessive knee flexion during walking over obstacles. Discussion This is the first report of cerebellar control of multijoint leg dynamics in humans and the first comparison of the extent of hypermetria in different movement contexts. Subjects with cerebellar damage often demonstrated leg hypermetria (an elevated foot height) when negotiating obstacles during walking and stepping. The hypermetria was primarily caused by excessive knee flexion and both the hypermetria and increased knee flexion were greater during stepping compared to walking. The duration of knee deceleration was also prolonged during stepping but not walking. The analysis of joint dynamics showed that cerebellar subjects produced knee muscle torques that overcompensated for the interaction and gravitational torques and disproportionately opposed deceleration, leading to excessive knee flexion and hypermetria. During the walking condition, the evidence indicates that the abnormal muscle torque was a voluntary strategy because control subjects produced very similar torque patterns when walking over a taller obstacle (a condition where peak foot heights and knee flexion angles were comparable between groups). During the stepping condition, cerebellar knee muscle torque patterns were more abnormal and unlikely to have been produced by a voluntary strategy alone, since they were not seen when control subjects stepped over a taller obstacle. There was a positive correlation between peak foot heights and average knee muscle torques in the cerebellar group during stepping, but no such correlation in the cerebellar group during walking or in the control group during either walking or stepping. This further supports the idea that the cerebellar subjects became hypermetric during stepping over obstacles because of an abnormal overproduction of knee flexor muscle torque. 161 Relating these results to our study, it is possible that walking over obstacles and stepping over obstacles invoke activity from different brain regions, and possibly different specific cerebellar regions. During locomotion, the CNS utilizes specialized circuitry in the brainstem and spinal cord dedicated to the control of walking. These areas are responsible for generating much of the basic locomotor pattern (Forssberg and Grillner 1973; Grillner 1975) and would likely have been involved in the obstacle walking condition. We hypothesize that the walking condition also involved more medial regions of the cerebellum. In contrast, during isolated voluntary limb movements such as the obstacle stepping condition, the locomotor pattern generators and the medial cerebellum were probably relatively less active, while the intermediate and lateral regions of the cerebellum would have been more heavily involved (Thach et al. 1992). Consequently, during the context of isolated voluntary limb movements, intermediate and/or lateral cerebellar regions appear to be required for the production of appropriate multijoint dynamics (Bastian et al. 1996, 2000; Topka et al. 1998b; Cooper et al. 2000). However, during the context of walking, intermediate and lateral regions of the cerebellum may be relatively less involved, while the medial cerebellum, brainstem, and spinal cord play a somewhat stronger role. In this case, the required circuitry for the generation of appropriate intersegmental dynamics controlling foot height might be at least partially contained within the brainstem and spinal cord. The primary role of the cerebellum during walking might therefore be less for the control of leg dynamics, but rather for the control upright posture and balance. In support of this hypothesis, we have previously shown that cerebellar damage produces locomotor abnormalities in humans that are closely related to deficits of balance but not to deficits of voluntary, visually guided leg movements (Morton and Bastian 2003). In our sample of cerebellar subjects, most had degenerative diseases that resulted in pancerebellar atrophy. Therefore, our data cannot help determine which specific cerebellar region(s) might have been most responsible for the hypermetria or the dynamic abnormalities. It is noteworthy that all of the subjects demonstrated hypermetria during walking over obstacles, but only some showed hypermetria during stepping over obstacles (see Table 1). This corresponds with our hypothesis that the hypermetria during walking over obstacles was more related to a voluntary safety strategy than an overt deficit in controlling knee dynamics. All cerebellar subjects were wary of hitting the obstacle and subsequently losing their balance during walking, so all showed hypermetria during this condition. But when balance demands (and the consequences of hitting the obstacle, e.g., falling) were minimized, as in the stepping condition, only some of the cerebellar subjects had clear deficits in controlling knee dynamics. In addition, we saw a relationship between the extent of limb ataxia (on clinical examination) and hypermetria during the obstacle stepping task, but no such relationship between limb ataxia and hypermetria during walking. Subjects who scored the worst on the heel-knee-shin test portion of the International Cooperative Ataxia Rating Scale (ICARS; Trouillas et al. 1997) were also hypermetric during the obstacle stepping task (CBL-1, -7, and -8). Likewise, of the four subjects who never showed hypermetria during stepping, three had no or only very minimal leg ataxia on the heel-knee-shin test (CBL-2, -3, and -6). On the other hand, hypermetria during walking appeared rather unrelated to the heel-kneeshin test of the ICARS, as subjects who had minimal or no leg ataxia according to the ICARS still showed frequent leg hypermetria when walking (CBL-2, -6). The ICARS scores thus seemed to be best related to hypermetria during stepping but not walking, further supporting the claim that hypermetria during walking reflected more of a voluntary strategy common to all subjects with cerebellar damage. Cerebellar control of multijoint leg dynamics Previous studies have shown considerable multijoint coordination deficits during voluntary arm movements in subjects with cerebellar damage (Goodkin et al. 1993; Bastian et al. 1996; Timmann et al. 2000; Goodkin and Thach 2003). Some of these deficits have been shown to be related to a specific inability to control interaction torques (Bastian et al. 1996, 2000; Topka et al. 1998b). In the present study, we showed that cerebellar hypermetria during voluntary leg stepping movements was primarily related to a failure to control excessive knee flexion. Knee joint dynamics were characterized by the inability to appropriately scale flexor muscle torque magnitudes (e.g., flexor muscle torques overcompensated for the interaction torque), leading to a prolongation of knee deceleration and resultant excessive knee flexion. However, we did not find considerable kinematic abnormalities at the other key joints of the leg that contributed to the hypermetria. Past studies in both humans and animals have shown that mechanical interactions between the knee and hip are exploited by the CNS to control leg movements. Active flexion at the knee is used to produce passive flexion at the hip during swing (Patla and Prentice 1995; McFadyen and Carnahan 1997; Eng et al. 1997; McFadyen et al. 1999). In addition, knee flexion can also produce a large dorsiflexor torque at the ankle that assists it in clearing the foot. In our study, excessive knee flexor muscle torque resulted in reduced knee extensor net torque. According to the mechanics described above, the reduced knee net torque should cause interaction torques at the hip and ankle to also be reduced. This was indeed the case (see Fig. 8A, B). Appropriate compensations for the reduced interaction torque magnitudes would have been to reduce flexor muscle torque production at both the hip and the ankle. At the hip, muscle torque was (inappropriately) actually slightly increased. At the ankle, however, compensation appeared to be taking place because the flexor muscle torque was slightly, though not significantly, reduced. Nevertheless, these moderate dynamic differences were 162 not enough to significantly alter the net torque magnitudes at either the hip or the ankle joint. Therefore, it is difficult to say that there were clear and meaningful deficits related to the prediction and control of interaction torques at all joints. Additionally, the knee joint hyperflexion did not appear to significantly affect ankle or hip joint kinematics. We report here moderate deficits in the control of foot height that probably reflect a compensatory strategy rather than a primary consequence of cerebellar damage when making an adjustment to the normal locomotor pattern. However, we should not conclude that the cerebellum is not required for intersegmental dynamic control of other aspects of the walking pattern. Our study examined only a portion of swing phase during a modification of the normal locomotor pattern (e.g., obstacle negotiation) in an attempt to find possible causes of leg hypermetria. Certainly other kinematic abnormalities exist during level and obstacle walking, such as irregular foot placement and abnormal stance limb joint-joint coordination patterns, which could in fact be caused by abnormal joint dynamics. These possibilities remain open and require further study. Conclusions We have shown that the amplitude of cerebellar leg hypermetria depends upon the context in which the leg movements are performed. The presence of abnormal joint dynamics also appears to be context dependent. During walking over obstacles, knee joint dynamics are moderately impaired and seem related to a voluntary safety strategy. For similar leg movements during stepping over obstacles, knee joint dynamics are clearly abnormal (and not related to a voluntary strategy alone), with a pattern of flexor muscle torque production that overcompensates for the interaction and gravitational torques. We suggest that during voluntary limb movements such as our stepping task and previously studied reaching movements, the intermediate and lateral regions of the cerebellum may be more important, and required for the production of appropriate multijoint dynamics. During walking, the medial cerebellum, brainstem, and spinal cord likely play a stronger role. In this context, the cerebellum may not be as necessary for the prediction and control of intersegmental dynamics controlling foot height, perhaps because the circuitry used for this during locomotion resides with the brainstem and/or spinal cord. Acknowledgements We would like to thank R. Bunoski for assistance with data collection, Drs. E. O��earn, S. Reich, R. Wityk, and D. Zee for patient referral, and Drs. P.S.G. Stein, W.T. Thach, and K. Thoroughman for thoughtful suggestions regarding this project. This study was supported by NIH grant HD040289 and the Foundation for Physical Therapy. References Armstrong DM, Edgley SA (1984) Discharges of nucleus interpositus neurons during locomotion in the cat. J Physiol 351:411432 Armstrong DM, Marple-Horvat DE (1996) Role of the cerebellum and motor cortex in the regulation of visually controlled locomotion. Can J Physiol Pharmacol 74:443455 Armstrong DM, Apps R, Marple-Horvat DE (1997) Aspects of cerebellar function in relation to locomotor movements. Prog Brain Res 114:401420 Bastian AJ, Martin TA, Keating JG, Thach WT (1996) Cerebellar ataxia: abnormal control of interaction torques across multiple joints. J Neurophysiol 76:492509 Bastian AJ, Zackowski KM, Thach WT (2000) Cerebellar ataxia: torque deficiency or torque mismatch between joints? J Neurophysiol 83:30193030 Botterell EH, Fulton JF (1938a) Functional localization in the cerebellum of primates II. Lesions of midline structures (vermis) and deep nuclei. J Comp Neurol 69:4762 Botterell EH, Fulton JF (1938b) Functional localization in the cerebellum of primates III. Lesions of hemispheres (neocerebellum). J Comp Neurol 69:6387 Brown TG (1911) The intrinsic factors in the act of progression in the mammal. Proc R Soc Lond 84:308319 Brown SH, Hefter H, Mertens M, Freund HJ (1990) Disturbances in human arm movement trajectory due to mild cerebellar dysfunction. J Neurol Neurosurg Psychiatry 53:306313 Chambers WW, Sprague JM (1955a) Functional localization in the cerebellum I. Organization in longitudinal cortico-nuclear zones and their contribution to the control of posture, both extrapyramidal and pyramidal. J Comp Neurol 103:105130 Chambers WW, Sprague JM (1955b) Functional localization in the cerebellum II. Somatotopic organization in cortex and nuclei. Arch Neurol Psych 74:653680 Cooper SE, Martin JH, Ghez C (2000) Effects of inactivation of the anterior interpositus nucleus on the kinematic and dynamic control of multijoint movement. J Neurophysiol 84:19882000 Corke PI (1996) A robotics toolbox for MATLAB. IEEE Robot Autom Mag 3:2432 Dichgans J, Diener HC (1985) Clinical evidence for functional compartmentalization of the cerebellum. In: Bloedel JR, Dichgans J, Precht W (eds) Cerebellar functions. SpringerVerlag, Berlin, pp 126147 Diener HC, Hore J, Ivry R, Dichgans J (1993) Cerebellar dysfunction of movement and perception. Can J Neurol Sci Suppl 3:S6269 Earhart GM, Bastian AJ (2001) Cerebellar gait ataxia: selection and coordination of human locomotor forms. J Neurophysiol 85:759769 Eng JJ, Winter DA, Patla AE (1997) Intralimb dynamics simplify reactive control strategies during locomotion. J Biomech 30:581588 Flament D, Hore J (1986) Movement and electromyographic disorders associated with cerebellar dysmetria. J Neurophysiol 55:12211233 Forssberg H, Grillner S (1973) The locomotion of the acute spinal cat injected with clonidine i.v. Brain Res 50:184186 Goodkin HP, Thach WT (2003) Cerebellar control of constrained and unconstrained movements. I. Nuclear inactivation. J Neurophysiol 89:884895 Goodkin HP, Keating JG, Martin TA, Thach WT (1993) Preserved simple and impaired compound movement after infarction in the territory of the superior cerebellar artery. Can J Neurol Sci (Suppl 3):S93S104 Grillner S (1975) Locomotion in vertebrates: central mechanisms and reflex interaction. Physiol Rev 55:247304 Hallett M, Massaquoi SG (1993) Physiologic studies of dysmetria in patients with cerebellar deficits. Can J Neurol Sci (Suppl 3): S83S92 Hollerbach MJ, Flash T (1982) Dynamic interactions between limb segments during planar arm movement. Biol Cybern 44:6777 163 Holmes G (1939) The cerebellum of man. Brain 62:130 Hore J, Wild B, Diener HC (1991) Cerebellar dysmetria at the elbow, wrist, and fingers. J Neurophysiol 65:563571 Hoy MG, Zernicke RF, Smith JL (1985) Contrasting roles of inertial and muscle moments at knee and ankle during paw-shake response. J Neurophysiol 54:12821294 Ito M (1984) The cerebellum and neural control. Raven Press, New York, pp 183188 Jansen J, Brodal A (1940) Experimental studies on the intrinsic fibers of the cerebellum. II. The corticonuclear projection. J Comp Neurol 73:267321 Marple-Horvat DE, Criado JM, Armstrong DM (1998) Neuronal activity in the lateral cerebellum of the cat related to visual stimuli at rest, visually guided step modification, and saccadic eye movements. J Physiol 506:489514 McFadyen BJ, Carnahan H (1997) Anticipatory locomotor adjustments for accommodating versus avoiding level changes in humans. Exp Brain Res 114:500506 McFadyen BJ, Lavoie S, Drew T (1999) Kinetic and energetic patterns for hindlimb obstacle avoidance during cat locomotion. Exp Brain Res 125:502510 Morton SM, Bastian AJ (2002) Does cerebellar damage impair control of interaction torques during walking over obstacles? Soc Neurosci Abstr 666.1 Morton SM, Bastian AJ (2003) Relative contributions of balance and voluntary leg coordination deficits to cerebellar gait ataxia. J Neurophysiol 89:18441856 Patla AE, Prentice SD (1995) The role of active forces and intersegmental dynamics in the control of limb trajectory over obstacles during locomotion in humans. Exp Brain Res 106:499504 Schwartz AB, Ebner TJ, Bloedel JR (1987) Responses of interposed and dentate neurons to perturbations of the locomotor cycle. Exp Brain Res 67:323338 Soechting JF, Lacquaniti F (1981) Invariant characteristics of a pointing movement in man. J Neurosci 1:710720 Stolze H, Klebe S, Petersen G, Raethjen J, Wenzelburger R, Witt K, Deuschl G (2002) Typical features of cerebellar ataxic gait. J Neurol Neurosurg Psychiatry 73:310312 Thach WT, Goodkin HP, Keating JG (1992) The cerebellum and the adaptive coordination of movement. Ann Rev Neurosci 15:403442 Timmann D, Horak FB (1998) Perturbed step initiation in cerebellar subjects. 1. Modifications of postural responses. Exp Brain Res 119:7384 Timmann D, Watts S, Hore J (2000) Causes of left-right ball inaccuracy in overarm throws made by cerebellar patients. Exp Brain Res 130:441452 Topka H, Konczak J, Dichgans J (1998a) Coordination of multijoint arm movements in cerebellar ataxia: analysis of hand and angular kinematics. Exp Brain Res 119:483492 Topka H, Konczak J, Schneider K, Boose A, Dichgans J (1998b) Multijoint arm movements in cerebellar ataxia: abnormal control of movement dynamics. Exp Brain Res 119:493503 Trouillas P, Takayanagi T, Hallett M, Currier RD, Subramony SH, Wessel K, Bryer A, Diener HC, Massaquoi S, Gomez CM, Coutinho P, Ben Hamida M, Campanella G, Filla A, Schut L, Timann D, Honnorat J, Nighoghossian N, Manyam B (1997) International cooperative ataxia rating scale for pharmacological assessment of the cerebellar syndrome. J Neurol Sci 145:205211 Udo M, Matsukawa K, Kamei H, Oda Y (1980) Cerebellar control of locomotion: effects of cooling cerebellar intermediate cortex in high decerebrate and awake walking cats. J Neurophysiol 44:119134 Winter DA (1990) Biomechanics and motor control of human movement, 2nd edn. Wiley-Interscience, New York, pp 5657 Yu J, Eidelberg E (1983) Recovery of locomotor function in cats after localized cerebellar lesions. Brain Res 273:121131