Brain Research 1988 - PDF document

Brain Res (1988) 73:648-658 Brain Research 1988 automatic postural responses: responses to horizontal perturbations of stance in multiple directions Moore, D.S. Rushmer, S.L. Windus, and L.M. Nashner Sciences Institute, Good Samaritan Hospital and Medical Center, 1120 NW 20th Ave., Portland, OR 97209, USA The effect of the direction of unexpected horizontal perturbations of stance on the organiza- tion requests to: Moore (address see above) that includes perturbation direction as a continuous variable. Key words: Unexpected postural Introduction Postural responses to unexpected perturbations of stance have been shown to be automatic and highly stereotyped in humans (i.e. Diener et al. 1984; Nashner 1977; Nashner et al. 1979) and cats (i.e., Rushmer et al. 1983, 1987). Responses to horizontal translations of the support surface in the anterior- posterior (A-P) direction involve activation of in the A-P direction, the ankle strategy is used to exert torque about the ankle and the hip strategy is used when torque about the ankle is insufficient to correct stance and the subject must depend upon hip generated shear force to restore balance. Postural responses composed of a single synergy can be observed, such as their "ankle synergy" or "hip synergy" however, Nashner and McCollum hypothesize that in practice, more com- plex patterns of muscle activation are formed by combining several synergies, for example when sub- jects adapt responses to changed support surface conditions (Horak and Nashner 1986). These responses are termed "mixtures". It is proposed that such mixtures are the combinations of more than one elemental synergy and that timing and latency varia- tions of the responses are due to segmental reciprocal delays, "a lower level segment' by segment interac- tion between individual muscle commands" (McCol- lum et al. p. 60, 1985). Such segmental mechanisms would prevent antagonist muscles acting on the same joint that participate in two different elemental synergies from coactivating. When animals or humans are exposed to unex- pected perturbations in either the anterior or poste- rior directions, motor responses that are appropriate for the perturbation direction are always selected (Rushmer et al. 1983; Moore et al. 1986). Thus, the organization of the postural response, as exemplified by the activated muscle groups, depends on the direction of the unexpected perturbation. However, past studies have only examined postural responses to horizontal perturbations in the sagittal plane. The question of how small changes in perturbation direc- tion affect the organization of postural movements has not been addressed using human subjects. In a previous paper, we have demonstrated that in the cat, which has only one strategy for response to perturbations in the A-P direction, organization of postural responses varies systematically as perturba- tion direction is changed (Rushmer et al. 1988). Does the organization of postural responses follow the same rules in humans, which have more complex strategies for postural responses? As perturbation direction is changed, will we see continuous varia- tions in response patterns or can the changes in responses be explained as different combinations of a few distinct synergies? To examine this problem, subjects were exposed to unexpected horizontal translations of the support surface while oriented at several different angles with respect to the platform motion. The results of the study demonstrate that amplitude and, in some cases, onset latency of each individual muscle's EMG activity vary as a continu- ous function of perturbation direction. A second finding is that the responses of proximo-axial muscles are influenced by perturbation direction differently than those of the distal leg muscles. Lastly, the results also suggest that postural response organiza- tion, i.e., the relationship between amplitudes and latencies of muscles active during the response, varies as a continuous function of perturbation direc- tion. Thus, if the idea of synergy as an elemental building block of activity is to be retained, the relations between muscles must be thought of as functions of several variables rather than fixed entities. Preliminary results of this study have been presented elsewhere (Rushmer et al. 1986; Moore and Rushmer 1987). and horizontal shear force were recorded from six normal, healthy subjects, between the ages of 21 and 33, as they stood on a moving hydraulically driven platform. The platform was controlled by an hydraulic servomotor and could be translated horizontally forward and backward. It con- sisted of two adjacent base plates, 20 cm by 42 cm. Strain guages mounted within each plate provided horizontal shear force meas- ures. For this study the platform moved 6 cm in 240 ms at an average velocity of 25 cm/s. To examine the effects of direction on human postural responses in the horizontal plane, it was necessary to perturb the subjects while they stood on the force platform at several different angles from the direction of the platform motion. This was achieved by having the subjects pivot on the platform at incre- ments of 15 ~ keeping their feet a constant distance apart (approxi- mately 6 in). Thus it was possible to present horizontal perturba- tions from 0 ~ to 360 o about the sagittal plane as shown in Fig. 1. Bipolar surface electrodes were used to detect muscle activity. The activity of up to 4 pairs of representative leg, thigh and hip muscles on the subjects' right side were analyzed. Three pairs of muscles were involved in responses to forward/backward horizon- tal translations and have been previously documented. They were: medial gastrocnemius (MG) and tibialis anterior (TA); biceps femoris of the hamstrings (HAM) and rectus femoris of the quadrieeps (QUAD); paraspinal at the iliac crest level (PARA) and rectus abdominus at the umbilicus level (ABDM). Because most of the perturbations used in this study contained a lateral component, it was necessary to also examine muscles that were active during hip abduction (tensor fascia latae, ABDC) and adduction (upper part of the hip adductors, ADDC), The myoelectric signals were amplified with cutoff frequencies of 70 and 2000 Hz, rectified and then low pass filtered (time constant 10 ms). When the electrodes were applied, the skin over the muscles was cleaned and electrodes were placed over the middle of each muscle belly, approximately 3 cm apart, center to center. A ground electrode was attached above the right lateral malleolus. During the experiment, subjects stood on the platform and were given several practice trials, in the backward and forward directions (0 ~ and 180~ to get accustomed to the platform motion. Subjects then performed 80 to 120 trials, which were presented in blocks of 10 trials, while oriented in 8 to 12 different directions in relation to the sagittal plane. Within each block of trials, there were 5 forward and 5 backward platform translations, randomly presented. The directions that the subject faced were also ran- domized. '~\F loaded sway " 90~ L" ~F ~ sway Schematic drawing showing the different angles the subjects faced in order to present horizontal perturbations from 0 ~ to 360 ~ about the sagittal plane. 0 ~ (anterior sway) and 180 ~ (posterior sway) perturbations resulted when subjects stood facing forward (F) and the support surface was translated backward (B) and F, respectively. Similarly, 90 ~ and 270 ~ perturbations resulted when subjects turned facing 90 ~ to the right and the support surface was translated B and F, respectively Data collected for all trials included the EMG signals, force measures and platform position. A total of 1 s of data was collected starting 150 ms prior to the onset of the perturbation. Signals were converted from analog to digital form, on-line, by a LSI-11/23 minicomputer, at a sampling rate of 500 Hz and stored for subsequent processing. The horizontal shear force was used to determine the onset of the perturbation. For each trial, onset latencies of muscle activity were determined by visual inspection and expressed with respect to the perturbation onset. Amplitude of the EMG response was determined by integrating the first 75 ms of muscle activity starting from onset of muscle activity (IEMG). Onset latency and IEMG measures were averaged for each direction, IEMG measures for each muscle were normalized by assigning the largest IEMG activity a value of 1 and expressing the IEMG values for all other directions as fractions of that maximum value. amplitude of each muscle's response varied systematically as the direction of the perturbation was changed. Figure 2 shows, for a single subject, averaged EMG responses of tibialis anterior for each of 16 different directions of platform motion. IEMG of the postural response was calculated for the 75 ms window defined by the vertical dashed lines. To demonstrate amplitude variation with perturbation direction, IEMG values were normalized and plotted on a polar plot (center, Fig. 2). This variation of IEMG with perturbation direction was defined as the muscle's "angular range of activation", a term which was first used by Buchanan et al. (1986) to describe the variation of elbow muscle EMG activity as a function of torque direction. Each muscle studied showed a unique angular range of activation for this set of horizontal transla- tions. Figure 3 shows polar plots of the angular range of activation for each muscle. The plots shown are averaged responses across 5 or 6 subjects, as indi- cated on the figure. The lines in the radial directions represent one standard deviation above the mean. Muscles could be divided into two groups based on their response characteristics. The first group showed small angular ranges of activation and their onset latencies remained relatively constant as per- turbation direction was altered. This group included medial gastrocnemius and tibialis anterior as well as quadriceps and will be referred to as "distal" mus- cles. The second group, the "proximo-axial" muscles, was comprised of the adductors, abductor, abdomi- rials and paraspinals. These muscles tended to show a larger angular range of activation and demonstrated bimodal IEMG distributions and/or variations in onset latency as perturbation direction changed. Unlike the other muscles, hamstrings showed consid- erable between subject variability. For some subjects hamstrings behaved like a distal muscle while for other subjects it behaved more like a proximo-axial muscle. Distal muscle responses. As shown in Fig. 3, both gastrocnemius and tibialis anterior had relatively narrow angular ranges of activation with very low between subject variability. Gastrocnemius was most active when the right leg participated in the correc- tion for anterior sway and when it was loaded as a result of lateral platform motion (270~176 As was observed in the cat (Rushmer et al. 1988), the angular range of activation was not oriented about the sagittal plane, as might be predicted if this muscle were primarily involved with responses to the A-P components of sway. The angular range of tibialis was more oriented about the sagittal plane (120~ ~ than that of gastrocnemius, although maximal activity was observed when the platform motion evoked posterior sway and loading of the right leg. The onset latencies for both muscles remained constant throughout the angular range of activity: the mean latency throughout the range for gastrocnemius was 101 + 8 ms and for tibialis was ~t " t t ," :' I l ~ . II ii , I DO ,~21o~ ~ lsoo/ . ,, : 100 msec 2. Polar plot representing the amplitude of the automatic postural EMG responses from tibialis anterior to horizontal perturbations of stance in 16 different directions, for one subject. EMG traces are the averages of 5 trials for each direction perturbation. Verti- cal arrows denotes onset of platform movement. The vertical dashed lines show the 75 ms window of integration. The polar plot is taken from the nor- malized amplitude values + 4 ms. Outside these directions, tibialis and gastrocnemius were relatively silent. The angular range of activation of quadriceps was unimodal, about the sagittal plane (120 ~ to 240 ~ and tended to overlap that of tibialis. The between subject variability of quadriceps activity level was relatively low. Quadriceps latencies varied little throughout the angular range of activity (mean latency 133 + 4 ms) and tended to lag those of tibialis by about 23 ms. muscle responses. angular ranges of activation of abdominals and paraspinals were bimodal, showing EMG activity in both the back- ward and forward directions (centered about both 0 ~ and 180~ The between subject variability observed for these two muscles was mainly due to peak activity occurring at slightly different directions for each individual subject. Peak activity for the paraspinals ranged between 0 ~ and 30 ~ in the anterior direction and between 180 ~ and 240 ~ in the posterior direction; the abdominal's peak activity occurred between 315 ~ and 0 ~ in the anterior direction and between 150 ~ and 180 ~ in the posterior direction. Despite these differ- ences between individuals, the shape of the angular ranges of activity were similar for all subjects. Onset latencies for responses of these two mus- cles varied considerably as a function of perturba- tion direction. Figure 4 shows amplitude and onset latency changes for paraspinals (Fig. 4A) and ab- dominals (Fig. 4B) plotted as a function of perturba- tion direction. For perturbation directions containing an anterior sway component (0 ~ abdominals activity began early (89 _+ 17 ms) and paraspinals became active later (160 + 29 ms). In contrast, for perturba- tions with a posterior sway component (180~ onset latency of abdominals was 184 _+ 32 ms and that of paraspinals was 100 + 20 ms. In both the paraspinals and abdominals, transitions of onset latency gener- ally occurred when platform motion was in the lateral direction (90 ~ and 270~ and activity in these muscles was relatively low. The increased between subject variability over these transition periods was possibly due to the difficulty in determining the onset latency when the activity level was low. The hip abductor showed a bimodal angular range of activation with low between subject variabil- ity. This muscle was most active for those perturba- tion directions which Ioaded the right leg and active to a lesser extent when the leg was unloaded (Fig. 3). The angular range was oriented about platform motion in the lateral directions. Hip adductors showed a broad angular range which extended from 30 ~ to 240 ~ , with maximal activity in perturbation directions which evoked posterior sway and unloaded the right leg. Onset latency variations with perturba- tion direction were also observed for these muscles (Fig. 5). Transitions in latency relationships occurred near the A-P directions; again there was an increase in between subject variability during the transition periods and when the activity level was low. Activity in adductors occurred early (range of 9%100 ms) for directions from 30 ~ to 90 ~ and late (range of 137-142 ms) from 210 ~ to 240 ~ . The abductor was active late (range of 139-151 ms) for directions from o ~ 3. Polar plots representing the average angular range of activation of the eight muscles examined in this study. Normalized IEMG (see text) is in the radial direction and perturbation direction is in the angular direction The solid lines in the radial direction indicate one standard deviation above the mean. The number of data sets used to calculate the mean is shown under the muscle name ~ to 150 ~ and early (range of 91-96 ms) for direc- tions from 240 ~ to 330 ~ . responses. mentioned above, ham- strings showed large between subject variability which appeared to be due to whether its response was similar to that of the distal group of muscles or to that of the proximo-axial ones. Averaged hamstrings activity showed a bimodal distribution centered about the sagittal plane. The greatest hamstrings activity was generally centered about platform motion in the anterior direction with lower activity -- 190 m E 160 ~ g 130 a. 100 70 190 E 160 ~ --~ 130 a Ill 100 70 45 90 135 180 225 270 315 0 (degrees) 0.6 "o 0.0 1,0 0.8 w 0,6 0.4 0.2 0.0 4. Normalized IEMG (dotted line) and onset latency (solid line) of paraspinals (A) and abdominals (B) plotted as a function of perturbation direction. Vertical lines represent _+ one standard deviation of the mean latency values when motion was in the posterior direction. The opposite result was obtained for one subject, where the greatest activity level was observed in the posterior direction. This accounts for the large stan- dard deviation of the averaged responses in the anterior and posterior directions. Hamstrings activity was lowest when platform motion was in the lateral directions (see Fig. 3). Two types of onset latency changes were observed. One type (N = 3) was similar to the distal muscle response, that is, as IEMG changed, there was little change in the onset latency of EMG activity. For example, one subject showed onset latencies of 111 _+ 13 ms at 0 ~ and 115 + 4 ms at 180 ~ For the other response (N = 2), onset latency changes were similar to those observed for proximo-axial muscle response. One of these subjects showed hamstrings response onset latency of 121 + 8 ms at 0 ~ and 83 + 8 ms at 180 ~ Apparently each subject grouped hamstrings either as a proximo- axial muscle or as a distal muscle and for the perturbations used in this study, this strategy did not change from one trial to the next. - 155 E 135 c o 115 0 '~ 95 155 r E 135 --~ 115 r D o 95 D 0.6 c~ O "13 0.4 ,0.0 "LO 0 45 90 135 t80 225 270 315 0 angle 0.6 ~3 O 0.0 5. Normalized IEMG (dotted line) and onset latency (solid line) of hip adductors (A) and hip abductors (B) plotted as a function of perturbation direction. Vertical line represent + one standard deviation of the mean latency values response organization. To show the relation- ships between muscles during postural responses, EMG responses were plotted as "musical scores", for each the directions examined. These are illustrated in Fig. 6. The magnitude of the muscle activity is indicated by the height of the triangles. Averaged onset latency is shown at the left edge of the triangle's base. The width of each triangle is arbitrar- ily set at 75 ms to denote the time during which the EMG was integrated. Activity of the muscles for each direction is shown across the figure. When presented in this manner, there appears to be a unique response organization for each of the direc- tions since the amplitude and timing relationships between the muscles and/or the muscle groups involved in each response are different for each of the perturbation directions. The EMG responses seen at 0 ~ were a distinct initial abdominal burst followed by gastrocnemius, hamstrings and paraspinals which were activated in a distal to proximal temporal order. The temporal relationships between muscles involved in the post- ural responses for perturbation directions from 300 ~ to 0 ~ were similar. However, although the temporal relationships between the muscles remained con- stant, their response amplitudes did not change proportionally. For example, while gastrocnemius was most active at 300 ~ and 315 ~ , hamstrings was just becoming active and was most active at 0 ~ Also, gastrocnemius dropped out of the response at 30 ~ while both hamstrings and paraspinals were still quite active. As the perturbation direction changed from 30 ~ to 120 ~ there was a gradual decrease in the activity of both hamstrings and paraspinals and there was little activity in the distal muscles. Abdominals also showed a decreased response amplitude with an increasing onset latency. Adductors activity level increased while the amplitude of the longer latency abductor responses decreased. For perturbation directions between 120 ~ and 225 ~ the ventral muscles were involved in the postural response. Tibialis anterior, quadriceps and abdominals responded in a distal to proximal tem- poral order which remained relatively constant for these perturbation directions. The relative response amplitudes varied considerably as direction was changed and did not change proportionally. At 180 ~ , paraspinal became involved early in the response and hamstrings and quadriceps were coactivated. Abduc- tor and adductors were also relatively active and appeared to cocontract in response to this perturba- tion, perhaps to stabilize the hip joint. These results suggest that the 180 ~ perturbation was more de- stabilizing than the translation at 0 ~ From 225 ~ to 300 ~ the posterior sway postural response component decreased, i.e. tibialis, quadri- ceps, abdominals and paraspinals activity levels decreased, and a more lateral component to the response became apparent. The abductor showed a gradual increase in amplitude and an earlier onset latency while the adductor response decreased and displayed a later onset latency, although activity in these muscles overlapped in time with a latency difference of only 30 ms. At 270 ~ the first muscle activated was the abductor. Interestingly, it was in this direction that the gastrocnemius response reap- peared and a temporal switch from an early to a later onset latency occurred in the abdominals. If a small number of discrete synergies are utilized by the nervous system in response to postural perturbations, the muscles involved should show tightly coupled amplitude and onset latency relation- ships as direction of the perturbation is changed. To test this hypothesis, linear regression analysis was performed on the EMG responses for each pair of muscles across the set of perturbation directions. The o ~ ~ ~ 90 ~ 120 ~ ~ ~ 180 ~ 210 ~ 2250 2400 270 ~ 300 ~ 315 ~ 330 ~ I .~Ir i I~ I : A. v  I .v - I QUAD . A I A. I : A. 200 msec PARA . ,~llll~ . . ..,N~ V v" ABDC A . : . A. .v I v I I qBer : 6. "Musical scores" showing the response relationships of the eight muscles examined in the present study for each direction of platform movement. Height of the triangles is the normalized IEMG response. Onset latency is shown at the left edge of the triangle's base. The width of each triangle is arbitrarily set at 75 ms to denote the time during which the EMG was integrated were generally low, which would be expected given Fig. 6. Several significant, positive correlations (P 0.05) were found for response amplitudes: tibialis on quadriceps (r -- 0.958), adductors on quadriceps (r = 0.805) and addSctors on tibialis (r -- 0.672). However, the onset latency correlations were insignificant for these muscle pairs (r = 0.103; r =-0.247 and r =-0.055, respec- tively). A significant correlation was found between the amplitude response of hamstrings and paraspinals (r = 0.631). As previously stated, two types of onset latency changes were found for hamstrings. Thus correlations were performed for each individual subject. For the subjects who showed little change in onset latency, no significant positive correlations were found. For the subjects who showed a variable hamstrings onset latency, significant correlations of r = 0.853 and 0.699 were found. study, which examined the effects of small changes in perturbation direction on human postural responses, revealed three significant findings. First, EMG activity, and in some cases, onset latency of individual muscle responses varied continuously as a function of perturbation direction. A similar result was obtained in an analogous experiment with cats (Rushmer et al. 1988). Secondly, a dramatic differ- ence in response characteristics between the distal and proximo-axial muscles was observed. We believe that this difference was due to the role that each group of muscles played in the control of posture as well as to biomechanical contraints. Lastly, as in the cat experiments, the results suggest that for each perturbation direction there was a unique postural response. While there were indications of constant temporal relationships between muscles involved in responses around the sagittal plane, the amplitude relationships varied continuously with perturbation direction. Thus it appears that perturbation direction is a key sensory variable in determining response organization and that the response organization may depend upon more factors than the summation of a few discrete synergies. of individual muscles - distal~proximo- axial differences. activity in each of the muscles studied varied in a systematic and continuous manner with perturbation direction. Such variation was observed in both the amplitude of the EMG bursts and, for some muscles, in the onset latencies for EMG activity. In general, the angular ranges of activation of the most distal muscles, medial gastroc- nemius and tibialis anterior, as well as hip adductor were not aligned about either the A-P or the lateral axes, while the other muscles examined in the present study were aligned about these two axes. Alignment on the A-P or lateral axes is a difference from similar studies in the cat, in which none of the muscles studied were aligned about these axes (Rush- met et al. 1988). An explanation for this discrepancy may be the methods of recording. In the cat indi- vidual muscles were identified anatomically and studied with indwelling EMG electrodes sewn into the muscles. The surface electrodes used in the present studies most likely provided summed EMG activity from more than one muscle within a muscle group (i.e., hamstrings and quadriceps) except perhaps for the records from tibialis anterior and medial gastrocnemius, which could be isolated with more certainty. Our data suggests that the operational rules for the action of proximo-axial muscles are different from those of the distal muscles. First, the angular ranges of the distal muscles were relatively narrow (180 ~ while the angular ranges of the proximo- axial muscles were broader and in many cases, had bimodal distributions. Second, both the agonist and antagonist of a proximo-axial muscle pair tended to be involved in the postural response. Lastly, when activated, the onset latencies of the distal muscles were relatively constant while the onset latencies of proximo-axial muscles varied dramatically as pertur- bation direction was altered. The differences in the response characteristics between the proximo-axial and distal muscles appear to be due to the different roles these muscles play in the control of posture. According to Smith and Zernicke (1987), a muscle can function to either control the limb dynamics (i.e., to produce move- ment about the joint) or to counterbalance interac- tive torques which are developed from mechanical interactions between limb segments. Thus, during the control of posture, the distal muscles may only function to produce movement about the ankles, and would not be subjected to interactive torques. The work of Buchanan et al. (1986) supports this conclu- sion, since none of the muscles that they studied showed any sign of playing the role of a stabilizer. The angular ranges of activation for the distal mus- cles in our experiments were similar to those reported by them for isometric elbow movement, a task where the muscles also did not have to deal with intersegmental torques. The proximo-axial muscles may serve both to produce movement and to counterbalance interactive torques. When hip movement was part of the pos- tural strategy, the burst of the first proximo-axial muscle controlled the limb dynamics by producing movement about the hip to bring the center of mass over the base of support. For example, from experi- menter observation it appeared that for some pertur- bations, particularly those involving lateral sway, the postural response involved movement about the hip. In this case, the right hip abductor responded first when the lateral perturbation loaded the right leg (270~ There were postural responses where no hip movement was observed but a proximo-axial muscle still responded first. Such a response was at 0 ~ when the abdominalis were the first muscles activated. Diener et al. (1988) also report a similar abdominal burst for this stimulus direction and size. Their kinematic data indicated that this postural response does not involve hip movement. We conclude that the initial burst of the abdominals is part of the postural response and functions to counterbalance the inertial loads that are generated as a result of the response. The present results suggest that response charac- teristics of hamstrings and possibly quadriceps may depend upon how destabilizing the perturbation is to the subject. When the perturbation is very destabiliz- ing, it may be necessary to activate these muscles early (i.e., hamstrings for perturbations about 180 ~ and quadriceps for perturbations about 0~ For example, an early response of hamstrings was observed for perturbations about 180 ~ A similar response was not seen in quadriceps for perturba- tions about 0 ~ since a 25 cm/s translation is much more destabilizing in the 180 ~ direction due to the biomechanics of the human foot. It is possible that if a more destabilizing perturbation was used, the response of both these muscles would be similar to the proximo-axial muscles while a less destabilizing perturbation would result in responses similar to the distal muscles. Referring back to Fig. 6, the initial burst of the proximo-axial muscle is always followed by a later burst of its antagonist muscle. According to McCol- lum et al. (1985), this late proximal muscle burst occurs after distal muscle activity and acts to brake the hip torque which follows the ankle torque. This braking function is similar to the role of the antago- nist burst in arm movement (i.e., Marsden et al. 1983; Wierzbicka et al. 1986). During step-tracking wrist movements in different directions (Hoffman and Strick 1986) and in the proximo-axial muscles in our own experiments, as the direction of movement changed, the muscles changed their role from agonist to antagonists through relative changes in onset latencies. Differences in action of distal and proximo-axial muscles have been described in other movement paradigms as well. Hoy and Zernicke (1985, 1986) reported distal/proximal differences for locomotion and the paw shake in the cat. During the paw shake (Hoy and Zernicke 1986), muscle moments at the ankle generated paw acceleration while muscle moments at the hip acted to counterbalance move- ments due to acceleration of the more distal segments as well as to maintain the hindlimb postural orienta- tion. Although these results were for on oscillatory movement, it is possible that similar mechanics are involved in the control of posture. Muscle response relationships. A goal of the present study was to test the hypothesis of Nashner and McCollum that automatic postural responses are organized by mixing a small number of elemental synergies. We theorized that support of their hypo- thesis might be indicated by relatively discrete changes in the organization of postural responses as the perturbation direction was systematically varied and boundaries between synergic actions were crossed. The continuous variation of response organization with perturbation direction observed in the present study does not support or refute the hypothesis that the automatic postural responses are simple mixtures of a few synergies. The data do show that the postural response organization is determined by sensory cues which include directionality. As long as perturbation direction is held constant, response organization is discrete. However, as perturbation direction is varied, response organization is domi- nated by continuous aspects. The results show that the response organization in directions near the sagittal plane did show rela- tively constant onset latency relationships between muscles that participated in the postural responses. The temporal relationships between muscles involved in Nashner and McCollum's ankle synergies are relatively constant for perturbation directions on or near the sagittal plane. From 300 ~ to 15 ~ abdomi- rials always responded first followed by medial gas- trocnemius, hamstrings and paraspinals. Similarly, latency relationships between tibialis anterior, quad- riceps, paraspinals and abdominals were relatively constant for perturbation directions from 165 ~ and 225 ~ . However, onset latency relationships between these muscles varied continuously for other perturba- tion directions. In addition, the response amplitude relationships between the same muscles varied con- tinuously over the entire range of perturbation direc- tions. While the latency relationships provide some evidence for discreteness in the postural responses, sharp transitions in latency and/or amplitude were not observed. The present data are not totally consistent with previous work. First, the postural response we observed at 0 ~ was similar to the ankle response observed by Horak and Nashner (1986). However, unlike the response described by these workers, a distinct abdominal burst was seen prior to the onset of activity in gastrocnemius, hamstrings and paraspi- nals. Others have also reported this initial abdominal burst (Diener et al. 1988). Second, assuming an average burst duration of 75 ms (Diener et al. 1988), we observed coactivation of agonist-antagonist pairs, both in the individual and in the averaged data, when both the agonist and antagonist were involved in the postural response. This coactivation is evident in Fig. 6 between hamstrings and quadriceps, and between abductors and adductors. Abdominals and paraspinals showed little coactivity, due to their large variation in onset latency. Thus, while segmental modulation of postural responses most certainly does occur, mechanisms such as the reciprocal delay mechanisms postulated by McCollum et al. (1985) do not seem to act to prevent cocontraction of agonist- antagonist pairs during postural responses. There are several reasons which may account for these dis- crepancies. The perturbation used by Horak and Nashner (1986) was 13 cm/s, which is slower than the 25 cm/s translation used in the present experiment. As well, past work has concentrated on anterior sway, with little emphasis on posterior sway. When cocontraction was observed in the present study, there was usually a posterior sway component to the perturbation. The muscle response organization appears to be dependent on similarities of biomechanical function of the muscles and highly dependent on the task. For example, during postural corrections, the amplitude of tibialis anterior and quadriceps covaried as direc- changed while those of medial gastrocnemius and hamstrings did not. Response coupling between these muscles does occur under different mechanical conditions. When subjects were required to produce an isometric force at the foot in directions about the horizontal plane, the activity of gastrocnemius and biceps femoris appeared to be similar (Wells and Evans Stuber 1986). Similarly, in the different biomechanical conditions of quadrupedal stance, the amplitudes of the cat's lateral gastrocnemius and vastus lateralis responses covary as a tightly coupled unit during horizontal perturbations of stance (Rush- mer et al. 1988). The present study does have limitations. First, six of the eight muscles examined were involved in flexion or extension and participated primarily in postural corrections in the sagittal plane. Largely because of difficulties in accessing involved muscles with surface electrodes, we only sampled from one abductor and one adductor of the hip and did not sample enough muscles that were primarily respon- sive to perturbations in the lateral direction to describe a lateral synergy. However, given the pres- ent findings, we would predict that the response amplitude and latencies of the additional muscles involved in a possible lateral synergy would also vary continuously with respect to changes in perturbation direction. Secondly, it is also possible that we did not examine enough muscles to be able to detect the contribution of synergies to the responses. For in- stance, if the subjects were mixing eight synergies, four for the A-P direction (hip and ankle, forward and backward) and four for the lateral direction, records from eight muscles would not allow delinea- tion of responses related to any single program. However, we believe that we can discount this explanation for the present results. Continuous varia- tion of muscle response organization with perturba- tion direction was also observed in cats (Rushmer et al. 1988) which, because of their biomechanics, have only two synergies in the A-P direction and probably a similar number in the lateral direction. Eight muscles were also studied in these experiments, enough to identify contributions from individual synergies, yet the results were similar to those reported here. While our data do not support a specific theoreti- cal mechanism for the production of automatic postural responses, they do indicate that, hke initial perturbation velocity (Diener et al. 1988), perturba- tion direction is a key variable for the determination of response characteristics. The response organiza- tion is always appropriate for a specific perturbation direction and is largely unaffected by any previously presented disturbances. Thus, perturbation direction may be processed differently than support surface size since the latter variable's affect on postural responses is strongly influenced by set and previous experience and is subject to adaptation over several trials. Acknowledgements. by NIH grants R01-NS 12661 and R01-NS 19484. S.P. Moore was funded in part by a post-doctoral fellowship from the Natural Sciences and Engineering Research Council of Canada. TS, Almdale DPJ, Lewis JL, Rymer WZ (1986) Characteristics of synergic relations during isometric contrac- tions of human elbow muscles. J Neurophysiol 56:1225-1241 Diener HC, Dichgans J, Bootz F, Bacher M (1984) Early stabilization of human posture after a sudden disturbance: influence of rate and amplitude of displacement. Exp Brain Res 56:126-134 Diener HC, Horak FB, Nashner LM (1988) Influence of stimulus parameters on human postural responses. J Neurophysiol 59: 1888-1905 Hoffman DS, Strick PL (1986) Activity of wrist muscles during step-tracking movements in different directions. Brain Res 367:287-291 Horak FB, Nashner LM (1986) Central programming of postural movements: adaptations to altered support-surface configura- tions. J Neurophysiol 55:1369-1381 Hoy MG, Zernicke RG (1985) Modulation of limb dynamics in the swing phase of locomotion. J Biomech 18:49-60 Hoy MG, Zernicke RG (1986) The role of intersegmental dynamics during rapid limb oscillations. J Biomech 19: 867-877 Marsden CD, Obeso JA, Rothwell JC (1983) The function of the antagonist muscle during fast limb movements in man. J Physiol (Lond) 335:1-13 McCollum G, Horak FB, Nashner LM (1985) Parsimony in neural calculations for postural movements. In: Bloedal J, Dichgans J, Pratt W (eds) Cerebellar functions. Springer, Berlin Heidelberg New York, pp 52-66 Moore SP, Rushmer DS (1987) Automatic postural responses of humans to horizontal perturbations of stance in different directions. North Am Soc Psychol Sport Phys Activity, Vancouver, BC Moore SP, Horak FB, Nashner LM (1986) Influence of stimulus anticipation on human postural responses. North Am Soc Psychol Sport Phys Activity, Scottsdale, Arizona Nashner LM (1977) Fixed patterns of rapid postural responses among leg muscles during stance. Exp Brain Res 30:13-24 Nashner LM, McCollum G (1985) The organization of human postural movements: a formal basis and experimental syn- thesis. Behav Brain Sci 8:135-150 Nashner LM, WooUacott M, Tuma G (1979) Organization of rapid responses to postural and locomotor-like perturbations of standing man. Exp Brain Res 36:463-476 Rushmer DS, Russell CJ, Macpherson J, Phillips JO, Dunbar DC (1983) Automatic postural responses in the cat: responses to headward and tailward translation. Exp Brain Res 50:45-61 Rushmer DS, Windus SL, Russell CJ (1986) Automatic postural responses in the hindlimb of the cat to horizontal perturba- tions of stance in different direction: variations in synergic organization. Symposium on neural control of limb move- ment. University of Washington, Seattle, pp A4 Rushmer DS, Macpherson JM, Dunbar DC, Russell CJ, Windus SL (1987) Automatic postural responses in the cat: responses of proximal and distal hindlimb muscles to drop of support from a single hind- or forelimb. Exp Brain Res 65:52%537 Rushmer DS, Moore SP, Windus SL, Russell CJ (1988) Automatic postural responses in the cat: responses of hindlimb muscles to horizontal perturbations of stance in multiple directions. Exp Brain Res 71:93-102 Smith JL, Zernicke RF (1987) Predictions for neural control based on limb dynamics. Trends Neurosci 10:123-128 Wells R, Evans Stuber N (1986) Electromyographic responses to the lower limb musculature in simulated postural and locomotor activities. North Am Congr Biomech, Montreal Wierzbicka MM, Wiegner AW, Shahani BT (1986) Role of agonist and antagonist muscles in fast arm movements in man. Exp Brain Res 63:331-340 Received August 10, 1987 / Accepted June 29, 1988