OF NE UROPHYSIOLOGY Vol. 39, No. 5, September 1976. Printed in U.S.A.

JOURNAL

Regulatory Actions of Human Stretch Reflex P. E. CRAGO, Department Baltimore, SUMMARY

J. C. HOUK,

AND

Z.

of Physiology, The Johns Maryland 21205 AND

HASAN

Hopkins

CONCLUSIONS

1. The stretch reflex in the elbow flexor musculature was studied in 23 human subjects. The subjects were required to establish an initial force equivalent to 10% maximum at a prescribed initial length; mechanical disturbances delivered at random times increased load force to 15% or reduced it to 5%. We measured arm force, displacement, and EMG (usually biceps) ; acceleration was calculated from displacement, and average respones from sets of 10 like trials. 2. Modification of the stretch reflex was studied by comparing average responses obtained with different instructions, but with the same disturbance. The usual introductions were “compensate for arm deflection” and “do not intervene voluntarily.” The initial response did not depend on instruction; changes in response that depended on instruction began abruptly after a latent-period which ranged from 70 to 320 ms (measured from force and acceleration), depending on conditions and subject. The latency became longer (lo-50 ms) and more variable when the subject did not know the direction of disturbance in advance. This and other observations indicate that modifications of the stretch reflex are not produced by servo actions. They are produced by triggered reactions, which occur at both short and long latencies and which have properties resembling the movements produced in a reaction-time task. 3. We confirmed that most subjects can suppress triggered reactions when the instruction calls for no intervention, leaving an unmodified reflex response. This response consists of a compliant deflection of the arm in the direction of the disturbance. 4. The compensatory actions associated with unmodified stretch (and unloading) reflexes were assessed from EMG responses of biceps. During a 300-ms transient phase, EMG changes were notably asymmetric when responses to symmetric disturbances were compared. Increased force, stretched biceps and produced a prominent increase in EM G, ‘whereas decreased force allowed biceps to shorten and produced Received for publication

September 3, 1975.

University

School

of Medicine,

either an EMG decrease of smaller magnitude or an actual increase. These asymmetric reflex actions produced quite symmetric mechanical responses (arm displacements and forces), which implies the existence of and compensation for nonlinear muscle mechanical properties. This result is discussed in relation to the hypothesis that the function of the stretch reflex is to compensate for variations in muscle properties, thus maintaining stiffness. 5. Effective control of muscle length or joint position does not result from servo action by the stretch reflex. Errors in position are corrected only when triggered reactions are superimposed on the reflex response. INTRODUCTION

The function of the stretch reflex in man is not well understood. It is usually assumed that reflex action should compensate for changes in the external mechanical load (cf. ref. 25, 36). This assumption is based on the hypothesis that muscle length, by way of feedback from spindle receptors, is the regulated property of the stretch reflex; a rigid regulation of length would prevent load changes from affecting movements or postures (27). However, several authors studying human subjects and other’primates (1, 10, 14,24, 26, 29, 32) have reported that length is not well regulated and that load compensation is poor, unless long-latency responses that are presumed to reflect supraspinal mechanisms are superimposed on the segmental stretch reflex. It has been suggested that these long-latency responses are mediated by a pathway from spindle afferents to the motor cortex, either by way of the cerebellum or more directly (10, 28, 31, 35, 37). Phillips (35) proposed that this pathway might constitute the afferent limb of a transcortical servo loop, an extension of the segmental stretch reflex, and other authors have supported this suggestion (10, 24, 30, 37, 41). However, this does not fit with the results obtained by Vallbo (39) who assessed the overall efficacy of muscle-length regulation from the relationship between spindle discharge rate and the force of voluntary contraction in man. A calculation based on quite conservative assumptions indi925

926

CRAGO, HOUK,

cated that feedback from spindle receptors, regardless of the central pathway, cannot result in a high stiffness. Vallbo suggested from this that muscle length is not well regulated by any servo system that uses muscle spindles as afferents of origin. Recently an alternative to the load-compensation hypothesis has been proposed (20). According to this theory the function of the stretch reflex is to compensate for variations in muscle mechanical properties (internal disturbances), rather than for changes in mechanical load (external disturbances). Experimental tests in decerebrate cats supported the hypothesis and showed further that the regulated property probably is not muscle length, but is a relationship between muscle force and length, the stiffness presented to external loads (33). One purpose of the present experiments was to test this theory in human subjects. The mechanical stiffness of skeletal muscle is known to depend on the direction of length change (22, 33); we therefore sought to determine whether or not reflex action, as measured by EMG (electromyographic) change, is appropriately asymmetric to compensate for this nonlinearity (8). We also wanted to evaluate the extent to which the responsiveness of the stretch reflex, as measured by stiffness, can be altered by supraspinal control mechanisms. The hypothesis that stiffness is a regulated property predicts that controlled changes in the gain of length and force feedback should alter the particular stiffness, which is then maintained by reflex action (33). Several authors have described differences in stretch responses which result when the instructions given to the subject are changed (3, 11, 16, 32). These differences occur at latencies which are shorter than the reaction times observed in response to taps applied to various places on the limb (16, 26); on this basis they have been attributed to controlled changes in the responsiveness, or gain, of the stretch reflex (16). Other authors have stressed the stereotypic nature of the mechanical responses attributable to the stretch reflex in man (4, 12, 13). Asatryan and Feldman (4) used the instruction “do not intervene voluntarily” to obtain what they considered to be reflex responses unmodified by secondary movements. They reported that responses to unloading were essentially invariant with this instruction, except for conditions resulting in a cocontraction of elbow antagonists, although electromyographic evidence for this was not provided. These authors did not study modifications of the type mentioned earlier, those which occur at latencies shorter than usual reaction times. The present experiments were performed on

AND HASAN the elbow flexor musculature in human subjects. In this sytem the long-latency response that has been attributed to a supraspinal mechanism has a minimal latency of 70 ms in force recordings (16). This is also the response component which is modified when the instructions are changed from “resist” to “let go” (16). The shortest reaction times for force development in response to taps on the limb were 90 ms in Hammond’s (16) study, and the mean latency was in excess of 100 ms. Our first objective was to determine whether modification of the “long-latency” response results from an actual change in reflex responsiveness or, alternatively, from the appearance of superimposed movements. Our results support the latter interpretation and since these delayed movements are events of abrupt onset, we will refer to them as triggered reactions. A second objective was to test the suggestion (4) that the instruction “do not intervene” reveals a stereotypic reflex mechanism. With some reservation we have been able to confirm this, and we will reserve the term “stretch reflex” to refer to this response. A third and major objective was to test the idea that the function of the stretch reflex is to compensate for variations in muscle properties. Our results provide evidence which favors this hypothesis. A fourth objective was to determine the extent to which human subjects are able to control the stiffness of the stretch reflex. In confirmation of work already cited (4) we found no changes in stiffness that were not accounted for by the mechanical effects of cocontraction or by superimposed movements. METHODS

We used the apparatus shown in Fig. 1 to study responses to mechanical disturbances that were applied to the arms of 23 adult male and female subjects. The subjects held a light-weight (250 g) handle which had strain gauges installed for measuring tension in the cable. The cable attached by way of a pulley (radius, 1.9 cm) to the shaft of a torque motor (Inland T-4036-B). The current delivered to the motor was controlled to produce an initial tension in the cable of 30-50 N, which represented about 10% of the maximal force that the subjects were capable of resisting. The two directions of mechanical disturbance were produced by equal but opposite changes in motor current, which amounted to 50-70% of initial current in different subjects; thus, the arm was never completely unloaded. The establishment of an initial force was important for two reasons. First, it provided a means of control over the initial excitability of motoneuronal pools, which seemed necessary to

HUMAN

STRETCH

REFLEX

921

motor

und

FIG. 1. Apparatus used to study responses to load change. The mechanical linkage between the subject and the motor was accomplished by having the subject grasp the handle which was attached to a pully on the motor shaft by way of a steel cable. The shoulder brace was included to minimize body movements. A computer was programmed to control the motor current (loadforce) and instruction lights and to samplethe following variables:

EMG from surfaceelectrodesover the biceus, arm force from a transducerin the handle. arm deflection from a potentiometer in the motor housing, motor current from the servo amplifier. distinguish controlled changes in reflex respon- diameter) separated by approximately 3 cm over siveness from other phenomena. Second, it fur- the belly of the muscle, with an indifferent elecnished a moderate level of tonic reflex activity trode placed laterally. In some cases we also that was suitable for studying and comparing recorded from brachioradialis or triceps. A speresponses to both loading and unloading. cially designed preamplifier was taped to the arm Our mechanical arrangement facilitated test- in a position which avoided movement artifacts. ing of the elbow flexor musculature in a comfort- EMG processing before sampling included recable and natural situation. The shoulder was in tification and filtering (100 Hz cutoff) to prevent contact with a brace which minimized body abasing. movement. Subjects were asked to keep the The experiment was conducted under compuhandle in line between the shoulder and the pul- ter control (PDP 1l/20). A ready light indicated ley and to use their wrist like a hinge. This ar- to the subject when it was time to begin a trial; rangement required the elbow flexor muscles to other lights defined an initial position of the arm support the full load produced by the motor, and which the subject achieved by either flexing or forearm supination favored biceps activation (5). extending the elbow. The angle of the elbow was Shoulder muscles were active in opposition to approximately 90” at the required initial position. gravity and in joint stabilization, but could not The linear displacement of the arm toward and relieve the load supported by elbow flexors. away from the shoulder was proportional to the (The resultant of shoulder moment was a force rotation of the motor shaft, which was measured perpendicular to the cable.) with a potentiometer. The disturbance was deliThe load mimicked both inertial and gravita- vered after the arm had remained within a small tional components of natural loads. The gravita- range (kO.16 cm) of the required initial position tional component was a force proportional to for a random ti,me (l-6 s). Sets of 20 trials were conducted in rapid succession. A rest period motor current; we will refer to this controlled variable as load force to distinguish it from arm was allowed between sets and subjects were alforce, which was monitored as cable tension. lowed to rest during a given set if they became (The mass of the handle was small in comparison fatigued, which seldom occurred. Records of load force, arm force, displacewith the equivalent mass of the arm, which was about I kg.) The inertial component of external ment, and EMG were sampled at 3- or 5-ms load, derived from the moment of inertia of the intervals with 12-bit resolution. Ensemble avermotor, was equivalent to a mass of 2.8 kg. If a ages and standard deviations were calculated load having this mass were lifted vertically, the after sorting into sets of like trials, typically gravitational force would be 27 N, which is ap- 10. Average EMG records were subsequently smoothed with a digital low-pass filter; the 3@ms proximately the initial force in our experiments. The presence of inertia prevented excessive ac- time constant was chosen to reflect the low-pass celerations and velocities, which tend to syn- characteristics of muscle (15, 16, 34). Accelerachronize the EMG into peaks and troughs (cf. tion traces were calculated by double numerical Fig. 9 of ref 33). Peak acceleration was about differentiation of displacement records. Prior to this operation, displacement samples were 6m/sz and peak velocity was about 0.2 m/s. The EMG activity of the biceps muscle was scaled up by 16 and filtered with a 3-point sliding measured by fastening two silver discs (8 mm in window to reduce quantization noise.

CRAGO, HOUK,

928

AND HASAN

RESULTS

We have compared the responses to load change obtained with several different instructions on different occasions, but the two which we used most frequently were “compensate for the deflection of your arm” and “do not intervene voluntarily to compensate for arm deflection.” In the first two sections which follow, we describe

differences

in the responses

Load

Force

c

Displacement

that de-

pend on prior instructions and make no assumptions

concerning

whether

unmodified

stretch

reflexes are part of these complex responses. Following this, we present evidence that the responses obtained after the request that the subjects make no intervention can, in selected cases, be regarded as unmodified stretch reflexes.

Responses to disturbances known direction Increases

in load force that were delivered

averages

of the responses

ob-

onsets, which we will refer to as “departures.” For the example shown in Fig. 2, the departures in acceleration, in ‘arm force, and in doubly filtered EMG all had latencies of approximately 70 ms, which suggest that they are analogous to the modifications in stretch response studied by Hammond (16) using a different pair of instruc-

tions. Departures in singly filtered EMG occurred at a latency of 60 ms. The unfiltered EMG was unavailable in this instance, but in other when it was available

Arm

Force

EMG

B

at

tained with the two instructions were superimposed in order to measure the latency at which differences in responses occurred. These instruction-dependent differences had abrupt

instances

t ion

of

random times deflected the arm away from the body (Figs. I and 2). The subjects had been instructed either to not intervene or to compensate; a sequence of trials with one instruction was followed by a sequence with the other. When the request was for no intervention, the maintained load force resulted in a maintained displacement of the arm in the direction of the disturbance. When the instruction was to compensate, the time course of displacement was more complex and the final value was often in a direction opposite to the direction of the disturbance; i.e., there was a maintained overcompensation, as illustrated in Fig. 2. This aspect of response cannot be attributed to servo action, since steady-state overcorrection is incompatible with stable operation, whereas the arm was quite stable. The ensemble

Accelero

we found

it too

variable to allow reliable estimates of the latenties of departures. The very earliest changes in

200

Time

400

600

( msec)

FIG. 2. Dependence of responses to increased load force on instruction. Heavy traces show averaged responses (n = 10) with the instruction “do not intervene” and light traces show averaged responses (n = 10) to the same disturbance with the instruction “compensate. ” For these records, the direction of disturbance was known in advance; when the direction was unknown (not shown), the latency at which an instruction-dependent difference (departure) in arm force occurred increased from 70 to 80 ms. The EMG was doubly filtered in A and singly filtered in B (30 ms time constant); acceleration was also filtered (15 ms time constant). Calibrations represent 10 N for load force and arm force, 2 cm for arm displacement, and 5 m/s* for acceleration.

singly filtered EM G began at 25 ms and did not differ for the two instructions. The latencies at which departures in acceleration occurred ranged from 70-165 ms in a group of eight subjects tested with the same paradigm. These subjects had latencies that ranged from 80 to 175 ms when the mechanical

disturbance

was

a decrease in load rather than an increase. For half of these subjects the latency was longer with load decreases; in the others, latency did not depend on the direction of the disturbance.

HUMAN

Responses to disturbances unknown direction

STRETCH

of

The responses described in the previous section were obtained with the subject fully aware of the direction of the impending disturbance. Hence, he always knew the direction of an appropriate response in advance. In the same subjects we studied the responses obtained when the direction of the disturbance was made random, which meant also that the direction of an appropriate response was not known in advance. This difference would not affect performance in a servo system; but, if the corrections were analogous to responses in a reaction-time task, one would expect the latencies to increase and become more variable due to the added element of choice (17, 23, 40). We found a definite increase in latency, and responses became more variable. The latency at which a departure in acceleration occurred increased by 10-50 ms in 11 out of 16 comparisons involving the 8 subjects already mentioned; in 4 comparisons, the increase was larger and in one there was no change. In a group of 21 subjects, the latency of acceleration departures with unknown disturbances ranged from 80 to 320 ms. The longest ones occurred in subjects who were told that a correct compensatory response was more important than a fast one; the shortest latencies were found when speed was stressed. However, there was considerable overlap between the two groups. It also appeared that latencies were shorter when larger disturbances were used, but this possibility requires further study. Plots of standard deviation versus time calculated from sets of displacement records indicated that variability of displacement was low (less than 10%) with either instruction during the latent period prior to the departure. An increase in variability (as much as threefold) which began at the latency of the departure was often observed with the compensate instruction, and usually not with the request for no intervention. Variability was more prominent with naive subjects than after a period of practice. The reasons for the late increase in variability became evident when individual responses were superimposed for comparison. Most of the responses obtained with the request for no intervention (NI) superimposed well on each other, as illustrated for two subjects in Fig. 3. Responses obtained with the request for compensation (C) also superimposed at early times, prior to the latencies at which they differed from the NI responses. A major cause of the variability at later times was a trial-to-trial difference in the

REFLEX

929

2

0

-iE-2

FIG. 3. Superimposed comparisons of individual responses to increased load force. Three records of arm displacement with the no intervention (NI) instruction and three with the compensate (C) instruction are superimposed for each of two subjects (A and B). Increases and decreases in load force (k 18 N in A and + 15 N in B) were delivered in a random sequence; only responses to increases are shown. They illustrate the instruction independence of response at early times, the abrupt onset and trial-to-trial latency variation of the modification (triggered reaction) associated with the compensate instruction and the lower variability with the no intervention instruction. Trace I in part B illustrates an inappropriate response, obtained with the compensate instruction.

timing of a delayed component of response, and it was clear that this component was responsible for the departures described earlier. The delayed component also showed some variation in time course and amplitude and, in occasional trials, it was initially in a direction opposite to the one which resulted. in compensation (trace I in Fig. 3B). These inappropriate responses suggest an error in a decision-making process and cannot be explained as the response of a servo system.

IdentiJication of unmodified stretch reflexes We concluded from the results already presented that the observed modifications of the stretch reflex, regardless of their latencies, were the result of triggered reactions, events which represent selected preprogrammed movements. Thus, one basis for the identification of unmodified stretch reflexes was the absence of triggered reactions. The latter were usually easy to recognize in individual displacement traces (cf. Fig. 3), although when they were small in amplitude and occurred during the initial deflection of the arm they were sometimes difficult to

CRAGO, HOUK,

930

detect. Another basis for identifying the unmodified response was a low variability which continued throughout the period of observation. The sensitivity of this test was improved by restricting our analysis to cases in which the direction of the disturbance was random since the modifications, when present, were more variable in this circumstance. Randomization of direction and distraction of the subject were both effective in delaying or inhibiting triggered reactions. Delaying them helped in their identification since triggered reactions were more easily distinguished when they occurred on the plateau of the stretch reflex (Fig. 3B) than when they occurred during the 100- to 200-ms transient phase (Fig. 3A). By using these various criteria and procedures, we were able to substantiate the assumption that responses which are relatively free of modification can be obtained with the instruction “do not intervene” (4). However, some subjects were not successful in suppressing triggered reactions, attesting to involuntary aspects of their control. The records of these subjects were not used in the analysis of reflex action.

AND HASAN nent E MG peaks at short latencies in response to increased load force. Instead, there was a broad elevation reaching a maximum at 100-200 ms (Fig. 4, heavy traces). This indicated that the monosynaptic pathway was not activated appreciably at its minimal latency, but did not exclude a participation in the later response along with polysynaptic pathways. The absence of a sharp, monosymiptic peak was attributed to the modest accelerations and stretch velocities achieved; as mentioned earlier, the abruptness of the disturbance was deliberately limited so as to prevent the EMG from becoming synchronized. The responses of most subjects showed some cyclical behavior before settling to a steady value, and there were correlated changes in force. In the example shown in Fig. 4A cyclic variations are prominent, whereas in Fig. 4B they are less noticeable. In none of our subjects did the periods of reduced activity represent actual silent periods, as confirmed by examining the records from individual trials. REFLEX

RESPONSES

TO

DECREASED

LOAD

Typical EMG responses to decreased load force are shown by the light traces in Fig. 4. EMG changes accompanying There was an initial, brief period of decreased reflex actions activity (sometimes absent, as in Fig. 4B), folThe earliest changes in EMG activity OC- lowed by a longer period during which activity curred at latencies ranging from 20 to 65 ms in a increased, frequently to values above the initial, group of 21 subjects. These latencies preceded predisturbance level; subsequently, the EMG those for departures in EMG for all subjects decreased to a value below the initial level. The although the latency of the latter was shorter or phase of increase always began while the biceps nearly the same for some subjects as the latency continued to shorten. Previous studies of the of the former was for others. For each subject it EMG responses to unloading have usually inwas clear that the first change in EMG rep- volved a complete, or nearly complete, removal resented the onset of reflex action, and this of the initial force on the limb. Under these change did not depend on whether the direction circumstances, the first E MG change is often a of disturbance was random or known. silent period that may be ended by a “terminal The results obtained after the instruction not motor volley” (2) or by a more complex pattern to intervene indicated that reflex action, once (38). Under our conditions of partial unloading, initiated, continued for the duration of the rec- we rarely observed a period of E MG silence of ord. This finding suggests that the long-latency sufficient length to be termed a silent period, and component of stretch response described in the the terminal volley was replaced by the broad INTRoDucTIoN is, in general, a composite re- increase in E MG activity described. sponse made up of a reflex action together with a The detailed features of the E MG responses triggered reaction. When we compared the extra differed between individuals even though the iniincrement in EMG attributable to a triggered tial mechanical conditions and the load change reaction with the increment attributable to reflex were approximately the same (compare Fig. 4A action, the two were sometimes of comparable and B). However, when a single subject was magnitude (compare heavy traces with the dif- examined on different days, the responses were ference between heavy and light traces in Fig. quite similar. This suggests that the differences 2). between subjects are genuine and are not due to The remainder of this section deals exclu- differences that might be produced by the difsively with EMG responses that were attributed ferent placements of the electrodes. to reflex action. FORCE.

ASYMMETRY REFLEX FORCE.

RESPONSES

TO

INCREASED

LOAD

None of our subjects showed promi-

CAL

LOAD

OF

EMG

symmetrical

RESPONSES

TO

SYMMETRI-

The EM G responses to changes in load were notably

CHANGES.

HUMAN

STRETCH

REFLEX

931

60-

409

20-

I

I

I

0

200

I

I

400

I

I

600

Time (msec)

O-

I

0

I

I

200

I

I

400

I

I

600

Time (msec)

FIG. 4. Comparison of symmetry of arm force, displacement, and EMG responses (means in all cases) to symmetric changes in load force for two subjects. The heavy traces represent responses to step increases in load force and the light traces represent responses to equal but opposite decreases. The EMG responses were singly filtered (30 ms time constant) and were calibrated in units of force. Both subjects (but especially B) showed force and displacement responses which were much more symmetric than were the EMG responses. Note, however, that the steady-state changes in EMG were more symmetric.

asymmetric in all of our subjects. The EMG always increased when the elbow flexors were stretched by loading; but, when these muscles were allowed to shorten by unloading, there was either

a decrease

of smaller

magnitude,

or an

actual increase. In the extreme case the asymmetry was so marked that the responses to unloading looked very much like the responseis to loading (Fig. 4B); in other cases the difference was less marked (Fig. 4A). We consider this asymmetry to be one of the characteristic features of the self-regulating pathways to and from individual muscles, a matter we will return to in the DISCUSSION. However, other possible interpretations required consideration. The first possibility was that the action of other muscles at the elbow joint complimented the asymmetry of the biceps EMG. We found, however, that the brachioradialis EMG either was similar in time course to that of biceps or did not change appreciably, and that the triceps EMG was undetectable during reflex responses in either direction. The initial flexor tonus, the limited amount of unloading, and the slowing of the movement by the inertial component of the

load were probably responsible for the absence of triceps activity. (The triceps was active in some subjects in association with triggered movements, and in other cases that will be considered later.) A second possibility was that EMG silence during unloading contributed to the asymmetry. However, silent periods were seen in only a few subjects; when present they lasted less than 20 ms and appeared only when the initial force was adjusted to a low level. The third possibility was that the relationship between EMG activity and contractile force was curvilinear, as reported to be the case for some subjects by Bouisset (6). However, monotonic curvilinearity cannot explain an increase in EMG above the initial value during shortening. Furthermore, the data in Fig. 4 are from two subjects with linear EMG-force relations.’ It 1 The EMG was calibrated by asking the subject to hold his arm at the same initial position with different steady loads. The steady-state relation between EMG and force at constant length was used to provide offset and multiplicative (scale) constants.

932

CRAGO, HOUK,

may be concluded that curvilinearity account for the asymmetry, although tribute to it in some subjects.

does not it may con-

Attempts to alter stiffness of stretch reflex We considered that the gain of the stretch reflex might be subject to presetting, such that the stiffness of the limb might be altered (INTRODUCTION). In six subjects we tried various instructions (e.g., “be rigid,” “resist,” “minimize arm deflection,” “cocontract”), but found that none resulted in any appreciable modification of response at times prior to the occurrence of a triggered movement except when the subject succeeded in cocontracting the elbow antagonists. We judged when cocontraction occurred either by direct recording from triceps or by observing an increase in the EMG of biceps when the initial force was unchanged. While one can readily cocontract when a joint is not initially loaded, it is more difficult to do this when the limb is preloaded, as it was in our experiments. Cocontraction doubled the stiffness in some cases, but the change was attributable to purely mechanical effects since we noted no appreciable alterations in the EMG responses other than the upward shift of the initial level. In the usual cases in which there was no apparent cocontraction, subjects accomplished the requested rigidity by producing a triggered reaction. We concluded that the responsiveness of the stretch reflex was not appreciably modified as a result of these different instructions to our subjects. DISCUSSION

Our results provide qualitative support for the hypothesis that the function of the stretch reflex is to compensate for variations in muscle properties, rather than for changes in load, and that the regulated property may be stiffness (20, 33). This interpretation is based on knowledge of the manner in which mechanical properties of muscle depend on the direction of length change. If a muscle activated by electrical stimulation is suddenly released from an isometric length, the force falls precipitously in a manner that is predicted by the force-velocity relationship (18).

(The velocity dependence of force during shortening is probably much more important than the length dependence at the velocities prevalent in our experiments (42).) Thus, the stiffness, calculated as the decrement in force divided by the decrement in length, is high and positive. The velocity dependence during lengthening is more complex and cannot be characterized by any

AND HASAN single force-velocity curve (22). However, the available data indicate that the initial stiffness of the muscle is high and comparable to that prevalent with release, whereas stiffness decreases markedly and may become negative when the amount of stretch exceeds approximately 1% of the rest length of muscle fibers (22, 33). As a result of these nonlinear features, the mechanical stiffness presented by active muscle fibers should be less during lengthening than during shortening, provided length change exceeds l%.*. 3 In our experiments, motor units in biceps that were active at the required initial force must have contributed a mechanical stiffness opposing length change, which was then modulated by motor output. The prominent increase in EM G that occurred when the biceps was loaded indicated that the output of the stretch reflex was appropriate to compensate for the expected reduction in muscular stiffness during lengthening. The decrease in EMG that occurred when the biceps was unloaded was smaller in magnitude, which suggested that motor output did not appreciably increase stiffness during shortening, when muscular stiffness would be expected to remain high. In fact, the EMG responses of many subjects actually increased during the transient phase of shortening, suggesting that motor output acted to reduce stiffness. The observed asymmetry of motor output thus supports the hypothesis that the actions of the stretch reflex compensate for variations in muscle stiffness. Compensatory action was quite effective in most subjects, for the displacements of the arm and the changes in arm force that resulted 2 The change in biceps length was estimated to be 0.14 times the wrist displacement based on the lever ratio given by Wilke (42) and an angle of 45” between the forearm and the direction of displacement. If we assume a typical displacement of 1.5 cm and a biceps muscle fiber length of 10 cm, the change in fiber length in our experiments would have been approximately 2%. For the purpose of this calculation we ignored the smaller changes in biceps length resulting from shoulder rotation. 3 It might be objected that these mechanical properties are special features of the animal muscles that have been studied and that human muscles behave quite differently. While this seemed unlikely, we searched for evidence that the elbow flexors in man presented less mechanical resistance to lengthening than to shortening. Cavagna and collaborators (7) demonstrated that the relationship between work and velocity during maximal voluntary contractions of the elbow flexors is asymmetrical with respect to lengthening and shortening. From their data it can be inferred that, when averaged over the cycle of length change investigated, there was less increase in force during lengthening than there was decrease during shortening.

HUMAN

from equal but opposite

changes

STRETCH

in load force

were approximately symmetric (Fig. 4).4

933

REFLEX

between

the body and the environment.

The im-

Phillips (35) that in primatesand mana transcor-

portance of compliance may be its ability to absorb the impact of a suddenchangein load, thus attenuating transmissionof the disturbance to the body and head (20). When a rigid control of

tical servo loop, or stretch reflex, dominates

position

Recently hypothesis

several authors have favored the which was originally put forth by the

is required,

it is available

by the produc-

segmental stretch reflex (10, 25, 30, 41). Our results bear on this question to the extent that

tion of short-latency triggered reactions. There are numerousreports indicating that the

we are able to compare

responsiveness, or gain, of the stretch reflex is modified when the initial conditions are changed

reflex action

in human

subjectswith that observed in decerebrate cats, which lack the postulated “transcortical servo loop. ” A qualitative comparison indicates that the actions of the stretch reflex are similar in the two cases; both show the rather

characteristic

form of asymmetry described in the previous section. The major difference between the

(e.g., ref 21, 25), but these modifications may be due to system nonlinearities rather than to a specific neural control of responsiveness. If there are brain mechanisms which act to alter the gain of the stretch reflex, it should be possi-

ble to demonstrate them when the initial force

stretch reflex in humans and in decerebrate cats is that the latter lack the superimposed move-

and length are controlled, as in our experiments. Our failure to find such changes agrees with the

mentswhich we refer to as triggered reactions. This obviously suggests,but does not prove, that the demonstrated transcortical pathway might be responsiblefor conducting or controlling short-latency triggered reactions, rather than mediating servo actions. This is an impor-

results reported by Asatryan and Fel’dman (4), and suggeststhat gain changemay not be a usual mechanismby which the stretch reflex is controlled. Our results indicated that triggered reactions cannot be attributed to the actions of a servo

tant distinction since the performance characteristics and neural mechanisms of the two types

system. At present we do not know how to characterize the neural systems which produce

of system are probably very different.

triggered reactions, except to indicate that there is probably a sequence of processes beginning tant difference between stretch reflexes in hu- with the detection of specific sensory cues and mans and in decerebrate cats. The phase of ending with the selection of appropriate reEMG increase during shortening is a feature that sponses.The presetting of these processes that has not been observed in decerebrate prepara- results when different instructions are given to tions (33; unpublishedobservations). Since ten- the subject must somehow establish different don organ discharge should be reduced by the criteria for detection and associatewith each an decreasedforce during shortening, a removal of appropriate preprogrammed movement comIb inhibition is probably an important factor conmand. Discovery of the neural mechanisms retributing to the EMG increase. The absenceof sponsiblefor these actions would probably lead this phasein the decerebrate may result from a to a better understanding of voluntary movelow gain in the Ib pathway, which is believed to ment control in general. be a characteristic of the decerebrate state (9, Our results did suggest one potentially

19, 21). The phase of EMG

impor-

ACKNOWLEDGMENTS

increase

during

.shorten-

We thank our colleagues in the Dept. of Physiology and Dr. E. Bizzi for their helpful comments. Dr. William H. Talbot designed the PDP-11 computer/ laboratory interface and assisted with computer prosignificanceof this result, it is important to con- grams. sider that a rigid cpntrol of muscle length is not We are grateful for the support of this work by the required for body stability; the only fundamental National Institutes of Health under Grant NS 11446 to requirementsare that forces be adjustedto coun- J. C. Houk and Program Grant NSO 6828. P. E. Crago was supported by a Postdoctoral Felterbalance external loads and that there be damping. The stretch reflex meets these re- lowship (NS 57240-01) from the National Institute of Neurological Diseases and Stroke. quirements while providing, a springlike property which imposes a compliant mechanical interface Present address of 2. Hasan: Dept. of Electrical Engineering,Indian Institute of Technology, New 4 Measured values of the static stiffness of the limb Delhi 29, India. ranged from 5 to 15 N/cm in different subjects and Present address of P. E. Crago: Dept. of Biomedical were not appreciably different for stretch and release. Transient properties of stiffness will be dealt with in a Engineering, Sears Tower E-B65, Case Western Reserve University, Cleveland, Ohio 44106. separate report.

ing is of specialfunctional interest sinceit represents a reflex action rather than opposing

assisting length change, it. In contemplating the

CRAGO,

934

HOUK,

AND

HASAN

REFERENCES 1. ALLUM, J. H. J. Responses to load disturbances in human shoulder muscles: the hypothesis that one component is a pulse test information signal. 2.

3.

Exptl. Brain Res. 22: 307-326, 1975. ALSTON, W., ANGEL, R. W., FINK, F. S., AND HOFFMAN, W. W. Motor activity following the silent period in human muscle. J. Physiol., London 190: 189-202, 1967. ANGEL, R. W., GARLAND, H., AND ALSTON, W.

17. 18.

19.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

Exptl. Neurol. ASATRYAN,

28: 230-242,

10: 925-935, BASMAJIAN,

1965.

J. V. AND LATIF, A. Integrated actions and functions of the chief flexors of the elbow. A detailed electromygraphic analysis. J. Bone Joint BOUISSET,

Surg.

39:

1106-l

118,

71: 479-494, FEL’DMAN,

967-981, 1974. GOTTLIEB, G. L., AGARWAL, STARK, L. Interactions between

G. C., AND voluntary and postural mechanisms of the human motor system. J. Neurophysiol. 33: 365-381, 1970. 16. HAMMOND, P. H. An experimental study of servo action in human muscular control. Proc. 15.

23. 24.

25.

Psychol.

3rd,

4: 1 l-26,

1960,

p. SO-

1952.

Ser. B 126: 136-195, 1938. T., JANKOWSKA, E., AND LUNDBERG,

33: 784-811,

1970.

G. C., RACK, P. M. H., AND WESTBURY, D. R. The mechanical properties of cat soleus muscle during controlled lengthening and shortening movements. J. Physiol., London 204: 461-

474, 1969. LEONARD, J. A. Tactual choice reactions. I. Quart. J. Exptl. Psychol. 11: 76-83, 1959. MARSDEN, C. D., MERTON, P. A., AND MORTON, H. B. Servo action in human voluntary movement. Nature 238: 140-143, 1972. MATTHEWS, P. B. C. Mammalian Muscle Receptors and Their Central Actions. Baltimore: Wil-

liams & Wilkins,

26.

27.

28.

29.

30.

1966. GOODWIN,

G. M. AND LUSCHEI, E. S. Effects of destroying spindle afferents from jaw muscles on mastication in monkeys. J. Neurophysiol. 37:

London, HONGO,

physiol. JOYCE,

578, 1966. FEL’DMAN,

A. G. Functional tuning of the nervous system during control of movement or maintenance of a steady posture. III. Mechanographic analysis of the execution by man of the simplest motor tasks. Biophysics 11: 766-775,

J. Exptl.

22.

1974.

A. G. Functional tuning of the nervous system with control of movement or maintenance of a steady posture. II. Controllable parameters of the muscles. Biophysics 11: 565-

Electron.,

J. C. The phylogeny of muscular control configurations. In: Biocybernetics, edited by H. Drischel and P. Dettmar. Jena, Germany: Fischer, 1972, vol. 4, p. 125-144. 21. HOUK, J. C., SINGER, J. J., AND GOLDMAN, M. R. An evaluation of length and force feedback to soleus muscles of decerebrate cats. J. Neuro-

1973. EVARTS,

E. V. AND TANJI, J. Gating of motor cortex reflexes by prior instruction. Brain Res.

Med.

HILL, A. V. The heat of shortening and the dynamic constants of muscle. Proc. Roy. SOC.,

1969. HOUK,

1957.

S. Emg and muscle force in normal motor activities. In: New Developments in EMG and Clinical Neurophysiology, edited by J. E. Desmedt. Basel: Karger, 1973, p. 547-583. CAVAGNA, G. S., DUSMAN, B., AND MAGARIA, R. Positive work done by a previously stretched muscle. J. Appl. Physiol. 24: 21-32, 1968. CRAGO, P. E., HASAN, Z., AND HOUK, J. C. EMG and mechanical responses to load perturbations applied to the human arm. Sot. Neurosci., St. Louis, 1974, p. 141. ECCLES, R. M. AND LUNDBERG, A. Supraspinal control of interneurones mediating spinal reflexes. J. Physiol., London 147: 565-584, 1959. EVARTS, E. V. Motor cortex reflexes associated with learned movements. Science 179: 501-503,

W. E. On the rate of gain of information.

Quart.

20.

1970.

D. G. AND FEL’DMAN, A. G. Functional tuning of the nervous system with control of movement or maintenance of a steady posture. I. Mechanographic analysis of the work of the joint on execution of a postural task. Biophysics

Conf.

A. The rubrospinal tract. II. Facilitation of interneuronal-transmission in reflex paths to motoneurones. Exptl. Brain Res. 7: 365-391,

Interaction of spinal and supraspinal mechanisms during voluntary innervation of human muscle. 4.

Intern. 199. HICK,

31.

32.

33.

1972, p. 546606. G. AND WATT, D. G. D. Observations on the control of stepping and hopping movements in man. J. Physiol., London 219: MELVILL

JONES,

709-727, MERTON,

1971.

P. A. Speculations on the servo-control of movement. In: Ciba Found. Symp. Spinal Cord, 19.53, p. 247-260. MEYER-L• HMANN, J., CONRAD, B., MATSUNAMI, K., AND BROOKS, V. B. Effects of dentate cooling on precentral unit activity following torque pulse injections into elbow movements. Brain Res. MONSTER,

94: 237-251,

1975.

A. W. Loading reflexes during two types of voluntary muscle contractions. In: Control of Posture and Locomotion, edited by R. B. Stein, K. G. Pearson, R. S. Smith, and J. B. Redford. New York: Plenum, 1973, p. 347-362. MURPHY, J. T., KWAN, H. C., MACKAY, W. A., AND WONG, Y. C. Physiological basis of cerebellar dysmetria, Can. J. Neurol. Sci. 2: 279-284, 1975. MURPHY, J. T., WONG, T. C., AND KWAN, H. C. Afferent-efferent linkages in motor cortex for single forelimb muscles. J. Neurophysiol. 38: 990-1014, NEWSOM

1975. DAVIS,

J. AND SEARS, T. A. The proprioceptive reflex control of the intercostal muscles during their voluntary activation. J. Physiol., London 209: 711-738, 1970. NICHOLS, T. R. AND HOUK,

J. C. Improvement in

HUMAN

STRETCH

linearity and regulation of stiffness that results from actions of the stretch reflex. J. Neurophysiol. 39: 119-142, 1976. 34. PARTRIDGE, L, D. Modifications of neural output signals by muscles: a frequency response study. J. Appl. Physiol. 20: 150-156, 1965. 35. PHILLIPS, C. G. Motor apparatus of the baboon’s hand. Proc. Roy. Sot., London, Ser. B 173: 141174, 1969. 36. STEIN, R. B. Peripheral control of movement. 37.

38.

Physiol. Rev. 54: 215-243, 1974. TATTON, W. G., FORNER, S. D., GERSTEIN, L., CHAMBERS, W. W., AND LIU, C. N.

G. The effect of postcentral cortical lesions on motor responses to sudden upper limb displacements in monkeys. Brain Res. 96: 108-113, 1975. TERZUOLO, C. A. AND VIVIANI, P. Parameters of

REFLEX

935

motion of EMG activities during some simple motor tasks in normal subjects and cerebellar patients. In: Cerebellum, Epilepsy, and Behavior, edited by I. Cooper, M. Riklan, and R. S. Snider. London: Pergamon, 1974, p. 173-215. 39. VALLBO, A. B. Human muscle spindle discharge during isometric voluntary contractions. Amplitude relations between spindle frequency and torque. Acta Physiol. Stand. 90: 319-336, 1974. 40. WELFORD, A. T. Fundamentals ofSkill. London: Methuen, 1968, p. 27-104. 41. WIESENDANGER, M., RUEGG, D. G., AND LUCIER, G. E. Why transcortical reflexes? Can. J. Neurol. Sci. 2: 295-301, 1975. 42. WILKIE, D. R. The relation between force and velocity in human muscle. J. Physiol., London 110: 249-280, 1950.