Proprioceptive Leg Reflexes in Cockroaches

Animals were filmed with a 64 frames/sec. 16 mm. movie camera while they walked or ran across a field marked with a reference grid. The films were examined using a projector having continuous or single-frame advance, and frame counter. The frequency of leg movements and the relative phase of the several legs were noted. Only a few results of the analysis are presented here. The significant remainder will appear elsewhere (Wilson, 1966).

Animals were fixed, ventral side up, and one leg was attached by means of wax to a controlled lever. The remaining legs, sometimes less one, were amputated at the coxa or trochanter. 50 μ wire, insulated except at the tip, was inserted into a leg muscle, most often the extensor tibialis, and connected to one side of a push-pull amplifier. A large reference electrode was placed in the hind gut. Muscle action potentials were displayed simultaneously with a record of the movement of the lever holding the leg. This movement was monitored by means of a photo cell and a partially interrupted light beam. The frequency response of the monitoring system was very high compared to that of the system under analysis. The lever was moved by a strong electromagnet energized by a variable signal generator and power amplifier. The mechanical system had a good frequency response up to only about 25 cyc./sec., but with sufficient energy could be driven to about 50 cyc./sec. with a useful amplitude.

Muscle potentials from one or two legs and lever movements were displayed on a storage oscilloscope triggered by the input generator. Weak responses were integrated by storing many repeated sweeps of the oscilloscope beam. This method provided a quick qualitative indication of strength of reaction and correlation and phase relationship. In addition to this visual examination, the movement and muscle potential records were analysed by two techniques of cross-correlation. In one method the intervals between muscle potentials were plotted over their times of occurrence or at the mid-intervals. Points for correlational analysis were computed by linear interpolation between these actual data points, thus generating a continuous function for comparison with the input. Conventional correlation functions were computed using a general-purpose digital computer.

For the second method, the muscle potentials were used to trigger a pulse-generator and the resulting train of pulses and the input function were analysed, on-line, using the Mnemotron CAT and Cross-correlator. By proper adjustment of the triggering level of the pulse generator it was possible to eliminate movement artifact from the motor-output function (see Fig. 1). If the muscle record had been used directly, spurious results would have been obtained since there was often some artifact observable even without averaging. At the higher frequencies movement artifact became a serious, and finally limiting, interference.

Most of the observations were made using Blatta orientalls or Periplaneta americana as the experimental animal. The principal results have been found to be characteristic also of a grasshopper, Trimerotropis pallidopenis.

Many observations similar to those of previous workers were made (see especially Hughes, 1952a, 1957; Wendler, 1964). In addition there were two fairly remarkable new findings. These are interesting in connexion with the later part of this paper, and will be presented without a complete discussion of the larger study of leg coordination from which they are excerpted (Wilson, 1966).

An adult Blatta orientalis, when running very fast, can move its legs at a frequency of at least 15 cyc./sec. and perhaps as rapidly as 20 cyc./sec. With the camera operated at 64 frames/sec. a complete leg cycle was recorded by only three frames. This camera was altogether insufficient for resolving details of the pattern of leg movements at the high speeds, but was just adequate for a crude measurement of the maximum frequency. The body temperature was not measured, but may be presumed to be higher than room temperature, due to the flood lamps. This high frequency of oscillation raises some question as to whether the leg reflexes can coordinate the leg activities.

At the higher running speeds the well-known alterations in stepping patterns following amputation of the middle legs do not occur (see Fig. 2). If the animal has only prothoracic and metathoracic legs it walks with a diagonal stepping pattern (such as many others have reported) only at moderate speeds. When the animal is excited to greater running speed the normal gait is used by the four remaining legs. The result is that both legs on one side step at the same time and the two sides alternate. This pattern is fully effective in the case of the cockroach, since at its higher frequencies of leg movement the body cannot fall to the ground between steps. For the more sluggish grasshoppers and the mole cricket (Stenopelmatus) locomotion nearly fails on a smooth surface because the body falls from side to side and traction is reduced. This non-adaptive pattern also throws some doubt upon the reflex coordination hypothesis.

If a cockroach leg undergoes forced sinusoidal movements along approximately the same plane as it uses for locomotion, a reflex motor output usually follows. The output may involve both the fast and slow motor axons. It may wane, or adapt (Fig. 1), or remain rather steady during long bouts of stimulation. Usually a response occurred on the first cycle of input. Sometimes many cycles of input were required before a response developed, but then it might be quite a strong one. The fast motor fibre activity always seemed to have a higher threshold and was therefore less dependable. However, most of the illustrations to be presented are of fast-type activity because the much greater amplitude of muscle potential provides relatively noise-free records. The output nearly always shows some temporal dependence upon the phase of the mechanical input, but this may be either very clear, as in Fig. 3, or more or less obscured by irregularity as in Figs. 7 and 9. Even in the case of highly irregular records, averaging hundreds of cycles demonstrated clear phase correlations. Fig. 4 shows correlograms generated by the Mnemotron equipment during stimulation at the same frequency, but at several amplitudes. The phase of the correlogram is independent of the amplitude of the input. Not indicated by this method of presentation is the fact that larger amplitudes of input give stronger responses which may also be more highly correlated so that the necessary averaging time is shorter.

If the amplitude of the sinusoidal mechanical input is held nearly constant and the frequency varied, a set of correlograms like those of Fig. 5 a may be obtained. Again the phase changes little if at all. The variation from run to run is not systematically related to changes in frequency. The alternative method of computation (see Methods) gave the same result. Consistent results were obtained over the range of 2–20 cyc./sec., probably the whole normal range for locomotion in these cockroaches. Apparently similar results could be obtained outside this range, but at lower frequencies the response was usually weak and irregular while at higher ones either movement artifact or mechanical damage to the leg interfered with data collection. Occasionally a preparation would continue to function at 40 cyc./sec. or more, giving a single properly phased muscle potential per cycle.

The input-output phase relationships were the same for many different preparations in which the same muscle was used.

A small number of preparations showed an anomalous behaviour. In the middle of the frequency spectrum the phase of the correlogram changed abruptly to a quite different value, whereas both below and above this frequency the phase was constant. In the only cases in which this behaviour was regularly repeatable (at least three) it was possible to relate it to the following mechanism. In these cases the recorded muscle unit fired at least twice per cycle at widely separated phase points (see Fig. 6). It occurred that one of these phase points was dominant at low frequency and the other at high frequency. The cross-over frequency was not constant from animal to animal or even for a single preparation. The significance of this is not understood. It could be that the ganglionic motor neuron pool undergoes two cycles of excitation during a single leg step-cycle even during normal locomotion, namely, one for each of the antagonistic sets, and that occasionally a neuron may fire during the wrong cycle. Occasional evidence of a similar phenomenon is seen in recordings of the activity of locust flight-motor units (unpublished observations).

Besides the clearly reflex activity of the leg muscles, seemingly spontaneous bursts of fast motor-neuron activity occur (Fig. 7 a). If these occur during the sinusoidal movement they may be influenced by the input (Fig. 7 b) or nearly override it (Fig. 7 c).

If one leg is moved sinusoidally and muscle records are made from the contralateral homologue, a crossed reflex may be observed. The phase of a particular muscle is approximately opposite that of the other side (Fig. 56). The response is not as strong as the ipsilateral one and it has not been possible to observe it over such a wide frequency range. However, over the range in which clear correlograms could be obtained the behaviour was constant with regard to phase. Rarely the response was in phase with the stimulated leg rather than in antiphase. This unusual response may be related to the peculiar double responses in the stimulated leg (see above).

Pringle (1940) found no intersegmental reflex. With the present method it has been possible to demonstrate weak intersegmental reflexes. It must be stressed that in most preparations effects were not observable or required long averaging. In the few cases in which marked intersegmental reflexes were seen they varied in sign even for the same leg combination. In general, it seems that they must be of only slight importance in co-ordinating the legs. The strongest intersegmental effect found was between a metathoracic leg and the ipsilateral mesothoracic one. Most often this was in an antagonistic relationship. Effects in more distant legs were either nil, positive, or negative. It is possible that these intersegmental effects are not due to direct motor reflexes, but that they result from activation and phasing of oscillatory activity in the next segment, since there are occasionally a few cycles of activity following cessation of the stimulus.

If an oscillatory input is applied to other parts of the body than the legs, well-phased motor output may occur over a similar range of frequencies. Tapping the caudal cercus or head at any frequency up to at least 20 cyc./sec. resulted in activity in the extensor tibialis muscles after a very small delay. From several paired recordings it is apparent that synchronous activity in many extensor muscles also occurs spontaneously at times, sometimes in repetitive bursts. This results in sudden strong extensor thrusts which could be adaptive in startle or escape manœuvres. Electrical stimulation of the nerve cord either in the abdomen or just anterior to the prothoracic ganglion can elicit a similar in-phase response in contralateral legs both after single shocks and during high-frequency input trains. The response to a single shock may be a burst of motor discharges. The latency may be only 15 msec, and is commonly about 20 msec. Under these stimulus conditions no tendency toward left-right oscillation could be found. This type of reaction may be the basis of the in-phase reflex effects between legs which normally alternate during walking.

Single cycles of square wave or sinusoidal movement of large amplitude were used to elicit minimum latency responses. The least values obtained were around 10 msec. Due to lag in the electro-mechanical drive system, it was not possible to determine exactly when the leg sense-organs were first stressed. Therefore, the reflex time was measured from the beginning of the electrical signal, so latency figures, if in error, are too large. In favourable records the whole sequence of activity can be identified. The earliest muscle potentials are probably due to slow fibre activation. The later, larger muscle potentials are preceded by visible nerve action potentials (see Fig. 8).

A few contralateral latency measurements were made. These were all longer. The minimum found was about 20 msec. Usually there was no response to a single input cycle and the response gradually increased in amplitude and decreased in latency with repetition. It is therefore possible that under more facilitated conditions the latency would be smaller.

In all of the cases in which both slow and fast responses could be observed, the slow motor neuron of a particular muscle had a lower threshold than the fast one. Fast activity was identified by larger amplitude muscle potentials and lack of facilitation. Anti-facilitation, or relative refractorin ss of responsiveness, occurred at the higher frequencies of discharge. Slow activity was identified by smaller amplitude potentials and often the presence of facilitation of the muscle action potential (see Fig. 9). It seems useful to think of the motor neurons of one muscle as being members of a neuron pool with those having larger axons also being of higher threshold (Hoyle, 1964). Hughes (1957) reports similar relationships during d.c. stimulation of the ganglia. During rhythmic stimulation the smaller muscle potential is activated first and only when stimulation is sufficiently intense is the largest fast motor neuron recruited (Fig. 10). Even though the axon producing the largest muscle action potentials probably has the higher conduction velocity, the smaller muscle potential occurs earlier in a cycle of activity (Fig. 10).

Occasionally during unstimulated activity both flexor and extensor motor units were activated during the same burst. Hoyle (1964) has similarly found simultaneous activation of antagonistic muscles in the legs of grasshoppers. This shows that even for the same leg antagonistic motor neurons are not connected in a simple, fixed reciprocal relationship.

The minimum latency of less than 10 msec, for the proprioceptive leg reflex of these insects suggests that the reflex arc is a rather simple one. Perhaps it is monosynaptic, but, of course, with many parallel input fibres converging on a few motor neurons. The 10 msec, includes several time-intervals which are probably relatively fixed, in addition to those which vary with stimulus velocity. The fixed delays should give rise to a phase lag as frequency increases, but this lag was below the level of resolution in the phase measurements. In sequence the various delays are (1) mechanical lag, (2) rise of sensory potential, (3) centripetal conduction, (4) ganglionic delay, (5) centrifugal conduction, (6) neuromuscular delay. (1), (2) and (4) should be proportional to stimulus velocity and therefore responsible for the variable latency necessary for constancy of phase.

The burst of motor activity during a single cycle of input showed no very regular pattern. At the higher stimulus frequencies only a few output pulses occurred during each cycle. It was not possible to identify a clipped sinusoidal modulation of pulse frequency as was found by Chapman & Smith (1963) for sensory spines on the cock-roach leg. In general, the comparison of this reflex activity with the studies on sensory units has not been very revealing, although the techniques are very similar.

The results agree with and extend earlier studies by Rijlant (1932a, b) and Pringle (1937) in which other input functions were used. Usually the ipsilateral reflex antagonizes the imposed movement and the contralateral one is reciprocal. All reflex outputs seemed to be in either of two phase positions with respect to the moved leg; that is, either positively or negatively correlated to the output of that leg. This limited diversity of phase relationships is relatively easy to understand as far as possible mechanisms are concerned, but it is not easily reconciled with reflex models for the co-ordination of limb-movement patterns, since the latter exhibit a wide range of phase relationships. The phase relationship for the reflex of one limb is such that the output tends to resist the input motion ; that is, output is some function of position. However, presumably due to adaptation at some point in the cycle, the response is not exactly in phase with the position, but rather is moved somewhat in the direction of being in phase with the velocity. The preparations varied as to whether the response was mainly antagonistic to position or velocity, but this variation was not a function of frequency.

Some observations on walking patterns in normal and limb-deficient animals call into question the adequacy of the proprioceptive reflex hypothesis of ambulatory co-ordination. It is shown here, however, that the reflexes have temporal characteristics which are consistent with their playing a significant role at all speeds of locomotion. In fact, it must ordinarily be that they do have an effect unless other parts of the nervous system can switch them out, an unappealing notion. More likely, in cases in which single motor units or whole motor complexes do not follow the proprioceptive command, a stronger command from another source is of overriding power. Thus, while in the apparatus, animals occasionally produced strong bursts of output not related to the timing of the lever movements. Also, at highest running speed evoked by intense exteroceptive stimulation, amputees utilized seemingly non-adaptive gaits which are not consistent with the reflex operation.

If the nervous events are capable of such high-speed operation, what limits the actual running frequency? Clearly, one limitation must be due to the relatively slow rate of muscle tension or length change and attendant mechanical processes. Twitch durations for the leg muscles are of the order of 40 msec, or greater (Usherwood, 1962). At only 15 cyc./sec. the single twitch would last more than half of the time for a whole leg cycle ; that is, simultaneity of antagonistic muscle action would be inescapable. The rise time of the twitch would occupy a significant fraction of the whole cycle and therefore introduce a phase shift of tension as frequency increased further. In fact, Horridge (personal communication) has found that the mechanical response to sinusoidal forced movements fails at frequencies much lower than those found here for the nervous components.

Several results suggest both positive and negative effects couple the activity of motor neurons which control antagonistic walking muscles. Similar effects are found between motor neurons used in both flight and walking in locusts (Wilson, 1962). Possibly related to these reflex effects are the findings of Hughes (1952ft), who applied direct current to the thoracic ganglia. He found that bilaterally symmetrical stimulation (electrodes arranged in the sagittal plane) resulted in symmetrical discharge in the motor neurons; that is, increase in either flexor or extensor activity on both sides, depending upon electrode polarity. On each side flexors and extensors showed reciprocal effects. Asymmetrical polarization of the ganglion (electrodes arranged on either side) resulted in crossed relationships between flexors and extensors of the two sides. Application of direct current to a motor nerve through the ganglion caused excitation or depression of both flexors and extensors innervated by that nerve. These labile interrelationships suggest that the motor neurons controlling different anatomical parts are not interconnected directly, but merely follow input which is diversely distributed to them according to activity. Proprioceptive feedback may, however, impinge directly on the motor neurons of the same leg.

Some generalities about the proprioceptive leg reflexes may be suggested. The movement of a leg results in a resistance reflex which is phasic-tonic; the reflex is negative and during oscillatory input the response leads the leg position. Response in other legs is weaker with distance, especially in other segments. Response at least in nearest neighbour legs (contralateral in same segment, ipsilateral in next segment) has a tendency to be in opposite phase. Phase variation between legs is all-or-nothing, being either positive or negative, and does not show the graded phase relationships which occur in walking.

1. Cockroaches have been found to run with stepping frequencies up to 20 cyc./sec. At the higher frequencies mesothoracic amputees do not exhibit the gait adaptations characteristic at lower frequencies.

2. Proprioceptive leg reflexes can follow input frequencies at more than 20 cyc./sec., without phase shift over the frequency spectrum. The myotatic reflex output shows some phase lead with respect to leg position.

3. The contralateral leg responds less strongly and is 180⁰ out of phase with respect to the response of the stimulated leg. Legs in other segments show even weaker responses which are variable in sign.

4. The nervous reflex effects are competent to influence co-ordination at all frequencies found in normal locomotion. The muscular abilities presumably set a limit to higher speeds of running.

I wish to thank Dr P. Timiras and the Mnemotron Corporation for the loan of the on-line computing apparatus. Other financial assistance was provided by the National Science Foundation (grant GB2116) and the National Institutes of Health (grant NB3927).