Proprioceptive Control of Posture in the Crayfish Abdomen^*

The crayfish abdominal musculature is segmented and bilaterally symmetrical. The extensor musculature in each half-segment consists, as does the flexor system (Kennedy & Takeda, 1965 a, b), of relatively massive deep muscles and thin superficial ones. It has been shown (Fields & Kennedy, 1965; Kennedy & Takeda, 1965 a) that the thin superficial extensors and flexors both give graded electrical and mechanical responses to repetitive stimulation of the motor nerves and maintain postural tension through tonic activity on the part of the motoneurones supplying them. The more massive deep muscles respond with a powerful twitch to single impulses in their motor nerves (Kennedy & Takeda, 1965 a; Abbott & Parnas, 1965). All available evidence indicates that, although the powerful contractions responsible for the swimming reflex involve the deep twitch muscles, most slower movements are accomplished by the thin, slowly contracting superficial muscles (Kennedy & Takeda, 1965 a).

Paralleling this division into tonic and twitch extensor muscles, the two muscle receptor organs have been shown to differ from each other both in their response to maintained stretch (Wiersma, Furshpan & Florey, 1953) and in the properties of their receptor muscles (Kuffler, 1954) which are arranged in parallel with the adjacent extensor muscles. The discharge frequency of the lateral MRO is maintained for up to several hours while the more medial receptor adapts completely within seconds. The muscle associated with the slowly adapting receptor (RM1) is thin, coarsely striated and develops tension in graded increments proportional to the discharge frequency of its motor nerve. The muscle of the rapidly adapting receptor (RM 2) is thicker and more finely striated than RM 1 and is capable of responding with a twitch to single impulses in its motor nerve. In addition, it has recently been shown (Fields & Kennedy, 1965) that RM 1 is supplied by branches of one or two of the five excitatory motoneurones supplying the superficial extensors, whereas RM 2 is generally supplied by a branch of one of the motoneurones to the deep twitch extensors.

These facts suggest that the two receptors provide the crustacean central nervous system with different types of information. In contrast to the rapidly adapting (phasic) receptor, the tonic receptor is capable of supplying the tonic proprioceptive input required for operation of the slow, postural muscles. In the absence of motor input the response of the phasic receptor is more directly related to rate of change of tension and/or length of its muscle than to the steady-state length of the RM. It may be that information concerning both change of muscle length and its first time-derivative is a general requirement for operation of motor systems; in fact, Bessou & Laporte (1962) have shown that the dynamic response of the primary ending of the vertebrate muscle spindle is related to rate of change of muscle length, whereas that of the secondary ending is more directly a function of instantaneous muscle length.

The experiments presented below demonstrate that the tonic receptor functions specifically in the segmental control of slow, postural movements and strongly suggest a separation of function between tonic and phasic MROs. A preliminary report of some of the findings has appeared elsewhere (Fields & Kennedy, 1965).

All experiments were carried out on Procambarus clarkii collected locally and maintained in tap-water tanks, in which the animals remained in good condition for several weeks. Three types of preparation were used. Most commonly, the abdomen was separated from the animal by a cut just anterior to the first abdominal segment, a procedure which minimized bleeding. Alternatively, intact animals or those in which the haemolymph had been replaced by van Harreveld’s (1936) solution were used. The parameters measured were not significantly dependent on the method of preparation.

Dissections were carried out with the preparations pinned dorsal side up on paraffin in a Perspex dish filled with van Harreveld’s solution. The extensor musculature was exposed by careful removal of parts of the dorsal exoskeleton and then the hypodermis, leaving the muscle attachments intact.

Wiersma et al. (1953) found that the nerve trunk supplying the extensor muscles, a dorsal continuation of the second root, splits into three major branches (see Fig. 1): the nerve to the profundus muscles (n.p.m.), the nerve supplying the MROs and the superficial extensors (n.r.m. or dorsal nerve), and the nerve supplying tactile hairs in the exoskeleton (n.c.s. or superficial nerve). In those experiments in which a controlled source of excitatory input to slow extensor motoneurones was desired, the superficial nerve—which contains tactile afferents found to be excitatory to these motoneurones— was dissected free for electrical stimulation, leaving its central connexions intact, and draped over a pair of micromanipulated platinum wires. Brief (0·2—0·4 msec.) square pulses were led through an RF isolation unit to these stimulating electrodes. By comparing the latency and stimulus strength required to elicit motoneurone activity before and after cutting n.c.s., the possibility of direct stimulation of motoneurones by current spread could be ruled out. When natural stimulation was used, a minimum of dissection was required to permit access to the extensor muscles for identification and recording.

Motoneurone activity was monitored by recording junctional events in muscle fibres with glass capillary microelectrodes suspended from a micromanipulator by a fine silver wire. These ‘floating’ microelectrodes remained in actively contracting muscle fibres without damage either to the fibres or to the electrode tips. The capillaries were filled with 3 M-KCI and had resistances ranging between 10 and 40 MΩ. Intracellularly recorded signals were led through neutralized capacity preamplifiers (Amatniek, 1958). Receptor activity was monitored by lifting the intact dorsal nerve up out of the solution into a pool of paraffin oil with a hook electrode of fine platinum wire. The signal in this case was led through a Grass P5 capacity-coupled preamplifier. In some experiments motoneurone action potentials were also recorded from n.r.m. ; in this manner both muscle and nerve potentials were displayed on a conventional dual-beam oscilloscope and recorded photographically.

In order to study the output of the MROs in as nearly intact a state as possible, and to correlate receptor output with abdominal position, the following procedure was employed. A completely intact crayfish was suspended by a carapace clamp, either in water or air, and a pair of small-diameter pins (Minutenädeln), with insulation removed only at the tip, were inserted directly through the tergite on one or both sides of a segment and led through a capacity-coupled preamplifier to the oscilloscope. The two pins were placed towards the centre of the half-tergite. Under these conditions the spikes of both fast and slow stretch receptors were large enough to be easily identified, and could be distinguished from one another by their response to manual flexion of the tail. The output of the stretch receptor was correlated with movements of the tail by simultaneously photographing, with separate cameras, the oscilloscope trace and the crayfish. The animal was illuminated with a stroboscopic light source, which was arranged in such a way that it also produced a reflexion ‘artifact ‘on the oscilloscope trace. This allowed correlation of each strobe frame with a segment of the oscilloscope record. All experiments were carried out at 10−15° C.

Results reported earlier (Fields & Kennedy, 1965) have shown that activity in slow extensor motoneurones is consistently evoked by stretching the receptor muscles with a fine wire hook, and en passant recording from the dorsal nerve established that this motoneurone activity is quantitatively correlated with discharge in the MROs. In the absence of other excitatory input to the motoneurones the relationship of firing frequencies in the sensory and motor axones was found to be approximately linear.

By separately stretching RM1 and RM2 it was shown that activity in the slow receptor alone is sufficient to evoke firing of ipsilateral slow extensor motoneurones but has no demonstrable effect on motoneurones supplying the deep extensors. Discharge of the fast stretch receptor, even at frequencies exceeding those which in the slow receptor are adequate to evoke motoneurone firing, produced no observable activation of either superficial or deep extensor motoneurones. Fig. 2 demonstrates the apparent lack of functional overlap of the two receptors. In this experiment most of the excitatory drive to the active motoneurone was produced by the slow receptor. The mechanical arrangement was such that stretching RM2 unloaded tension on RM 1 ; thus fast and slow receptors were activated alternately. Excitatory junctional potentials occurred in slow extensor motoneurones only in response to RM 1 discharge.

Although there is no direct evidence for any central effect of the fast MRO except self-inhibition (Eckert, 1961 a), it seems likely that its reflex function is associated with control of the phasic deep extensors. The deep extensors are twitch muscles, incapable of developing the fine gradations in tension required for slow changes in tail position ; and, in contrast to the superficial extensors, they do not show the tonic activity required for maintenance of a set position of the abdomen. No reliable method of centrally activating fast extensors was found and consequently the role or the phasic MRO in co-ordination of abdominal movements has not been established.

Much more detailed information was obtained concerning the central effects of the tonic receptor. It was observed that the response of a motoneurone to discharge in the tonic MRO, evoked by mechanical stretch of RM1, was highly dependent on other sources of excitatory input. By electrical stimulation of tactile afferent fibres In n.c.s. the level of central excitability could be raised high enough so that single impulses in the slow MRO produced single motoneurone discharges (Fig. 3). However, under these conditions, the motoneurone followed receptor impulses on one-to-one fashion only over a limited frequency range, dropping out at frequencies greater than about two to three impulses per second. The latency from MRO spike to the following motoneurone spike under conditions of one-to-one following was about 30 msec. Even under conditions of frequency division (i.e. more than one receptor impulse per motoneurone impulse), the latency from the immediately preceding receptor impulse to the motoneurone impulse remained at this value, provided drive from other sources did not activate the motoneurone (see Fig. 3). However, this latency figure gives little information about the number of interposed synapses since the conduction velocity of fine fibres in the ganglionic neuropile is unknown.

In several experiments intracellular recordings were made in large numbers of superficial extensor muscle fibres in one segment while RM 1 was stretched and the output of the tonic MRO was recorded. It was observed that MRO discharge consistently activated only one size of junctional potential in each fibre. It has been established, however, that the superficial extensors are supplied by five excitatory efferent fibres as well as an inhibitor (Kennedy, Evoy & Fields, 1966). By recording extracellularly from the muscle it was found that receptor discharge evoked a single large muscle potential. It therefore appears certain that MRO discharge activates only one of the five ipsilateral slow excitatory motoneurones in its segment.

This observation could be explained in two different ways. The MRO might have anatomical connexions with only one segmental motoneurone. Alternatively, the MRO might project to the other motoneurones, but their level of excitability could be lower than that of the one activated by MRO discharge, i.e. the other motoneurones could constitute a sort of ‘sublimai fringe’. The results of several kinds of experiments support the first interpretation. (1) It was possible to activate motoneurones by electrical stimulation of interneurones isolated from connectives of the ventral nerve cord. This was accomplished by dorsal dissection of the extensor and part of the flexor musculature posterior to the site of recording, thus exposing the ventral nerve cord dorsally. Small bundles of fibres were then separated from the cord and electrically stimulated. Stimulation of certain bundles could activate several motoneurones at the same time; RM 1 was then intermittently stretched and relaxed. The effect of MRO input under these conditions was to increase the frequency of impulses in only one of the active motoneurones. (2) Even in the rare cases when an increase in discharge frequency of two motoneurones was observed to accompany MRO activity, the effect was considerably more pronounced in one than the other. (3) It was observed in many preparations that isolation of one of the ganglia in the abdominal cord from the rest of the CNS, by cutting the connectives anterior and posterior to the ganglion, resulted in a marked increase of spontaneous efferent activity in the second root, including the slow extensor motoneurones. Utilizing such isolated preparations, it was possible, by matching spike amplitudes in the nerve trace with muscle junctional potentials, to identify unambiguously three spontaneously active units and to correlate their activity with the discharge frequency of the slow MRO. Fig. 4 graphically demonstrates that segmental MRO input affects only one of the three motoneurones and that the central excitability of the motoneurone (as measured by level of spontaneous activity) is not the factor that determines which one will be so activated. This evidence is not quite conclusive since the study utilized an isolated ganglion; extra-segmental connexions could conceivably mediate MRO effects, even on motoneurones in its own segment.

There is evidence that the motoneurone activated by the MRO has properties that distinguish it from the other slow extensor motoneurones. First, it has a wide distribution to slow extensor muscle fibres. In several preparations, serial penetrations of numerous muscle fibres in all components of the superficial extensors revealed that without exception they were innervated by a motoneurone which could be activated by stretching RM1. In most preparations these fibres showed spontaneous activity, and by correlating the spontaneous junctional potentials with spikes simultaneously recorded en passant in the dorsal nerve it was established that this spontaneously active fibre was identical to that activated by the MRO. In most but not all cases this fibre also had the lowest threshold among the motor axons to tactile stimulation.

The question of whether this fibre innervates RM 1 has been examined, and the results suggest that it does not. In the numerous experiments in which RM1 was stretched to produce motoneurone activity there was no case in which a sudden increase in receptor output followed the first recorded junctional potential, though such an increase was generally seen in response to electrical stimulation of motoneurones which branch to supply RM 1 (see Fig. 5). Furthermore, while low-intensity segmental tactile stimulation was quite effective in activating the motoneurone also driven by the tonic MRO, such stimulation did not generally cause an increase in MRO discharge frequency. In two experiments the motoneurones activated by the MRO was clearly identified in en passant nerve recordings by correlating its spike with the junctional potentials evoked by MRO discharge. When this motoneurone was driven by low-level stimulation of n.c.s., there was no increase in the impulse frequency of the tonic MRO. It thus seems clear that if this motoneurone does supply RM 1 its contribution to tension development is minimal.

Other segmental influences of slow MRO input were sought but not found. In several preparations the inhibitor axon was spontaneously active, and could be identified by correlating hyperpolarizing junctional potentials in the muscle fibres with spikes of a specific amplitude in the dorsal nerve record. In other cases the inhibitor was selectively driven by electrical stimulation of ‘command’ interneurones isolated in the ipsilateral 5−6 connective of the abdominal nerve cord (cf. Kennedy et al. 1966). In no case did MRO input have a clear-cut effect on impulse frequency in the inhibitory nerve, although in some preparations a slight, transient decrease in discharge frequency of the inhibitor accompanied maintained activity in the tonic MRO.

In several experiments activity in the slow flexor motoneurones was recorded while the RMs were stretched. This was accomplished by dissecting the MROs free of the extensor musculature and then exposing the flexor muscles by a separate ventral dissection ; alternatively, the MROs were left intact and a contralateral dorsal dissection was extended through to the superficial flexor muscles. In neither case was any effect of MRO discharge observed upon flexor motoneurone activity. Apparently discharge of the tonic receptor has a rather limited segmental effect, namely, the excitation of one slow extensor motoneurone and self-inhibition via the accessory nerve reflex (Eckert, 1961a). This conclusion is consistent with Eckert’s (1961a) finding that stretching the RMs activates just two efferent fibres in the dorsal nerve.

The distribution of reflex excitation produced by the tonic receptor was studied by recording junctional potentials in ipsilateral superficial extensor muscle fibres in adjacent segments and contralaterally in the same segment. It was found that stretching RM1 evoked ipsilateral motoneurone activity in both the next anterior and next posterior segments but had a negligible contralateral effect in the same segment. This is a somewhat unexpected finding in view of the fact that the contralateral extensors will be synergistic to ipsilateral extensors in all abdominal movements, whereas the ipsilateral extensors in adjacent segments may act as antagonists in some movements.

By simultaneous recording of ipsilateral motoneurones in segments 2, 3, and 4 it was determined that, in response to a given stretch of RM 1 in segment 3, the increase in impulse frequency of the activated motoneurone in segment 3 was 10 times greater than that in either segment 2 or 4. This segmental specificity contrasts sharply to that displayed by the accessory nerve reflex described by Eckert (1961a); in this reflex the accessory nerves of adjacent segments are activated to the same extent as those in the reference segment.

In order to interpret segmental interactions a study of the effect of contraction of the extensors in adjacent segments on MRO discharge was undertaken (cf. Wiersma et al. 1953). It was found that stimulation of motoneurones to the deep extensors in adjacent segments caused a phasic output in both phasic and tonic receptors, and that, generally, contraction of fast extensors in the posterior segment was more effective than contraction of those in the anterior segment. Repetitive stimulation of slow extensor motoneurones in both adjacent segments resulted in a gradual increase in tonic MRO output which attained a final frequency dependent on rate of stimulation.

MRO discharge was monitored in a completely intact crayfish while abdominal movements were simultaneously photographed (see Methods). Fig. 6 (Plate 1) shows that receptor output sometimes varied widely with small changes in abdominal position ; conversely, there was often no receptor output even though the abdomen varied extensively in degree of flexion. Receptor output sometimes actually increased when the segment shortened; this is opposite to the response expected on anatomical considerations. It was repeatedly observed that if the abdomen was manually flexed from a set position it returned to that position, often with remarkable precision. The slow MRO discharge increased when the tail was manually flexed, and was progressively silenced as the initial position was re-approached. Fig. 6 (Plate 1) shows that active extension was preceded by a burst in the tonic MRO which became silent as the new position was approached, as one would predict if the RMs were activated along with the extensor muscles. Experimental evidence such as that of Fig. 7 (Plate 2) demonstrates clearly that receptor output is more directly a function of displacement from some set abdominal position than of absolute degree of flexion of the abdomen. In Fig. 7(Plate 2) abdominal position in frame 3 is very nearly the same as that in frame 4 but MRO activity is much greater at 3 than at 4. In 3 the position was attained by manual displacement of the abdomen from a centrally set position, whereas in 4 the abdomen is in a ‘desired ‘position, i.e. there is no artificial displacement. These results seriously damage the concept of the MRO as a simple length-measuring device and suggest instead that it may operate as an error detector, signalling deviation of abdominal position from some ‘desired’ position.

The analysis of the innervation pattern and reflex effects of the tonic MRO as described above indicates that the following sequence of events takes place in naturally occurring extension of the tail. When the segmental slow extensor motoneurone(s) which supply RM 1 are activated, the muscle of the tonic MRO contracts and the receptor discharge increases; this in turn feeds back as excitation to the most widely distributed of the slow motoneurones. As the extensor muscles shorten, tension in RM 1 decreases, resulting in a decreased frequency of firing in the sensory neurone. The validity of this postulated sequence is supported by results showing that tactile stimulation may cause a transient rise in the frequency of receptor discharge which in turn evokes motoneurone discharge (Fields & Kennedy, 1965). Since the impulse frequency of the MRO is a linear function of tension, which is in turn directly related to the initial length of RM1 (Krnjević & Van Gelder, 1961), the magnitude of the sensory discharge fed back in response to motoneurone activity will depend on the initial length of RM1.

Merton (1950, 1953) was the first to describe the role of the muscle spindle in terms of servo theory. Although it may be questioned whether this approach adds anything that could not be intuitively understood (see Granit, 1962), it provides a simple descriptive framework for discussion and may lead to insights that are not intuitively obvious. The experiments described above strongly indicate that the slow stretch receptor functions as part of a follow-up length servo system similar to that described by Merton. The simplified block diagram of Fig. 8 illustrates the relationships of the major components of the system. It is similar to a schematic of the mammalian spindle system devised by Partridge & Glaser (1960). Presynaptic drive to the slow extensor motoneurone(s) that supply RM 1 determines the set point The output, tail position (in this case, lengthening of the reference segment), represents a balance between the various inputs including segmental motoneurone activity (negative, i.e. shorten the reference segment) and the opposing forces of flexor contraction, gravity and contraction of extensors in adjacent segments (positive, i.e. lengthen the reference segment). The stretch receptor acts as an error detector which responds to differences between the set point (at which its output is zero) and the actual tail position; the magnitude of this error is, of course, given by the impulse frequency of the MRO. MRO discharge activates one of the extensor motoneurones, resulting in shortening of the segment and reduction of the error. By altering excitatory input to the motoneurone supplying RM1 the set point can be varied, allowing the servo loop to operate over a wide range of abdominal positions. Such a system enables the crayfish to adjust the force of slow extensor contraction to any level that might be required to compensate for changes in load or, as pointed out by Barlow (1961), muscle power (i.e. resulting from fatigue or contraction of synergists). Such a system also seems to be operating in certain amphibian muscles, in which the same motoneurones innervate both intrafusal and extrafusal muscle fibres. Katz (1949) observed that when the motoneurones were stimulated spindle discharge was diminished if there was no load on the muscle but was increased instead when the muscle was sufficiently loaded.

The finding that the excitation fed back to the slow extensor motoneurones via MRO discharge does not affect the motoneurones supplying RM1 is of theoretical importance, since if MRO output caused contraction of RM 1 the error signal would be distorted. Further, this type of positive feedback, although diminished by concomitant extensor contraction tending to unload the receptor, could lead to instability of the servo mechanism. This problem is, of course, obviated in mammals since their receptor muscles possess independent (gamma) innervation; although activation of alpha and gamma motoneurones is parallel for most inputs (see Eklund, von Euler & Rutkowski, 1964; Granit, 1955), spindle discharge apparently has no consistent effect on gamma motoneurones (Diete-Spiff & Pascoe, 1959).

In this regard it is significant that the motoneurone activated by the tonic MRO has the lowest threshold to tactile stimulation and apparently supplies every superficial extensor muscle fibre except the receptor muscle. Among the consequences of this arrangement are : first, that the wide distribution and greater excitability of the motoneurone activated by the MRO increases the gain of the servo loop without sacrificing stability; and second, that the arrangement allows filtering of ‘noise’, since transient minor input will cause only a transient shortening of the segment and will not activate the servo loop. Finally, the tonic extensor motoneurones can serve an ‘arousal ‘function, since tactile stimulation almost anywhere on the exoskeleton will excite them. If the servo loop is then brought into action by higher centres, the MRO input will summate with the tactile input either centrally or at the neuromuscular junction.

As required by the theory discussed above, Eckert (1961b) demonstrated that the RMs contract when abdominal extension occurs, but that receptor discharge does not generally accompany the movement unless the receptor muscles are clamped so as to impose isometric conditions. Eckert also reported that when the crayfish actively extended the abdomen manual flexion resulted in a far greater receptor output than if the abdomen had been ‘relaxed’ at the beginning of imposed flexion. In the relaxed state slow receptor discharge did not begin until the abdomen was at 50 % full flexion whereas if the animal had actively extended its abdomen the slow MRO began to discharge at the initial stages of flexion. These latter observations are consistent with the role of the tonic MRO as an error detector and, together with the experimental evidence described in the present work, contradict Eckert’s (1961b) own hypothesis that MRO discharge is directly related to degree of abdominal flexion.

According to Eckert (1961b) MRO output inhibits extensor motoneurones during flexion. He cites as evidence records showing that impulses in the dorsal nerve are inhibited during flexion of the tail. However, he gives no direct evidence that these impulses do, in fact, represent activity in slow motoneurones. It is unlikely that all amplitudes of spikes recorded in the dorsal nerve distal to the n.p.m. branch correspond to extensor excitors, since it has been found in the present experiments that the largest efferent fibres in the dorsal nerve are, in decreasing order of size, the slow extensor inhibitor, the motor nerve to RM2, one of the slow extensor excitors and the accessory nerve. In any case, the dorsal nerve (n.r.m.) does not contain motor fibres to the deep extensors (which run in n.p.m.) and it is likely that these are the muscles participating in the extension phase of the swimming reflex with which Eckert was concerned.

Inhibition of extensors during flexion may occur; but it is not a direct result of discharge in the slow MRO. A more attractive alternative is that this inhibition, if it does occur, is due to unloading of stretch receptors in the ventral nerve cord (Hughes & Wiersma, 1960) which are known to exert a strong excitatory drive upon slow extensor motoneurones (Kennedy et al. 1966). Adding to this effect would be the activation of the accessory nerve, which should in theory inhibit the slow extensor motoneurones by reducing MRO discharge frequency. A centrally originating burst in the accessory nerve has been shown to accompany active flexion (Eckert, 1961 b). The present demonstration that MRO discharge excites extensor motoneurone(s) suggests a plausible function for the centrally originating accessory nerve burst, which would serve to shut off the servo loop and prevent resistance to flexion.

Reflex self-inhibition by the MROs seems more significant when considered at the intersegmental level. Eckert showed that reflex inhibition of MRO discharge is as powerful in adjacent segments as in the reference segment. Since the reflex excitation of motoneurones due to MRO discharge is only a tenth as great in adjacent segments, it is obvious that the spread of reflex inhibition to MROs is greater than the spread of reflex excitation to adjacent slow extensor motoneurones. It has been demonstrated that contraction of slow extensors in a given segment loads adjacent MROs and this, in the absence of reflex inhibition, will activate adjacent superficial extensors. The contraction of these adjacent extensors would both oppose shortening of the reference segment and tend to activate extensors even further removed from the reference segment. This coupling effect would be accentuated by the spread of the tonic receptor’s excitatory effect to adjacent segments. However, the powerful spread of reflex MRO inhibition via the accessory nerve would be expected to counteract this coupling effect and would contribute to the independence of segments and flexibility of the abdomen.

1. The reflex role of the crayfish abdominal muscle receptor organ (MRO) was studied in intact preparations under relatively physiological conditions by electrophysiological and photographic methods.

2. By separately stretching the receptor muscles of the two receptors in each half-segment it was found that discharge of the tonic MRO is sufficient to evoke activity in the tonic superficial extensors but has no effect on the deep twitch extensors. No effect on either muscle group was produced by activity in the phasic MRO.

3. Discharge of the tonic MRO was found to activate only one of the five motoneurones supplying the superficial extensors. This motoneurone is generally ionically active and has a low threshold to tactile input. Evidence is given that this motoneurone does not innervate the receptor muscle of the tonic MRO.

4. The excitatory effect of tonic MRO output on ipsilateral superficial extensor motoneurones is only one-tenth as great in adjacent segments as in the same segment. The effect contralaterally in the same segment is negligible.

5. Simultaneous electrical recording of tonic MRO output with photographic recording of abdominal position proves that the response of the tonic MRO is a function of the difference between actual segment length and some centrally determined ‘desired’ length. Evidence is presented supporting the concept that the tonic MRO functions as an error detector in a follow-up length servo control system which governs the posture of the crayfish abdomen. A model of this system is presented and discussed.

The author acknowledges the help of Drs Donald Kennedy and William H. Evoy in some of the experimental work and in criticizing the manuscript, and is grateful to Mrs Philip Hanawalt for technical assistance.