The Central Nervous Control of Complex Movements in the Uropods of Crayfish

Specific, uniquely identifiable interneurones that produce complex motor patterns occur in several invertebrate central nervous systems. The giant fibres of crayfish and other macruran decapods, for example, trigger motor output in many segments to muscles that achieve rapid tail flexion. Wiersma and his co-workers (Wiersma, 1952; Wiersma & Ikeda, 1964) have demonstrated a similar release of defensive behaviour and of swimmeret beating in crayfish. In two preparations, the crayfish and the mollusc Tritonia, stimulating single cells elicits equally well-co-ordinated responses involving large numbers of motoneurones (Kennedy, Evoy & Hanawalt, 1966; Willows, 1968).

The influences of such ‘command fibres’ upon a motor system has been extensively analysed in the case of a slow flexor and extensor muscles of the crayfish abdomen (Kennedy, 1966; Evoy & Kennedy, 1967; Kennedy, Evoy, Dane & Hanawalt, 1967; Atwood & Wiersma, 1967). Single command interneurones activate assemblies of motoneurones in a reciprocal fashion within a single segment, and usually produce output in a series of adjacent segments. Elements having a similar action may differ from one another in the identity of the individual motoneurones they drive, or in the segmental ratio of the output they produce.

The present paper extends this kind of analysis to the more complex motor system that controls the movements of the terminal abdominal appendages, the uropods and telson. The neuromuscular system of these appendages has been described (Larimer & Kennedy, 1969); the tonic muscles are reasonably accessible and regulate movements in the three axes of extension/flexion, promotion/remotion, and rotation, as opposed to the single one of flexion/extension. This provides an opportunity to discover how combinations of movement are centrally programmed. Furthermore, in contrast to the bilaterally symmetrical output that invariably characterizes abdominal movements, the appendages frequently move asymmetrically for ‘steering’ actions. We have tried to establish how central interneurones control these two additional complexities, and have also dealt with specific units that produce rhythmical cycles of motor output. The properties of one such interneurone will be described in detail, and evidence will be provided that its effects are independent of sensory feedback.

Crayfish were maintained and prepared for experiments as described in the previous paper (Larimer & Kennedy, 1969). After the uropod and telson muscles and the nerves innervating them had been exposed, the dissection was extended anteriorly to the second or third abdominal segment and one or more connectives were de-sheathed with fine forceps. Activity in two to four of the nerves to the uropod muscles was monitored with suction electrodes; alternatively, electromyograms were recorded from the muscles themselves with suction electrodes. Fine filaments were isolated from a de-sheathed connective and placed over paired platinum hooks in oil for stimulation. Suction electrodes could be attached to an appropriate region of a more caudal connective in order to monitor the evoked activity descending toward the sixth ganglion. In some of the experiments described below this procedure allowed us to ascertain immediately that stimulation of a single fibre was producing the motor output, since the monitoring electrode detected the recruitment of a single all-or-none descending action potential when the stimulus voltage reached the threshold for evoking the response.

A Beaulieu camera was used to take motion pictures of some movements evoked by command-fibre stimulation. It was operated at a speed of approximately 24 frames per second; a photodiode placed behind the shutter provided a synchronizing signal for one oscilloscope trace that allowed frames on the filmed record to be correlated with specific points on the electrical record of motor output. To reconstruct movements, tracings were made of single frames from the cine film projected through a time-and-motion-study projector equipped with a frame counter.

In some respects the effects of command interneurones on uropod motor systems resemble those of other command elements upon the tonic extensors and flexors of anterior abdominal segments. The mixture of innervation in the various nerves prevented us from making complete comparisons of the outflow to antagonistic sets of muscles. However, root 6 of the last abdominal ganglion supplies flexor muscles almost entirely, and various command fibres evoke discharge in this root that is reciprocal to the discharge in the portion of root 2, ganglion 5 which supplies the tonic telson extensor muscles (see, for example, records B, C, Fig. 3).

Reciprocity is also normally preserved within the population of efferent neurones serving a single muscle. All the tonic muscles of the uropods receive peripheral inhibitory axons, and these axons are usually active only during periods when discharge of the motoneurones to the same muscle has been centrally suppressed. Figure 1, for example, is a record of activity in the anal dilator muscle made with a small suction electrode. The command fibre stimulated in this experiment was one which produced a shifting pattern of motor output that alternately excited and inhibited flexor muscles. In the record, bursts of activity in motoneurones (producing excitatory junctional potentials (e.j.p.s)) alternate with repetitive discharge in the peripheral inhibitor (producing inhibitory junctional potentials (i.j.p.s)). Other examples of such internal reciprocity are to be found in Fig. 3, records B and C, where the extensor inhibitor (largest spike, trace 4) is reciprocally related to the discharge of the smaller excitatory axons, and in Fig. 2, where the excitatory axon to the dorsal rotator muscle (lower traces) is active in record B and the inhibitor alone responds in records C and D. Whatever peripheral inhibitory axons were identified and recorded in these experiments, their activity was found to be reciprocally related to the discharge of agonist exciters. It is, however, often the case that inactivation of a particular muscle is attained exclusively by central means, i.e. by suppression of excitatory outflow without discharge of the peripheral inhibitory axon.

Most of the movements accomplished by the tail appendages in freely moving animals are combined movements; that is, they depend upon the action of two or more sets of muscles that move the appendage in different planes. Some combinations are, of course, more common than others; it is therefore possible to characterize as ‘semi-agonists’ muscles which are usually synergists but do not have the same action. The rotator muscles, for example, tend to be synergistic with the flexors. Two modes of control of combined movements by command interneurones could be envisioned. Several command elements, each one specific for a given plane of movement, could be activated simultaneously; alternatively, single command fibres might release activity in ensembles of motoneurones to produce the combined movement.

We have evaluated these alternatives by simultaneously recording from two or more motor nerves to muscles that are not strict antagonists. Figure 2 illustrates a simple experiment in which activity in the anal dilator muscle (a flexor) was recorded with a suction electrode while discharge in the nerve to the dorsal rotator muscle was monitored simultaneously. Figure 2A is a record of spontaneous activity; 2B-D show the responses to three different command fibres isolated in the same preparation. The interneurones were stimulated at a constant frequency of 75/sec. in each case, and the records are samples taken at least 2 sec. after the stimulation was begun, so that the motor ouput had reached a constant level. In B both the rotator and the flexor were strongly excited. In C both were inhibited; the smaller axon discharging in the dorsal rotator record was identified as the peripheral inhibitor. In D there is a selective action: the rotator is inhibited, but the flexor is weakly excited. This finding was confirmed several times with these two muscles; the effect of central commands was usually of the same sign and their relationship was therefore normally synergistic, but this synergism was occasionally broken by command elements that inhibited one muscle and excited the other.

An analysis involving additional motor output channels is illustrated in the experiment of Fig. 3. Here simultaneous records were taken from the sixth root of ganglion 6, which serves primarily flexor muscles (trace 1); the third root of ganglion 6, which serves promotors and remotors (trace 2); the posterior branch of the first root of ganglion 6, which innervates the dorsal rotator (trace 3); and the second root of ganglion 5, which goes to the telson extensors. Record A shows spontaneous activity; records B-D illustrate steady-state responses to stimulation (75/sec.) of three different command interneurones isolated from the same preparation. The flexor muscles reciprocate reliably with the extensors in B, C, and D. Excitation of motoneurones in the promotor/remotor group may accompany drive to the flexor group (B) or to the extensors (C); but different sets of motor elements may be responding in the two cases, since separation of promotor and remotor axons was not possible. Excitatory output to the dorsal rotator muscle could occur in the presence (B) or in the absence (D) of flexor or promotor-remotor drive. These results make it clear that single command elements provide for a related series of actions in different motor channels.

The additional complexity of movements involving the uropod muscles relates to a question raised earlier by experiments on postural movements of the abdomen. The demonstration that fine filaments isolated from a given region of the connective produce the same motor output from preparation to preparation suggest that a single interneurone with a specific central location is involved, but—unless the movement is a unique one and the isolation has been achieved repeatedly—they are not completely convincing. With the postural system in the abdomen it was possible to isolate single interneurones at two points with motor outputs between them, and to show by monitoring the activity at one point that impulses in one axon alone could release the behaviour (Kennedy, Evoy & Hanawalt, 1966). This technique is not feasible with the uropod, motor system because of the terminal location of the sixth abdominal ganglion.

In a few cases, however, we were able to work with command fibres near one surface of the 2−3 connective, and to monitor descending activity in the appropriate region of the 5−6 connective with a fine suction electrode. An experiment in which the output resulted from activity in a single interneurone is shown in Figs. 4 and 5. Figure 4 shows tracings of projected moving picture frames; the film was taken while electrodes on the caudal connective were used to record electrical activity. Figure 5 A shows that tactile stimulation delivered to the telson produced impulses in a single unit in the isolated filament (upper trace, I) and that these spikes were correlated with spikes in the 5−6 connective (lower-trace, C). This procedure established the amplitude of discharge in the cord record that was associated with the active unit in the isolated filament. The filament was then stimulated at constant frequency and gradually increasing voltage while the movement was simultaneously photographed. The arrow in the record of Fig. 5B shows the point at which an impulse of the appropriate amplitude was evoked in the cord record; on the slow time-base its appearance is marked by an increase in the apparent duration of the stimulus artifacts. As the voltage was further increased during the remainder of the record no additional recruitment occurred. The amplitude of the driven descending spikes can be clearly measured at the end of the record, since their tops form a notch on the large stimulus artifacts. A vertical line above the record indicates, for comparison, the amplitude of the cord spikes identified and measured in A. The two are equal.

The frame in which the movement could first be detected in the accompanying motion picture was 460 msec, after the point marked by the arrow in Fig. 5 B. Such a value is reasonable in view of the fact that the movement involves tonic muscles, which characteristically show a long latency. The movement itself involved flexion, some rotation and promotion of the ipsilateral exopodite, and was clearly produced by a single interneurone. That interneurone, furthermore, responded to sensory stimulation in the abdominal region (cf. Kennedy, 1968). In another case a single element was shown to produce motor output to the dorsal rotator muscle and to an unknown number of others.

To understand the control of motor output by central command interneurones it is desirable to work with a set of identified cells that meet certain criteria. The output should be as complex as possible, both temporally and spatially, in order (1) to establish the uniqueness of the triggering central element, and (2) to place maximum constraint upon the connexion models that might be drawn. At the same time the output must be highly stereotyped. One should be able to de-afferent the system completely, and the central element that releases the response must be available for repeated isolation.

All these criteria are met by a unit that we have identified in area 85 of the abdominal connectives (Wiersma & Hughes, 1961). This interneurone reliably releases a cyclical pattern of movements that involves activity in all motor roots supplying the uropod and telson muscles. Figure 6 shows a series of tracings of the tail appendages during constant-frequency (75/sec.) stimulation of the command fibre. The movement consists of a series of extensions and flexions of the tail fan at gradually declining frequency, accompanied by much more prominent cycles of promotion/remotion and rotation. The movements involve the uropods of both sides, which are usually in phase at first but later become independent. With prolonged stimulation the movements occur at a lower frequency, and phase relationships within a cycle also change somewhat. Periods of stimulation lasting less than 10 sec. are usually followed by a terminating movement that begins when the stimulus is turned off. This off-effect consists of promotion and more delayed rotation. A similar output is often produced when stimulation has been maintained for very long periods and the frequency of movement has declined drastically.

These movements correlate well with electrical records of the motor discharge to appropriate groups of muscles. Such records also provide a more objective measure of consistency. Figure 7 A is a continuous record of discharge in roots 6, 3, and 1 of ganglion 6 and root 2 of ganglion 5 (traces 1−4, respectively) in response to continuous command-fibre stimulation at 75/sec. The record has been cut so that corresponding portions of the three successive response cycles shown are approximately aligned vertically; stimulation was stopped at the point indicated by the downward arrow.

The records shown were taken after about 3 sec. of stimulation, when the repetition rate of the discharge cycle had reached a relatively stable (though gradually declining) value. At this point the cycle length was approximately 750 msec. A convenient reference point for describing the structure of a single cycle is the peak of discharge of the large unit in trace 4, identified as the extensor inhibitor. It goes through two frequency maxima, the second at about twice the frequency of the first; after the second peak the discharge terminates, and a smaller, excitatory neurone begins to fire. The inhibitory output to the extensors accompanies inhibition in the output to the dorsal rotator muscle, a period of suppression of most units of the promotor/remotor complex of root 3, and a brief excitation of flexor motoneurones in root 6. This cycle is replicated with remarkable fidelity during a given train of constant-frequency stimuli to the interneurone, and also upon successive stimulations. When stimulation is stopped (record A₃) the subsequent period of discharge in the promotor/remotor nerve is somewhat stronger. Activity in the dorsal rotator motoneurone reappears later, after the end of record A₃, and is then of higher frequency and longer duration than in a single cycle.

Figure 5 B shows a single cycle that followed the onset of stimulation by about the same length of time as did those in 5 A. It is, however, from a different experiment, which took place on another day and involved another animal. The command fibre was isolated from the same region of the connective, and recording conditions from the nerves were as identical as one could make them using suction electrodes. The record is remarkably similar to those shown in A₁₋₃, though the excitatory units in the extensor trace (4) begin firing during the peak of the inhibitor discharge, instead of following it. Such breakdowns in reciprocity are uncommon, but do occur in the homologous tonic extensor supply in more anterior segments (Kennedy, Evoy & Fields, 1966; Fields, Evoy & Kennedy, 1967). The accompanying pause in the excitatory discharge to the dorsal rotator muscle is shorter, and the decrease in outflow to the promotor/remotor muscles is more clearly a reciprocal event, involving discharge of a single, larger unit during the silent period of the other neurones. The excitatory burst to the flexor muscles is also earlier and less distinct. The features in this list which do not involve an alteration in phase or latency relations are probably attributable to differences in recording conditions, which can never be made precisely alike. Those involving temporal changes are certainly real, but they are at the level of detail rather than of basic structure.

It was important to determine whether the motor output was influenced or conceivably even determined by reflex responses to afferent feedback from the periphery. Most of the experiments, like that shown in Fig. 7, were performed using en passant recording techniques; under these conditions motor output reaches the muscles. If the appendage is not restrained by being pinned, it moves in the fashion indicated in Fig. 6. Both pinned and unpinned preparations were employed for experiments like that in Fig. 7, and no important differences in output were noted between them. This result indicates that sensory feedback from a moving appendage is not required for regulation of the motor output; but obviously some degree of feedback reaches the central nervous system from the periphery even when muscles are contracting against restrained appendages. For this reason, in several experiments the motor output to the uropod muscles was recorded before and after severing all roots to the sixth ganglion distal to the recording sites, de-afferenting the anterior segments, and cutting the ventral nerve cord one connective rostral to the de-afferented portion. Figure 8 shows such an experiment, and also illustrates that the homologous interneurones on the two sides of the cord have equivalent effects. Record A shows the responses to stimulation of the command fibre in area 85 right, and B the responses, after the de-afferentation procedure, to stimulation in area 85 left. The preparation is the same one that supplied the record in Fig. 7B. The output generated by the two interneurones is essentially identical, and it is unaffected by the elimination of sensory feedback.

The interneurone responsible for this behaviour is extremely constant in its location, and has been isolated more than 20 times in these experiments. On several occasions attempts were made to split the bundle responsible by dissection, and it never was possible to fractionate the effect even when the remaining filament was extremely fine. The constancy of effect and of location, and the sharpness of the threshold for output, leave little doubt that a single cell with complex, highly specific output connexions is responsible for the motor output. The alternative—that several fibres must be stimulated together—is contra-indicated by the fact that a large number of isolations and sub-dissections of bundles in the appropriate area always produced the same output.

In a number of cases an active filament was recorded while natural stimuli were delivered to the animal in an attempt to discharge the interneurone. All of these attempts failed; small bundles with low electrical thresholds for producing the motor output never responded to sensory stimulation, suggesting that the interneurone has its input in more rostral segments or requires a highly complex set or sequence of caudal inputs. Since the motor output affects both sides, we sought to determine whether the interneurone on each side of the nerve cord made symmetrical output connexions, or whether the effect was transferred between the partner fibres by an excitatory cross-link in the sixth abdominal ganglion. Both fibres were isolated in a rostral connective and placed on hook electrodes, and each was shown to produce the characteristic motor output upon stimulation. Activity in one of the filaments was then recorded while the other was stimulated repetitively at a strength greater than that required to produce the output. No activity was evoked in the homologous contralateral fibre by this procedure. Furthermore, the effect of stimulating both command fibres together was shown to be greater than that due to stimulating either one alone. These results show that the bilaterally homologous interneurones lack a strong excitatory interconnexion in the sixth or more anterior ganglia, and that the output connexions made by each are symmetrical and impinge upon the same final motor elements.

The postural abdominal muscles always respond to command-fibre stimulation with an output that is bilaterally symmetrical, though roots to the swimmerets in the same segments were shown to respond asymmetrically to the same command fibres (Evoy & Kennedy, 1967). Since the uropods frequently participate in steering movements, the left and right appendages are often positioned differently in normal behaviour, and we have therefore analysed command-fibre influences upon specific homologous motor pathways of the two sides. We have concentrated on the nerves supplying the two dorsal rotator muscles, since they are accessible and because rotation is a prominent component of steering reactions.

Figure 9 shows the responses of left and right dorsal rotator nerves to five different I command fibres, isolated sequentially in the same preparation and stimulated at the same frequency (75/sec.). The spontaneous discharge frequency of the larger, excitatory axon (record A) was approximately equal on the two sides; the upper trace in each record (L) is from the left side, and the lower (R) from the right. The position of each command fibre in a right or a left rostral connective is indicated by the letters R or L above each record. The effects show almost all possible symmetry relations for excitation and inhibition. Command fibres can produce ipsilateral excitation and contralateral inhibition (E) or the reverse (F); symmetrical inhibition (D), nearly symmetrical excitation (C); or contralateral excitation without ipsilateral inhibition (B). Inhibition of a given side can be accomplished with (F) or without (E) activity in the peripheral inhibitory axon. The set of effects illustrated is undoubtedly not exhaustive, and a complete catalogue would probably include the missing reciprocal actions of records B and F.

When clearly antagonistic pairs of muscles are compared, our results on the uropod motor system resemble those obtained earlier from the postural abdominal muscles (Evoy & Kennedy, 1967). Antagonists behave in a reciprocal fashion under drive from a variety of command fibres. The basis for this reciprocity is not known; it is sufficiently consistent, however, to suggest that inhibitory interconnexions at the moto-neurone level might be responsible for it. Strong connexions of this kind have been directly demonstrated only in a much smaller visceral ganglion, the stomatogastric (Maynard, 1966). Whether such networks also arrange for the control of opposing skeletal muscles is an important question, but direct evidence is difficult to obtain. Antidromic stimulation of purely motor roots in more anterior segments does not produce the crossed inhibition that would be predicted from the presence of such connexions, but the antidromic impulses may have failed to invade the sites where the cross-linkages are located (Evoy, Kennedy & Wilson, 1967).

The consistency with which certain muscle pairs are driven by command fibres is at variance with a proposal about motor control made by Hoyle (1964). He found that locusts making apparently similar locomotor movements often displayed differences in the EMG patterns recorded by implanted wire leads, and suggested that different outputs might be employed to produce the same behavioural result. Sensory feedback from that movement would be required to match a centrally stored template, but the output channels themselves could be used in various ways as long as the matching conditions were satisfied. Our experiments deal with different animals under different conditions; but they argue strongly for the view that a fixed set of motor output connexions—a ‘motor score’ (Wilson, 1968) activated by specific interneurones—is employed instead.

Inhibitory and excitatory axons innervating the same tonic uropod muscle displayed reciprocity in response both to natural stimuli and to command-fibre drive, as did those of the postural abdominal muscles. This finding expands the generality that crustacean peripheral inhibitory axons function in the absence of concurrent excitation, and that pre-synaptic inhibitory mechanisms are therefore unnecessary. The opener muscle of the dactyl (Bush, 1962; Wilson & Davis, 1965) is still the lone exception to this rule.

Our experience with command fibres suggests that certain pairs of muscles usually are synergistic (e.g. the rotators and flexors; Fig. 2). These relationships, however, are not rigid; the synergism is occasionally ‘broken’ by a particular command fibre, and they therefore cannot readily be explained on the basis of fixed connexions made entirely at the motoneurone level. The situation resembles that described by Wilson (1962) in locusts, where a ‘bifunctional’ muscle may be the synergist of another muscle in one behaviour pattern and its antagonist in another. In these cases the connexions necessary for the appropriate motor outputs are probably made at a level pre-synaptic to the motoneurones. If they are connexions made by the command fibres themselves, then different processes of the same cell must be capable of excitatory and inhibitory actions. Double-action interneurones of this kind are known in molluscan ganglia, and complex networks based upon them have been worked out (Kandel et al. 1967). Alternatively, several sets of pre-synaptic ‘driver’ networks, each programming a specific combination of motor outputs, might intervene between the command interneurones and the motor elements (Evoy & Kennedy, 1967; Kennedy, 1968). We have no further evidence on which to locate the patterning connexions.

The experiments on different muscle groups, however, show clearly that ‘combined’ movements—those involving simultaneous action in several planes of movement—are already coded for in the connexions made by single command interneurones. The execution of such complex movements does not require activity in a whole set of interneurones, each connecting with a special fraction of the motor output. Instead, movements utilizing a diverse combination of muscle groups are often triggered by single interneurones. This finding is not surprising in view of the complexity of other behaviour patterns which have been obtained in the same animal with single-cell stimulation: the defensive reflex (Wiersma, 1952), the swimmeret rhythm (Wiersma & Ikeda, 1964) and the uropod rhythm. Our results suggest, however, that such connexions—instead of being restricted to a few especially significant complex behaviour patterns for which special triggers are adaptive—are, in fact, a regular feature of the control of muscle. The repertoire of behavioural combinations possible may be larger than the number of command fibres, requiring that interneurones often be used in groups. In principle, however, a wide variety of combined movements may be produced by single-cell stimulation, and conversely, most interneurones tend to affect a wide variety of outputs.

The complex and highly stereotyped output of the interneurone repeatedly isolated from area 85 shows that an entirely central rhythm may be triggered by a single cell. Recent studies on a variety of arthropod behaviour patterns have supported the view that cyclical output is often the product of purely central rhythmicity, rather than the result of recurring reflexive drive. The case has been made rigorously for insect flight, where an appropriate motor output involving 80 motor neurones and repeated about a dozen times each second will be generated by a totally de-afferented ganglion if its level of excitation is high enough (Wilson & Wyman, 1965). In terms of number and diversity of output channels the behaviour pattern described in the present paper is of about this level of complexity. That it can be produced by single, identified cells suggests that whole locomotor patterns may quite generally be initiated and maintained by the activity of individual interneurones. The constancy of connexion from animal to animal in the interneurone we have studied is remarkable, and certainly adequate for the reliable operation of a critical behaviour pattern. The example we have presented here is novel in that the internal fine structure of cyclical motor output due to a single cell withstands the test of de-afferentation. The same is probably true of the swimmeret system, where Davis (1969) has identified the role of specific motor units in the cycle of activity; single central neurons undoubtedly can release the pattern (Wiersma & Ikeda, 1964).

The experiments on combined movements, on symmetry of output and on the identified ‘rhythmic’ interneurone all lead to a substantial extension of our earlier conclusions about command fibres. From the experiments on the abdominal muscles, it was clear that single interneurones could activate or inhibit a number of motoneurones in reciprocally related sets, and that they could operate in several adjacent segments. These actions, however, required only that flexor and extensor groups of efferent neurones be acted upon differentially at a constant level of pre-synaptic drive as long as the command fibre was discharging. A variety of simple connexion models could adequately explain all the phenomena we observed: for example, a ‘flexor driver’ neurone, impinging upon flexor exciters and the extensor inhibitor, could be coupled by reciprocal inhibition to an ‘extensor driver’ with the converse efferent connexions. A flexor command fibre would innervate the flexor drivers of several adjacent segments, with the synaptic efficacy of the connexion varying systematically between segments. Accessory connexions from command fibres to specific moto-neurones could be proposed (Evoy & Kennedy, 1967) to account for the selective actions of some command elements.

To this simple version we would now have to add the following embellishments: (1) In order to account for the variable utilization of synergists, we must propose several sets of ‘drivers’ having different sorts of coupling between them. This coupling could be reciprocal inhibition where strict antagonists are involved, but might be weakly excitatory where the muscles are usually synergistic. (2) The great variety of motor output combinations we have observed requires either a very large number of ‘driver ‘interneurones, or a hierarchy of them, or a situation in which there is a more modest number of driver interneurones and certain command fibres make by-pass connexions directly with some motoneurones. (3) The complex temporal and spatial organization of the rhythmic output suggests that there must be complex interconnexions between groups of neurones postsynaptic to the command fibre. In view of the considerations discussed earlier these are not likely to be at the motoneurone level, and this argues further for one or more layers of interpolated ‘driver’ interneurones. Apart from some proposals for the locust flight system (Wilson, 1966; Wilson & Waldron, 1968), there have been few attempts to specify what sorts of interconnexions might be important for a complex temporal pattern involving several channels. Inhibition is probably involved, and an examination of the pattern itself suggests that the interconnexions must link several different sets of elements. (4) The findings on symmetry of output in homologous motor nerves indicate that command fibres must vary in the lateral distribution of their connexions. Some clearly affect only the ipsilateral or contralateral side, though most have an influence upon both. It is not reasonable to infer that bilaterally symmetrical output normally depends upon crossconnexions between homologous bilateral interneurones. If that were the case then all outputs of a given command fibre should be identical on the two sides. This is clearly not so in some instances (cf. Evoy & Kennedy, 1967), and the rhythm-producing interneurone has also been shown to lack cross-connexions. Thus the spread of excitation from command fibres to their contralateral outputs must depend upon trans-ganglionic interneurones or upon motoneurone branches that cross the midline. Such ‘crossing’ elements must, in turn, make very selective connexions with different command elements in order to account for the diverse symmetry patterns that exist.

Supported in part by grants from the U.S. Public Health Service (NB-02944, D.K. and NB-0542, J.L.) and the U.S. Air Force Office of Scientific Research (AFOSR 334-68, D. K.). We thank Mrs Philip Hanawalt for her unfailing technical assistance, and Dr Donald Wilson for helpful suggestions.