Crayfish Escape Behaviour: Commands for Fast Movement Inhibit Postural Tone and Reflexes, and Prevent Habituation of Slow Reflexes

An understanding of interactions among the systems underlying different behavioural elements of an animal is crucial to any explanation of the behaviour of that animal. This principle was one rationale for the study of interactions among spinal reflexes (Sherrington, 1906), and among oscillatory behavioural elements like fin movements and respiration (von Holst, 1937, 1939). Ethologists emphasize that interactions also occur during more complex behaviour. For example, displacement activities, threat behaviour and aspects of courtship and other displays have been interpreted as consequences of inhibition and disinhibition among behavioural systems (e.g. Iersel & Bol, 1958; Rowell, 1961). It may be that neural mechanisms that mediate behavioural interactions are common to behaviours of differing complexity.

Study of behavioural interactions is now being made at the cellular level (see, for example, Stein, 1971; Burrows, 1975a, b; Koester et al. 1974; Kovac & Davis, 1977; Mulloney, 1977). The present research investigates interactions at the cellular level between two neuromuscular systems in the crayfish abdomen: the postural flexor system and the escape flexor system. Both systems act at the same points but are behaviourally incompatible in that they act at strikingly different speeds.

The neural basis of each behaviour is already known in some detail. The crayfish abdomen consists of six segments hinged to move mainly in the sagittal plane. The postural flexor muscles consist of thin sheets of slow muscle. In each half segment they are innervated by a thin nerve containing five small, tonically active motor neurones and a peripheral inhibitor. The fast flexor muscles that power the ‘tailflip’ of the escape response are massive, spiralling muscles, innervated in each half-segment by six to ten large motor neurones and a peripheral inhibitor (Takeda & Kennedy, 1964; Selverston & Remler, 1972; Mittenthal & Wine, 1978). The large flexor motor neurones are silent except during tailflips.

The two flexor systems are very different in both their time constants and frequency of use. Peak tension in fast flexors has a latency of 50 ms, whereas the onset of tension in the tonic muscles has a latency of about 100 ms (Roberts, 1968b) and peak tension may require several seconds (Evoy, Kennedy & Wilson, 1967). In adults the tailflip motor circuitry may be inactive for days, depending on the animal’s environment, while in contrast the most active motor neurones in the tonic system can fire 12−17 impulses/s (Kennedy, Evoy & Fields, 1966) or over a million impulses a day.

These contrasts and the apparent independence of the parallel motor systems have led to independent investigations of them. Only two brief attempts to discover interactions have previously been reported. In one, antidromic stimulation of fast flexor motor neurones had no effect on the tonic motor neurones (Evoy et al. 1967), but we now know that antidromic impulses may fail to invade the synaptic regions of neurones (Mulloney & Selverston, 1972). In the other study, impulses in the giant interneurones that command tailflips were reported to cause a late, weak discharge in tonic motor neurones in addition to their earlier and more powerful excitation of the fast ones (Roberts, 1968b).

In this paper we study the response of the postural flexor system to stimulation of the tailflip command neurones. Unlike Roberts (1968b), we find that the main effect is inhibitory and is directed at multiple levels of the postural circuitry, including receptors, central neurones, and muscles. The command neurones have been previously shown to produce multiple level inhibition of the fast flexor and extensor systems (Roberts, 1968a, b; Wine, 1971, 1977a, b; Wine & Mistick, 1977; Wine & Hagiwara, 1978; Krasne, Wine & Kramer, 1977).

Experiments were performed on crayfish (Procambarus clarkii) measuring approximately 7 cm from rostrum to telson. The animals were obtained commercially and maintained in tanks at approximately 18–20°C. Each crayfish was immobilized by slowly cooling to 3–5°C, after which the abdomen was isolated and pinned to a dissection dish filled with cold, oxygenated van Harreveld’s solution (van Harreveld, 1936), in which the bicarbonate buffer was replaced with Trizma at pH = 7·2. A strip of the ventral cuticle and ribs was removed to expose the ventral nerve cord. The main artery and excess connective tissue on the ventral surface of the nerve cord were removed, and the cord isolated by cutting all roots. The roots were kept as long as possible, especially the superficial branches of the 3rd roots (R3S), which contain the axons of the tonic flexor motor neurones. The cord was transferred to a Sylgard-lined preparation dish and pinned out for recording. For muscle recordings, the cord was left in the abdomen and the main branches of the 3rd roots, which contain the axons of the phasic flexor motor neurones, were cut to prevent twitching. The ventral cuticle of a segment was removed to expose the tonic flexor muscle fibres.

We used glass suction electrodes for extracellular recording and stimulation, and glass microelectrodes, filled with either 3 M-KCI or 4 M-K acetate for intracellular recording and stimulation.

The axons of the command neurones (Wiersma, 1938) that mediate the crayfish escape response comprise a pair of lateral giant (LG) axons and a pair of medial giant (MG) axons. They take up the whole dorsal surface of the nerve cord, each axon and its sheathing being about a quarter of the diameter of the cord. Selective stimulation was ensured by using a suction electrode with a tip diameter smaller than the axon, by using just threshold stimuli, and by monitoring the cord for impulses from other neurones evoked below the command neurone’s threshold. Intracellular stimulation was used on two occasions to confirm the extracellular results.

After finding that stimulating the MG and LG axons had equivalent effects on the tonic motor system, we used the LG axons in most experiments because they are larger than the MG axons in the abdomen and so are easier to stimulate selectively. When we wanted to restrict the giant axon impulse to one portion of the nervous system, however, we used the MG axons because, unlike LG axons, they are not cross-connected in the abdomen. Each MG axon has almost symmetrical bilateral outputs (Selverston & Remler, 1972). This permitted us to transect one MG and then compare its effects with an intact MG (Roberts, 1968 a). LG axons are cross connected in every abdominal segment, so that an impulse initiated anywhere in them reaches both sides of all ganglia.

Single impulses in any giant axon usually produced weak and variable effects on the slow flexor system, but reliable effects were seen with brief trains. We adopted as a standard stimulus a train of five 0·1 ms pulses at frequencies of 165 to 200 Hz (i.e. 20–25 ms trains), and adjusted the stimulus intensity to a level that just drove the giant axons one-to-one. The use of exaggerated stimuli may be necessary to reveal connexions in depressed nervous systems such as the isolated abdominal cord, where even monosynaptic electrical transmission appears to be weakened (cf. Furshpan & Potter, 1959). However, trains of the length and frequency we used have been recorded in intact, freely moving crayfish in response to natural stimuli (Wine & Krasne, 1972).

Extracellularly recorded impulses of the six axons in the nerve leading to the slow flexors could be assigned to the individual axons on the basis of recorded amplitude (Fig. 1). Soma size, axon diameter and extracellular spike size are directly related; the cells are labelled according to increasing size from fi to f6 (Kennedy & Takeda, 1965; Wine, Mittenthal & Kennedy, 1974). The second largest cell, f5, is the peripheral inhibitor. In isolated abdominal nervous systems, all of the slow flexor efferents are tonically active except f6 (Kennedy et al. 1966).

Recordings were made conventionally and photographed on an oscilloscope screen. Histograms of tonic motor impulses, with sample times of 10 or 25 ms, were constructed by hand. Samples of spontaneous activity of the motor neurones were taken before each trial and were used to establish baseline activity. In any given preparation, confusion can occur between the identification of fi and f2 and of f3 and f4 since the impulses are similar in both amplitude and activity. Therefore, we sometimes combined the data from either pair of these neurones. This was justified since in all the cases in which all six efferents were clearly distinguishable, fi and f2 responses were very similar, as were the responses of f3 and f4.

The most obvious effect of LG impulses on the tonic system is to excite the peripheral inhibitor (f5). Excitation of f5 by brief LG trains was quantified by stimulating the LG axons to produce a train of five impulses, as described in the Methods. The inhibitor, which has a low rate of spontaneous activity in the isolated nervous system, was recruited during the stimulus train (Figs. 2, 3) and in one preparation was usually fired by a single LG impulse. The inhibitor was excited in every preparation studied, although the strength of the effect varied (Fig. 2). Even when excitation was strongest, the latency to an f5 impulse was too long to be consistent with an exclusively monosynaptic connexion (Fig. 3 B).

To be certain that direct stimulation of an LG axon alone was sufficient to drive f5, we impaled an LG axon with a microelectrode and stimulated it with intracellularly injected current. The peripheral inhibitor was still excited (Fig. 3C), just as with extracellular stimulation.

Identification of f5 was usually based on its spike size, rate of activity and reciprocal firing pattern relative to the tonic flexor motor neurones, but positive identification was obtained in two preparations by making intracellular recordings from the slow muscle fibres concomitant with the extracellular recordings from the nerve (Fig. 3D).

Tatton & Sokolove (1975) demonstrated that f5 inhibited the tonic flexor motor neurones, so it is not surprising that LG impulses, producing f5 activity, inhibit these motor neurones (Fig. 4). Using the same stimulation and recording procedure, inhibition was seen for all motor neurones except f6, which was not spontaneously active. The pair of smaller neurones (f1 and f2) was more weakly inhibited than the pair of larger cells (f3 and f4), but within each pair the cells were inhibited to the same degree. Therefore, in Fig. 4 the data for f1 +f2 are combined and plotted separately from the combined data for f3+ f4. As shown in the inset, the inhibitory link between f5 and the tonic flexors is the most parsimonious way to explain the results. However, it will be shown in a later section that an additional inhibitory pathway is present.

A delayed period of increased motor neurone activity follows, inhibition (Fig. 4). Delayed excitation could result either from excitation arriving via a slow, polysynaptic pathway, or from postinhibitory rebound (see Perkel & Mulloney, 1974, Ref. 6). A delayed period of increased activity is not evident when the motor neurones are inhibited via antidromic stimulation of f5 (Tatton & Sokolove, 1975), which would seem to rule out inhibitory rebound with such stimulation. However, we used orthodromic stimulation which elicited bursts of impulses and activated an additional inhibitory pathway (see below), thus producing more profound inhibition. Hence the experiments are not comparable and the possibility of inhibitory rebound cannot be dismissed. Delayed excitation following inhibition may have been what Roberts (1968b) observed in his experiments. The inhibition itself would have been easy to miss if the background rate were low. However, it is also possible that early inhibition and late excitation are separable events.

The peripheral inhibitor (f5) shows a period of decreased activity for 60–100 ms after being excited by the LGs (Fig. 3); a period of decreased activity in f5 also occurs when it is stimulated antidromically (Sokolove & Tatton, 1975). The tonic flexor motor neurones also show decreased firing after being stimulated antidromically (Sokolove & Tatton, 1975). In general, then, the tonic discharge rate of the flexor efferents is ‘elastic’ and rebounds from either excitation or inhibition.

Although in intact animals the LG axons normally fire only single impulses or brief bursts (Wine & Krasne, 1972), we tested the effects of prolonged trains of LG impulses on the tonic flexor system, since prolonged trains have frequently been used to study the effects of single interneurones in the crayfish cord (e.g. Wiersma & Ikeda, 1964; Kennedy, Evoy & Hanawalt, 1966; Larimer & Kennedy, 1969). When trains of impulses were elicited in the LG axons, f5 was excited throughout the cord and the tonic flexor motor neurones were almost completed inhibited. This is illustrated in Fig. 5 A which shows a continuous recording from a tonic flexor root in response to a 2·2 s long, 100 Hz LG train. Following stimulation, the tonic motor neurones remained silent for about 0·5 s and then resumed firing at almost twice the rate they showed before stimulation. We also stimulated the LG axons with repetitive bursts at interburst frequencies of 10 Hz, which is within the range of tailflip frequencies seen during swimming (Fig. 5 B). This regimen also excited f5 and completely inhibited the tonic motor neurones. Since available evidence suggests that both giant and non-giant escape commands activate similar inhibitory pathways (Wine, 1977 a; Krasne et al. 1977), the experiment shown in Fig. 5 B suggests that the tonic flexors will be inhibited throughout an escape swimming sequence.

Is the pathway involving f5 the exclusive inhibitory pathway between the giant axons and the tonic flexor motor neurones? To find out, we took advantage of the weak excitation of by the giant axons in our experiment and observed the effects of CG bursts that were just short enough to be subthreshold for elicitation of f5 impulses.

In these experiments we monitored both sides of the ganglion, to be sure that f5 was not activated on either side. The results show that activation of f5 is not necessary for central inhibition of the motor neurones by giant axon impulses (Fig. 6). The histograms for f3 and f4 (Fig. 6 A) show strong inhibition, but it is somewhat briefer than usual. Fig. 6B shows that f5 was not excited in most trials, although a slight increment above baseline can be detected when the responses are summed over 90 trials. Thus, the giant axons inhibit the tonic flexor motor neurones via two pathways, one involving f₅ the other not.

An impulse in any one of the four giant axons recruits a large number of interganglionic interneurones, so that both a burst of impulses ascending the cord and a descending burst can be recorded in any connective of the isolated cord for approximately 50 ms after the giant axon impulse (Wine, 1971). These ‘corollary discharge’ interneurones presumably serve many functions, including mediation of the extensive inhibitory effects observed within the escape circuitry (Roberts, 1968a; Kennedy, Calabrese & Wine, 1974; Krasne & Bryan, 1973; Wine & Mistick, 1977; Wine, 1977a, b). It therefore seemed important to determine if corollary discharge interneurones could also mediate inhibition of the tonic motor system.

The paradigm consists of stimulating one section of a transected medial giant (MG) axon while recording tonic motor activity in a ganglion on the other side of the transection (see Methods, and the inset in Fig. 7). Any effect seen on the other side of the cut axon must then be mediated by interganglionic interneurones that were recruited by the MG impulse before it reached the cut.

Impulses in a transected MG axon drove f5 and inhibited the tonic flexor motor neurones in a ganglion on the other side of the transection (Fig. 7), showing that corollary discharge interneurones are sufficient to mediate inhibition. Inhibition occurred both caudal to the block when the rostral portion of an MG axon was stimulated (data not shown), and rostral to the block when the caudal portion was stimulated. We did not test for inhibition in ganglia several segments away from the block, so we cannot judge the length of the relevant corollary discharge interneurones. Also, we did not attempt to see if inhibition could still be obtained in this situation without recruiting f5

A crayfish frequently responds to light touch by flexing its abdomen. Flexor tone is also increased when isolated abdomens are touched (unpublished observations). In isolated cords, an electrophysiological analogue of the postural response can be produced in flexor motoneurones by electrical stimulation of segmental sensory nerves. We quantified the response in order to investigate interactions of escape commands with tonic flexor reflexes.

The reflex response to stimulation of a sensory root is primarily the result of increased firing in f3 and f4 (Fig. 8 A, B), the response in f3 and f4 being weaker, later, and more variable (Fig. 8C, D). The reflex can be obtained by stimulating any abdominal sensory root, but in our experiments the stimulus was always delivered to a 2nd root caudal to.the root being recorded, and was limited to ganglia 2 through 4. The stimulation did not drive the inhibitor, f5, and motor neurone f6 was recruited in only one preparation.

When stimulation of the sensory root was preceded by a brief train of LG impulses, the evoked response in f3 and f4 was reduced by about 55% (Figs. 9, 10). The following experiments indicate that this inhibition is not merely a trivial consequence of the previously demonstrated pathways that inhibit tonic activity, and that neurons afferent to the flexor motor neurones must also be inhibited by the LG impulses.

The following experiments were designed to demonstrate three points: (1) postural reflexes are depressed by repeated stimulation; (2) depression is site-specific, and is therefore likely to be the result of synaptic depression of the primary afferents; (3) depression is prevented by LG-evoked inhibition. All of these effects have previously been demonstrated within the LG escape system (Zucker, 1972b); Krasne & Bryan, 1973), and we shall argue that the parallels between the escape and postural systems suggest that some afferent pathways are shared between them.

When repeated stimulation causes a reflex to wane, transmission is decreasing somewhere in the pathway from receptors to effectors. Because reflexes almost always are evoked by highly convergent sensory input, transmission decrements that occur early in the reflex pathway will result in depression that is restricted to narrow sensory fields, while decrements at later stages will give rise to generalized depression. For example, repeated stimulation of a sensory root at frequencies above 0·2 Hz caused the reflex discharge of tonic flexor motor neurones to decline (Figs. 11–13). If repeated stimulation led to depression as a result of accumulating refractoriness of the motor neurone, we would expect to see complete generalization of depression: that is, responses to all sensory roots would be depressed no matter which one had been stimulated originally. Conversely, if reflex waning were due solely to depression of the synapses made by receptors, then no generalization would be observed: one set of receptors could be stimulated until the reflex was completely depressed, and then stimulation of neighbouring receptors would evoke a full response.

Our results show that depression of tonic flexor reflexes is site-specific. Two kinds of experiments were run to test for generalization. In the first experiment, we stimulated a 2nd root with three shocks spaced at 4-second intervals and recorded the decline in the reflex discharge of f5 and f4 in two different ganglia. We then repeated this test just 4 seconds after the 3rd of three shocks to a contralateral 2nd root, and found no sign of generalization whatsoever, although significant depression of the reflex occurred during each set of shocks (data not shown).

A more rigorous test for generalization was obtained by comparing the responses of tonic flexors to stimulation of each half of a 2nd root that had been divided surgically (Fig. 11). In this experiment reflex responsiveness was tested by five shocks delivered at 3-second intervals to one of the halves of the root. In the test for generalization of depression, five stimuli were first delivered to the upper half root, then the sixth through tenth stimuli, still at 3-second intervals, were delivered to the lower half root (Fig. 11 A). As a control, the lower half of the root was stimulated alone, after at least a two-minute rest (Fig. 11 B). The entire regimen of test and control sessions was repeated 18 times. Fig. 11 shows that each of 5 stimuli caused significant depression of the reflex (first vs. last response: P < 0·001). But, when the test (Fig. 11 A, solid points) and control (Fig. 11B) responses are plotted together (Fig. 11C) the curves are seen to be superimposable. Thus, prior depression of the reflex by stimulating one half root has no effect on the reflex caused by stimulating the other half root. A comparison of receptive fields of afferents in the root (Wine & Hagiwara, 1977) and of receptive fields of sensory interneurones (e.g. Wiersma & Hughes, 1961) makes it almost certain that afferents in the two halves of the root converge on many of the same sensory interneurones. Thus, the lack of generalization strongly suggests that the site of reflex depression is the first central synapse made by the afferents.

In these experiments we compared the amount of depression obtained by repeated activation of the postural reflexes with the depression seen when reflexes were inhibited by prior activation of the LG axons. The logic of this paradigm, which was first introduced by Krasne & Bryan (1973; see also Bryan & Krasne, 1977 a, b), is that inhibition will diminish depression if it is directed at a site prior to any depression-prone sites in the response pathway.

A trial consisted of either a single shock to a sensory root, or a 20 to 30-millisecond train of 5 LG impulses immediately followed by a shock to a sensory root. ‘Habituation’ sessions comprised shocks to the root at 10-second intervals; ‘protection’ sessions were identical except that on the first two trials the shocks to the roots were preceded by five LG impulses. Each of six animals was run through 20 sessions separated by 5-minute intervals. Half the sessions were habituation and half protection, given in random order. For all conditions, the flexor motor response was defined as the total number of impulses in the 500-millisecond period immediately following the stimulus. We stimulated either a 2nd root (which is mainly sensory) or a 4th root of the last abdominal ganglion, which is sensory except for a single efferent.

The results are presented separately for flexor motor neurones f3, f4, and f6 in Fig. 12. Repetitive stimulation of a sensory root caused rapid depression of the evoked motor neurone response (Fig. 12A-C, solid lines); the third response was significantly depressed relative to the first (one-tailed t-test for differences between first and third responses: f3, P < 0·01; f4, P < 0·025; f6, P < 0·01). Motor neurones fi and f2 did not show significant reflex activation and so could not be tested for depression.

Depression was prevented for motor neurones f3, f4, and f6 by the preceding trains of LG impulses (Fig. 12, dashed lines). The difference between the last trial on protection and control runs was significant for all cells (P < 0·025). When considered with our results on site-specific depression, the protection effect strongly suggests that LG-evoked inhibition is acting presynaptically on the terminals of afferents involved in postural reflexes.

An alternative explanation for the above results is that the LG stimulus might act to sensitize or dishabituate the reflex. A control experiment for a single animal is shown in Fig. 13. Habituation and protection runs are as before, but a control condition was added in which the first two stimuli to sensory roots were followed, i s later, by a train of 5 LG impulses (triangles and dotted line). The main finding is that the protection effect cannot be explained as dishabituation: the response on the third trial following two on which LG impulses preceded the root shock (dashed line) is significantly elevated relative to the response on the third trial following two on which LG impulses came after the root shock (dotted line).

LG trains do enhance the reflex, as shown by the upward shift of the response line. We have not examined this effect further, but the data show that sensitization is specific to the reflex -there is no increase in spontaneous activity caused by escape commands (Fig. 13, inset).

The postulated inhibitory pathways from the escape command neurones to the tonic flexor system are shown in Fig. 14. Solid lines indicate monosynaptic connexions and dashed lines indicate functionally defined pathways. Pathway 1 is based on data in Figs. 2, 3, 4, and 7; Pathway 2 is based on data in Fig. 6; and Pathway 3 on results by Tatton & Sokolove (1975, their Fig. 4A). These are minimal pathways.

Tonic reflexes and interactions of escape commands with them can be explained most parsimoniously by assuming that the same mechano-sensory system which excites the LGs (Zucker, Kennedy & Selverston, 1971; Zucker, 1972 a) also excites the tonic flexor motor neurones. There is no direct evidence that the same afferents go to both systems, but all indirect evidence is consistent with the hypothesis. The parallels between the two systems are: (A) both respond to stimuli such as water-borne vibrations that excite the pit hairs (Wiese, 1976); (B) both show low-frequency depression with repeated stimulation (for LGs see Krasne, 1969; Zucker, 1972b; Wine, Krasne & Chen, 1975); (C) for both, depression is site-specific (see Zucker, 1972b); (D) for both, protection against the depression that occurs with repetitive stimulation is afforded by prior activity in the LG axons (Krasne & Bryan, 1973; Wine et al. 1975). Finally, shared sensory input to phasic and tonic systems is not unprecedented, since it has recently been demonstrated that the muscle receptor organs or stretch receptors (Alexandrowicz, 1951; Wiersma, Furshpan & Florey, 1953) project to phasic as well as tonic systems (Wine, 1977C). For these reasons, some of the sensory pathways to the LG are assumed to be shared with the tonic motor circuits, and Pathways 4 and 5 (Krasne & Bryan, 1973) become relevant to tonic reflexes.

Pathway 4 presynaptically inhibits the depression-prone chemical synapses between afferents and sensory interneurones (Krasne & Bryan, 1973; Kennedy et al. 1974) and prevents or diminishes depression. Pathway 5 (Krasne & Bryan, 1973) would enhance the inhibitory effect exerted against reflex activation of the tonics, but should not influence depression in this system. The implications of these pathways are considered below.

Our chief findings are that the postural and escape systems interact in two ways: escape commands inhibit the postural system at all levels from afferents to muscles, and circumstantial evidence suggests that afferent circuits are at least partially shared. We did not test for reciprocal connexions from the postural system to the escape system.

It is surprising that the giant axons are the first identified interneurones shown to influence the slow flexor system. Although many axons can, when stripped from the connectives, drive or inhibit the slow flexors (Kennedy et al. 1966; Evoy & Kennedy, 1967), neither the structure of these cells nor the sensory pathways to them are known. Furthermore, they have usually been stimulated with long trains at high frequency to produce their effects. When the LGs are similarly stimulated, they also produce a clear-cut effect on the slow flexor system (Fig. 5 A) and might be defined in terms of that system if their function as escape command cells were unknown.

Even short bursts of LG impulses inhibit the postural control systems, yet, as was mentioned earlier, the independence of the two neuromuscular systems has long been assumed. In fact, the two motor systems are relatively separate functionally, but this independence is not incidental it is ensured by the extensive central interactions outlined above and previously (Wine, 1977b).

We have not performed any experiments to assess the functional significance of these inhibitory connexions, but we propose that by suppressing reflexes and tone in the postural system the animal increases the efficiency and precision of escape swimming. This proposal depends on the following observations and assumptions: (1) We assume that command-derived inhibition occurs during the repetitive tailflips of non giant escape swimming, which often follows the initial, giant-mediated response (Wine & Krasne, 1972). Evidence supporting this proposal is that the fast flexor motor neurones, which are active during both giant-and non-giant-mediated tailflips, are the origin, via their connexions with corollary discharge interneurones, of at least some command-derived inhibition (Wine & Mistick, 1977; Wine, 1977 a; Krasne, Wine & Kramer, 1977, and unpublished observations). (2) All available evidence (Evoy et al. 1967; Roberts, 1968b) suggests that the postural muscles are too slow to be used synergistically with the fast flexor muscles, and so would usually be out of phase with the fast flexors during swimming. (3) We assume that conflict between slow and fast muscles might introduce significant variability into the movements in spite of the disproportionality in muscle bulk. (In this regard, it should be pointed out that the fast extensors are not especially massive; furthermore, during many tailflips only the thinner, lateral divisions of the extensor muscles are active [J. J. Wine & G. Hagiwara, unpublished results].) If these arguments are correct, the escape systems should also inhibit the slow extensor system. Preliminary evidence suggests that they do (Wine, 1977b; van Lunteren, 1977).

While the above reasoning offers a rationale for inhibition in general, it does not explain why inhibition should be directed at multiple levels of the postural system (Fig. 14). In considering this problem, we first note that no convincing function for peripheral inhibition in the postural system has ever been suggested, so that it is not profitable even to speculate on the significance of Pathway 1. However, we can suggest plausible functions for the remaining pathways. Inhibition of the motor circuits Pathways 2 and 3) is necessary because the motor circuits are spontaneously active, inhibition of afferent pathways to the postural motor circuits (Pathways 4 and 5) will, as already indicated, selectively block reflexes. During escape, inhibition of sensory’ pathways simply reinforces the effect of inhibition at other levels, but it seems likely that under other circumstances Pathways 4 and 5 can be activated selectively, in order to block reflexes while permitting unimpeded activation of motor circuits by central commands.

An additional problem is raised by Pathway 4, which has the special property of protecting the sensory synapses against use-induced depression (Krasne & Bryan, 1973; Bryan & Krasne, 1977a, b). Until now, the significance of both depression and protection of these synapses has been specifically related to habituation of escape behaviour (e.g. Wine, Krasne & Chen, 1975). However, the commonality of the sensory pathways and the consequent response generality of both habituation and protection cause us to view these phenomena from a new perspective. Habituation is usually used in reference to a specific response, such as escape responses in crayfish, gill withdrawal in Aplysia, or leg flexions in the cat. However, in all these systems the depression-prone synapses are either the afferent terminals or stages just beyond the terminals (see Kandel, 1976). Thus the changes are primarily sensory ones and will affect various types of behaviour to the extent that the sensory pathways are shared. The importance of this view is that the functional significance of both synaptic depression and afferent inhibition may elude us if we view them as response-specific.

As our knowledge of different neural networks increases, we can expect to find ever more complex interactions among them. The well-studied escape system is a useful focus for such studies, because its extensive motor involvement and great rapidity demand that it have many inhibitory pathways to ensure its priority. The high priority of escape behaviour is likely to be a general phenomenon: Davis, Mpitsos & Pinneo (1974) have shown that escape interrupts every other behaviour they looked at in the marine mollusc Pleurobranchaea, and widespread inhibitory effects have been shown to emanate from the swimming ‘command neurones’ in Tritonia (Getting, 1977; Taghert & Willows, 1978). However, a crayfish does not always escape if presented with an adequate stimulus (see Davis et al. 1974). In fact, the decision to escape is subject to powerful inhibition that sometimes makes it virtually impossible to evoke a tailflip (Krasne & Wine, 1975). Thus, the likelihood that a response will occur, and a response’s ability to interrupt other types of behaviour if it does occur, are separate aspects of behavioural interactions.

We thank Grace Hagiwara and Cecilia Bahlman for preparation of the manuscript, and F. Krasne, D. Mistick and P. Lennard for criticisms of earlier versions of the manuscript.

The research was supported by National Science Foundation Grant BNS 75-17826 and by a National Science Foundation predoctoral fellowship to J.Y. K.; J.J.W. is an Alfred P. Sloan Research Fellow.