Swimming Patterns and the Activity of Identified Equilibrium Interneurones in the Shore Crab, Carcinus Maenas

In many motor patterns, cyclical activity has been shown to be dependent on functionally connected groups of central interneurones and motoneurones (Selverston, 1980; Roberts & Roberts, 1983). However, few general principles are known about how such cyclical activity may be modulated by sensory feedback or controlled by the nervous system. In arthropods particularly, the variety of identified proprioceptors and other sensory receptors may help demonstrate how sensory elements are necessary and how they operate with their clearly defined inputs to control motor outputs in some types of behaviour (Bush & Clarac, 1985).

Swimming in crabs like Carcinus maenas is a rhythmic motor behaviour which can be extensively modified by sensory information, and, is thus a good model to study sensory modulation of cyclical motor patterns. Living in a turbulent environment near the sea shore, the crab must constantly be able to modify swimming tasks to compensate for water disturbances. A number of anatomically and physiologically described sense organs are thus intensively stimulated; these include water current detectors, leg proprioceptors and exteroceptors, and the equilibrium sensory system. The latter consists of a pair of symmetrical organs, the statocysts, which comprise fluid-filled canal systems analogous in many respects to the semicircular canals of vertebrates (Sandeman & Okajima, 1972; Fraser, 1981). Consideration of this sensory system as a model should therefore be of value in understanding how the vertebrate equilibrium system functions to modulate motor activity. Information from the equilibrium system is carried by large, identified premotor interneurones which are known to be active during swimming (Fraser, 1974b, 1982) and which can be subjected to a wide range of analyses.

Swimming behaviour in crabs was first described by Bethe (1897a,b) as a raising of the last pair of pereïopods (P5) over the dorsal carapace before cyclical out-of-phase rotations of these appendages. This behaviour is easily elicited by lifting the crab from the substratum (Schafer, 1954), tilting the animal head-down, or dropping it through the water (Fraser, 1974b). In a rigidly tethered animal, swimming may occur spontaneously, in response to tilt or when a stream of bubbles from an aerator is released underneath the animal. Although the crab Carcinus maenas is not considered a good swimmer, since its swimming simply reduces its sinking rate rather than provides positive lift, occasionally it can swim vigorously enough to rise from the substratum by beating all of its legs (Hartnoll, 1971). During an earlier study involving intracellular recording and dye injection from oesophageal connective interneurones, results were obtained which suggested the involvement of equilibrium interneurones in swimming behaviour (Fraser, 1974a).

In this paper, using chronic preparations in which nervous elements can be clearly identified and studied during behaviour, we describe for the first time the full swimming pattern in terms of sequences of muscular activity from all the muscles of a leg. Simultaneous extracellular recordings from one of the two oesophageal connectives enabled us to define the involvement of the equilibrium interneurones during the behaviour pattern. In addition, three surgical manipulations (ablation of statocysts, autotomy of pereïopods and cutting a connective) were used to modify the equilibrium interneurone activity and thus to help in understanding how these sensory interneurones may act to control the observed motor outputs.

Male intermoult crabs, Carcinus maenas, of carapace width 6−8 cm were used throughout the study.

A rod was attached to the crab by means of a stirrup screwed into the dorsal carapace above the gill chamber and secured with dental cement (Fig. 1A). The rod was connected to an apparatus designed to move the whole animal in such a manner that it never contacted the tank. Two types of rotational movement were used: (i) horizontal rotations to set the animal in one of three vertical planes, either to the right (+45°) or to the left (− 45°) with respect to the sagittal plane (0°) (see diagrams in Fig. 6 and Table 1) ; and (ii) head-up (HU) or head-down (HD) tilts in one of these three vertical planes (see Fig. 1A).

Electromyographic activity (EMG) was recorded using, for each muscle, a pair of Teflon-coated 120-µm diameter silver wire electrodes glued to the cuticle and inserted via a small hole near the muscle apodeme. Recordings were made from up to four muscles simultaneously, while allowing the crab to swim unrestrained; the remotor muscle burst was used as a reference for the cyclic rotation of the pereïopod.

To monitor the interneurone activity extracellularly, an anterodorsal part of the carapace was removed with care, so as not to damage the median artery to the brain. Anterior and lateral gastric mill muscles on one side were cut and the stomach was displaced to allow visualization of the connective. Either an electrolytically sharpened steel wire or a Teflon-coated silver wire was implanted into or close to the connective. The piece of carapace was then replaced and the wound sealed with dental cement.

Imposed movements in one plane were monitored by means of a potentiometer and battery to record angular displacement. Nerve and muscle activity were amplified, filtered (bandpass 800−3000 Hz and 30−300 Hz, respectively) and recorded on a multitrack FM tape recorder. Selected activities were displayed on a Gould electrostatic recorder or digitized and analysed using a Nascom 3 microcomputer.

The swimming pattern was analysed statistically with a MINC 11.23 computer for burst duration organization and spatially reconstructed with photographic shots taken during the behaviour. These data were used to make a schematic quantification of the spatial rotatory movement of the pereïopod (Fig. 3), with the simplifying assumption of uniform equal forces from each muscle. Each vector represents the fraction of the relative duration (burst duration/cycle period) and the direction of movement produced by each of the four muscles within a cycle, either independently (remotor, promotor, depressor, levator) or as the result of contraction in combination (remotor + depressor, remotor + levator, promotor + levator).

To separate the different types of sensory input to the equilibrium interneurones and to study the effect of the interneurones in the behaviour, three methods were used: (i) leg autotomy to diminish the sensory inputs from the leg proprioceptors; (ii) statocyst ablation to reduce principal inputs to the interneurones; (iii) cutting of one connective to turn off ascending and descending neuronal activities.

The general swimming behaviour described by Bethe (1897a,b) and Hartnoll (1971) involves cyclic sculling movements of the last pair of pereïopods (fifth leg or P5) and, during strong bouts, legs 2−4 beat at the same frequency.

In tethered crabs, swimming occurred spontaneously or could be elicited by tilting the animal head-down in a pitch plane. In both cases, the temporal organization of movements was the same and, for convenience, both types are considered throughout this study as swimming. Sequences of triggered behaviour lasted from less than a second to several minutes with leg oscillation frequencies ranging from 0·5 to 4 Hz.

When swimming was elicited by tilting the animal head-down (HD) by a vertical angular displacement of 30°, the behavioural sequence started with both fifth pereïopods being brought upwards over the dorsal carapace under synchronous contraction of the remotor muscles (Fig. IB). Even though swimming behaviour showed antiphase rotations of fifth legs, few cycles (2−5) were necessary for the right and left remotors to fire in opposition (Fig. IB). In typical swimming (Fig. 1C), antagonistic muscles of a leg segment displayed alternate bursting (i.e. remotor and promotor muscles) as did similar muscles of each pair of pereïopods (i.e. remotor muscles). The maximum occurrence for the beginning of one remotor burst in the period of the other is at 0·5 (Fig. ID).

All leg muscles were active during swimming behaviour, as shown in a statistical analysis performed to display the full motor pattern within a rotation of the leg (Fig. 2A). The muscle burst duration was expressed as a fraction of the period defined by the cyclical activity of the remotor. For each segment, the antagonistic muscles did not fire simultaneously, although burst overlapping could occur. As a result of these muscular contractions, two types of movement occurred during each cycle: (i) the rotation of the appendage caused by the proximal leg muscle contractions; (ii) the orientation of the scull, formed by the three last segments of the pereïopod, in the best propulsive position. In the last segments, the bender and closer muscles burst together and the stretcher burst with the opener, these two sharing the same excitatory motor units (Wiersma, 1961). When the leg was remoted (which caused an elevation of the pereïopod) and extended, the scull was brought up with no resistance against the water; as the leg was promoted and flexed, maximum pressure on the water was applied, slowing down the animal’s sinking. Schematic drawings in Fig. 2B show the spatial positions and orientation of the different segments of the leg during the course of a rotatory cycle. In the proximal part of the pereïopod, one major feature of the pattern was the timing of depressor and levator muscle bursts (Fig. 2A). Although they fired antagonistically, most of their activity occurred during the remotor burst and the resulting association of the activity of the four basal muscles caused the pereïopod rotations that are typical of swimming behaviour.

As in all Portunidae, morphological adaptation for swimming involves modification of the orientation of the basal joint axes with respect to the thorax (Hartnoll, 1971). Each joint has only one degree of freedom and thus allows the remaining distal segments to move in only opposite directions within a plane. Body and thoracico-coxal (T-C) axes lie parallel in the fifth pereïopod (Fig. 3A), and thus the remotor burst elevated the leg relative to the thorax. In contrast, elevation of the leg was controlled by levator bursts in the other pereïopods. The spatially oriented movements produced by the basal muscles were represented by vectors coding time and direction of movement, with the assumption that the same duration of activity leads to equal displacement for each basal muscle (Fig. 3B). This demonstrates how individual and combined bursting of the thoracic and coxal muscles can produce cyclic motion of the appendage.

During strong bouts of activity all the other legs could be involved. However, the fifth pereïopods were the first to start and the last to stop. Recordings of the homologous muscles (promotor) in all the legs of one side of the crab showed (Fig. 4) that head-down tilt in the pitch plane produced in the promotor of the P5 an EMG burst which then continued to fire cyclically. Following slight initial activation, promotors of legs 2−4 were recruited to be active in a metachronal fashion, with the wave spreading from anterior to posterior. It can be seen that a phase-dependency between muscle bursts in legs 2, 3 and 4 exists, but variation in the responses suggests that several patterns of phase relationship are possible, and that no simple resolution is possible without more precise analysis.

All tilt-evoked responses could be obtained in the dark, and in crabs with their eyes blackened, although in this case it was noticeable that the behaviour was less easy to sustain. Covering the anterior ventral part of the carapace with plasticine, to eliminate water-current-sensitive hairs, caused a drop in excitability, though full swimming could still be triggered.

Apparently the five pairs of pereïopods (Pl−P5) are involved in different ways in swimming. The chelipeds (Pl) are not used in the behaviour, the P5 rotates and the others (P2−P4) may or may not beat.

In a crab with only its fifth pair of legs, the full swimming pattern as described above was present, demonstrating the importance of these legs relative to the other pereïopods. Although such animals are not really found in the wild, often one to several pereïopods are missing. As brachyurans, crabs exhibit autotomy, which is an autosectioning of an injured leg at a preformed breakage plane. This discards all but the two basal segments distal to the basi-ischiopodite. In addition, autotomy suppresses most of the sensory feedback arising from the pereïopod, but leaves intact two main proximal proprioceptors: the thoracico-coxal muscle receptor organ (TCMRO; Bush, 1981) and the coxo-basal (C-B) chordotonal organ (CB; Mill, 1976).

To define the involvement of the information arising from the equilibrium system in tilt-evoked swimming it was necessary to minimize leg sensory inputs. As shown in Fig. 5, autotomizing pairs of pereïopods changed swimming motor outputs. With progressive autosectioning of the fifth (Fig. 5A), the fourth (Fig. 5B) and the third (Fig. 5C) pairs of legs, cyclical activity of the basal leg muscles disappeared in the autotomized appendages but remained in the intact ones. In the remaining stump, EMG firing of T-C and C-B muscles consisted of a tonic discharge with occasional non-patterned, low cyclical bursting activity.

In a crab with only its fifth pair of pereïopods intact (P5 crabs), the typical swimming pattern was present while a maximum of leg sensory information had been suppressed. Such animals were used in the experiments described below.

Movements of the animal in the gravitational field are detected by the sensory elements of the statocysts which comprise the equilibrium system. Each statocyst is formed by two nearly closed orthogonal canals, one vertical and one horizontal. The vertical canal plane of the statocyst lies at −45° for the left and +45° for the right vertical canal with respect to the sagittal plane (0°). When the crab is tilted in the pitch plane (0°), sensory elements of both vertical canals are stimulated; when the animal is rotated in a horizontal plane by 45° before the tilt is imposed, only the sensory elements of the vertical canal in the plane of tilt are affected (Fraser, 1977).

In P5 crabs, typical EMG recordings from right and left remotors showed swimming elicited by head-down (HD) tilt in the three planes (Fig. 6). A double form of asymmetry in the behaviour can be seen: (i) in the initiation of remotor bursting and (ii) in the strength of EMG activity of remotors. HD tilt in the plane 0° (Fig. 6B) evoked strong bursts in both remotors as well as symmetrical beginning of cycling in the two muscles before rotations in opposition were displayed. When the crab was rotated horizontally before tilt (− 45°, Fig. 6A; +45°, Fig. 6C), asymmetrical swimming occurred; the remotor bursts of the leg contralateral to the stimulated statocyst vertical canal were stronger and started before the other remotor activity. In this case the first leg to start usually showed prolonged beatings compared to the other leg. If a series of tilts was made the number of remotor bursts per HD displacement tended to decrease.

The delay between HD displacement and the first spike in the first remotor to start was analysed (Fig. 7). The animal was rotated from −45° to +45° in a horizontal plane before being tilted five times in each vertical plane. The leg contralateral to the stimulated statocyst started first, with a delay of about 250-350 ms, compared to 550−650 ms for the ipsilateral leg. In both cases, the delay between the HD movement and the muscle activity is too long to be mediated by descending monosynaptic pathways.

Sensory neurones from the statocysts project in the brain onto specific interneurones. For vertical displacements, stimulation in the plane of each vertical canal leads to the excitation of a group of three equilibrium interneurones whose somata are located in the ipsilateral side of the brain (Fig. 8A). As shown in Fig. 8B, cell A is known to detect HU displacements and sends information to the thoracic ganglion via an axon in the contralateral connective; ipsilateral cells C and D detect HD movements, but the axon of C travels in the ipsilateral connective while the axon of D runs through the contralateral connective (Fraser, 1974a,b). The activity of these six interneurones leads to different firing patterns in the connectives with respect to the direction of tilt (HU or HD) and to the plane in which tilt is performed (−45°, 0°, +45 ° ; Fig. 8C). However, the activity of only one of the two connectives is sufficient to characterize the different patterns of firing of the equilibrium interneurones. Implanted electrodes near one connective allowed us to distinguish the extracellular activity of one to several equilibrium interneurones (Fig. 8C). In general the action potentials generated by the interneurones during movements code integrated angular acceleration, i.e. angular velocity (Fraser, 1977), and often generate a prolonged burst after the displacement. No other interneurones responding to tilt in this way are known or are even visible in extracellular recordings from the oesophageal connective in Carcinus.

In previous work, intracellular recordings have been made from cells A, C and D) in crabs whose legs were free to move. Some results from this study along with brain anatomy of cells A and C from Procion yellow and cobalt filling have already been published (Fraser, 1974b). Further results are given here.

Electrical stimulation (50−100 Hz) of cell C during Procion yellow filling with long-duration current pulses caused swimming movements, consisting of bilateral raising of the fifth pereïopods followed by asymmetrical oscillatory beating, with the contralateral fifth leg beating more strongly. It also caused backward and forward movements of more anterior legs. Eight successful penetrations of cell C were made, leading to unique staining of a unit in the connective, with two cobalt fills showing complete brain anatomy. Penetrations were also made in preparations where the connective was cut posterior to the intra-axonal recording site and a suction electrode was used to record descending activity. This sort of preparation confirmed that the extracellular spikes recorded from cell C (Fraser, 1974b) and the intracellular records were from the same unit. It also confirmed that the cell fired at a frequency of approx. 100 Hz during current application for Procion yellow injection. Although this sort of experiment confirmed singular stimulation of the penetrated cell, it is difficult with the relatively high currents (up to 2µA) used for dye injection to be sure that other cells were not also being stimulated. It is also possible that this sort of stimulation could lead to antidromic activation of other cells which could be contributing to the observed behaviour. Electrical stimulation with metal microelectrodes and stimulation of small bundles of axons from the part of the connective containing the cell C axon elicited swimming movements. More than 40 intracellular penetrations were made from cell A. Electrical stimulation of cell A during dye injection caused swimming but no adoption of the swimming posture as seen with cell C stimulation. In three of the penetrations followed by electrical stimulation, no overt behaviour occurred. No clear behavioural effects were noted on current injection into cell D. These experiments suggest that cells A, C and D are involved in swimming, but in the absence of complete monitoring of the activity in other cells it was impossible to say that swimming is uniquely caused by activation of any one cell. No monitoring of leg movements other than direct visual observation was used during this study with intracellular electrodes.

During swimming in the sagittal plane (0°) muscle activity was recorded along with the right connective activity (Fig. 9A,B). Using the different firing patterns of the equilibrium interneurones in the three planes, a window discriminator was set to discriminate the activity of cell C (Fig. 9C). This was then either fed to an instantaneous ratemeter (Fig. 9A) or to a digital computer to show the instantaneous frequency of firing of cell C (Fig. 9B). The mean firing frequency of cell C between and during triggered swimming is displayed in Fig. 9D. In both spontaneous or HD-tilt-evoked swimming (Fig. 9A,B) cell C was active, and its firing frequency increased just before the behaviour and remained elevated throughout the swimming bout. Recordings of cell D (data not shown) under the same conditions were similar to those shown for cell C.

When swimming was evoked by tilting the animal HD in each of the three vertical planes, −45°, 0°, +45° (Fig. 10), not only did the equilibrium interneurones show different firing patterns, but also the remotor muscles exhibited both (i) symmetrical (0°) or asymmetrical (—45°, +45°) onset of bursting and (ii) variation in the strength of their activity as recorded in the EMGs. For HD pitch in the sagittal plane (0°, Fig. 10B), the two interneurones (C, D) of each side were excited and the onset of the remotor bursts were synchronous; in the sagittal plane, cell A fired during HU tilt. HD tilt at −45° (Fig. 10A) stimulated only the right equilibrium interneurones and cell C alone was recorded in the right connective, while the left pereïopod (L5) started bursting. HD tilt at +45° (Fig. 10C) activated the left equilibrium interneurones (LI in Fig. 8A) and only cell D was recorded in the right connective while swimming started with the right pereïopod (R5).

In addition to these interneurones, giant cell 4 (Fraser, 1974a) responded to water currents during movements although it was silent during tilt in the air. Most of the time cell 4 fired during HU and HD tilts and gave a maintained discharge (Fig. 10).

Swimming bouts could be evoked nearly always by HD tilts and occasionally by HU tilts. Bubbling air under the crab elicited the behaviour, and gave large discharges of activity in both cells C and D, not by stimulation of the statocysts but presumably via leg proprioceptor feedback.

To demonstrate whether or not interneurones C and/or D are important during swimming, we have removed most of the proprioceptors (by autotomizing legs) and/or ablated one or both statocysts. As shown in Fig. 11, full symmetrical swimming evoked by HD tilt in the pitch plane occurred for animals with or without statocysts when all legs were present (Fig. 11A,C) as well as in P5 crabs with intact statocysts (Fig. 11B). However, if all pereïopods but the fifth were autotomized and the statocysts were ablated, swimming and interneurone discharges were abolished (Fig. 11D). Interneurones C and D are clearly important in the generation of swimming.

To delimit the role that cells C and D play in eliciting swimming, one of the connectives was cut. Because the axons of the ipsilateral interneurones C and D travel through different connectives, it is possible, by cutting one of the connectives and ablating one statocyst, to limit descending activity to a single interneurone and hence to see its effect on swimming. In addition, since leg joint proprioceptors can drive swimming via the equilibrium interneurones, we have reduced the sensory information as near as possible to only the statocyst inputs, by autotomizing all but the fifth pereïopods. However, this experiment is limited by the fact that sectioning a connective also reduces other ascending and descending activities and usually weakens the evoked responses to tilt. Tests were performed on a group of P5 crabs with the right connective cut or intact, tilted in the three planes −45°, 0°, +45°; the same tests were repeated after ablation of one statocyst. The responses were scored and are displayed in Table 1.

When both statocysts were present, tilt in the asymmetric planes (− 45°, +45°) triggered the contralateral leg to the stimulated statocyst, while both legs were equally activated for tilt in the sagittal plane (0°). If the right connective was cut, stimulation of the right statocyst (−45°) led to virtually no activation of the left pereïopod, while stimulation of the left statocyst (+45°) still activated the right leg. In this case, only the activity of left cell C interneurone was still descending towards the thoracic ganglia.

These data suggest that the equilibrium interneurones C are responsible for the onset of the P5 leg responses, and that their firing activates the contralateral fifth pereïopod. To confirm the validity of this hypothesis, ablation of one of the two statocysts was made and the tests were repeated. The hypothesis predicts: (i) that when the right statocyst is present alone, virtually no activation of the right fifth pereïopod should be seen, and the contralateral leg should be activated at 0° and —45° while the right connective is intact or cut; (ii) that when the left statocyst is present alone, opposite effects should be seen. The data presented in Table 1 confirm that the activity of the equilibrium interneurones C is necessary to obtain the onset of the fifth leg activity in response to HD tilts, and that descending information which travels via the ipsilateral connective down to the thoracic ganglion should activate neural elements controlling swimming of the contralateral leg.

In P5 crabs, tilt in the plane of one statocyst vertical canal elicited swimming, which started with the contralateral P5 leg. This initiation of behaviour was symmetrical in its expression and one or the other fifth leg could move first when appropriate sensory elements of the contralateral statocyst were affected by displacements. However, in quite a number of cases we did not find this expected behaviour: i.e. in some crabs HD tilt in the plane of the vertical canal of one statocyst evoked a strong contralateral leg response, but stimulation of the other statocyst vertical canal led to at most a weak effect in the other contralateral leg. Animals showing such a type of behaviour were crabs which had lost one or more legs by natural autotomy. Furthermore, these crabs when left on a substrate displayed a turning tendency comparable to crabs with cut connectives (Bethe, 1897b): turning clockwise for left and anticlockwise for right connective cut.

To show this asymmetrical behaviour, crabs with one or two legs missing on one side were selected. Only those showing a black seal to the autotomized stumps (indicating autotomy had taken place more than 2 days previously) were used. All remaining legs except fifth legs were autotomized. Each crab was gently lifted up and let fall to the bottom of the tank. This stimulated locomotor movements. Turning tendency was scored. Each crab was tested 16 times. Table 2 shows results obtained from four crabs with right legs autotomized, four with left legs autotomized and, as control, four crabs with no previous autotomy. When a crab had old autotomized right legs, it tended to turn anticlockwise; in contrast, it turned clockwise if left pereïopods were missing. The development of the turning bias was not immediate, as indicated by the lack of bias in freshly autotomized crabs.

The major aim of this paper was to investigate the swimming behaviour of Carcinus in relation to the involvement of the equilibrium interneurones. This was done by using different types of stimulus such as tilt in the plane of a vertical canal of one statocyst which fires the equilibrium interneurones in a defined way.

In addition, swimming behaviour has been described, for the first time, in terms of motor outputs, so that it may be compared with walking behaviour, the other well-known motor pattern displayed by the same appendages (Clarac & Coulmance,. To produce adequate organized motor outputs in any behaviour, an animal needs to be able to collect and integrate a certain amount of data, such as the direction of displacement of its body in space or the position of all the leg segments in relation to each other, to the animal and to the environment. Regulation of motor acts is known to involve some of the numerous sense organs (see Bush & Laverack, 1982; Clarac & Barnes, 1985; Evoy & Ayers, 1982; Mill, 1976; Sandeman & Okajima, 1972; Vedel & Clarac, 1976) which continually send information to the central nervous system. In this study, special attention was given to a set of polymodal premotor interneurones, the equilibrium cells (Fraser, 1974a,b), which receive their main inputs from the sensory cells of the statocysts and from some of the leg proprioceptors.

Carcinus maenas has a specific swimming pattern identical with respect to muscle burst organization in spontaneous swimming and in swimming evoked by HD tilt in a vertical plane. Swimming is performed by typical out-of-phase sculling rotations of the fifth pair of pereïopods (P5). However, both fifth pereïopods display identical muscle patterns within a leg rotation.

Other crabs, Portunus holsatus (Hartnoll, 1971) and Callinectes sapidus (Spirito, 1972) use this type of swimming merely to maintain their position in the water. They also display sideways orientated swimming in which muscle patterns are different in the two fifth legs within a leg rotation. Asymmetrical movements of pereïopods have been well described in sideways walking in crabs. Each side acts differently: one leads and the other trails as the animal respectively pulls or pushes on its appendages to move its body (Clarac & Coulmance, 1971; F. Clarac, F. Libersat, H.-J. Pflüger& W. Rathmayer, in preparation). A major difference between walking and swimming can be demonstrated by analysing muscle activity. All the muscles of the fifth legs are active in swimming and thus all joints are used, allowing satisfactory rotations of the pereïopod. In sideways walking, only muscles operating joints controlling dorsoven-tral movements are involved. However, in the less common form of oblique walking all the muscles can be used, but the motor pattern in this case differs greatly from the former locomotor patterns. The temporal organization of EMG bursts in the two commonest types of locomotory behaviour, swimming and sideways walking, shows how dissimilar they are. In swimming, remotor and promotor bursts are essential, as are bursts in the depressor (firing just at the beginning of the remotor burst) and the levator (active during the second part of the remotor and the beginning of the promotor activities). This pattern of discharge in these four muscles causes the rotation of the appendage. In sideways walking, remotor and promotor show very weak or even no activity (F. Clarac, F. Libersat, H.-J. Pfliiger & W. Rathmayer, in preparation), but levator and depressor bursts are very important in relation to the activity of the flexor and extensor muscles operating at the mero-carpopodite joint.

A striking point is that muscles innervated by a given set of motoneurones can be involved in motor patterns which differ greatly, with respect to both muscle burst duration and timing relationships. Since this part of the pathway is common, it is necessary to suggest that a certain number of nervous elements must be specific to each type of behaviour.

Swimming is an example of a rhythmic motor behaviour in which movements of the legs and the body in a fluid environment will cause proprioceptive feedback. Hydrodynamic receptors, leg sense organs and the equilibrium system all contribute to the regulation of the behaviour. However, the activity of the giant cell 4 interneurone, which collects inputs from hydrodynamic structures of the anterior part of the cephalothorax, does not seem necessary during swimming.

Two types of leg sense organs are involved in different ways. First, leg proprioceptor discharges are used in the fine control of motor outputs in order to produce an appropriate coordinated behaviour. However, during a rhythmic behaviour such as swimming, sensory information does not vary much from cycle to cycle unless a leg hits an obstacle, or when the angular position of the animal’s body is altered and as a consequence produces a redistribution of forces over the leg system. Second, cuticular exteroceptors, the funnel-canal organs (FCO; Gnatzy, Schmidt & Rornbke, 1984) need to be silent for the animal to swim, whereas their rhythmic firing in walking is used for step regulation (Zill, Libersat & Clarac, 1985). When an animal is swimming, electrical stimulation of the FCO afferents inhibits the remotor bursting and thus stops the behaviour, exactly as happens when the legs of the animal contact the substratum (Bévengut, Libersat & Clarac, 1986). It is possible that these sense organs have a role analogous to that played by contact receptors in insect flight, since neither of these types of behaviour can be elicited when leg contact is present (Ritzman, Tobias & Fourtner, 1980).

For the animal, the maintenance of balance is necessary for any coordinated behaviour. Body movements are detected mainly by the equilibrium system, provided in crabs by the statocysts. Each statocyst comprises two orthogonal canals disposed in such a way that any displacement in one of the three orthogonal planes of space is encoded unequivocally by differential stimulation of their sensory elements (Sandeman & Okajima, 1972). Thus, the equilibrium system with a pair of statocysts detects the six possible directions of movements in the gravitational field. We have studied only the effect of vertical displacement. Variation of the plane of tilt (+ 45°, − 45°, relative to the sagittal plane 0°) leads to asymmetrical initiation of motor outputs in swimming. However, if a crab is tilted HD at +45° or −45° and it is not swimming, the legs are positioned asymmetrically. This spatial orientation of the legs resembles that seen in typical righting responses.

Equilibrium reactions involve, in addition to the statocyst organs, sensory receptors in the legs (Neil, 1985) as well as visual inputs (Reichert & Rowell, 1986). In the shore crab, swimming can be elicited when the eyes are covered, but sensory information from the statocysts and leg proprioceptors converges onto four large polymodal interneurones identified in each side of the brain (Fraser, 1974a,b, 1975). Each one codes for one direction of movement in one of the three orthogonal planes of space, and forms part of a negative feedback loop involved in swimming behaviour (Fraser, 1982).

In this work, only the equilibrium cells encoding for vertical displacements (cells A, C and D) were studied in tilt-evoked swimming. In Carcinus maenas, HD tilt in a vertical plane evokes firing of cells C and D as well as swimming if the crab’s pereïopods do not contact the substratum; however, HU tilt in a vertical plane, firing cell A, can lead to different forms of swimming. In the crab Scylla serrata, swimming-like behaviour is elicited by H U tilt while little reaction is seen for HD tilt (Fraser, 1975). In tilt-evoked swimming in Carcinus, two distinct phases can be distinguished. First, the initiation of swimming behaviour can occur when leg exteroceptor activity is absent and equilibrium interneurones are firing. However, the onset of behaviour in tilt-evoked swimming is characteristic. If both statocysts are stimulated (0° to the sagittal plane) swimming starts with both fifth legs. Tilt in the plane of only one statocyst vertical canal always causes asymmetrical onset of swimming, starting with the contralateral leg to the stimulated statocyst. Second, the continuation of the behaviour displays typical alternating activity of pairs of symmetrical muscles as well as of the two fifth pereïopods. This could be due to the leg proprioceptor inputs being integrated not only by local interneurones but also by the equilibrium cells and causing the long-lasting discharge of the equilibrium interneurones throughout the swimming sequences. Autotomizing legs Pl−P4 or ablating statocysts decreases the behaviour and the bursting activity in the interneurones. When the legs are autotomized and the statocysts are ablated, both the tilt-evoked behaviour and the interneuronal activity are abolished.

In rhythmic behaviour such as insect flight (Wilson, 1961; Robertson, 1986), the concept of central pattern generators is now being reconsidered, taking into account the role of sensory afferents in the patterning of motor outputs (Wendler, 1983; Pearson, Reye & Robertson, 1983). Pearson (1985) concluded that it is necessary to include central and sensory neurones in the functional network of cells producing rhythmical motor outputs to give a correct description of the behaviour. In crabs, sensory integrators like the equilibrium interneurones might be considered as part of the functional network in the locomotory system.

Giant interneurones in cockroaches (Ritzmann et al. 1980) can trigger walking if leg contact exists and flight if it is absent. In the same way, the crab equilibrium interneurones are involved in swimming and walking (Fraser, 1982), and can provide a possible mechanism for a central drive of these behaviours as trigger interneurones. In the crayfish Procambarus, four pairs of directional equilibrium interneurones control leg movements and uropod steering but combined activity of these cells is necessary to ensure bilaterally organized movements: a single interneurone can neither initiate nor trigger the equilibrium reactions since each one connects with only some of the motoneurones involved (Takahata & Hisada, 1982a,b). In Carcinus, the detailed connections between the equilibrium cells and the motoneurones are not yet known. Our experiments suggest that in swimming behaviour, cell C activity is strongly involved in the onset and the continuation of the behaviour, and that firing of cell C alone, when the other equilibrium interneurones are silent, leads to the same results. However, its effect does not seem to occur by monosynaptic pathways since a long latency exists between the onset of cell C firing and the onset of the leg muscle activity. Although electrical stimulation of interneurone C reliably causes swimming, the possibility of simultaneous activation of other cells by antidromic activation or current leakage cannot be ruled out. It seems best to consider interneurone C as an important element in the cellular network controlling swimming.

In swimming behaviour, both fifth pereïopods have the same motor pattern, although out-of-phase. Previous autotomy of one or several legs leads to an asymmetrical expression of the behaviour in experiments performed on crabs with only fifth pereïopods, one side displaying the expected motor pattern while the other shows only little muscular activity. One possible explanation for this asymmetry is that in the absence of a leg, the remaining pathways increase in strength to compensate for those lost. This may underlie not only the turning tendency of crabs but also the failure to record bilaterally symmetrical rotations of pereïopods in experiments where crabs with previously autotomized legs were used.

This work has been supported by a grant from the European Science Foundation (ETP/TW/475). We thank Dr P. Dickinson and Dr S. L. Hooper for critically reading the manuscript.