Motor Patterns for Horizontal and Upside-Down Walking and Vertical Climbing in the Locust

Walking in a complex environment requires continuous adjustments of the underlying motor patterns to changing substratum conditions. In spite of the wealth of information that has been gathered about the central neurones and proprioreceptors involved in walking movements of insects, there are few descriptions of the activity of muscles during different kinds of walking.

Most studies on insect walking have concentrated on cockroaches (Wilson, 1966; Pearson and Iles, 1973), locusts (Hoyle, 1976) and stick insects (Bässler, 1983), which in most cases were tethered or walking on treadmills to allow controlled manipulation of sense organs (Wilson, 1965; Usherwood et al. 1968; Cruse et al. 1984; Bässler, 1983; Büschges et al. 1994) or intracellular recordings from central neurones (Godden and Graham, 1984; Ramirez and Pearson, 1987; Wolf, 1992). In such experiments, however, the question remains open as to which kind of walking or locomotory behaviour the recorded motor patterns really represent, bearing in mind that motor neurones and muscles involved in walking can also be recruited during kicking, swimming, flight, stridulation and righting (Pflüger and Burrows, 1978; Heitler and Burrows, 1977a,b). In this respect, the interpretation of pharmacologically induced motor patterns recorded from isolated nervous systems is also difficult (Ryckebusch and Laurent, 1993) without comparable data from freely behaving insects. For instance, the reciprocal muscular activity in a deafferented cockroach limb corresponds to ‘righting behaviour’ rather than to ‘walking’ (Zill, 1986). Thus, recordings from unrestrained moving animals are also clearly important in the study of the nervous control of coordinated movement.

Previous studies of free-walking insects (locusts: Hoyle, 1964; Runion and Usherwood, 1966; Usherwood et al. 1968; Usherwood and Runion, 1970; Godden, 1975; Laurent and Hustert, 1988; Periplaneta americana: Pearson, 1972; Pearson and Iles, 1973; Delcomyn and Usherwood, 1973; Krauthammer and Fourtner, 1978; Carausius morosus: Cruse and Pflüger, 1981; Gryllus bimaculatus: Laurent and Richard, 1986) have mainly concentrated on horizontal walking and, thus, have not examined changes in the motor patterns due to different environmental requirements such as climbing or walking upside-down hanging from a branch. Furthermore, they have only addressed the activity of a few selected leg muscles. In locusts, for example, recordings have been made from the flexor and extensor tibiae muscles during free walking (Burns and Usherwood, 1979). In the metathoracic leg, these two muscles also play important roles in jumping, swimming and kicking (Pflüger and Burrows, 1978), and in this segment alone do they control retraction and protraction of the leg during a step cycle (Burns and Usherwood, 1979). In the pro-and mesothorax, muscles of the more proximal leg joints, especially those of the subcoxal–trochanteral and the coxal–trochanteral joints, are clearly more important for controlling swing (protraction) and stance (retraction) (Burns, 1973; Hustert, 1983). In spite of this, these proximal leg muscles have only been recorded in the foreleg of Gryllus bimaculatus during tethered walking on a styrofoam ball (Laurent and Richard, 1986) but never in locusts.

This study examines the motor patterns of metathoracic subcoxal muscles during various different walking situations. Since the metathoracic leg in particular is involved in many different motor behaviours, such as jumping, kicking and swimming (Pflüger and Burrows, 1978), and some of the coxal muscles are even used during flight (Wilson, 1962; Ramirez and Pearson, 1987), the hind subcoxal joint should be considered as a ‘multifunctional’ joint. Furthermore, most of the coxal muscles respond to stimulation of leg sense organs (Bräunig and Hustert, 1985a,b), which suggests that they are able to adjust their activity to changing substratum conditions. Our results present a complete account of the motor patterns of the musculature of this joint and its functional contribution when unrestrained locusts walk horizontally, climb vertically or walk hanging upside-down from a branch.

Adult locusts, Schistocerca gregaria Forskål, of both sexes were taken from our crowded laboratory colony at the university of Berlin. Differential electromyographic (EMG) recordings were obtained by inserting a pair of 70 cm long steel wires (diameter 40 µm), insulated except for the tip, into a muscle through small holes in the cuticle and fixing them with beeswax. The extra load of approximately 100 mg could be neglected when compared with changes in load due to a daily food intake of 0.5–2 g. Muscles were named and numbered according to Snodgrass (1929) and Albrecht (1953; see Fig. 1), and their activities were recorded when the animals walked freely on a horizontal surface, climbed a vertical wall or walked hanging upside-down from a branch (diameter 1.5 cm). Surfaces were the same for each orientation. The motivation to walk and, furthermore, to walk in a straight line could be increased by reducing visual input by covering the compound eyes and ocelli with a mixture of a two-component dental adhesive (Scutan) and a black powder. This is in agreement with earlier observations on locusts (Moorhouse et al. 1978) and cockroaches (Spirito and Mushrush, 1978). A comparison with intact, untreated animals showed that the ‘blindfolding’ had no measurable effect on the recorded motor patterns. The myographic results presented below are based on at least 10 recordings from each muscle in different animals, each of which performed all three walking tasks (walking sequences were between 50 cm and 70 cm long). All recordings presented in one figure are from one animal; different figures are from different animals.

Recordings were made from up to four muscles simultaneously and at room temperature (20–25 °C). The exact location of the electrodes was determined post mortem either by dissection or by the Prussian Blue method (Grundfest et al. 1950). To detect possible cross talk in the EMG patterns, recordings were made simultaneously from neighbouring muscles. The recordings were free of cross talk if the electrodes were placed in the middle of the muscle. All evaluated EMG patterns were free of cross talk with the exception of the recordings from M130. Owing to the very small size of this muscle and its difficult anatomical location, recordings were never completely free of cross talk. Therefore, we were not able to distinguish different units in recordings of M130 (Fig. 6), but, since this muscle is innervated by only one excitatory motor neurone (Fig. 3), we are confident that we have correctly evaluated the phase of the step cycle during which the one motor unit of M130 was recruited.

In addition, 15 animals were video-taped simultaneously with the electromyographic recordings and the video recordings were analysed frame by frame (25 frames s⁻¹; or 50 half-pictures s⁻¹) to measure exact protraction and retraction times. Synchronisation of video and electromyographic recordings was achieved by giving an acoustic signal on the audio channels of both recorders. Since both were video systems, which had very similar running times and, therefore, the timing error was neglectable.

The subcoxal joint can be considered to be a ‘ball and socket joint’ with its pleural articulation acting as the pivot. Fig. 1A,B shows all the muscles that can be seen when the animal is fixed on its back with the ventral cuticle removed. Fig. 1C is a schematic drawing of the inner view of the metathoracic subcoxal joint showing all the muscles investigated. The muscles are attached to the coxa at four different points, classified here as anterior (M118, M121 and M125), dorsal episternal (M126), posterior (M119, M120, M122, M123 and M124) and ventral sternal (M130) insertion points. For anatomical reasons, it can be assumed that the first and the second pronator-extensor (M127 and M128) and the depressor-extensor of the hind wing (M129), which all insert at the dorsal part of the coxa (not shown in Fig. 1C), do not lift the joint during walking. The arrangement of M126 is more complex than described previously (Snodgrass, 1929; Albrecht, 1953). It inserts on the dorsal margin of the episternum and on the anterior face of the pleural ridge and is typically fan-shaped. Additionally, it has a third prominent bundle which inserts anteriorly, close to the spiracle at the meso/metathoracic border. This part thus lies on top of M125, the other abductor of the coxa (Fig. 1C).

Electrical stimulation of each muscle revealed how it moves the coxa. Fig. 1D shows schematically how each coxal muscle involved in walking moves the joint (similar inside view to that of Fig. 1C). The resulting positions of the coxa and the femur as seen from the outside are shown in Fig. 1E. The anterior rotator (M121) pulls the anterior part of the coxal joint ventrally and towards the thorax (Fig. 1D,E). Thus, M121 is a good candidate for support of the swing phase or protraction (the phase of the step cycle during which the leg moves forward relative to the body and the ground; Delcomyn, 1971). The contractions of the posterior rotators (M122, M123 and M124) cause an analogous movement at the posterior insertion point (Fig. 1C,D). They thus create a force which simultaneously pulls the leg towards the thorax and presses it towards the substratum (Fig. 1E), which is probably required during the stance phase or retraction (during which the leg moves backwards relative to the body but not relative to the ground; Delcomyn, 1971). The second remotor (M120) pulls the posterior part of the joint dorsally (Fig. 1D), which is best described as a posterior levation of the coxa (Fig. 1E). Contractions of the abductors (M125 and M126) produce a force which can be divided into two vectors. The first is a levation of the anterior coxa. The second is an abduction of the posterior, ventral part of the joint, which is mainly realized by contractions of M125 (Fig. 1D), whereas M126 mainly supports the anterior levation of the coxa (Fig. 1D). Thus, each abductor lifts and abducts the ventral coxa (Fig. 1E). A contraction of the adductor coxae (M130) moves the ventral, sternal insertion point towards the thorax (Fig. 1D). Because the pleural articulation lies on the dorsal side of the coxa, the resulting movement is an adduction of the coxa (Fig. 1E).

The promotor (M118) and the first remotor (M119) are omitted from Fig. 1D,E, as they are not recruited during walking (see below). However, electrically induced contractions of M119 cause a posterior levation of the coxa (as described for M120) and those of M118 cause a similar movement at the opposite side of the joint, best described as an anterior levation.

Fig. 1F shows the minimum and maximum angles between the lateral margin of the body and the long axis of the femur as seen from a dorsal view (α), between the femur and the tibia (β) and between the long axis of the animal and the long axis of the femur as seen from a lateral view (δ) for each walking situation investigated. The measured angles are very similar for horizontal and upside-down walking, but differ during vertical climbing (based on a Student’s t-test). In the latter situation, the leg is abducted more strongly (as indicated by the lower value of α_min, Fig. 1F) and the whole position of the leg shifts posteriorly. Thus, both the angle between the femur and the tibia (β) and the angle between the long axis of the animal and the long axis of the femur(δ) increase (Fig. 1F).

Since the hind femur is held almost parallel to the body during walking, protraction and retraction movements can be correlated with flexion and extension of the tibia. Thus, most of the propulsive force is due to the extension of the tibia, whereas the coxa mainly functions to lift the joint during protraction (Burns, 1973). Simultaneous myographic recordings and video recordings of 100 steps of each of the three walking situations investigated revealed that, during horizontal walking and vertical climbing, the flexor tibiae (M136) burst can be used as a good indicator of protraction. Hence, we define one step as the time from the start of one flexor burst to the start of the next flexor burst. M136 was recorded simultaneously with the subcoxal muscles in each case and was taken as a reference against which to determine their phase relationships.

In all animals tested, the step duration is shortest during horizontal walking (range 200–1300 ms; mean ± S.D. 500±120 ms) and longest during vertical climbing (range 1000–4000 ms; mean 2000±950 ms). The mean step duration during walking upside-down on a branch is 900±600 ms. The duration of protraction is remarkably constant in all walking situations (Fig. 2A–C), but still shows a weak correlation with step duration (horizontal walking, r=0.4; vertical climbing, r=0.53; upside-down walking, r=0.46; the probability that non-correlated distributions would show these correlation coefficients is lower than 0.001, based on a Student’s t-test for significance). The duration of retraction strongly correlates with the step duration (horizontal walking, r=0.98; vertical climbing, r=0.97; upside-down walking, r=0.99, Fig. 2A–C). This shows that, in all three walking situations, an increase in walking speed is mainly achieved by a marked decrease in retraction time. This corresponds with earlier observations (Burns, 1973) on horizontal walking in locusts.

In the electromyograms, different motor units can easily be distinguished by their spike amplitudes, here defined as small and large motor units of the respective muscle (see Figs 4–10). For a given muscle, this differentiation into different units is consistent in all animals investigated, even if the electrode positions are altered. It is generally assumed that, in electromyograms, the large-amplitude units are caused by fast and the small-amplitude units by slow motor neurones (see the fast and slow extensor tibiae recordings of Hoyle and Burrows, 1973). The innervation patterns of all muscles investigated are shown in Fig. 3. In most recordings, except those of M120 (see below), the number of distinguishable motor units corresponds with the number of excitatory motor neurones in a muscle.

Fig. 4 shows the activity of the flexor tibiae (M136) during five steps of horizontal walking (Fig. 4A, lower trace), two steps of vertical climbing (Fig. 4B, lower trace) and one step of upside-down walking (Fig. 4C, lower trace). Since M136 is innervated by at least nine excitatory motor neurones (see Fig. 3), we were not able to distinguish between single identified motor units. However, the recordings reveal that, during horizontal walking, a large (fast) motor unit is recruited during protraction; this unit fires only sporadically during vertical climbing and never during upside-down walking (Fig. 4). During vertical climbing, a very small unit is active during retraction; this unit usually does not fire in the other two walking situations (Fig. 4B, lower trace). During upside-down walking, the onset of the flexor burst coincides with the beginning of protraction, but activity lasts significantly longer than actual protraction (because of this, the duration of protraction during upside-down walking was measured from video analysis). This implies that the flexor and extensor tibiae co-contract during early retraction.

Figs 4–10 show representative pairs of recordings from the flexor tibiae (M136) and all other coxal muscles during each of the three walking situations. Fig. 12 summarizes the results of all the recordings. The activity patterns of each investigated muscle in each of the different walking situations are described below.

Simultaneous recordings of the first abductor (M125) and the flexor tibiae (M136) of one representative animal in the three different walking situations are shown in Fig. 4. M125 is innervated by three excitatory motor neurones (Fig. 3). This corresponds with the number of motor units observed in the myograms. During horizontal walking, the largest unit, unit 1, is active exclusively during retraction, whereas a smaller unit, unit 2, bursts only during protraction (Fig. 4A, upper trace). The smallest unit, unit 3, is mainly active during retraction and is partly masked by unit 1. Thus, latencies and burst durations were measured only for the other two units. During vertical climbing, all motor units of M125 start to fire at the beginning of protraction, with the burst ending during late retraction (Fig. 4B, upper trace). During upside-down walking, both smaller units are masked to a great extent by the large one. In this walking situation, M125 is active only at the transition between retraction and protraction (Fig. 4C, upper trace). This is exactly the phase of the step cycle during which M125 is silent during horizontal walking and vertical climbing.

The second abductor (M126) is also innervated by three motor neurones (see Fig. 3). Correspondingly, we recorded three different motor units in our EMGs (Fig. 5D). Since the very small unit, unit 3, fires irregularly during the whole step cycle, we were not able to measure its latency or burst duration. We assume that its most likely function is to maintain a certain muscle tonus. In contrast to the first abductor (M125), units 1 and 2 of M126 are always active during protraction in all three types of walking situation. The main difference between horizontal walking (Fig. 5A, upper trace) and both vertical climbing and upside-down walking (Fig. 5B,C, upper traces) is that, for the latter walking situations, unit 1 always shows a burst of at least six muscle potentials, as opposed to a maximum of three potentials during horizontal walking (500 steps from 10 different animals; the average number of unit 1 potentials is 1.7).

Representative recordings of the adductor (M130) are shown in Fig. 6. Because of its anatomical location and its very small size (Fig. 1A), the quality of the recordings suffers from movement artefacts which are strongest during fast horizontal walking sequences. The changes in the amplitude of the muscle potentials (Fig. 6) are probably caused by changes of the electrode positions during contractions of the muscle. Nevertheless, activity in the one excitatory motor neurone that innervates M130 (see Fig. 3) occurs exclusively during retraction, irrespective of the walking situation.

The anterior rotator (M121) is divided anatomically into an anterior and a posterior part (M121a and M121p, respectively; Fig. 1A), each innervated by different motor neurones (Hoffmann and Pflüger, 1990). Careful electrode placement allows independent recordings from each part of M121. Fig. 7 shows that, in all walking situations, the small unit of M121a (middle traces) is tonically active during the whole step cycle. Although M121a clearly shows changes in its spiking frequency (Fig. 7A–C, middle traces), we were not able to measure whether these changes are dependent upon phase or walking speed because they occurred very irregularly (even within one animal) and the data were not consistent when averaged over 10 animals. During horizontal walking, the large phasic unit of M121p fires sporadically during protraction (Fig. 7A, upper trace). During vertical climbing, recruitment of this unit is stronger and it bursts during protraction (Fig. 7B, upper trace). This observation is similar to the motor pattern adjustment of the large unit of M126. Dramatic changes in the motor patterns of M121p become obvious during upside-down walking (Fig. 7C, upper trace). Activity of the large unit continues into the second half of retraction, the phase of the step cycle during which it is never recruited during horizontal walking and vertical climbing (Fig. 7B,C, upper traces).

Each posterior rotator muscle (M123 and M124) is innervated by three motor neurones (Fig. 3). Irrespective of the walking situation, all motor units of M124 are active for most of the duration of retraction (Fig. 8A–C, upper traces). Only during upside-down walking does activity sometimes start during late protraction (Fig. 8C).

M123 shows two clearly distinguishable motor units during horizontal walking and vertical climbing. Like those of M124, both are exclusively active during retraction (Fig. 9A,B, upper traces). During upside-down walking, the onset of the burst overlaps with the previous protraction and activity ends during the beginning of the following protraction. Furthermore, M123 sometimes shows a third muscle potential of very large amplitude (Fig. 9C, unit 3, marked with an arrow) during upside-down walking. This unit was recorded in three of ten animals, in which it occurred in 10 % of the steps (300 steps measured) irrespective of the walking speed.

The remotors (M119 and M120) and the promotor (M118) insert at the coxa and at the wing base. They can thus move the wings and the subcoxal joint and are accordingly often referred to as ‘bifunctional’ (Wilson, 1962; Bräunig and Hustert, 1985a,b; Ramirez and Pearson, 1987). During flight, they all function as wing elevators (Wilson and Weis-Fogh, 1962), but during walking we have found that only the second remotor (M120) is active. This muscle is innervated by six excitatory motor neurones (Fig. 3). During horizontal walking and vertical climbing, at least two units of M120 are active. In both walking situations, these units are active during the transition phase between retraction and protraction (Fig. 10A,B, upper traces). During horizontal walking, the large unit, unit 1, fires one or two potentials per step, whereas during vertical climbing it bursts. Dramatic changes in the motor patterns become obvious during upside-down walking (Fig. 10C, upper trace). Here, M120 bursts during the transition between protraction and retraction and during early retraction, which is exactly the opposite phase of the step cycle compared with its activity patterns during the other two walking situations.

During flight, M120 is activated once or twice per wing-beat cycle (Fig. 10D). However, for any one animal the amplitudes of these extracellularly recorded potentials during flight are 2–3 times larger than those of any of the units recorded during walking (compare upper and lower traces, Fig. 10D). It thus appears that different motor neurones are recruited for M120 during walking and during flying.

Neither M118 nor M119 shows activity during any of the walking situations investigated. Fig. 11A,B shows simultaneous recordings of the flexor tibiae and each of these muscles during horizontal walking. In Fig. 11C,D, recordings from the same animals during flight are shown. In contrast to all investigated muscles which are active during walking, M118 and M119 are not innervated by a common inhibitor (CI) neurone (Fig. 3).

Fig. 12 summarizes our data in a ‘muscle activity score’. We averaged the burst durations and latencies (with respect to the onset to the flexor tibiae burst) of each muscle within each walking situation over at least 100 steps obtained from at least 10 different animals and normalized them to the step duration (period).

In the present paper, we show that the motor patterns are relatively constant for a given walking situation, but are markedly altered in different conditions, such as horizontal walking, vertical climbing and walking upside down (Fig. 12). Combining information about the joint’s anatomical and structural properties (Fig. 1) with the results from the myographic recordings (Figs 4–10) reveals the role of the subcoxal joint during walking and its functional adjustments to different load and gravity distributions in the respective walking situations. This information can now be used in future studies as a reference for assessing the contribution of sense organs to different walking motor patterns and for interpreting naturally or pharmacologically induced motor activity in deafferented and restrained preparations.

Walking speed was slowest during vertical climbing (0.5 steps s⁻¹), which is not surprising since here, in contrast to horizontal and upside-down walking, the animal’s weight has to be moved in the opposite direction to gravity. The average step frequency during upside-down walking (1.5 steps s⁻¹) is significantly lower than during horizontal walking (2.5 steps s⁻¹). This is probably associated with the greater amount of tonic muscle work, including co-activation of antagonists (see below), that is required to maintain a correct posture during upside-down walking.

The duration of protraction as a function of the period (step duration) varies significantly in the three different walking situations, being 20±4 % for horizontal walking, 12±6 % for vertical climbing and 28±6 % (mean ± S.D.) for upside-down walking (see Fig. 12). During upside-down walking, an increased proportion of the swing phase may be required for the processing of sensory information to ensure that the tarsi grasp the branch securely. This is supported by the results of Pearson and Franklin (1984), who showed that locusts search for a ‘secure footing’ when walking on rough terrain and, therefore, increase the variability of the angles between the different leg limbs and the thorax. The lower proportion of protraction time during vertical climbing is in agreement with observations on the stick insect. Here, retraction is prolonged and the next swing phase is delayed when the animal walks up a slope (Spirito and Mushrush, 1979) or when a small clamp is attached to the trochanter (Bässler, 1977), so that (presumably) campaniform sensilla signal that the leg is still being loaded during late retraction (load usually decreases the further the leg is moved posteriorly). This indicates that, when walking against an increased load (such as during vertical climbing), force may be exerted so that the proportion of retraction is increased.

In all walking situations, changes in step duration show a strong correlation with changes in the duration of retraction, but only a weak correlation with the duration of protraction (Fig. 2). This corresponds to earlier observations on lobsters (Clarac and Chasserat, 1983), scorpions (Pearson, 1981), stick insects (Graham, 1972), locusts (Burns and Usherwood, 1979) and cats (Pearson, 1976). Our experiments show that this common principle of increasing the stepping frequency mainly by reducing retraction duration is used not only during horizontal walking but also during vertical climbing and upside-down walking (Fig. 2).

In disagreement with earlier findings in Melanoplus differentialis (Thomas) (Wilson, 1962), and in the homonomous mesothoracic muscles of locusts (Ramirez and Pearson, 1987), the promotor (M118) and the first remotor (M119) are not active during any of the walking situations investigated but only during flight (Fig. 11). In contrast to the mesothoracic leg, where swing and stance movements are mainly due to promotion and remotion of the coxa, these movements correspond to flexion and extension of the tibia in the hind leg. Thus, activity of M118 and M119 might not necessarily be required during walking. This is further supported by the results of Bräunig and Hustert (1985b), which show that these muscles do not react to the stimulation of coxa–trochanteral mechanoreceptors. Furthermore, in contrast to all subcoxal muscles investigated involved in walking, neither M118 nor M119 is innervated by a CI neurone (Fig. 3). The innervation by a CI neurone is probably required to enable fast motor units to move the joint during protraction by reducing the muscle tonus and the joint’s ‘stiffness’ (Wiens, 1989; Rathmayer, 1990; Wolf, 1992), which are necessary to stabilize the joint during retraction (see below).

The only muscle of the metathoracic subcoxal joint that is active during both walking and flight is the second remotor (M120). Since the amplitudes of motor units observed during flight are much larger than those measured during walking (Fig. 10D), we assume that M120 is excited by different sets of motor neurones which are active in only one of the two types of locomotory behaviour. This might explain why M120 receives six excitatory motor neurones in contrast to the three that innervate the unifunctional muscles M118 and M119 (Fig. 3). This observation contradicts the findings of Ramirez and Pearson (1987) on motor neurones innervating the homonomous mesothoracic muscles. They observed that, without exception, all motor neurones supplying bifunctional muscles were rhythmically active during flight and walking.

In all recordings of subcoxal muscles (Figs 4–10) we were able to identify different motor units by their spike amplitude. The number of motor units in a given muscle generally corresponds to the number of excitatory motor neurones innervating the muscle, which enables us to discuss the contribution of ‘slow’ and ‘fast’ units to different walking tasks. In a given walking situation, the motor patterns are remarkably constant (Fig. 12, small standard deviations), irrespective of the walking speed. This contrasts with observations on cockroaches, where in coxal depressor and levator muscles a large unit produced more potentials during faster walks (Pearson, 1972). Our results on Schistocerca gregaria demonstrate that changing mechanical conditions, due to a slightly different posture of the leg and to different loadings of the joint, but not to the walking speed, are mainly responsible for changes in the bursting patterns.

Since M136 is innervated by at least nine excitatory motor neurones (Fig. 3), we were not able to distinguish between single identified motor units, but the recordings reveal which types of motor units are preferentially used in the different walking situations. During horizontal walking, large phasic units are required for generating fast, strong contractions of the flexor tibiae muscle. The very small motor units, which are active during retraction in vertical climbing (Fig. 4B, lower trace), might function to maintain a constant muscle tonus to cope with the increased load in this walking situation. Motor units of very large amplitude are always active during horizontal walking, only sporadically active during vertical climbing and never active during upside-down walking (Fig. 4). As mentioned earlier, the latter walking situation requires the most accurate placing of the leg, and thus it might be necessary to recruit slow motor neurones in order to increase the tonic part of contraction.

During horizontal walking and vertical climbing, the flexor burst corresponds exclusively to protraction, whereas during upside-down walking we observed co-activation of the flexor and the extensor tibiae during early retraction. This probably produces a certain ‘stiffness’ in the femur–tibia joint, which might be necessary to maintain a constant distance between the body and the substratum. A similar type of joint stabilization, due to co-activity of antagonistic muscles, was described by Krauthammer and Fourtner (1978) for slow walking sequences of the cockroach.

For any one walking situation, some coxal muscles generate joint movements, others stabilize the joint for appropriate transduction of the propulsive force to the substratum and some can fulfil both functions by recruiting different units at different times during a step. Within a given walking situation, the recruitment of motor units is remarkably constant but it alters as soon as the walking situation changes.

During protraction, the coxa has first to be lifted and abducted, so that the tarsus does not drag on the substratum during the flexion of the tibia and, second, to be pulled anteriorly. Such movements are realized by the second remotor (M120), the abductors (M125 and M125) and the anterior rotator (M121; Fig. 1D,E). During horizontal walking and vertical climbing, all these muscles are active (Fig. 12). The main difference between these walking situations is that, during climbing, the coxa has to be moved in the opposite direction to gravity. Therefore, the large units of M126 and M121 are recruited more strongly (Fig. 12). During upside-down walking, the protraction movements are basically the same but it becomes unnecessary to lift the coxa actively since the leg can automatically be lifted (relative to the thorax) by gravity. Therefore, M120 is not recruited during protraction (Fig. 12) but during retraction. In this walking situation, the thorax has to be kept at a constant distance from the branch and the joint posture must therefore be stabilized and sometimes corrected during retraction. The posterior levation of the coxa caused by contractions of M120 pulls the joint, and simultaneously the thorax, towards the branch. The smaller units of M120 might function to hold the coxa in a fixed ‘stiff’ position, whereas the large unit, which either bursts or spikes sporadically (Fig. 10), corrects this posture.

During retraction, the coxal muscles serve two main functions. First, the joint has to be pulled back and adducted until it reaches the posterior extreme position. These movements are realized by the posterior rotators (M123 and M124) and by the adductor (M130; Fig. 1D,E) which, in all walking situations, are active during retraction (Fig. 12).

Second, the joint has to be held in a fixed position during retraction to ensure that the powerful extension of the tibia can be transduced into forward motion. In this respect, Snodgrass’ hypothesis (1929) that the hind coxa is held relatively stiffly during a step cycle is interesting. This ‘stiffness’ during retraction may be caused by tonically firing units of muscles inserting on the coxa at opposite ends. Since the thoracico-coxal joint has ball-and-socket characteristics, with the pivot at the dorsal part (see Fig. 1D), one could assume that muscles inserting ventrally, anteriorly and posteriorly of this point are activated at the same time in order to hold the joint posture fixed. This is precisely the case. During horizontal walking and vertical climbing, an antagonistic force to that of the retractor muscles (M123, M124 and M130) described above is created by contractions of the small unit of the anterior rotator (M121a) and the large unit of the first abductor (M125; Fig. 1D,E), both of which are tonically active during retraction (Fig. 12). During upside-down walking, both abductors (M125 and M126) are exclusively active during early protraction. They function to abduct the leg so that the tarsi can be loosened from the branch before the tibia is flexed. Thus, in this walking situation, the first abductor is not used to produce joint ‘stiffness’ but, instead, the large unit of the anterior rotator is recruited during retraction and thus it co-contracts with its antagonists M123 and M124 (Fig. 12C). This co-activity might again be necessary to produce a certain ‘stiffness’ in order to stabilize the joint.

In all walking situations, this ‘stiffness’ has to be reduced towards the end of retraction before the start of the next protraction. In locusts, the three CI neurones are active shortly before and during protraction (Runion and Usherwood, 1968; Burns and Usherwood, 1979; Wolf, 1992). They probably enable fast motor units to move the joint by reducing the muscle tonus and its ‘stiffness’ (Wiens, 1989; Wolf, 1992). The observation that different units of a muscle fire antagonistically during leg movements, for instance the small and large units of M125 and M121 (Figs 4, 7), emphasises the difficult task of naming the subcoxal muscles according to their function.

We conclude that the motor control of the subcoxal joint musculature is actively adjusted to meet the different requirements of different walking tasks. This emphasises the significance of proximal leg joints for walking in a complex terrain. In any one walking situation, some motor units of coxal muscles serve to realize joint movements and others to stabilize the joint for appropriate transduction of the propulsive force to the substratum; some can fulfil both functions. Within a given walking situation, the recruitment of the motor units is remarkably constant, but it alters as soon as the walking situation changes. Thus, different units of one muscle can serve quite different functions, which can all independently be adjusted to changing substratum conditions.

The results presented in this paper form part of the diploma thesis of C.D. The support of H.J.P. by the DFG (Pf 128/6-4) is gratefully acknowledged. We thank Professor M. Burrows (Cambridge) and Dr P. A. Stevenson (Berlin/Leipzig) for many helpful comments and lively discussions. We also thank Ms Phil Maher for help with the English translation.