Stick Insect Locomotion On A Walking Wheel: Interleg Coordination of Leg Position

Interleg coordination is an intrinsic feature of insect locomotion. Numerous studies of different insect species have described temporal features of this coordination in the form of preferred phase relations and timing restrictions (for reviews, see Wilson, 1966; Bowerman, 1977). These temporal features have been used to construct model networks which postulate specific kinds of interleg interaction (Wendler, 1968 ; Baess-ler & Graham, 1978; Graham, 1977; Cruse, 1981). Nevertheless, the intersegmental interactions which can be demonstrated in dissected preparations are quite weak in comparison to intraleg reflexes (Wilson, 1965; Delcomyn, 1971; Graham & Wendler, 1981).

The apparent weakness of interleg reflexes in dissected preparations, plus the realization that intraleg reflexes vary considerably between quiescent and active insects (Baessler, 1976; J. Dean, personal observation), led to our undertaking a further, detailed analysis of the stick insect. This analysis was designed to take advantage of an automated method for registering leg position (Wendler, 1978). Applying regression techniques, the analysis began by seeking relationships among different walking parameters which might explain some of the variability apparent in individual parameters. The first part of this report describes one such relation which serves to influence the position of leg placement at the end of protraction. The second part of the report shows how this behaviour is affected by externally imposed manipulations. The results confirm and extend findings by Cruse (1979), who used a different approach which confined his investigation to single, discontinuous steps. The third part of the report examines some of the proprioceptive factors involved in this interleg coordination.

Data were collected from adult female stick insects (Carausius morosus) walking along a light walking wheel (radius 16·4cm, rim width 3·0cm, inertial mass equivalent to 1·5gm). For each experiment, an insect was glued along its dorsal surface to a metal rod which was then clamped at a fixed distance above the wheel. Experiments were conducted in dim light. If the animals did not walk spontaneously, they were stimulated to do so either by puffs of air or light touches to their abdomens.

The anterior-posterior position of each leg was sampled at 10 ms intervals using a photoelectric method described in detail elsewhere (Wendler, 1978). Basically, this method employs a photocell to record the times at which each femur interrupts a light beam which moves alternately on each side of the animal along a path parallel to the body axis. The resulting measurement corresponds to the tangent of the angle between femur and body axis as projected onto a plane perpendicular to the light beam This measurement principally records movement of the coxo-thoracic joint; it approximately proportional to the anterior-posterior tarsus position as long as the Temur does not rotate and the lateral position of the tarsus is constant, as it is when the tarsus is always placed on the edge of the wheel surface. However, it has been shown for free walking animals that the relation between tarsus position and femur angle is not linear for extreme leg positions (Baessler, 1972). In our measurement units, the proportionality factor also changes with the lateral position of the light beam: this position varied from one animal to another but remained constant for each animal. A second photoelectric sensor attached to the wheel axis measured the absolute rotation of the wheel in 1 mm units.

These data on wheel motion and leg position were recorded on tape (15 ips FM) in a pulse-coded format and processed later by a PDP 11/40 minicomputer. Using the computer, protraction and retraction endpoints were defined semi-automatically for each leg. (Under the experimental conditions described here, these two endpoints correspond closely with the Anterior and Posterior Extreme Positions - AEP and PEP - used by other authors, e.g. Baessler, 1972, Cruse, 1976, but they were not defined as the extreme leg positions.) Subsequently, regression and phase relations were computed by combining as needed: (1) the position and timing of the protraction and retraction endpoints of interest, (2) the corresponding position of other legs, and (3) the phase of any imposed periodic stimulus. Steps were excluded from this analysis if the period of the right rear leg differed by more than 20 % from its value in either the preceding or the subsequent step. The multiple regression programme incorporated IBM SSP subroutines. These subroutines calculated regression coefficients b and their standard errors as well as a Student’s t for the null hypothesis: b = 0. In addition, they provided the simple correlation coefficient r for one independent variable or the multiple correlation coefficient R for more than one independent variable. R², or r² for simple regression, is equal to the fraction of the total variance in the dependent variable which can be attributed to the regression.

Leg restraint was achieved by placing the distal femur through a small wire ring (diameter 1·5 mm) clamped in a manipulator. This arrangement allowed tarsus and tibia to move freely but the femur was held high enough to prevent the tarsus from contacting the wheel. The tarsus was not provided with a substrate; this omission was necessary in order to avoid the disruption caused by the tendency of adjacent legs to grasp such a support instead of the wheel, a tendency which itself reflects the phenomenon to be investigated.

Motion was imposed on a leg by loosely attaching the distal femur to a balsawood arm connected to a pen-motor (Brush Co.). The femur was held high enough to prevent tarsus contact with the wheel while the pen-motor rotated the leg forward and backward around the coxo-thoracic joint. The vertical axis of rotation for the balsawood arm was positioned directly over this joint. The axis of normal joint rotation is slightly inclined from vertical (Graham & Wendler, 1981), but it was necessary to use a vertical axis in order to prevent the apparatus from interfering with the measurement of leg position. This arrangement sufficed to move the tarsus in a horizontal arc just above the substrate and within its normal path during stepping.

Modification of sensory input centred on two sets of mechanoreceptive hairs: (1) four hair rows located on the caudal surface of the coxa (Baessler, 1965) and (2) hairplates cxHPv and cxHPd located on the cranial margin of the coxa (Wendler, 1964 : here, the labels trHP, cxHPv, and cxHPd are introduced to replace the original designations BF1, BF2, and BF3 of Wendler, 1964; this change is made for the sake of clarity and increased uniformity of terminology, e.g. Pringle, 1938; Pflueger, Braeunig & Hustert, 1981). Depending on joint position, individual hairs in each set are successively covered and deflected by folds of joint membrane. Retraction of the coxa increases the number of deflected hairs in the hair rows; protraction does the same for the hairplates. Phasic input from these hairs was eliminated by first exposing the full set of hairs and then applying a thin layer of wax/rosin mixture to fix the hairs in an upright position. This method necessarily introduced a potential mechanical impediment to the normal movement of the joint.

The results presented here focus primarily on factors influencing the endpoint of leg protraction. A secondary theme concerns consequences of this spatial coordination for other step parameters. Three types of results will be presented in turn: (1) the results of a regression analysis applied to data from intact, unrestrained animals - an analysis which examined naturally occurring variation in protraction endpoint; (2) changes in protraction endpoint caused by restraining one leg or by subjecting it to sinusoidal movement; and (3) results of modifying sensory inputs from leg proprioceptors thought to mediate these effects.

The regression analysis began with a number of spatial and temporal parameters in a search for regular features of stick insect walking. The clearest result of this search concerned step by step variation in protraction endpoints. The initial finding was a positive correlation between protraction endpoint and its starting position (= endpoint of the preceding retraction).

For the front legs, this was the most common relation but it was not very consistent. Correlation coefficients reached 0·77 but were under 0·40 in most cases and were not significant for about a third of the walks examined.

For rear and middle legs, further tests with additional factors showed that variation in their protraction endpoints was more strongly and consistently correlated with another parameter - the momentary position of the adjacent anterior leg at the time protraction ended. For example, the position of the middle leg at the time that the rear leg completed its protraction typically accounted for between 16% and 75 % of the total variance in the rear leg protraction endpoint (r values between 0·4 and 0·85 ; e.g. Fig. 1 and Table 1 : Intact values). In our measurement units for the animals in Table 1 (Intact values) and six others not tabulated, the slope of the regression of rear leg protraction endpoint on simultaneous middle leg position was generally between 0·65 and 0·8; these values remained much the same when the measurements were converted to degrees of arc between femur and body axis. The relation between middle and front legs was weaker than that between rear and middle legs. Variation in the position of the front legs accounted for only 4 % to 49 % of the total variance in the protraction endpoint of the ipsilateral middle leg (r values between 0·2 and 0·7; e.g. Table Intact values plus six additional animals not tabulated); regression coefficients were between 0·2 and 0·4 in most walks.^*

Despite these appreciable correlations, the link between protraction endpoint and the simultaneous position of the forward leg is not immediately obvious during normal walking because both these parameters vary within quite narrow ranges. The ratio of the standard deviation in step endpoint to mean step amplitude is usually between 8 % and 15 %. The relation does become obvious when for some reason the target leg is in an unusual position, for example when omitting a protraction (Fig. 2).

Correlation alone does not establish a causal link between variables. However, Fig. 2 aptly illustrates the temporal sequence in which the relevant leg positions are established and thereby suggests the direction of the underlying coordinating influence.

As the caudal leg protracts towards the adjacent forward leg, the latter is nearing the end of its support phase in a stance which typically was assumed more than half a step cycle previously. Thereafter, the position of this target leg is coupled to that of all the other supporting legs and cannot be varied independently. This temporal order implies that the extent of the protraction is controlled by signals related to the position of the more forward leg.

This hypothesis was tested by imposing different positions on one middle leg as observing the stepping of the unrestrained legs, especially that of the ipsilateral rear leg. In the unrestrained animal, the normal position of the middle leg when the rear leg completes protraction is slightly in front of its retraction endpoint. Restraining the middle leg behind the mean retraction endpoint caused the ipsilateral rear leg to shorten its protraction (Fig. 3: PRO3 for fB = fixed back versus C = control). Observation of the animals during the experiment revealed that the rear leg was not blocked by the middle leg. In most steps, the rear leg completed its protraction and stepped down without hesitation and without touching the middle leg. Only occasionally did the rear leg rotate and stroke down across the distal tibia of the restrained middle leg, sometimes repeating the stroke before stepping down behind the middle leg and beginning retraction.

Moving the middle leg forward caused the rear leg to protract farther forward, but extreme forward positions elicited smaller shifts in rear leg protraction than did intermediate positions (Fig. 3: PRO3 for middle leg conditions fM and fF, fixed middle and forward, versus fB, fixed back).

These changes in protraction endpoint occurred without parallel changes in retraction endpoint. On the contrary, for three of the six animals the rear leg retraction continued farther to the rear when the middle leg was fixed farther forward. In all animals, the net result was an increase in mean step amplitude for the ipsilateral rear leg (Table 2). This amplitude change for rostral versus caudal middle leg restraint amounted to 20–40% of the control step amplitude. This increase in amplitude usually was accompanied by an increase in mean protraction duration (Table 2).

Restraint of a middle leg also affected the stepping of the ipsilateral front leg. Restraint in the caudal part of the middle leg’s range elicited changes which differed from one animal to another. In four of the six animals, both the retraction and protraction endpoints shifted caudally by similar amounts, leaving step amplitude unchanged (Fig. 3). The other two animals showed a rostral shift of protraction endpoint which exceeded the shift in retraction endpoint, leading to an increase in step amplitude. Although protraction amplitude did not decrease for any animal, its mean duration decreased for all six.

Changing the position of the restrained middle leg elicited parallel changes in both protraction and retraction endpoints of the ipsilateral front leg for five of the six animals. Step amplitude was unaltered in all, but protraction duration remained shorter than that of control walking.

Such a differential change in the step amplitudes of ipsilateral rear and front legs must also alter step timing. When one leg decreases its step amplitude relative to that of another, it must either spend more time in protraction or it must step more often. Table 2 shows that for most animals protraction duration decreased for the smaller amplitude steps taken during caudal middle leg restraint; therefore, it can be expected that the rear leg must step more frequently. This expectation was realized: the ipsilateral rear leg stepped at a higher frequency relative to the ipsilateral front leg when the middle leg was restrained in caudal positions and this change was reversed when the middle leg was moved rostrally (Table 3).

Restraining one middle leg in different positions did not systematically affect the step characteristics of the contralateral legs. As a result, changes in the step amplitude and step frequency of the ipsilateral rear leg were reflected in its step frequency relative to the contralateral rear leg (Table 3: L3/R3). Normally the two rear legs make roughly the same number of steps. However, restraining one middle leg in a caudal position increased the relative step frequency of the ipsilateral rear leg (Table 3 : L3/R3). This change was reduced or reversed when the restrained middle leg was placed in a rostral position. Since the step amplitude of the ipsilateral front leg did not change with the position of the restrained middle leg, there was at first no reason to expect changes in its step frequency relative to the contralateral front leg. Nevertheless, this pair did show alterations opposite to those of the two rear legs (Table 3 : L1/R1). Evidently, the alteration in rear leg stepping, plus possible direct influences of the restrained middle leg, created distortions which altered the step ratios of several other leg pairs. On the contralateral side, the middle and rear legs continued to step virtually one to one. Relative to these two legs, the contralateral front leg in three animals slightly increased its step ratio for caudal middle leg restraint. However, most of the change in step ratio for the front legs apparently reflected an alteration relative to the contralateral side as a unit.

The previous experiments revealed discrete changes in leg coordination when one leg was fixed at different positions for each walk. A further set of experiments demonstrated the continuous nature of these changes by altering the position of the restrained leg within a walk. For these tests, a single middle or front leg was moving sinusoidally through its step range. Two frequencies of imposed movement were tested : one was about one third that of normal stepping, the second was considerably slower. As expected from the previous results, movement of one middle leg at both frequencies elicited parallel shifts in the protraction endpoint of the ipsilateral rear leg (Fig. 4A, B). However, the retraction endpoint of the rear leg showed smaller, less consistent, shifts. In some step sequences, the rear leg shortened its retraction and made quicker, smaller steps when the middle leg was to the rear. At other times, its retraction endpoint moved in parallel with the protraction endpoint and step period remained nearly constant. This continual variation in step amplitude caused changes in timing indicated by the failure of the rear leg to step one to one with the other unrestrained legs.

Imposing a slow sinusoidal movement on one front leg elicited a similar pattern of changes in the ipsilateral middle leg (Fig. 4C). As expected the protraction endpoint of this leg followed the momentary position of the front leg. The retraction endpoint showed parallel shifts of lesser magnitude, leading to an expansion of step amplitude for rostral positions of the front leg. Unlike the result for movement of the middle leg, these oscillations in middle leg step amplitude and period were replicated in the remaining legs and all five unrestrained legs maintained one to one stepping.

The previous results have established that a leg adjusts its protraction endpoint according to the location of the adjacent forward leg. To accomplish this adjustment, the central nervous system (CNS) needs information on the position of the target leg and possibly that of the protracting leg as well. Information on the coxal angle is available from two groups of external proprioceptors located on the coxa itself : four sows of mechanoreceptive hairs spread along its caudal surface (Baessler, 1965) and two hairplates found near the cranial, proximal margin (Wendler, 1964). (Internal chordotonal organs may provide additional information on joint angle, Hustert, 1978; Braeunig, Hustert & Pflueger, 1981.) These external proprioceptors are known to function in intraleg control loops. Ablation of the hairplates results in over-extension of protraction (Wendler, 1964); ablation of the hair rows, in conjunction with bending the hairs of the hairplates, disturbs the proper completion of retraction (Baessler, 1977). Although Baessler noted frequent overstepping by the rear leg adjacent to an operated middle leg, neither of these previous studies explicitly measured changes in adjacent, unoperated legs.

The last set of experiments examined interleg coordination of stepping when input from the external proprioceptors was altered. For each animal, proprioceptive input was modified in two stages. In the first operation (01), wax was applied to the four coxal hair rows; in the second (02), additional wax was added to cover the hairplates cxHPv and cxHPd. Three possible functions of the sensory input were tested.

The first function to be considered was whether sensory information from the target leg is involved in setting the average protraction endpoint of the adjacent caudal leg. The hypothesis was as follows. Fixing the caudal hair rows in an upright position (01) should remove an input which normally signals increasing retraction. Therefore, the CNS should attribute to that leg a position forward of its actual position. Based on this faulty assessment, an operated middle leg should retract farther to the rear and the ipsilateral rear leg should protract farther forward. Only the second part of this prediction was consistently fulfilled. In all six animals, the rear leg’s protraction endpoint shifted forward (Fig. 5: PRO3 for 0l versus C), despite the fact that the corresponding position of the middle leg generally remained unchanged (Fig. 5: upper axes). The rear leg’s retraction endpoint moved forward in all six animals. Step amplitude did not show consistent changes but protraction duration was significantly increased in four animals (Fig. 6: L3 duration for 01 versus C). The predicted caudal shift in the middle leg’s retraction endpoint only occurred for three animals. However, the middle leg’s protraction endpoint moved caudally in five of the six animals. The net effect was a shortened step amplitude in all six animals (Fig. 6: L2 amplitude for 01 versus C). Nevertheless, in three of the six, the operated leg required a longer time for this smaller protraction (Fig. 6 : L2 duration for 0l versus C). Removal of hair row input alone sometimes interfered with the proper termination of retraction. In these steps, the operated middle leg failed to release the wheel upon reaching a posterior position. Instead it was pulled free by a build-up of force from the other legs (Fig. 7).

Subsequent immobilization of the rostral hairplates (02) introduced contradictory signals in that absence of activity from the hairplates normally indicates caudal leg positions whereas the continuing absence of activity from the caudal hair rows indicates forward positions. It was expected that this second operation would counteract any bias induced by the first operation and cause middle and rear legs to revert towards their normal step relations. For all four animals, the mean protraction endpoint of the operated leg shifted forward relative to values for 01 as expected (Wendler, 1964) ; the retraction endpoint moved forward for three animals (Fig. 5 : PRO2 and RET2 for o2 versus 0l). Extreme overstepping, which follows ablation of the hairplates alone (Wendler, 1964), did not occur. Step amplitude was slightly increased in all four animals and protraction duration remained longer than control values (Fig. 6). The ipsilateral rear leg reacted as expected in three of the four animals: its protraction Endpoint shifted caudally despite the rostral shift in the corresponding position of the middle leg (Fig. 5: PRO3 for 02 versus 01). In all four animals, rear leg protraction duration was longer than in control walking (Fig. 6) and retraction endpoints remained forward of control values (Fig. 5).

The second feature of interleg coordination likely to depend upon control signals from the target leg was the step by step coordination of protraction endpoint. For the same animals described above this feature was studied using both a regression analysis to test within the small range of variation in normal walking and leg restraint to test a wider range of target positions. In four of the six animals, immobilizing the four caudal hair rows alone was sufficient to reduce the correlation between middle leg position and rear leg protraction endpoint (Table 1 : 0l values). In the two exceptions, subsequent immobilization of the hairplates reduced the remaining correlation (Table 1: 02 values). After the fixation of the hair rows, large changes in the leg’s position caused by restraint near the forward and rear margins of its step range were still conveyed to the rear leg (three animals tested). After subsequent fixation of the hairplates, even these large differences in middle leg position had little or no effect on the protraction endpoint of the rear leg (two animals tested).

The third feature of interleg coordination to be investigated concerned the contribution of the protracting leg’s own proprioceptors to its adjustment to different target positions. This function was tested using the animals described above by examining the correlation results for the leg pair in which the operated middle leg served as protracting rather than as target leg. In two additional animals, proprioceptors on a rear leg were modified. Results varied among the eight animals but half continued to show a significant correlation between protraction endpoint of the operated leg and the position of the adjacent forward leg (Table 1 : five of eight animals after 0l, four of six animals after 02). For the two animals with modified rear leg out, the protraction endpoint of this leg was also examined in relation to restraint of the ipsilateral middle leg. Neither operation eliminated the adjustment of the re¹ leg’s protraction endpoint in response to restraint of the ipsilateral middle leg in extreme positions. However, signals from the two sets of receptors interacted in opposite ways with signals from the target leg. Absence of hair row input opposed large rostral protractions and decreased the change in endpoint resulting from a fixed change in the position of the middle leg (Fig. 8: 01 fB and fF). Subsequent immobilization of the hairplates removed this bias and permitted larger than normal changes in protraction endpoint for the same change in middle leg position (Fig. 8: 02 fB and fF).

The results described here demonstrate that during locomotion the endpoint of middle and rear leg protraction is guided by the instantaneous position of the adjacent forward leg. Our method of recording leg position only measured the angles of the coxo-thoracic joints in the target and protracting legs. However, the function of this guiding response is only clear when its effect on tarsus placement is considered. Observation of animals walking on the wheel as well as inspection of the footprints of free-walking insects shows that each trailing leg steps to a position slightly behind that of the leg in front of it : middle and rear legs in turn step close to the footprints of the ipsilateral front leg (Wendler, 1964). Cruse (1979) demonstrated an active mechanism for this phenomenon. He began with a stick insect standing with one middle leg on an unmoving platform and the other five legs on a treadwheel ; he then recorded the position to which the tarsus of the ipsilateral rear leg moved in its initial protraction. By altering the position of the platform in relation to the insect’s body, he demonstrated that the rear leg adjusted both the lateral and longitudinal aspects of its protraction in order to place its tarsus a constant distance behind that of the middle leg. Our measurements demonstrate that this same guiding response occurs during continuous locomotion. During walking, a forward leg typically is nearing the end of its support phase when the leg behind it finishes protraction; therefore, the effect of the adjustment in protraction endpoint is to direct each leg to a location where the more rostral leg has already found a supporting substrate.

In our experiments, the protracting leg did not compensate fully for changes in target leg position. With few exceptions the regression coefficients were less than one ; they were much less than one for the relation between middle and front legs. These coefficients suggest that compensation is less than complete; however, this interpretation of the results is subject to three qualifications. First, our method recorded only the protraction of the coxo-thoracic joint; this is only one of the movements involved in tarsal placement. Our measurement is strictly proportional to longitudinal tarsus position only as long as there is no rotation of the leg and the tarsus steps to the same lateral position. Neither condition is strictly obeyed. Baessler (1972) found that for free walking stick insects, tarsus position is linearly related to femur angle only in the middle range of step movement. Deviations can be expected for extreme positions, such as the protraction and retraction endpoints measured in our experiments. Second, the curvature of the wheel, the restriction of lateral movement of the body, and the removal of the need for the insect to support its own weight may all reduce the compensation which might normally be expressed. Third, each leg normally avoids stepping on or in front of the leg ahead of it; this is a reasonable precaution to avoid obstructing the ensuing protraction of the forward leg. If the protracting tarsus does contact the adjacent leg, it immediately withdraws and moves to the rear (Graham, 1979). Therefore, the errors in compensation can be larger in the caudal direction. The degree of compensation as measured by a regression analysis will be reduced because the variance is not constant for all target locations.

Deviation from complete compensation was also evident in the experiments in which a middle leg was restrained in extreme points of its range. Here the method of restraint may well have hindered optimal performance. The lack of restraint on the tibia allowed considerable variation in the tarsus position which probably is the actual target of the protracting leg. Moreover, the target leg distal to the femur did not signal the presence of any substrate. Nevertheless, partial compensation did occur and the experimental conditions should be amenable to neurophysiological investigation of the underlying neural mechanism. The alternative method of providing the target leg with a small platform has the disadvantage that the very guiding response under study brings the posterior leg onto the platform and thereby disrupts regular walking.

Information on the angle of protraction at the coxo-thoracic joint is provided by two sets of coxal sensory hairs. These sense organs participate in both intra-and intersegmental control pathways. As might be expected from their positions on opposite sides of the coxa, they have antagonistic influences in both pathways. In the intact animal, absence of activity from the caudal hair rows is associated with forward positions of the target leg and forward placement of the protracting leg, while absence of activity in the rostral hair plates is associated with posterior positions. By successively immobilizing the sensory hairs in a layer of wax, we hoped to leave the sense organs intact while biasing the total input to the CNS so as to signal an incorrect leg position. In retrospect, this bias might better have been accomplished by shaving the sensory hairs. With our method, the presence of the wax between the moving joint surfaces may have contributed to the large variation in results from one animal to the next.

The intrasegmental pathway has been investigated previously. Signals from the hairplates actively oppose exaggerated forward protraction (Wendler, 1964) while signals from hair rows and hairplates together act to terminate retraction properly (Baessler, 1977). The present results largely agree with these previous findings. Two differences probably relate to differences in method. First, immobilization of both hairplates and hair rows did not lead to the extreme overstepping found by Wendler (1964) after ablation of the hairplates alone. Second, applying wax to the hair rows did not elicit the consistent caudal extension of retraction which Baessler (1977) obtained by ablating these hairs and simultaneously applying paste to bend the hairs of the hairplates forward. Apparently this false signal from the hairplates must be present and sufficiently large to achieve a consistent result (U. Baessler, personal communication).

The intersegmental signals derived from the coxal proprioceptors perform two related functions. One is to monitor target leg position to enable the step by step control of the protraction endpoint of the adjacent caudal leg. During normal walking, the target position only varies over a small, caudal fraction of the full step range. The caudal hair rows are important for monitoring this variation, since immobilizing the hairs on the target leg usually decreased the compensation shown by the protracting leg. The cranially located hairplates do signal larger changes in target position which extend into the forward part of the leg’s range.

The second intersegmental function is properly speaking another facet of the step by step regulation. It is expressed as a bias in the mean protraction endpoint of the protracting leg. Experimental removal of hair row input from the target leg resulted in a rostral shift in mean protraction endpoint of the adjacent caudal leg but did not usually alter the corresponding position of the target leg. This combination sometimes produced overstepping of the kind observed by Baessler (1977). Additional immobilization of the hairplates caused this protraction endpoint to revert to more caudal positions and at the same time caused the target leg position to shift forward. This latter change may indicate that the hairplate input contributes to a position dependent signal affecting the timing of protraction by the caudal leg. One possible interpretation would be that the absence of hairplate input, which normally signals posterior positions of the target leg, is one factor promoting protraction of the adjacent leg.

Both these intersegmental influences probably are conveyed by interneurones rather than by direct interganglionic projections of the sensory hairs. In cockroaches, the coxal hairplate is known to project into segmental interneurone pools (Pearson, Wong & Fourtner, 1976). In locusts, the only intersegmental sensory projections are those of the proximal chordotonal organs (Hustert, 1978; Braeunig et al. 1981). The latter authors report unpublished results showing that the projection pattern in the stick insect is similar. Preliminary cobalt labelling in our laboratory has not shown any intersegmental projections from hairplates or hair rows. The chordotonal organs described by Hustert (1978) are capable of registering coxal protraction. These organs were left intact in our experiments but they apparently are incapable of sustaining a guiding response to any significant degree.

In the absence of signals from its own hairplates and hair rows, a leg can still adjust its protraction endpoint to changes in the position of its target leg. Intra-and interleg influences interact to determine protraction endpoint. Wendler (1964) showed that signals from the hairplates act to damp forward protraction and to resist both external and internal disturbances tending to move the leg forward. The present results show that one of the internal signals subjected to this damping is derived from the position of the adjacent forward leg. The influence of this particular signal is limited to one part of the leg’s step - the endpoint of protraction. In cockroaches, an analogous function is performed by the coxal hairplate (Wong & Pearson, 1976) ; this hairplate projects both onto leg motoneurones and onto segmental interneurones (Pearson et al. 1976). Integration with intersegmental signals could occur at either level.

This control signal arising from the forward adjacent leg represents a spatial coordinating signal which passes from front to rear. This direction is opposite to that proposed in most models for temporal coordinating signals of insect walking (Wendler, 1968; Graham, 1977; Wilson, 1966), but Cruse (1981) has shown the feasibility of a model based on front to rear flow of temporal cues. His model incorporates both temporal and spatial parameters of leg movement; it includes the spatial coordination demonstrated in the present results, but this feature is not a major determinant of the temporal pattern.

During normal walking, the small changes in protraction amplitude due to variations in the position of the forward leg are normally accomplished without a measurable effect on protraction duration or step rate. When the spatial changes are made more extreme, the consequences for temporal coordination become more evident. This interaction of spatial and temporal parameters is dramatically revealed when one front leg is subjected to a slow, continuous change in position (Fig. 4C). Two further examples appear in the comparison of walking for different positions of middle leg restraint. Changes in the mean protraction amplitude of the ipsilateral rear leg are directly coupled to changes in mean protraction duration and inversely coupled to changes in step rate. An additional role for sensory input in modifying the speed of protraction is suggested by the increase in protraction duration following removal of sensory hair input.

In summary, this report has discussed adjustments in leg placement during locomotion which are guided by the position of the forward adjacent leg. However, this is only one circumstance in which it may be necessary or advantageous for the movement of one leg to be adjusted according to the position of another. Therefore, the CNS should be able to monitor the position of each leg and to integrate this information into the control of the other legs in a manner appropriate to such different tasks as locomotion, cleaning, and searching. One circumstance in which this plasticity might be sought is when a stick insect walks backwards. With the change in direction, a middle leg searching for a substrate at the end of its swing needs to be guided not by the ipsilateral front leg but by the rear leg. Backward walking is unusual in the insect’s normal life ; it is more common on a walking wheel but does tend to be slow and irregular. In one instance in which a stick insect made enough regular steps to allow a regression analysis, the placement of the middle leg at the end of its swing was significantly correlated with the momentary position of the ipsilateral rear leg.

This plasticity of interleg coordination may explain the weakness of interleg reflexes which have been found in restrained and dissected preparations. The influence of one leg on the legs adjacent to it is not an invariant factor; it must change according to the behavioural requirements. During locomotion it may only be active during a specific part of a leg’s step cycle.

This research was supported by a grant from the Deutsche Forschungsgemein-schaft (We 719). The authors wish to thank Dr D.E. Forsythe for reading the manuscript, and Dr H. Scharstein and U. Horsmann for the construction of the walking wheel and the initial computer programming.