Interlimb Coordination During Slow Walking in the Cockroach: I. Effects of Substrate Alterations

The neural control of insect walking has been the subject of several recent studies, centred on two levels of organization. One is the control of movements within a single limb, and the contributions of central pattern generators and proprioception to intralimb coordination. This work has been recently reviewed by Pearson, Fourtner & Wong (1973). Another level of organization, that of inter-limb coordination, has been studied by Hughes (1952), Wilson (1966), and Delcomyn (1971), in the cock-roaches; Wendler (1966) and Graham (1972), in stick insects; and Burns (1973), in locusts and grasshoppers. In this work, two basic gaits have been observed. During fast walking, the gait generally consists of an alternating tripod as observed by Wilson (1966). During slow walking, however, a second gait is observed in the stick insect (Graham, 1972). In this gait, segmental pairs of limbs are in approximate alteration while ipsilateral leg groups move in a metachronal (3-2-1) sequence. Studies of other arthropods, such as the tarantula (Wilson, 1967) and scorpion (Bowerman, 1975), have indicated that the metachronal gait is commonly found in these groups.

High-speed walking in the cockroach ( > 10 steps/s) involves rapid leg movements and minimal adjustment to substrate or other environmental variables. During slow walking, factors such as loading, substrate angle, texture, and local variations would have a more significant effect on the animal’s leg movements and stability, and could, through feedback control, produce alterations in walking behaviour. Thus, in this report, we examine slow walking under a variety of conditions. Previous workers have suggested that a non-tripod gait is used at low speeds (Hughes, 1952; Wilson, 1966; Delcomyn, 1971), and this work was undertaken to quantify the slow gait and to determine which parameters of the walking behaviour are affected when substrate conditions are altered. A second aim was to obtain behavioural data as a basis for future neurophysiological studies of the mechanisms controlling interlimb coordination in the cockroach.

Adult cockroaches (Periplaneta americana) were used in all experiments. Free-walking animals were filmed in a large arena (70 × 70 cm) with a horizontal, paper-covered surface, which was tilted to 45° to record walking on an incline. Ball-walking was used to obtain long, uninterrupted, sequences of walking. The animals were restrained by a wire which was waxed to the pronotum, and attached to a counter-weighted balance arm, so that the cockroach supported only its own weight and was free to move in the vertical direction. This apparatus was suspended over a styrofoam ball (9 cm in diameter) which rested on an air stream within a large funnel (Carrel, 1972). A cardboard sheet with a central hole was placed over the ball to deflect the air stream away from the walking surface. Thus, the cockroach could walk and rotate the ball beneath. The air stream supported the ball and allowed rotation in any direction. A series of markings on the ball were used to distinguish between straight walking and attempts to turn.

In all cases, animals were filmed from above at 64 frames/s, and sequences containing long bouts (6–40 steps) of slow (less than 10 steps/s), straight walking were later analysed using a stop-action projector. Data taken from films (frame numbers of touch-down and lift off points for each leg) were analysed to determine the retraction and protraction stroke durations for each leg, as well as parameters of interlimb coordination among all six legs. The terms protraction and retraction are used here to describe a complete leg movement cycle, consistent with earlier studies. However, it should be noted that leg movements during walking are far more complex than a simple protraction/retraction, involving a coordinated series of angular movements of all leg segments.

The coordination data are utilized in two ways. First, an interlimb phase is calculated, using leg lift-off times as a reference. Within a movement cycle of leg A, one can determine the time at which leg B begins a cycle (lifts off). The fraction of the leg A cycle at which this occurs is the phase of leg B in leg A (e.g. a phase of 0 or 1·0 indicates synchrony of leg movement, while leg alternation would yield a phase of 0-50). This measure reflects coordination with respect to complete leg cycles, but gives no information concerning the relationship between the protraction/ retraction durations of each leg. A second measure, which we refer to as interval, is defined as the time between touchdown of a reference leg and lift-off in an adjacent leg (Pearson & Iles, 1973). Interval data, when combined with interlimb phase, allows us to determine the exact relationship between legs and to reconstruct an average leg movement pattern. Statistical procedures used for analysis of phase data were those developed for circular distributions (Batschelet, 1965). A total of 200–400 steps of walking, obtained from 3-6 individuals, were used for each set of observations.

For some experiments, cockroaches were blinded by surgically interrupting the optic tract and/or optic ganglia on both sides through incisions at the medial surface of the eyes. During this procedure, the antennae were amputated close to the base. After collecting walking data, dissection of the fixed heads confirmed that the optic tracts or ganglia had been bilaterally interrupted.

Movement data for normal free-walking at slow speeds are quite similar to those previously reported by Delcomyn (1971) and Pearson & Iles (1973). Both intralimb (protraction and retraction durations) and interlimb (phase, interval) parameters are essentially the same as those found earlier.

All data reported here are in the range of 2–7 steps/s. At frequencies of less than 2 steps/s, the cockroach walks very erratically, with frequent stops and turns, and with continuous searching movements of the head, antennae, and mouthparts. A stepping frequency of 2–7 steps/s corresponds to a velocity of approximately 5–20 cm/s (Delcomyn, 1971). As walking speed increases in this range, there is a large, non-linear decrease in the duration of retraction, and a smaller decrease in the duration of protraction, with these parameters approaching equal duration at high frequency. Since the protraction/retraction ratio encompasses both of the above measures, it will be used in the comparison of walking under other conditions. The protractiOn/retraction ratio for normal free-walking is shown in Table 1 and Fig. 1. All individual leg data is an average over six legs. It should be noted, however, that the detailed sequence of movements made by a leg differ markedly between thoracic segments. Therefore, although we will be concerned with times of touch-down (contact with substrate) and lift off (loss of contact), the sequence of movements at individual leg joints necessary to produce this behaviour can differ markedly between legs. These differences are also pointed out by force measurements reported by Cruse (1976) for the stick insect.

The interlimb coordination (gait) during slow free-walking in the horizontal arena does not vary with walking speed. At frequencies of 2–7 steps/s, the gait utilized is the same as that seen during more rapid walking (i.e. the alternating tripod gait with all adjacent leg pairs maintaining an average phase relationship of close to 0-50). The gait data for normal free-walking is shown in Table 2. The average phase relationship for ipsilateral leg pairs (L1/L2, R1/R2, etc.) is 0·47 ± 0·06 C.S.D., while the average phase for contralateral leg pairs (L1/L2, L2/R2, etc.) is 0·49 ± 0·07 C.S.D. The interval times for normal free-walking are long and variable, with ipsilateral pairs averaging 86 ± 65 msec S.D. and contralateral pairs averaging 96 ± 75 msec S.D.

Thus, the gait of a normal free-walking cockroach is characterized by an alternating tripod, with considerable variability in the durations of protraction and retraction in all legs.

This situation was originally designed to provide a restrained preparation for neural recordings during normal walking. However, as described in detail below, the walking behaviour under these conditions is significantly different from normal.

As can be seen in Table 1 and Fig. 1, the protraction/retraction ratio for ball-walking is similar to normal. They are not significantly different in either magnitude or slope and, over the range of slow walking speeds, the leg movement pattern is indistinguishable from that seen in normal free-walking animals.

Interlimb coordination during ball-walking is quite different from the normal patterns. At frequencies below 7 steps/s, the phase relationship between ipsilateral leg pairs is lower than the 0·50 phase expected for a tripod gait (Table 3). In fact, there is a continuous change in phase with increasing walking speed, ranging from 0·32 at 3 steps/s up to 0·46 at 9 steps/s (Fig. 2). At any given step frequency, the phase is similar for all leg pairs in all animals measured. Therefore, even at very low speeds, coordination among ipsilateral legs is rigidly maintained, implying that it is actively controlled, rather than simply a degeneration of the tripod gait. Contralateral leg pairs also display a different phase relationship during ball walking. In this case, the phase of all three pairs is greater than 0·50 over the range of 3·9 steps/s (Table 3, Fig. 2), and there is no change with increasing frequency. The lateral phase asymmetry seen in these data is consistent for all individuals and is similar to that described by Graham (1972) for free-walking stick insects.

As might be expected, the interval data for individuals during ball-walking is also quite different from that during normal locomotion (Table 3). The intervals for ipsilateral leg pairs are very short and constant ⁠. In addition, we have never observed two adjacent ipsilateral leg pairs simultaneously off the substrate (i.e. negative interval) in over one thousand measurements. Contralateral interval data are also shorter than normal in the left-after-right direction ⁠,. reflecting the lateral phase asymmetry seen during ball-walking.

Thus, the gait of a ball-walking cockroach is very much like the gaits of other arthropods, with ipsilateral legs moving in a metachromal 3-2-1 sequence. As walking speed increases, there is a smooth transition to an alternating tripod gait, with all leg pairs at a phase close to 0·50. Contralateral leg pairs alternate asymmetrically when walking on the ball, which is also reflected in the interval data. There are factors involved in ball-walking which could account for these changes in gait. First, it is likely that the styrofoam ball, riding on an air cushion, alters the normal load against which the animal must work to propel itself forward. Second, since the animal does not move, the upper portion of its visual field remains stationary, thereby altering visual feedback which may control aspects of walking. These two factors are tested in the following sections.

The lateral asymmetry, seen during ball-walking but not during free-walking, could be caused by the conditions under which the animal walks. One possibility is that there exists an inherent asymmetry in the walking control system which, under normal conditions, is masked by visual feedback. To test this hypothesis, surgically blinded animals were allowed to walk freely in the horizontal arena. The general behaviour of these animals is somewhat different from normal. In addition to the expected lack of response to visual stimuli, there is an increase in locomotor activity which is highly regular and ‘mechanical’ in appearance. Blinded animals tend to walk continuously and at a constant speed for long periods of time.

This robot-like walking is also reflected in the data collected from these animals. The protraction/retraction ratio for blind animals is virtually constant over the frequency range of 2–7 steps/s (Table 1, Fig. 1). Thus, individual leg movements for blind animals are regular and the relationship between protraction and retraction components of stepping is constant over a wide range of stepping frequencies.

Interlimb coordination in blind animals is also regular and invariant (Table 4). The phase relationships for both ipsilateral and contralateral leg pairs are indistinguishable from the alternating tripod gait, and also from the data on normal free-walking. The intervals for ipsilateral leg pairs are intermediate between free-walking and ball-walking, but more invariant than either ⁠. For contralateral leg pairs, the interval data is similar in length to ball-walking, but again more invariant ⁠.

Thus, free-walking behaviour in blind animals appears to be a rigid, invariant tripod gait with each leg maintaining a fixed protraction/retraction ratio. There is no indication of lateral asymmetry under these conditions.

Ball-walking data suggest that variations in gait could be due to alterations in the load against which the cockroach must work to propel itself forward. To examine this possibility, cockroaches were allowed to walk freely on a surface inclined at an angle of 45°. Sequences selected for analysis include only straight walking directly up or down this plane.

The protraction/retraction ratios are similar for walking either up or down the ramp (Table 1, Fig. 1). In both cases, the changes which occur with changes in walking speed are similar to those seen in normal free-walking. They are, however, considerably depressed (longer retraction/shorter protraction) compared to level walking.

Interlimb coordination during walking on an inclined plane is varied and complex. For both up-and down-hill data, there are asymmetries among both ipsilateral and contralateral leg groups. During up-hill walking, phase data for ipsilateral leg pairs are close to normal except for the R2-R3 pair (0·53) which is significantly different (P < 0·001) from all other pairs (Table 5). Contralateral phase data in this case are similar for pro- and mesothoracic pairs and have a lateral asymmetry similar to that seen during ball-walking. The metathoracic pair, however, is significantly different (P < 0·001) from both normal and ball-walking data. The interval data for up-hill walking are somewhat shorter than normal for all leg pairs. Since leg R is common to both leg pairs showing an altered phase, it can be assumed that this single leg has a unique behaviour during up-hill walking. Otherwise, the cockroach has a normal ipsilateral phase and laterally asymmetrical contralateral phase under these walking conditions.

During down-hill walking, the interlimb coordination is similarly complex (Table 6). In this case, leg Li has a unique pattern. Ipsilaterally, the L1/L2 pair phase is significantly different (P < 0·001) from other pairs but indistinguishable (P > 0·05) from normal. Contralaterally, the prothoracic (L1/R1) pair under these conditions has a much lower phase than other contralateral leg pairs, and this phase is similar to L3/R3 for up-hill walking. Other ipsilateral leg pairs have a phase relationship which is lower than normal and similar to that found for ball-walking. Likewise, the remaining contralateral leg pairs are laterally asymmetrical in the same direction as was seen during ball-walking and up-hill walking. The interval data under these conditions is near normal. Thus, when walking in a down-hill direction, the gait pattern is similar to ball-walking, with the exception of leg L1, which has a unique relationship to the remaining legs.

From the data presented here, it is obvious that there is no single gait for cock-roach walking, and that both individual leg movements and interlimb coordination are dependent on the conditions under which the data is collected. These obvious differences in walking behaviour under different conditions are largely unique for each situation.

During normal free-walking, the alternating tripod gait is observed at all walking speeds over 2 steps/s, confirming the findings of earlier workers (Wilson, 1966; Delcomyn, 1971). The movement pattern for blind free-walking animals is also a tripod gait, but with less variability than is seen in normal animals. This is especially true for the individual legs, where the protraction/retraction ratio is constant in blind animals. These data imply that the visual and antennal systems play a significant role in the control of normal locomotion, presumably in the determination of speed and direction. Thus, even in normal, straight-walking animals, there might be continuous, subtle changes in the speed and direction of walking which would then be reflected in the variability of phase data and the changes in protraction/retraction ratio. The original purpose of the blind walking experiment was to determine whether there is an inherent lateral asymmetry in the walking control system which is normally masked by visual input. The fact that there is no asymmetry in the gait of blind animals suggests that the asymmetry seen under both ball-walking and inclined-plane walking conditions is due to other factors.

Under the artificial conditions imposed by ball-walking, there are significant changes in gait at slow speeds, including a metachronal ipsilateral coordination and a lateral gait asymmetry. These changes could be due to an alteration in the load against which the cockroach must work to propel itself forward. To examine these loading effects, one can compare ball-walking with inclined-plane walking. If, in ball-walking, there is a decreased loading, then this should be comparable in some respects to down-hill walking. If, on the other hand, the loading on the ball is greater than normal, it would be similar to up-hill walking. For ipsilateral leg groups, the average phase for up-hill walking (0·48) is close to normal (0·47) while that for down-hill walking (0·43) is intermediate between normal and ball-walking (0·37). The contralateral coordination in all three situations of decreased load, there is a change in ipsilateral phase away from the tripod and toward a more metachronal gait. When load is increased, there is little or no change in ipsilateral phase on the 45° incline tested.

There are also changes in individual leg movements during walking on an inclined plane. The protraction retraction ratio for both up- and down-hill walking is consistently lower than normal, while changes with frequency resemble those found for normal and ball-walking. When an animal walks on an inclined plane, its postural stability is decreased, since a vertical line through its centre of gravity intersects the substrate at a point closer to the organism’s points of substrate contact. In other words, an animal slowly walking under these conditions is more likely to become posturally unstable. One mechanism which would tend to decrease this instability would be to decrease the protraction/retraction ratios of all legs, thus increasing the time a leg is in a support position. Such changes are seen in both up and down-hill walking, but not in ball-walking, where there is no change in stability.

Another interesting finding of the inclined-plane is that, in each case, there is a single leg which has unique phase relationships to adjacent legs both along and across the body. During down-hill walking, for example, leg Li has unusual phase relationships with both L2 and R1. Similarly, leg R3 during up-hill walking has an unusual coordination with R2 and L3. In both cases, the ipsilateral phase is significantly higher than remaining pairs, while the contralateral phase is significantly lower. Since, in these situations, the leg in question is on the extreme down-hill end of the animal, a strut or support function might be suggested in both cases. This single leg, moving at a somewhat different phase than the rest, would tend to increase substrate contact time and, thus, add stability during slow walking.

The lateral asymmetry observed during ball-walking and inclined-plane walking, is not an inherent property which is masked by visual feedback, and is similar in both direction and magnitude for all three experimental situations. Since the function of such an asymmetry is not readily apparent, it could be merely a by-product of the metachronal ipsilateral gait which occurs under these conditions. This would be consistent with the observations of Graham (1972), where a lateral asymmetry was present in conjunction with the metachronal gait in the free-walking stick insect.

In discussing possible neural mechanisms for the control of walking under these five diverse conditions, one can begin with certain assumptions. First, consider a set of six hemisegmental oscillator mechanisms of the type proposed by Pearson (1972), interconnected through reciprocal protraction-stroke inhibition (Wilson, 1966). To this system can be added three types of sensory influence. At the most local level is proprioceptive input from the legs. This would include information on substrate contact, joint angles, muscle states, and loading. Central effects of this input could be both intra- and inter-segmental, utilizing a sensory tape mechanism (Hoyle, 1964), such that its effects would be seen only under abnormal conditions. A second sensory influence would be from the gravity-detection system. This input would exert a system-wide effect on posture and locomotion, tending to compensate for changes in orientation of the body with respect to gravity. Finally, we can add an overall locomotion control input, utilizing extroceptive information from the visual, antennal, cereal, and other sensory systems. This control input would influence the initiation speed, and direction of walking.

Examining the five walking conditions, we begin with the simplest case: a blind animal walking on a smooth horizontal surface. Under these conditions, proprioceptive and gravity input would be ‘normal’, while control input would be minimized due to the lack of visual and antennal influences. In this situation, the rigid, alternating tripod gait would result. During normal free-walking, there is additional control input from the eyes and antennae. This input would act on the system as a whole, producing a change in the protraction/retraction ratios of all legs and the observed increase in gait variability.

If we now consider ball-walking, we must add an altered proprioceptive influence due to the change in load. This abnormal input might produce the altered gait seen under ball-walking conditions (i.e. the metachronal ipsilateral coordination and asymmetrical contralateral gait) with little change in intralimb parameters. If the change in proprioceptive input were more local, as when walking on irregular substrate, one would expect to see more restricted compensations in the walking pattern. The situation for walking on an inclined-plane is the most complex. Here we are dealing with both changes in loading (proprioception) and changes in body angle (gravity detection). Under these conditions we see, during down-hill walking, the basic, decreased-load gait (as in ball-walking), with the addition of postural compensations in the form of decreased protraction/retraction ratios and gait changes in certain legs. For up-hill walking the increased loading is presumably not significant and we see only the postural compensations.

The walking system of the cockroach is, therefore, very complex. It is capable of adaptation to the conditions imposed by the environment by changing both its limb movement patterns and its gait. Future work on this system must be designed such that the walking pattern under study is carefully defined, or the situation carefully specified.

This work was supported by grants from the Ohio University Research Committee (OURC no. 556) and the Ohio University College of Osteopathic Medicine to C. P. Spirito. We would like to acknowledge R. Nichols for technical assistance and Dr C. Kaars for critical comments on the manuscript.