J. Exp. Biol. (1971), 54.

With 6 text-figures
Printed in Great Britain


Department of Biology, University of Oregon
(Received 8 September 1970)

Much attention is currently being directed toward the problem of control and
co-ordination of leg movements during locomotion in insects (Wilson, 1965, 1966;
Runion & Usherwood, 1966, 1968; Pearson & lies, 1970). Yet, a detailed, quantitative description of the leg movements of insects during walking does not exist for
the most common experimental animals. Wendler (1964) supplied such a description
for leg movements in the stick insect, but this insect is little used in physiological
studies of locomotion. A description of locomotion in cockroaches was given by Hughes
(1952), but it is in general a qualitative one, and no longer provides the depth of detail
required by the physiologist.
The present work was undertaken to provide a firm foundation of behavioural data
on which to base hypotheses concerning the mechanisms of control and co-ordination
of leg movements (Delcomyn, 1971). The main finding has been that the gait used by
the cockroach, Periplaneta americana, the alternating tripod or alternating triangle
gait, is essentially the same over nearly the entire range of observed speeds of locomotion. Only at very low speeds is there any significant deviation.

All animals used in this study were intact adult American cockroaches, P. americana
L. Each subject was placed in an oblong Perspex box for observation. The bottom of
the box was lined with graph paper to give a reference grid for movements, and its
walls were coated with petroleum jelly to prevent the animal from escaping. The
cockroaches were photographed with a Hycam rotating-prism, high-speed motion
picture camera (Red Lake Labs, Santa Clara, California) at 200 or 500 frames per
second as they moved freely along the length of the box. The camera was arranged so
that four to seven full steps of each leg could be captured on the film. Lighting was
provided by two Photoflood lights, one above each end of the box. These lights were
turned off periodically for brief periods during the photographic sessions to prevent
overheating of the animal.
The films were examined with the aid of an analytical motion picture projector
(L-W Photo, Inc., Van Nuys, California). Analysis consisted of counting the number
of frames occupied by protraction and retraction (definitions below) during each
cycle of movement, for each leg. The parameters of interest, protraction, retraction,
* Present address: Department of Zoology, University of Glasgow.

Dragging the body is a mode of behaviour characteristic of cockroaches in poor condition. The maximum speed observed here is similar to that reported by others (Hughes. These diagrams illustrate typical sequences of leg movements at stepping frequencies from 1-5 to 23 Hz. shows that the substitution is a straightforward one. 2 and 3. and are as follows: Protraction: the forward movement of a leg relative to the body and the ground. a measurement which is not possible under some experimental conditions.444 FRED DELCOMYN cycle duration. 1952. The movement of an animal's legs during locomotion can be visualised with the aid of diagrams of stepping patterns. the long narrow structure of the running box and the petroleum jelly on the walls combined to produce a large number of relatively straight runs through the field visible to the camera. and of freeing the experimenter from the necessity of measuring directly the animal's rate of progression. 1965). the lighting was arranged so that the ends of the box were not as well lit as the centre. and relative phase positions were calculated from the frame data with the aid of a digital computer. such as those shown in Figs. The nearly linear relationship between average frequency of leg movement and rate of forward progression (Fig. Lag: the time from the beginning of protraction of one leg to the beginning of protraction of another. Graphs of the pattern of leg movements (e. The cockroaches did not appear to be excessively disturbed by the lights. protraction/retraction ratios. In addition. which tended to increase the probability that the negatively phototactic cockroach would continue through the well-lit centre rather than stop in mid-run. RESULTS Cockroaches do not generally run straight for long distances. Fig. The range of speeds observed was about 2-80 cm/s (at a maximum temperature of 290 C). since the parameters under these conditions often bore different relationships to frequency of leg movement than they did during steady locomotion. and also estimated by measuring the frequency of leg movement in Hertz (Hz). 1). frequency of leg movement was substituted for rate of forward progression as the independent parameter in this study. Phase: the lag between two legs divided by the cycle duration of the leg which moved first. In the present study the animals' rates of forward progression were measured directly from the films in cm/s. Steps during which the animal abruptly changed speed or direction were excluded. 1966). The latter measure has the dual advantage of allowing cycle-by-cycle plotting of parameters. For these reasons. Wilson. The terms used in describing the results are similar to those already established in the literature (Hughes. However. In addition.g. they tended to carry their bodies well off the floor of the box. 2) were drawn for each walking sequence and used as the basis for editing the computer output. No movement relative to the ground. the time relation between the movements of two legs. i. and show qualitatively the behaviour of both the individual step . Retraction: the backward movement of a leg relative to the body. for after 10-15 m m m the box some moved about quite slowly.e.

90 r 70 S. Each point represents a single continuous run. but it falls more rapidly than protraction as frequency of leg movement increases. and the inter-leg timing parameter. This may be seen in the diagrams of stepping patterns (Figs. 1. and in Fig. This differential rate of decrease is also reflected in the protraction/retraction (p/r) ratio (Fig. The relationship between p/r and frequency is nearly the same for each of the legs.e. 5). 50 < 30 10 10 20 Average frequency of leg movement (Hz) 30 Fig. but they do so at different rates. The parameters of the regression lines and other statistics for the data for each leg are shown in Table 1. protraction and retraction. The behaviour of each of these parameters as a function of the frequency of leg movements is described in detail in the following sections. unrestrained animals. . 2 and 3). the rate at which p/r increases as stepping frequency increases is essentially the same for each of the legs. which increases linearly with increases in frequency of leg movement. i.The locomotion of the cockroach 445 parameters. The average speed of locomotion as a function of the average frequency of leg movement. Individual leg movements The durations of both protraction and retraction decrease as the cockroach moves at progressively higher speeds. Retraction is nearly three times longer than protraction at the lowest frequency of leg movement. All data are for intact. until at the highest observed stepping frequencies it is often shorter than the latter. 4. phase. There is no significant difference between the slopes of any of the distributions.

B.--m --m .) .• -i 0-2s Fig.•1 H mi'-m• -m .- L2fL1 . This difference between the mean pjr of R2 and L 2 and each of the remaining R3 .-m . dotted lines retraction..- §-m .• . no... 2. Ri R2 66 64 R3 62 Li L2 58 65 63 0-6922 0-8113 07031 07239 0-026510-0035 0-0358 + 0-0033 00344 ±00045 00308 ±0-0039 0-0279 ±0-0031 00336 ±0-0033 L3 c-7457 0-7901 y int.e.E. R3. Table 1. Frequency of leg movement about 22 Hz. each of which form a triangle. Each mean pjr corresponds to a frequency of leg movement of about 14-5 Hz (except R i .m.• .- 1.» .•1 •H» • L (Ms B L31- . Notice the simultaneous movements of the two sets of three legs. L2 and L3.. Li. Ri. the solid bars representing protraction.446 FRED DELCOMYN However.. those of the middle legs (R.. Frequency of leg movement about 9 Hz. correlation coefficient.2 and L2) being lower than those of the remaining four legs (Table 1).. Correlation and regression parameters of the p/r ratio distribution for each leg Leg n r b±s. Stepping patterns of rapidly moving intact animals. regression coefficient (slope)±its standard error.• .R2 . 139).• . The legs on the right (R) and left (L) sides of the body are numbered from front to rear.- i. R2. 6±s. Each row represents the movements of the indicated leg..• R11- -•- L3..-m- L2 . y int. r. 0456 0199 0-386 O-344 0-301 0330 Mean Plr 0-825 0-713 0889 0788 0-704 o-8n («. ordinate intercept of regression line. A... The diagram reads from left to right.« . The conventions established by Wilson (1966) have been followed.• . the mean p/r ratios of each of the legs do not show such uniformity. of observations.

2 and L2. Periplaneta always uses the alternating tripod gait. Timing of leg movements Except at the very lowest speeds of locomotion (stepping frequencies less than about 3 Hz). A. the regression lines for R2 and L2 are depressed vertically compared to those from the remaining legs (Fig. L i . The alternating movements of the two triangles of legs are still maintained at this low frequency of leg movement (4 Hz). B. Nevertheless. 5). Protraction durations have now become significantly shorter than retraction. 2B and 3 A. R3 or L3. 2 and 3. and might be termed 'uncoupled' alternating triangle. a great deal of variation in thepjr ratio at any speed. this phenomenon may be seen quite clearly in the R2 steps in Figs. Conventions are as in Fig. Stepping patterns of slowly moving intact animals. Clearly. . There is no significant difference between the mean ratios of legs R. Since all the slopes are the same. Although there is. This may be seen qualitatively in the stepping patterns shown in Figs. 3.The locomotion of the cockroach legs is significant statistically to at least the 97-5 % level of confidence in each case. thus necessitating the statistical treatment. This means that the middle legs are off the ground for shorter periods at any given speed R3 -1 R2 - R1 • L3 L2 L1 0-2Ss 0-5s Fig. resulting in periods during which all six legs are on the ground at once. Graphically. the leg movements are rarely synchronous. of locomotion than are any of the four other legs. 2. therefore. nor between those of any of the remaining four legs. the overall consistency of the stepping pattern is also clear. in fact. The alternating triangle pattern has begun to break down at this frequency of leg movement (1-5 Hz). the three legs comprising one tripod do not usually begin protraction at exactly the same instant. that is. the p/r ratio at any given frequency of leg movement is lower for legs L2 and R2 than for legs Ri.

legs in a single body segment. as a function of the frequency of leg movement (of leg L3). t stepping frequencies of about 3-4 Hz. since phase is simply a 600 520 440 360 ~Z 280 U 200 s 120 o «fP Ofl. 6A). » . the rate of change is different for the two. At very low walking speeds. as shown in Fig. 40 • » . This change in phase at low talking speeds occurs only in ipsilateral legs. 6). This is not surprising. 4. do not show any reduction at low walking speeds. . leasure of the timing relationships between the legs. 3 B.448 FRED DELCOMYN This consistency is reflected in the distributions of phase values. rather than at exactly the same time. which in turn determine the gait mployed during walking. The phases of contralateral leg pairs. of course. While both components occupy less time at higher frequencies of leg movement than at lower ones. Except at the extreme low end of the range of walking speeds. protraction (filled circles) and retraction (open circles). as rould be demanded for a rigid alternating triangle gait. lat is. Statistical tests support this statement (Table 2). The duration of the two components of each step. *» ' • 8 6 10 14 18 Frequency of leg movement (Hz) 22 J 26 Fig. since in no case is there a significant correlation between phase and frequency of leg movement when phase values corresponding to frequencies less than 5 Hz are omitted. the protraction of each front and middle leg egins earlier than usual relative to the ipsilateral rear legs. The fact that the average phase of L1 on L3 and of R1 on 13 is 0-954 indicates that each of the front legs generally begins protraction slightly arlier than the corresponding rear legs. so le corresponding phase values are reduced (Fig.. there is no significant change in phase as a function of the frequency of leg movement (Fig.

Thep/r ratio of legs R3 (A) and R2 (B) as a function of the frequency of leg movement.0 . 1879. Mean phase y int.) 1-8 r 1-8 r 1-4 1-2 1-2 0-8 0-8 0-4 0-4 15 21 3 15 21 Frequency of leg movement (Hz) Fig. Description of the behaviour of another parameter of the timing of leg movements. current attention has turned to investigations of mechanisms by which patterns . R I L2. L 3 Li. This finding suggests the possibility that the mechanism(s) responsible for the phasing of leg movements may be different for contralateral leg pairs than for ipsilateral legs. Correlation and regression parameters of the phase distribution of leg pairs. Morgan. Demoor. Leg pair r (ri) ( > 5 Hz only) L2. DISCUSSION While early work on insect locomotion had been concerned primarily with the correct description of leg movement (e. 5. is implicit in the description of phase given above.0 5 4 3 (63) b ± s. but there is between the vertical displacements. 1886.Rz —0.g.R 3 L3. not shown) protractions are generally shorter at any given speed of locomotion than those of the remaining legs. 0-00500 + 0-00121 — 0-00042 + 000081 0-00294 + 0-00101 0-00010 + 0-00091 0-471 0882 0499 0425 0-508 000148 + 000099 0520 0510 0-500 0-00212 + 0-00092 0-469 O-934 0-464 0-499 O-954 °-459 0-00079 ± 0-00090 0-00142 + 0-00129 — 0-00036 + 0-00102 0-482 °954 O-493 0467 (Symbols as in Table 1.0 0 6 6 8 (62) 0-3446 (64) 0-0141 (64) -01828(66) 0-2862 (61) 0-1467 (57) .1172 (53) 0-2586 (47) -0-1293 (53) 0-2207 (52) 0-1045 (52) Li. which is simply a reciprocal function of frequency of leg movement. R3 Ri. lag must bear a constant relationship to cycle duration. 1890).The locomotion of the cockroach 449 Table 2. See also Table 1 and text.L2 — O-222I (56) O-O3O8 (50) . There is no significant difference between the slopes of the distributions. with best-fit regression lines shown.E.0 . L 3 Li. Carlet. R3 Ri. R2 r(«) O-II28 (62) O-4953 (55) . lag. Phase is simply lag divided by cycle duration.1 6 3 8 (54) R2. indicating that R2 (and L2. Since phase is constant over nearly the entire range of locomotor speeds.0 0 8 8 4 (48) .

Davis. B.g. extension retraction. reflecting the breakdown of the alternating triangle gait below this speed. There is an apparent tendency for the phase to fall at leg movements below about 3 Hz. Their results show an increase in both extensor and flexor burst durations as the burst period increases (i.• • a. Pearson & lies.1 o o o _l 12 I 15 I 18 0-7 r 0-5 0-3 0-7 0-5 0-3 U 0 3 Fig. although the latter relationship is a weak one. Their findings fit reasonably well with the behavioural observation that as leg-cycle time increases. Hughes. as the frequency at which the bursts occur decreases). Wilson.FRED DELCOMYN 45° A 1-2 e 10 r o O ° o ooc .e. 0-4 • I I I 21 24 27 I I L I 6 9 12 15 18 21 Frequency of leg movement (Hz) I 24 I 27 _ i _ __i I n. suggesting that the production of these bursts may be under the control of a central programme generator. 1971). Open circles. (The horizontal lines in the graphs are i-o and 0-5 reference lines only. so that the duration of a burst of activity in any one flexor . 1966. •• • oo o o 0-8 0-6 * i. filled circles. Phase as a function of the frequency of leg movement during walking. Pearson & lies (1970) have shown that alternating bursts of flexor and extensor motoneurone activity can be obtained in partially de-afferented preparations of Periplaneta. 1970. flexion causes protraction.) The weak relationship between flexor burst duration and burst period would be accounted for if few or none of the flexor muscles were active during the entire period of protraction (cf. A. filled circles. Delcomyn. the lobster swimmeret system. R3. phase of L3 on R3. phase of L i on L 3 . There is no fall in phase at low speeds with these contralateral leg pairs. The present work is especially relevant to some of these.) of leg movements might be generated (e. Open circles. phase of L i on R i . phase of L2 on L3. (In the leg they examined. 6. 1957. 1968). 1965. the duration of both protraction and retraction increase.

that is. examination of his results suggests that Blatta walks more like Periplaneta than Carausius. so that the gait of this insect does not change over most walking speeds. 1957) on the cockroach.. 279). 1). resulting in increasing phase between them. Thus. The data reported above are also relevant to Wilson's (1966) model of insect Wilson pointed out that with three conditions fulfilled. 3 A above). 1952.g. The basic rhythm is the alternating triangle gait (cf. the relationship between flexor activity and burst period would not necessarily be similar to that between protraction and leg-cycle time. over the whole range of walking speeds. and (3) phase between contralateral leg pairs in each body segment remained constant at 0-5. and is a continuous function of speed during very slow locomotion. Blatta orientalis. an insect in which gait is affected by speed of walking only at very low speeds. 1966. 4 and 5 above. over the whole range of speeds. But these grade insensibly into the normal rhythm with an increase in speed. R3. gait is independent of the speed of walking at medium and fast speeds. None of the phase values which can be calculated from his data are substantially different from those for Periplaneta shown in Figs. Carausius morosus. Mechanisms of locomotion in stick insects and cockroaches need not be as different as the above descriptions of their patterns of locomotion might suggest. Fig. but protraction is constant. Conditions 1 and 2 clearly do not hold for Periplaneta. The phase of leg pairs within a single body segment (e. Hughes states: ' Cockroaches use very nearly the same rhythm (R1. and stays constant at higher speeds. Wilson has suggested (personal communication) that Periplaneta might be considered a 'fast walker'.. the speed at which it appears 29 EXB 54 . 1957) suggest that non-alternating tripod gaits in Blatta also appear only below this stepping frequency (equivalent to about 5-7 cm/s. . L2 and R2) is independent of speed. 1957).The locomotion of the cockroach 451 muscle would not accurately reflect the duration of protraction. Thus. etc. increasing as speed of walking increases. and: ' The same basic rhythm of leg movements is found at all speeds of cockroach movement. as in the cockroach. and thus support this notion. Wendler. 1966) on the stick insect. In this case. However. Wilson based his model primarily on the work of Wendler (1964. These conditions were that as speed of walking increased: (1) protraction remained constant. as shown in some of the graphs of leg movements appearing in his papers (Hughes. The data of Hughes (1952. but phases of ipsilateral leg pairs drift. in this insect gait is a function of the speed of locomotion. . although at the very slowest ones different rhythms may be observed. While the data are not complete for low-speed walking. (2) lag between ipsilateral legs remained constant. 280). The locomotion of Carausius differs from that of Periplaneta in several respects. they do suggest a change of gait as the animal slows down. Carausius is a much more slowly moving animal. The results of Hughes (1952) on Blatta have been interpreted similarly (Wilson. As in Periplaneta the pjr ratio increases and retraction decreases with increasing speed of walking. gait is a direct function of velocity. and their results would be fully compatible with the behavioural data. In Periplaneta the data suggest that at stepping frequencies below 3-4 Hz. 1966). and while the alternating triangle gait is used only at the highest walking speeds. see Fig. L2. an insect's gait would be a direct function of its speed of locomotion." (p. and Hughes (1952.) at all speeds from 1 to 15 cm/s [about 1-8 H z ] . in Blatta. ' (p.