Jumping strategies and performance in shore bugs (Hemiptera, Heteroptera,Saldidae)

Jumping is a common means of escaping rapidly from predators or increasing the speed of locomotion. Many insects that jump can also fly and the intimate interactions between these two movements are expressed in the following ways. Jumps can frequently lead directly to flight and wing movements may stabilize the trajectories of a jump by reducing rotation of the body once airborne. Opening the wings, however, will increase the frontal area of the insect and therefore slow the take-off velocity through increased drag(Bennet-Clark and Alder, 1979). The wings can either start to move before take-off, or their movements are delayed until take-off has been achieved(Brackenbury and Wang, 1995; Card and Dickinson, 2008; Hammond and O'Shea, 2007; Pond, 1972). The propulsion for jumping in insects is typically provided by rapid movements of the legs,or less often by movements of other parts of the body, all involving extensive mechanical, muscular or neural specializations to accomplish this demanding form of locomotion. Accomplished exponents, which propel jumping by rapid movements of their hind legs, are found in the Siphonaptera, fleas(Bennet-Clark and Lucey, 1967; Rothschild and Schlein, 1975; Rothschild et al., 1975; Rothschild et al., 1972);Orthoptera, bush crickets (Tettigonidae)(Burrows and Morris, 2003) and locusts (Acrididae) (Bennet-Clark,1975; Godden,1975; Heitler,1977; Heitler and Burrows,1977a; Heitler and Burrows,1977b); Coleoptera, flea beetles (Alticinae)(Brackenbury and Wang, 1995);and Phasmatodea, stick insects (Burrows,2008). By contrast, in the Collembola, springtails jump by rapidly extending their terminal abdominal appendages(Brackenbury and Hunt, 1993),whereas movements of the whole abdomen are used in the Hymenoptera by some ants (Baroni et al., 1994; Tautz et al., 1994), in the Phasmatodea by a stick insect (Burrows and Morris, 2002) and in the Archaeognatha by Petrobius(Evans, 1975). In the Coleoptera, click beetles (Elateridae) move the prothorax against the mesothorax (Evans, 1972; Evans, 1973; Kaschek, 1984), and in Hymenoptera trap-jaw ants (Odontomachus) rapidly close their mandibles against the ground, or an approaching object, to propel themselves upwards or backwards (Patek et al.,2006).

The ability of insects to jump reaches its zenith in the Auchenorrhyncha,one of the four sub-orders of the Hemiptera, and in particular by members of one of its numerous families, the Cercopidae. Froghoppers (spittle bugs) have the best jumping performance of any insect described so far, accelerating their bodies in less than 1 ms to a take-off velocity of 4.7 ms^–1, experiencing a force of some 550 g(Burrows, 2003; Burrows, 2006a). This outstanding performance is achieved by using a catapult mechanism in which force is developed by the slow contraction of huge thoracic muscles and the stored force is then released rapidly(Burrows, 2007c). A second family, the Cicadellidae or leafhoppers, are also accomplished jumpers(Burrows, 2007a; Burrows, 2007b) with one group having long hind legs and another short hind legs although both achieve comparable take-off velocities (Burrows and Sutton, 2008), but which are considerably slower than those of froghoppers. The other three sub-orders of the Hemiptera also contain proficient jumpers. In the basal Coleorrhyncha at least one extant species jumps (Burrows et al., 2007). In a second sub-order, the Sternorrhyncha, one family, the pysllids or jumping plant lice, are, as their colloquial name implies, well known for their jumping but their performance is only currently being investigated (M.B.,manuscript in preparation).

The fourth sub-order of hemipterans, the Heteroptera, contains a wide diversity of bugs, but only two families have species that are reported to jump. This paper analyses the jumping mechanisms of Saldula saltatoria, a member of one of these families, the Saldidae(Polhemus, 1985), which lives on the muddy banks of freshwater and is known colloquially as a shore bug. Its hind legs, which it uses to propel jumping are shorter than those of other,supposedly non-jumping species in the same family, and have few specialisations compared with froghoppers. Two distinct jumping strategies are used. First, the hind legs are accelerated rapidly about the coxo-trochanteral joints while the wings remained closed. Second, the same rapid movements of the hind legs are used, but before they lose contact with the ground, the wings are opened and begin to move in the wing beat cycle. This study therefore investigated whether the early opening of the wings will increase drag and therefore slow the take-off velocity, or can contribute power and therefore increase the take-off velocity. Are jumping and flying co-ordinated to achieve the highest take-off velocity or the most stable trajectory after take-off? What therefore are the relative benefits of these two jumping strategies?

Adult Saldula saltatoria (Linnaeus 1758) were collected from the muddy edges of ponds in northern Cambridgeshire, England. Our experience in collecting was reminiscent of that of Butler(Butler, 1886) who commented`Much hunting yields but few Saldas (= Saldula), and he who desires to cultivate the virtue of patience, cannot do better than try a day's Salda-hunting'. They belong to the order Hemiptera, sub-order Heteroptera and to the family Saldidae.

Sequential images of jumps were captured at rates of 5000 frames s^–1 and with an exposure time of 0.05 ms with a Photron Fastcam 1024PCI high speed camera [Photron (Europe) Ltd, Marlow, Bucks., UK]. The images were fed directly to a laptop computer. The insects were manoeuvred with a fine paint brush into position in front of the camera pointing directly toward the middle of a chamber 80 mm wide, 80 mm tall and 10 mm deep at floor level and widening to 25 mm at the top. The floor, which was horizontal or a few degrees from the horizontal, was of high density foam. Within the chamber, Saldula was free to jump in any direction but the shape of the chamber meant that most jumps were in the image plane of the camera. Detailed measurements of changes in joint angles and distances moved were made from jumps that were parallel to the image plane of the camera, or as close as possible to this plane. Calculations show that jumps that deviate to either side of the image plane of the camera by ±30° would result only in a maximum error of 10% in the measurements of leg joint, or body angles. Sequences of images were analysed with Motionscope camera software (Redlake Imaging, Tucson, AZ, USA), to determine the position of points on the body or hind legs, or with Canvas X (ACD Systems of America, Miami, FL, USA) to determine angular changes. A point on the body that could be recognized in successive frames and was close to the centre of mass, as determined by balancing the insect on a pin, was selected for measurements of the velocity and trajectory of the body and is indicated in Fig. 7D. The time at which the hind legs lost contact with the ground and the insect became airborne was designated as t=0 ms so that different jumps could be compared and aligned. The time at which the hind legs started to move and propel the jump was also labelled and the time between these two events therefore defined the period over which the body was accelerated in a jump. Peak velocity was calculated as the distance moved in a rolling three point average of successive frames (a 0.6 ms moving time window). Acceleration was also measured from a rolling three point average during the acceleration period,with the peak value given in Table 3. Photographs and anatomical drawings were made from both live and preserved specimens. Data are based on 34 jumps by seven Saldula saltatoria recorded at 24°C. Seven museum (University Museum of Zoology, Cambridge, UK) specimens of each of the following five species of saldidids were measured to determine their leg and body lengths: Saldula littoralis (Linnaeus 1758), Saldula scotia (Curtis 1835), Saldula pallipes (Fabricius 1794), Saldula c-album (Fieber 1859) and Chiloxanthus pilosus (Fallen 1807). Measurements are given as means ± standard error of the mean (s.e.m.). Movies, captured at 5000 frames s^–1 and with an exposure time of 0.05 ms, of a jump by Saldula with wings closed and of a jump with wings open are available as supplementary material Movie 1 and Movie 2, respectively.

The adult Saldula used here had a mass of 2.1±0.09 mg (mean± s.e.m.) and a body length of 3.5±0.09 mm (N=7; Table 1). The antennae were approx. 1.5 mm long or about 45% of the body length. The hardened front wings covered the whole abdomen dorsally and were darkly coloured with some patches of white. By contrast, the hind wings were thin, membranous and white. The prominent sucking mouthparts extended ventrally from the head as far posteriorly as the coxae of the middle legs and were encased in a curved,pointed and sclerotised sheath.

The front legs of Saldula were 1.8±0.08 mm long(N=7), the middle legs 2.0±0.06 mm, and the hind legs 3.2±0.08 mm, so that the ratio of leg lengths was 1 front: 1.2 middle:1.8 hind (Table 1). In total the hind legs represented 89.6±3.26% of the body length whereas the front and middle legs represented 50.3±3.32% and 58.2±3.24%respectively. The longer hind legs resulted largely from their tibiae which at 1.6±0.02 mm (N=7) were 93% longer than the middle and 138%longer than the front tibiae. By contrast, their femora were 1.0±0.04 mm (N=7) and only 25% longer than the middle and 46% longer than the front femora. Expressed as the cube root of body mass, the hind legs have a ratio of 2.5, similar to that of long-legged leafhoppers(Burrows and Sutton,2008).

The legs of S. saltatoria were compared with those of five other species within the family Saldidae that apparently do not jump but have a similar body shape and length (Table 1). The ratios of the lengths of the three pairs of legs were similar to those of S. saltatoria, but relative to the length of the body all had longer hind legs. This suggests that the length of the hind legs does not predict jumping ability.

Jumping was powered by the rapid and simultaneous depression of both hind legs about their coxo-trochanteral joints, with the large depressor muscles located in the thorax. Each hind coxa of S. saltatoria was large relative to the size of the metathorax(Fig. 1B, Fig. 2) and could rotate forwards and backwards through a small angle relative to a reference point on the mesothorax. The coxae of the two hind legs were apposed to each other at the midline (Fig. 2A). The ventral surface of a hind coxa was sculpted to accommodate the femur when the trochanter was levated and the hind leg was swung forward into its preparatory position for jumping (Fig. 2A,B).

A hind trochanter pivoted with the coxa about a ventral(Fig. 2) and a dorsal articulation in which curved horns of the trochanter inserted into sockets on the coxa. It could be levated (flexed) and depressed (extended) through an angle of ∼140° in a plane approximately parallel with the underside of the body. The large tendon of the trochanteral depressor muscle attached to the medial rim of the trochanter. The main body of the depressor muscle was located in the thorax and was large relative to the trochanteral levator muscle, which was entirely within the coxa.

A hind femur could move through a small angle about the trochanter. Its femoro-tibial joint showed no outward specialisations for jumping(Fig. 2A). A hind tibia was longer than the femur and had a semi-circular array of spines on its ventral surface at the tibio-tarsal joint. A tibia could extend and flex through an angle of ∼160–170° in the same plane as the levation and depression movements of the trochanter about the coxa. A tarsus also had arrays of spines pointing ventrally (Fig. 2A), which, like those at the tibio-tarsal joint, seem appropriately placed to increase traction with the ground when jumping.

Saldula used two strategies for jumping. In nine of the 34 jumps analysed (26%), the wings remained closed and were not moved either before or at take-off (Figs 3, 4), and in 25 jumps (74%) they were opened before take-off and then flapped (Figs 5, 6). The jumping data (take-off velocity, acceleration, take-off angle and body angle at take-off) from the seven individuals were normally distributed and analysed using parametric statistics.

In the jumps in which the wings were closed at take-off, no data are available as to whether flight is assumed at some later point in the jump trajectory. All jumps were performed in the same arena and occurred spontaneously once the insect had been manoeuvred into position in front of the camera. There were no obvious stimuli or circumstances that were associated with one jumping strategy rather than the other. Similarly, there were no detectable trends in performance, or in the use of the wings during a sequence of jumps by an individual shore bug.

In this first strategy for jumping, the hind legs executed three distinct phases of movement while the wings remained firmly closed and were not used either during the launch into take-off in a jump or initially when airborne(Figs 3, 4). The three phases of movements by the hind legs were:

First, in preparation for a jump, the hind legs were drawn fully forwards by levation of their coxo-trochanteral joints so that the femora lay firmly pressed into the hollowed regions of the ventral coxae. The tibiae were also flexed about the femora but the two segments were not closely apposed to each other along their length. The result of these movements, which lasted up to a few hundred milliseconds, was that the tips of the tarsi of the two hind legs were placed on the ground just outside the lateral edges of the body and therefore did not touch each other (Fig. 4).

Second, the hind legs were held motionless in this fully levated position for periods up to 400 ms. During this time, movements of the front and middle legs adjusted the angle of the body relative to the ground and also changed the azimuth orientation of the body, thus determining the azimuth direction of the jump.

Third, the rapid jump movement itself resulted from a powerful depression of the coxo-trochanteral joints of both hind legs at the same time, and an accompanying extension of the femoro-tibial joints (Figs 3, 4). The movements of the coxo-trochanteral joints could most clearly be seen in the ventral views of the body (e.g. Fig. 4). No differences in the start of the movements of the two hind legs could be detected with the resolution (0.2 ms) available in these sequences captured at 5000 frames s^–1. The movements will thus be called synchronous within this constraint of resolution. The time from the first movements of the hind legs until they lost contact with the ground at take-off, a period that defined the time over which the insect was accelerated in a jump, was 3.97±0.111 ms (N=9). During this period a trochanter progressively depressed at rotational rates of 25,000 deg s^–1 with the result that the femur moved downwards and backwards relative to the body (the most obvious movement of the hind leg when viewed from the side, as in Fig. 3), and with the tibia also extending progressively. The continuing thrust generated by these movements moved the body forwards and raised it progressively from the ground. The consequence of the body being raised was that both the front and middle legs lost contact with the ground before the hind legs. During the latter part of the jump, propulsion could therefore only be provided by the hind legs. Even in the earlier part, the middle and front legs showed no consistent movements to suggest that they were providing much propulsive force. Take-off occurred when the coxo-trochanteral and femoro-tibial joints of the two hind legs were almost fully depressed and extended, respectively. After the tarsi had lost contact with the ground the momentum of the movements of these two joints resulted in the tibiae crossing beneath the abdomen (Fig. 4).

The same sequence of movements of the hind legs also characterised this second strategy for jumping, but it was preceded by opening movements of both pairs of wings at variable times before take-off. The wings were first held elevated and after a variable period were depressed and elevated in a flight pattern which began 12.8±0.58 ms (range 8.4–21 ms; N=25)before take-off (Figs 5, 6). By contrast, the time from the first movement of the hind legs until the insect became airborne was 3.86±0.055 ms (N=25) and these movements were always accompanied by a forward and upwards displacement of the body. To compare this acceleration time with that in jumps when the wings were not moved, a single sample t-test was used because the sample sizes for the two jumping strategies was different (Table 2). Comparing the smaller sample of jumps when the wings were closed with the mean of larger sample when wings were opened shows that times are not different at the 5% level (t₇=1.5, P=0.177). There is thus no significant difference in the acceleration times of Saldula when jumping with or without the contribution of wing movements. At the point of take-off in different jumps, the wings were in the depression (Fig. 5) or elevation (Fig. 6A) phase of the wing beat cycle and at different positions relative to the body in either phase (Fig. 6B). This indicates that there was no strict temporal relationship between the cycle of the wing beat and the timing of the propulsive movements of the hind legs in a jump.

Jumping using the first strategy in which propulsion is generated by the hind legs while the wings remain closed, propelled Saldula to a take-off velocity of 1.28±0.119 m s^–1 (nine jumps by four Saldula; Fig. 7A, Table 2). In the second strategy in which the wings were opened and moved, the take-off velocity achieved was 1.29±0.032 m s^–1 (25 jumps by the same four Saldula) (Fig. 7B, Table 2). The best performance in either strategy was a take-off velocity of 1.8 ms^–1. The peak velocity was attained just before take-off but was usually maintained for the few milliseconds that the insect remained in the frame of the camera (Fig. 7A,B).

The take-off angle when the wings remained closed and the jump was powered only by leg movements was 52±6.5 degrees (N=9) and when wing movements were also used the take-off angle was 58±1.7degrees(N=25; Table 2). The attitude of the body relative to the ground was 47±4.7 degrees(N=9) when legs alone were moved compared to 54±2.6 degrees(N=25) when the wings were also moved. Using the same statistical test as used above in comparing the acceleration times, none of the parameters of the jumps in the two strategies differed significantly at the 5% level[take-off velocity (t₆=0.962, P=0.373), take-off angle (t₅=1.377, P=0.227) or body angle at take-off (t₅=1.304, P=0.249)]. The initial trajectories of the jumps were also similar when either strategy was used(Fig. 7C,D), but would be expected to diverge because the wings should enable greater heights and distances to be achieved.

Jumping performance was calculated from the data obtained from the high-speed images (Table 3). The average acceleration over the whole of this period was 335 m s^–2 (average of 32 jumps with wings open or closed) rising to 529 m s^–2 in its best jumps so that it would experience forces up to 54 g. The energy required to achieve these performances (mean or best) was 1.8–3.4 μJ, the power output was 0.5–1.0 mW and the force exerted was 0.7–1.1 mN. Assuming that, as in froghoppers (Burrows,2007c) the mass of the muscles generating propulsive movements of the hind legs represents about 11% of body mass, then the power per mass of muscle is 4500 W kg^–1. This far exceeds the power that muscle could produce by direct contraction(Alexander, 1995; Vogel, 2005b) and indicates that contractions must begin before the jump with energy being stored and then released suddenly.

This investigation shows that shore bugs use the same rapid and simultaneous movements of both hind legs in two distinct strategies when jumping; in the first, the jump is propelled by rapid depression movements of the trochantera of both hind legs accompanied by extension of both tibiae, and in the second these same leg movements are preceded by opening of both pairs of wings and subsequent elevation and depression movements in the wing beat cycle before take-off is accomplished. The main propulsive forces are applied to the trochantera by depressor muscles located in the metathorax. The hind femora are thin indicating that the extensor tibiae muscles are correspondingly small and cannot therefore contribute greatly to the forces generated in jumping. Secondly, both hind legs are beneath the body, like a froghopper which propels its jumps with large trochanteral depressor muscles,and not arrayed along the side of a body, as in a grasshopper which propels its jumps with large extensor tibiae muscles.

Where does the jumping performance place Saldula among other Hemipterans and amongst other insects that power jump by movements of the legs. Saldula accelerates its body in 4 ms to a take-off velocity in its best jumps of 1.8 m s^–1. The acceleration of froghoppers(Auchenorrhyncha, Cercopidae) is more than four times higher and the take-off velocity (4.7 ms^–1) 2.6 times greater. In long-legged leafhoppers (Auchenorrhyncha, Cicadellidae), which like Saldula have hind tibiae that are longer than the femora, the mean acceleration times range from 4.4 to 6.4 ms. Saldula matches the best take-off velocities of four species of cicadellids that range from 1.6 to 1.85 m s^–1, being bettered only by Aphrodes at 2.9 m s^–1 (Burrows,2007b). Its performance also exceeds that of Hackeriella(Coleorrhyncha, a sister sub-order to the Heteroptera), which has a mean acceleration time of 2 ms and a best take-off velocity of 1.5 m s^–1 (Burrows et al.,2007).

Compared with other insects, the acceleration time of Saldula is longer than that of a flea (Siphonaptera) but its take-off velocity is higher(Bennet-Clark and Lucey, 1967). Its best take-off velocity is comparable to that achieved by flea beetles(Coleoptera, Alticinae) (Brackenbury and Wang, 1995) and many bush crickets (Orthoptera, Ensifera,Tettigoniidae) (Burrows and Morris,2003), but falls short of that achieved by heavier insects such as the false stick insect Prosarthria teretrirostris (Orthoptera,Caelifera, Proscopiidae) with a mass of 280 mg and which takes 30 ms of acceleration to achieve a take-off velocity of 2.5 m s^–1, and 1–2 g locusts (Orthoptera, Caelifera, Acrididae) which accelerate in 20–30 ms to a take-off velocity of 3.2 m s^–1(Bennet-Clark, 1975).

Saldula should jump a distance of 170 mm and reach a height of 50 mm when there is no contribution from active wing movements when using its mean take-off velocity of 1.3 m s^–1 and its mean body angle at take-off of 52°. These distances would rise to 320 mm and 105 mm,respectively, if it used its best take-off velocity of 1.8 m s^–1. This indicates that Saldula could jump to a height equivalent to 30 times its body length. On this measure it is again outperformed by both froghoppers and fleas (both more than 100 times body length), but almost matches the 40 times achieved by a short legged cicadellid(Ulopa) of similar mass and length(Burrows and Sutton, 2008). These estimates for Saldula, however, take no account of the wind resistance that is likely to be offered by a small body moving at such high velocities. Vogel has estimated that the froghopper Philaenus which has a mean mass of 12 mg and a mean length of 6.1 mm(Burrows, 2006a) would lose some 25% of its jumping range because of drag, a smaller flea beetle 40% and an even smaller flea 80% (Vogel,2005a). Given its size and mass a shore bug would thus be expected to achieve only about half its calculated range.

Saldula has few specialisations for jumping particularly in comparison with the most adept jumpers within its order, the Hemiptera. The length of the hind legs themselves is not remarkable when compared with non-jumping members of its family. Although the proportions of the three pairs of legs are similar in both jumping and non-jumping species of this family,the hind legs are shorter relative to the length of the body than those in the non-jumping species. Saldula has hind legs that are 90% of the body length and relative to the cube root of body mass, have a ratio of 2.5, close to the value for long-legged leafhoppers(Burrows and Sutton, 2008),but greater than for froghoppers with proportionately shorter hind legs. The length of the hind legs does not, however, influence take-off velocity when jumping is powered by a catapult mechanism(Burrows and Sutton, 2008). Saldula requires a power output per mass of muscle of 4500 W kg^–1 to jump and this is unlikely to be generated by direct muscle contractions acting on the long lever arms of the hind legs(Alexander, 1995; Vogel, 2005b). Instead the implication is that the trochanteral muscles must contract slowly in advance of the jump, store force, probably in skeletal structures, and then release it suddenly in a catapult action. This is the strategy used by froghoppers(Burrows, 2003; Burrows, 2006a; Burrows, 2006b; Burrows, 2007c) and leafhoppers (Burrows, 2007a; Burrows, 2007b) and by some other insects as diverse as fleas(Bennet-Clark and Lucey, 1967)and locusts (Bennet-Clark,1975). The advantage of long legs may therefore lie in the lower ground reaction forces exerted compared with insects that have shorter hind legs, enabling them to jump effectively from the mud on which they live.

None of the structures found on the proximal joints of the hind legs of members of two families of auchenorryhnchan bugs so far described and thought to be specialisations for jumping, were found in Saldula. For example, the two hind coxae although enlarged and apposed at the midline have no linking structures similar to the fields of microtrichia in froghoppers(Burrows, 2006b), or the press-studs (poppers) in long-legged cicadellids(Burrows, 2007a; Emeljanov, 1987). They thus more closely resemble the coxae of coleorrhynchans(Burrows et al., 2007) or short-legged cicadellids (Burrows and Sutton, 2008). Similarly there are no obvious proprioceptors in the same positions on the coxa as in either froghoppers(Burrows, 2006b) or cicadellids (Burrows, 2007a)that could signal engagement of a femur with the hollowed ventral region of a coxa, or movements of the coxa relative to the metathorax. This suggests that the control of force in the build up to a jump is less critical than in their relatives that achieve higher take-off velocities. Finally, there are no structures like those in froghoppers(Burrows, 2006b) which are covered in microtrichia and restrain a femur against a coxa while the trochanteral depressor muscles contracts to generate the force required for a jump.

Little information is available on the natural ecology of these insects to suggest why they might jump. They live and forage for food on the exposed surface of the muddy shores of fresh water ponds and other larger expanses of fresh water. They feed by sucking out the contents of dead insects trapped in the mud. This life style may thus expose them to predation by other animals and in particular birds, from which their camouflaged colouration may provide insufficient protection. A rapid jump that leads directly to flight may thus be an important survival strategy. This is supported by the result presented here that in almost three quarters of the jumps recorded, the wings were opened and then elevated and depressed in the wing beat cycle before the hind legs had lost contact with the ground and the insect become airborne. In the remaining quarter of the jumps the hind legs alone were moved rapidly while the wings remained firmed closed and covered the body. Neither strategy produced a faster take-off so what advantage accrues from using one strategy rather than the other?

Are wing movements used to stabilize the body once airborne or to ensure a smooth transition from jumping to flying? The bodies of many jumping insects rotate about either the longitudinal or transverse axes (or sometimes even both axes) once airborne following a jump. This means that they could potentially lose energy because of these rotations and also increase the probability that the jump will end in an uncontrolled landing. The similarly-sized cicadellid Ulopa loses as much as 13% of its kinetic energy to rotation as it spins with wings closed at frequencies up to 89 Hz about the transverse and longitudinal axis of its body after take-off(Burrows and Sutton,2008).

Any improvement in stability from opening the wings will, however, be at the potential expense of increased drag and thus diminished jump height and distance which may be a costly solution if jumping to escape a predator. Estimates of the drag likely to result when the wings of Saldula are opened during take-off indicate that the deceleration would be only about a tenth of the acceleration provided by the forces generated by the hind legs. These estimates are made by assuming that the wings are elliptical plates of known area moved at the maximum take-off velocity measured, through air with a density of 1.3 kg m³. Because drag is proportional to the square of velocity, it will be lower at the slower velocities at the start of a jump when the body is being accelerated. Take-off velocity should not therefore be reduced by the opening of the wings, as is observed, but stability could be improved. The body of Saldula is flattened dorsoventrally and may thus offer resistance to rotation, explaining why a jump with the wings closed is also stable. Keeping the stiff front wings closed means that they help with streamlining the body and the reducing drag. This presumably explains why some insects, including the fastest and most powerful jumpers like froghoppers(Burrows, 2003; Burrows, 2006a), tolerate rotation and a potentially unstable landing. In flea beetles, opening and then moving the wings stabilises the body against spinning and results in better targeting and a much higher proportion of jumps ending with a feet-first landing (Brackenbury and Wang,1995). For shore bugs, the body was stable and there were no indications of rotation even when the wings remained closed over the first few milliseconds of a jump that were available for analysis. Wing movements might,however, provide more sustained stability later in the trajectory particularly in the face of cross currents or turbulence.

The movements of the wings during the acceleration of the body in a jump could provide a smoother and faster translation into flight. If the opening of the wings were delayed until the insect were airborne, there could be two impairments to performance. First, the forces experienced at the peak velocity experienced at take-off might impede the opening the wings, particularly the membranous hind wings. We know that the drag forces from the wings are small relative to those exerted on the body by the legs, but we do not know how they compare with the forces generated by the muscles moving the wings. Second,opening the wings while airborne might impede the forward and upward momentum of the jump with a potentially disastrous outcome if the movement is for escape. The strategy observed in Saldula is to establish the flight pattern before take-off and then have the jump lead to a smooth transition to flight and maintenance of forward speed. Wing movements also commonly accompany or even precede the leg movements of a jump in the small but long-legged cicadellid Empoasca, but less frequently in larger members of the same family (Burrows,2007b). The wing movements of Empoasca, sometimes preceded the leg movements of a jump, but the first depression movement began only after take-off. In flea beetles, the downward movement of the wings is said to be `exactly synchronised with hind-leg extension at take-off'(Brackenbury and Wang, 1995)and in mantids strong coupling between leg extensors and wing depressors is inferred from single still images taken during take-off(Brackenbury, 1990; Brackenbury, 1991). A jump also often launches an adult locust into flight with the wings generally opening 15-35 ms after take-off, but sometimes they open before this(Camhi, 1969; Pond, 1972). The implication is therefore that there should be a close link between the interneurons initiating and controlling flying and jumping, but few natural stimuli that elicit jumping or kicking in locusts have been used to test this linkage. Moreover, many jumps do not lead to flight. Recordings from locust flight motor neurons show that their activity in a flight pattern is initiated at variable times before or after the time that a kick (not a jump) by both hind legs is released (Pearson et al.,1986). The conclusion from intracellular recordings from individual neurons is that those interneurons that trigger jumping are not involved in initiating flying, but those that maintain the linkage between these two movements remain to be identified. The lack of coupling between the phase of the wing beat cycle and the release of the power developed by the hind legs for jumping in both shore bugs and locusts suggests that there is little aerodynamic advantage in linking the two motor patterns more closely;it is sufficient that the pattern of wing movements is established around the time of take-off when the jump is a launch into flight.

I am very grateful to Roger Northfield for help in collecting these bugs,to Greg Sutton and Steve Rogers for many enlightening discussions and much help, and to other Cambridge colleagues for their many helpful suggestions during the course of this work and their comments on the manuscript.