Jumping mechanisms and performance of snow fleas (Mecoptera, Boreidae)

The flightless adults of snow fleas (or snow scorpion flies), Boreus, are active only during the colder winter months and will often walk or jump on snow, or on the underlying moss on which they feed, at temperatures of –3 to +3°C. Marshall notes that they ‘often jump straight up when you approach them for a close look, landing back in the snow with their appendages folded against the body so that they resemble inanimate specks of dirt on the snow’ (Marshall, 2006). Their ability to jump is said to be unique among Mecoptera (Whiting, 2002), but it enables them to escape from predators and to traverse snow, upon which walking is difficult. An analysis of their jumping mechanisms is, however, lacking despite their being thought to be the closest extant relatives of the fleas (order Siphonaptera). This close relationship is indicated by molecular studies of four genetic markers (Whiting, 2002; Whiting et al., 2008) and by anatomical and other phenotypic features (Grimaldi and Engel, 2005). To the criteria that can be used to support this inferred phylogenetic relationship, this paper now adds details of the jumping mechanisms of Boreus. Some of these mechanisms are clearly in common with those of fleas (Bennet-Clark and Lucey, 1967; Rothschild and Schlein, 1975; Rothschild et al., 1975; Rothschild et al., 1972; Sutton and Burrows, 2011), while the intriguing differences suggest a clear evolutionary line.

Jumps in insects are usually propelled by the rapid movements of a pair of legs, although other parts of the body may be used by some groups. For example, in Collembola a jump is propelled by extension of an abdominal appendage (Brackenbury and Hunt, 1993; Christian, 1978; Christian, 1979) and a click beetle jack-knifes its body at the joint between the prothorax and mesothorax (Evans, 1972; Evans, 1973; Kaschek, 1984). In some ants (Baroni et al., 1994; Tautz et al., 1994) and a stick insect (Burrows and Morris, 2002), movement of the abdomen adds to the propulsion from the legs.

Where the legs are used, it is most commonly the single pair of hindlegs that propel jumping, although small flies such as Drosophila use their middle legs (Zumstein et al., 2004). The hindlegs of jumping insects are arranged mechanically in one of two ways; in locusts and fleas the hindlegs move in planes laterally displaced on either side of the body, whereas froghoppers, leafhoppers and planthoppers (Hemiptera, Auchenorrhyncha) use an undercarriage arrangement in which they move in the same plane beneath the body. The mechanics of these different arrangements impose particular constraints upon the jumping mechanisms, resulting, for example, in the elevation and azimuth directions of a jump being controlled in different ways (Sutton and Burrows, 2008; Sutton and Burrows, 2010).

A further difference in the jumping mechanisms results from the use of different sets of leg muscles to generate the required forces. In insects such as locusts (Bennet-Clark, 1975; Heitler and Burrows, 1977) and flea beetles (Brackenbury and Wang, 1995), the extensor tibiae muscles in the hind femora generate the force, but in fleas (Bennet-Clark and Lucey, 1967) and in plant-sucking bugs (Auchenorrhyncha), force is generated by the trochanteral depressor muscles within the thorax (Burrows, 2007b; Burrows, 2007c; Burrows and Bräunig, 2010). The power in the first example comes from rotation of the tibiae and in the second from rotation of the trochantera. Whichever muscles are used, the same demands exist for high take-off velocities and short acceleration times, and this means that a catapult mechanism has to be used because the legs are short. A catapult mechanism allows the power-producing muscles to contract slowly and store energy in distortions of the skeleton, which can then be released suddenly to power the jump. The energy stores are diverse but a role for the elastic protein resilin (Weis-Fogh, 1960) has been implicated in fleas (Bennet-Clark and Lucey, 1967) and demonstrated in froghoppers (Burrows et al., 2008) and planthoppers (Burrows, 2010). A known exception to the use of a catapult mechanism is in bush crickets, which have very long hindlegs that provide sufficient leverage for direct muscular contractions to propel a jump (Burrows and Morris, 2003).

This paper analyses the jumping mechanisms of the snow flea, Boreus, to understand how it fits into the emerging picture of the general principles that underlie jumping in insects. Of particular interest is whether these mechanisms can also shed light on the evolutionary relationships between this group of insects and the fleas. A brief report (Edwards, 1987) has suggested some of the mechanisms that might be used. The anatomy of the thorax of Boreus (Fuller, 1954) shows that the legs are arranged at the sides of the body and details of its thoracic musculature (Fuller, 1955) indicate that large trochanteral muscles in the thorax may power jumping. Both features are shared by true fleas. High speed images of jumping presented here show that the two middle legs and the two hindlegs move together to power jumping. Paralleling the use of four legs, four resilin pads are revealed that are associated with each middle leg and hindleg, but not with the front legs. In contrast, in fleas the hindlegs are the sole provider of power for jumping and only two pads of resilin are found associated with them.

Adult male and female Boreus hyemalis (Linnaeus 1767) were caught in January and February of 2008–2010 in pit fall traps laid in sandy soil beneath moss near Santon Downham and Lakenheath, Norfolk, England. They were maintained in the laboratory for a few days feeding on their host moss at temperatures of 4–5°C. They belong to the order Mecoptera (scorpion flies) and to the family Boreidae that consists of 24–26 species in three genera of Holarctic insects (Grimaldi and Engel, 2005). Colloquially they are known as snow fleas, but so too are springtails (Collembola) such as Hypogastrura nivicola. They are also, and less confusingly, known as snow scorpion flies. The results are based on an analysis of 28 adult Boreus.

Sequential images of jumps were captured at rates of 5000 s^–1 with an exposure time of 0.1 ms using a single Photron Fastcam 1024PCI high speed camera [Photron (Europe) Ltd, West Wycombe, Bucks, UK], and were fed directly to a laptop computer. The camera, with a 100 mm micro Tokina lens, pointed directly at the middle of a glass chamber 80 mm wide, 80 mm tall and 10 mm deep at floor level and widening to 25 mm at the top. The floor of the chamber was of high density foam. Boreus was free to jump in any direction (see supplementary material Movies 1–3 for jumps viewed from the side, above and behind, respectively) but the shape of the chamber constrained most jumps to the image plane of the camera. 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. Jumps that deviated to either side of the image plane of the camera by ±30 deg were calculated to result in a maximum error of 10% in the measurements of joint or body angles. Sequences of images were analysed with Motionscope camera software (Redlake Imaging, Tucson, AZ, USA), or with Canvas 11 (ACD Systems of America, Miami, FL, USA). To allow different jumps to be aligned and compared, the time at which the hindlegs and middle legs lost contact with the ground and the insect became airborne was designated as t=0 ms. The time at which these legs started to move and propel a jump was also determined so that the time between these two events defined the period over which the Boreus actively accelerated. Peak velocity was calculated as the distance moved in a rolling 3 point average of successive frames. One-hundred and nineteen jumps by 18 Boreus (12 males and 6 females) were captured at temperatures of 23–24°C and analysed to determine jumping performance. Measurements are given as means ± s.e.m.

The external anatomy of the legs was examined in intact Boreus and after fixation in 70% alcohol or 50% glycerol. Dried specimens were mounted on specimen holders, sputter coated with gold and then examined in a Philips XL-30 scanning electron microscope. To reveal the presence of the rubber-like protein resilin, dissected Boreus were viewed through Olympus Mplan 10×/0.25 NA, and LUCPlanFLN 20×/0.45 NA objective lenses, under ultraviolet (UV) or white epi-illumination on an Olympus BX51WI compound microscope. UV light from an X-cite series 120 metal halide light source was conditioned by a Semrock DAPI-5060B Brightline series UV filter set (Semrock, Rochester, NY, USA) with a sharp-edged (1% transmission limits) band from 350 to 407 nm. The resulting blue fluorescence emission was collected at wavelengths from 413 to 483 nm through a dichroic beam splitter.

Adult Boreus are flightless, so their forms of locomotion are restricted to walking and jumping. Walking was affected by the low temperatures at which Boreus lives in the winter. At 8°C walking speeds of 14 mm s^–1 were measured but at 3°C speeds fell to as low as 1 mm s^–1. In adult females, both pairs of wings are greatly reduced, and in adult males they are modified to form dorsally protruding and backwardly curved hooks (Fig. 1) which are used to grab a female in mating. All parts of the body are darkly coloured. The head in both sexes has a large ventrally pointing rostrum, or beak, with biting mouthparts at the end, and prominent articulated antennae (64±2.9% of the body length, mean ± s.e.m., N=6 males) that are often rested on the ground, particularly at the start of a jump. Females have a large curved ovipositor protruding prominently from the posterior of the abdomen. Females weighed 4.2±0.37 mg (N=9), whereas males weighed only 2.9±0.28 mg (N=10) (Table 1). Females had a body length of 3.6±0.26 mm (N=7), excluding the length of the ovipositor, compared with males at 3.4±0.2 mm (N=13). Females are therefore significantly heavier (t-test, t₁₇=2.881, P=0.01) than males but are not significantly longer (t-test, t₁₈=0.574, P=0.73), when the ovipositor is discounted.

The mesothorax and metathorax are fused to form a strong and rigid cuticular box supporting the middle legs and hindlegs, whereas the prothorax supporting the front legs is separately articulated (Fig. 2A, Fig. 3A). The boundary between the episternum and epimerum of both the mesothoracic and metathoracic segments is marked by an indentation. This represents the pleural ridge that extends from the wing hinge dorsally to the articulation of the coxa ventrally. The surfaces of the thoracic cuticle and the legs have a dense covering of fine hairs ranging from 15 to 35 μm in length. The front legs are the shortest and in a male are 3.5 mm long, whereas the middle legs are 4.4 mm long and the hindlegs are the longest of all at 5.4 mm (Table 1). Relative to the front legs this gives a ratio of leg lengths in males of 1:1.3:1.7 (front:middle:hind) and in females of 1:1.2:1.7. In males, the middle legs represent 140% and the hindlegs 187% of the body length, while in females the values are smaller when the length of the ovipositor is excluded. Proportional to each other, to the length of the body and to the body mass, this makes the leg lengths of the snow flea similar to those of a flea (Table 1).

The legs move in two separate planes on each side of the body with the body slung between them. The middle and hind coxae of a male are each about 0.6 mm long and 0.2–0.25 mm wide and articulate with the fused mesothoracic and metathoracic box (Fig. 2B,C and Fig. 3A,B). These articulations allow small forward and backward rotations of these coxae. The front coxa, in contrast, is somewhat shorter. The hind and middle trochantera of a male are each about 0.2 mm long and articulate through an angle of about 140 deg with their respective coxa (Fig. 2D). The anterior and posterior hinges of these articulations mean that levation of the joints swings the more distal segments of each leg dorsally and forwards (Fig. 3A). Accordingly, when fully levated, as in preparation for a jump, both femora are pressed against the lateral surfaces of the fused epimera and episterna, with the femoro-tibial joints pointing dorsally. The movements of the coxo-trochanteral joint would appear to be monitored by arrays of proprioceptive hairs (Fig. 2D, Fig. 3B–D). One group of 8–10 hairs, 40–50 μm long is associated with a dorsally pointing cuticular protrusion (Fig. 2D). Other groups are associated with the two articulations of each of the middle legs and hindlegs. On the middle leg, there is a group of about 20 hairs, 10–12 μm long on both sides of each articulation (Fig. 3B,C). On the hindlegs, there is a similar arrangement of hairs though some of them may be up to 20 μm in length (Fig. 3B,D).

The joint between the trochanter and femur consists of a dorsal and ventral articulation that allows the femur to move through only a small angle. The hind femur of a male is 1.4 mm long but only 0.12 mm in diameter at its widest point, and the middle femur is 1.1 mm long and about the same diameter (Fig. 2E). There is therefore little space in either the middle or hind femora in which to accommodate an extensor tibiae muscle capable of generating force that might make a substantive contribution to jumping movements. The femoro-tibial joints of both hindlegs and middle legs consist of a lateral and a medial articulation that allow movements of some 170 deg. A hind tibia is 1.5 mm long but only 0.07 mm wide. The longest parts of the middle legs and hindlegs are the tarsi, which in a male are 1.5 and 2.0 mm long, respectively. Arrays of ventrally pointing spines are present at the articulation of the five tarsal segments.

High speed images of jumps were captured from different angles by a single camera (Figs 4, 5, 6, 7) so that the orientation and movements of the legs could be interpreted in all three dimensions. The majority of the jumps were viewed from the side (male in Fig. 4, female in Fig. 5; supplementary material Movie 1), supplemented by views from above (Fig. 6 and supplementary material Movie 2) and behind (Fig. 7 and supplementary material Movie 3) to give further information about the angles of femora relative to the body and of the femoro-tibial angles. The following description combines information from all views, while the angular changes of the leg joints during a jump are plotted in Fig. 8. The sequence of leg movements in jumping was observed to be the same in males and females.

Jumps occurred from a standing start, or followed a period of walking after a short and variable pause. All were preceded by movements of the middle legs and hindlegs into their most levated positions by rotation about the coxo-trochanteral joints. These levation movements moved the femora into either a vertical or a slightly forward-pointing position so that proximally their medial surfaces were closely pressed against the epimera and episterna of the fused mesothorax and metathorax (Figs 4 and 5). The femora of the middle legs and hindlegs were parallel to each other and were sometimes splayed laterally from the body (Figs 6 and 7). Their femoro-tibial joints protruded above the dorsal surface of the thorax (Figs 4 and 5) and were flexed so that the tibiae and tarsi were drawn forward. The degree of flexion of the femoro-tibial joints varied in different jumps indicating that full flexion was not a pre-requisite for jumping. The front legs, in contrast, adopted different angles prior to different jumps. The antennae were usually held elevated before a jump so that they pointed anteriorly. In preparation for most jumps they were moved downwards so that the distal third of their segments rested on the ground (Figs 4 and 5). The whole body adopted a curved position.

At the start of a jump, the hind tarsi were level with, or just behind, the tip of the abdomen and lateral to each side of the body. The middle legs adopted a similar but more anterior position, with their femora and tibiae parallel to the respective segments of the hindlegs (Fig. 6). The first detectable propulsive motion of a jump, as viewed from the side, was a simultaneous backwards movement of the four middle and hind femora (Figs 4 and 5). These changes were produced by depression of the coxo-trochanteral joints with little rotation evident at the trochantero-femoral joints, with the result that a trochanter and a femur seemed to act as a unit. The tibiae were then extended about the femora so that the body was progressively raised from the ground and the insect was propelled forwards. The antennae initially remained in contact with the ground as the middle legs and hindlegs depressed and extended, pushing the body upwards and forwards. The continuing depression and extension movements of the coxo-trochanteral and femoro-tibial joints of the middle legs and hindlegs then raised the front of the body further so that the curvature of the antennae became less pronounced as fewer of their segments contacted the ground. The movements of the middle legs and hindlegs proceeded in unison with all the particular segments of one leg moving in parallel with the corresponding segment on the other leg joints until they reached almost full depression and extension. At this stage, the middle and hind tarsi were level with, or behind, the tip of the abdomen. The tarsi remained in their same positions outside the lateral limits of the body. The front legs made no obvious movements that could apparently contribute substantive force to the jump. Indeed, the front legs lost contact with the ground only 1–3 ms after the first movements of the other legs and sometimes as much as 4 ms before take-off (Fig. 5). Take-off was marked by the middle and hindlegs losing contact with the ground. After take-off, the tibiae of both the middle legs and hindlegs crossed over and during the initial part of the trajectory trailed behind the body.

The parallel movements of the middle legs and hindlegs were revealed clearly when changes in the angles of their joints were plotted (Fig. 8). The analysis given here is of a jump that was close to the mean performance by male snow fleas. The coxo-trochanteral joint was the first to move, as indicated by changes in the angle of the femur relative to the body in dorsal views (Fig. 8A), or by the coxo-trochanteral joint itself when viewed from the side (Fig. 8C). The first movements of these joints in the hindlegs and middle legs occurred within the same frame (resolution 0.2 ms) in all jumps analysed. Changes in the angle of this joint in the two middle legs and the two hindlegs then proceeded in parallel until they reached full depression at take-off. In contrast, the angle of this joint in a front leg changed little throughout the acceleration period of a jump (Fig. 8A). The initial depression of the coxa was also followed by changes in the femoro-tibial angles as the thrust was applied by the tarsi to the ground in moving the insect forwards (Fig. 8B). A sustained extension of the tibia about the femur was delayed by 2–3 ms relative to the start of the coxo-trochanteral joint movements. Then the femoro-tibial joints of the four middle legs and hindlegs were extended in parallel until take-off (Fig. 8B,D). After take-off, there was sometimes (Fig. 8D) a slight initial flexion of these joints. In contrast, the femoro-tibial joints of the front legs, in addition to their coxo-trochanteral joints, showed no changes consistent with their contributing substantially to propulsion in any jumps.

The acceleration period of a jump, measured from the first observable trochanteral depression to the time when all legs had lost contact with the ground, was 6.6±0.33 ms in males and 6.6±0.17 ms in females (mean of means of 8 males and 6 females performing 87 jumps) (Table 2). These values are not significantly different (t-test, t₁₂=0.164, P=0.872).

The angle of the body relative to the ground was set by movements of the front legs that preceded any propulsive movements by the middle legs and hindlegs. At take-off the body angle, represented by a line joining the tip of the abdomen with the dorsal surface of the head, was 14.0±2.58 deg (range 11–33 deg) in males and 16.2±2.33 deg (range 10–26 deg) in females (mean of means of 9 males and 6 females performing 83 jumps) (Table 2). These values again are not significantly different (t-test, t₁₃=0.58, P=0.572). The initial trajectory of the jump in males was 39.6±2.06 deg (range 32–47.5 deg, mean of means for 7 males performing 29 jumps) and in females was 52.2±4.55 deg (range 41.6–65.5 deg, mean of means for 5 females performing 19 jumps). These figures are significantly different (t-test, t₁₀=2.81, P=0.018). The take-off velocity, measured as a three-point rolling average in a 2 ms period preceding take-off, was 0.8±0.02 m s^–1 in males (mean of means of 29 jumps by 7 males, range 0.72–0.85 m s^–1) and 0.7±0.04 m s^–1 in females (mean of means of 19 jumps by 5 females, range 0.64–0.85 m s^–1). These values are not significantly different (t-test, t₁₀=1.827, P=0.98). The fastest jump was made by a male that achieved a take-off velocity of 1 m s^–1 with an acceleration of 161 m s^–2, which required an energy output of 1.1 μJ and a power output of 0.18 mW, and exerted a force of 0.35 mN or about 16 times its body weight (Table 2 gives figures for mean and best performances of both sexes and the formulae used in the calculations).

The maximum distances jumped at 21.5°C and at less than 10°C were not significantly different (paired sample t-test, t₅=1.963, P=0.107). The time taken to prepare for a jump has, however, been reported to be longer at lower temperatures (Edwards, 1987). The mean distances jumped by males at 21.5°C were 100 mm in males and 85 mm in females (mean of means for 20 jumps by 4 Boreus of each sex).

Once airborne the body remained quite stable. In the pitch plane the rate of rotation, measured from 2 ms before take-off to 2 ms after take-off, was 5.4±0.61 Hz or 1900 deg s^–1 (mean of means of 12 Boreus performing 40 jumps with a minimum of 3 jumps by each, range 2.1–9.5 Hz or 750–3400 deg s^–1). In a 100 mm long jump with a take-off velocity of 1 m s^–1 and an elevation of 40 deg describing a parabolic trajectory, the body is predicted to rotate a maximum of 0.7 times (or by 250 deg, range 100–450 deg) before landing. Only four jumps were recorded with rotation in the roll plane (mean 43 Hz, range 13–62 Hz) and only one jump yawed at 1 Hz.

From the kinematic analysis of the jumping movements, the key movements of both the middle legs and hindlegs were a levation followed by a rapid depression of the coxo-trochanteral joints that provided the propulsive power. Most of the space within both the mesothorax and metathorax is occupied by huge trochanteral depressor muscles [dvm II – musculus dorsoventralis mesothoracis; dvm III – musculus dorsoventralis metathoracis (Fuller, 1955)], which are the largest muscles in the body (Fig. 9A,B). These muscles arise from the mesothoracic and metathoracic notum, and attach to tendons that extend through their respective coxae to insert on the dorsal rims of the trochantera. In contrast, the homologous muscles moving the trochantera of the front legs are much smaller. Muscles levating the trochantera into their cocked position ready for jumping are smaller than the depressors used for propulsion and are located in the coxae (Fuller, 1955).

Assuming that the depressor muscles generating the propulsive movements of a jump represent about 11% of body mass, as in froghoppers (Burrows, 2007c), the power per mass of muscle ranges from 450 W kg^–1 for the best jumps of females to 740 W kg^–1 for the best jumps of males (Table 2). In Boreus, the muscle mass may be greater than in froghoppers because two pairs of legs (the middle and hind) propel jumping. Nevertheless, these values are close to the maximum active contractile limits of striated muscles, which range from 250 to 500 W kg^–1 when operating at high temperatures (Askew and Marsh, 2002; Ellington, 1985; Josephson, 1993; Weis-Fogh and Alexander, 1977). In contrast, Boreus jumps readily at temperatures near 0°C, suggesting that direct contractions of the muscles would be unlikely to be able to power jumping. Some energy must be stored before a jump and then released rapidly in a catapult-like action.

A search was therefore carried out for sites where energy might be stored, with a particular focus on the occurrence of resilin, which in fleas (Bennet-Clark and Lucey, 1967) has been implicated in energy storage for jumping. In froghoppers (Burrows et al., 2008) and planthoppers (Burrows, 2010), resilin has been clearly shown to be present at the sites where energy is stored for jumping. Under illumination with specific wavelengths of UV light, four patches of bright blue fluorescence were found, two in the mesothorax and two in the metathorax, each associated with a particular middle leg and hindleg (Fig. 10). No blue fluorescence was found at the equivalent sites in the prothorax associated with the front legs. The fluorescence emanated from a crescent-shaped area of cuticle (Fig. 10B) at the articulation of the heavily sclerotised dorsal head of the pleural ridges of the mesothoracic and metathoracic segments (indicated externally by the black arrows in Fig. 2A) with the wing hinge. The fluorescence was highly localised and was not present in surrounding skeletal structures or other tissue. The blue fluorescence diminished in intensity if the pH of a saline solution was made acidic and recovered when it was returned to its normal pH. In alkaline saline the intensity increased but subsided to its previous level when returned to a saline at its normal pH. Such specific, blue fluorescence and its reversible pH dependence are two key signatures of the elastic protein resilin (Andersen and Weis-Fogh, 1964; Burrows et al., 2008; Neff et al., 2001).

Snow fleas, Boreus, propel their jumps by rapid depression and extension movements of their middle legs and hindlegs, achieving take-off velocities of up to 1 m s^–1 (mean of 0.8 m s^–1 in males and 0.7 m s^–1 in females) and forward distances of about 100 mm, equivalent to some 30 times their body length. Once airborne, the body remains relatively stable with only low rates of rotation in the pitch plane. Jumps are often strung together in a sequence, each with an erratic shift in direction and ending in a cataleptic posture, suggesting that they subserve an escape function. Jumping also gives a marked improvement to the speed of locomotion when compared with natural walking speeds, particularly at low temperatures. In these insects, jumping is all the more remarkable because the adults only live for a few weeks, emerging in the middle of a northern hemisphere winter and often jumping from the surface of snow.

The middle legs and hindlegs are slung from the sides of a strong box formed by the fusion of the mesothorax and metathorax, while the front legs are separate. Both the middle legs and hindlegs are longer than the body, with the middle legs representing 140% and the hindlegs 187% of the body length in males. How do these relationships compare with those in other insects that propel jumping with their legs (Tables 1 and 2)? Jumping insects can be divided into two broad functional categories according to the way the hindlegs are oriented. These different arrangements have significant mechanical consequences for the way force is delivered to the ground and the control of jumping (Sutton and Burrows, 2008; Sutton and Burrows, 2010).

In the first group are insects such as froghoppers, leafhoppers and planthoppers, with their legs slung beneath the body. All their jumps are powered by trochanteral depressor muscles. In this group the proportions of the hindlegs relative to the other legs are low except in the long-legged cicadellids. Relative to body length, the hindlegs are much shorter than those of snow fleas and relative to body mass they have a lower index.

In the second group are insects such as snow fleas Boreus, true fleas, locusts and bush crickets, which all have legs that move at the sides of the body. Power for jumping is generated by trochanteral depressor muscles in snow fleas and fleas, but by extensor tibiae muscles in Orthopterans and flea beetles. In fleas such as the hedgehog flea Archaeopsyllus erinacei (Table 1) the proportions of the legs relative to each other and relative to the length or mass of the body are similar to those in snow fleas. In Orthoptera the hindlegs are proportionately longer than the other legs, but relative to body length they are shorter, and relative to body mass they have indices in the same range as snow fleas. The exceptions are bush crickets, which have hindlegs that are much longer relative to body length, and male Prosarthria, which have a much higher index relative to body mass.

What sets snow fleas apart from all other members of this second group is that all 119 of jumps recorded in this study were propelled by the middle legs and hindlegs acting together. No jumps were observed in which the hindlegs acted alone. The first movements of the four legs appeared to occur at the same time, as determined by the capture rate of the images, which gave a resolution of 0.2 ms. No detectable pattern emerged to suggest that one pair of legs moved consistently before the other pair, or that one leg of a particular pair moved before the other. During the acceleration phase of a jump, the angular changes of the coxo-trochanteral joints, which are key to the propulsion, were matched in all four legs. Similar and progressive angular changes in the femoro-tibial joints also followed in all four legs. No observations indicated that the timing of the movements of the four legs had to be as closely controlled as for the hindlegs in planthoppers (Burrows, 2010). The mechanics of the legs also do not suggest that uncontrollable spins would result, as in planthoppers, if the legs did not act precisely at the same time. Delivering closely timed movements of four legs nevertheless poses challenges for the underlying neural control mechanisms, and there is no possibility that simple mechanical linkages between the legs could simplify the coordination as in planthoppers.

There are at least three possible reasons why snow fleas need to use both middle legs and hindlegs to propel jumping.

First, the use of four legs will distribute the ground reaction forces over a larger area provided by the four tarsi. These forces are further reduced by the 6 times longer acceleration time of a snow flea jump compared with that of a flea. Lower ground reaction forces should enable jumping either from more compliant surfaces or from soft surfaces such as snow. A similar outcome is found for jumping leafhoppers, a few species of which have long hindlegs while others have short hindlegs (Burrows and Sutton, 2008). Both use a catapult mechanism for jumping and achieve similar take-off velocities, but it takes longer to accelerate long hindlegs, so they will exert lower ground reaction forces. Like the snow flea, this should enable take-off from more compliant surfaces.

Second, using four legs gives a stable platform for take-off. If the centre of mass is between the middle legs and hindlegs, their torques will be in opposition. A consequence of this is that the rate of spin of the body once airborne should be low. The kinematic analysis confirmed this by showing that the mean rate of rotation in the pitch plane was only 5 Hz and rotation in the roll plane was rare. In contrast, fleas spin in the pitch plane at rates as high as 40 Hz (G. P. Sutton and M.B., unpublished observations), froghoppers at up to 80 Hz (Burrows, 2006) and pygmy mole crickets at 190 Hz (Burrows and Picker, 2010).

Third, two pairs of legs may be the only way of generating enough power to launch a snow flea into the air. The legs are thin and the femora contain little muscle mass so that most of the power must come from the trochanteral muscles in the thorax. Again, the volume of one segment of the thorax may mean that the mass of muscle that can be accommodated is insufficient to generate the necessary power.

Only a few other insects and some spiders use two pairs of legs to propel jumping, and the advantages in doing so are probably similar to those of snow fleas. The ant, Myrmecia nigrocincta, propels its jumps by extension of its middle legs and hindlegs (Tautz et al., 1994), judged to be simultaneous by images captured at intervals of 2.5 ms (400 frames s^–1, or 12.5 times slower than in the present study). Another ant, Harpegnathos saltator, is said to use both middle legs and hindlegs synchronously to power a jump, but only recordings from leg nerves and muscles, shown at a slow time scale, are available to support such actions during fictive movements (Baroni et al., 1994). A second analysis, however, showed that the hindlegs of this ant moved to full extension while the middle legs remained flexed (Tautz et al., 1994). The middle legs then extended to propel the ant to take-off. It is hard to extract information from these papers that would allow a direct comparison with the jumping performance of snow fleas. The acceleration period is given as 15–25 ms and the take-off velocity is estimated to be from 0.49 (Baroni et al., 1994) to 0.7 m s^–1 (Tautz et al., 1994), but it is unclear how either value was measured. Similarly, the calculations of acceleration and energy requirements are not clear enough for comparison (Baroni et al., 1994). The stick insect Sipyloidea sp. ‘Thailand 8’ throws the mass of its abdomen forward and pushes off the ground with its thin middle legs and hindlegs in a jump that reaches a take-off velocity of between 0.6 and 0.8 m s^–1 (Burrows and Morris, 2002). Some jumping spiders such as Salticus scenicus (Parry and Brown, 1959) and Cupiennius salei (Weihmann et al., 2010) also use two pairs of legs to propel jumping.

There are four reasons for suggesting that a catapult mechanism must be used by snow fleas to propel their jumping. First, the length of the legs and the short period over which acceleration is applied indicate that insufficient energy could be generated by the direct contraction of muscles working against levers. Second, the energy requirements for a jump are toward the extremes of what striated muscle could generate (Askew and Marsh, 2002; Ellington, 1985; Josephson, 1993; Weis-Fogh and Alexander, 1977). A further factor is that snow fleas must often jump at temperatures around 0°C, but the quoted estimates of the performance of striated muscle were made at their much higher, optimum operating temperature. It is, however, not known whether the muscles of snow fleas are adapted to work more efficiently at low temperatures. If snow fleas were to operate at 30°C and to have a slightly higher mass of jumping muscle, then energy storage would not be necessary. In snow fleas the energy store could be used to deal with reduced muscle performance at low temperature, in contrast to other insects that use their energy stores to surmount the problem that even their maximum possible muscle performance would not be sufficient to propel jumping. Third, jump distance is not greatly affected by temperature, but a jump powered by direct muscle contractions would be temperature dependent. For snow fleas, the period during which the catapult is loaded would be expected to be temperature dependent, and this has been reported (Edwards, 1987). Fourth, resilin is present at the junction of the pleural ridges with the hinges of the wings of the middle legs and hindlegs, but not of the front legs. This suggests that energy is stored by contractions of the muscles powering the movements of the middle legs and hindlegs. The soft and elastic resilin together with the hard cuticle of the skeleton could provide a mechanism for storing the energy generated by the slow contractions of the trochanteral depressor muscles. The combination of these two materials, as in the energy storage mechanisms of froghoppers (Burrows et al., 2008), provides a means of bending the hard cuticle through small distances while the resilin ensures that it does not shatter under deformation, and that the whole structure returns rapidly to its original shape after a jump. If a catapult is used, then there must be a mechanism that allows the trochanteral muscles to contract and energy to be stored without movements of the hindlegs, and which then allows the stored energy to be released suddenly. This mechanism is not known in snow fleas.

Does this analysis of jumping behaviour and its associated mechanisms illuminate the phylogenetic relationships between fleas and snow fleas, which are currently placed within different orders? A comparison of the jumping mechanisms of snow fleas, analysed in this paper, and true fleas (Bennet-Clark and Lucey, 1967; Rothschild and Schlein, 1975; Rothschild et al., 1975; Rothschild et al., 1972; Sutton and Burrows, 2011) reveals both similarities and differences, with the latter indicating adaptations to different habitats. The muscles that propel a jump are the same trochanteral depressors that lie in the thorax. The two insects achieve similar take-off velocities but, once airborne, a snow flea spins less than a flea because the use of two pairs of propulsive legs provides a more stable platform for take-off. The acceleration time of a snow flea in a jump is about 6 times longer than that of a flea and this means that it exerts less ground reaction force. Both use a catapult mechanism that requires the storage of energy in advance of the jump. Resilin is present at the articulation of the wing hinge with the heavily sclerotised dorsal head of the pleural ridge in both insects and is assumed to play a key role in energy storage. The clear difference is that fleas propel jumping with just the hindlegs whereas snow fleas use both the middle and hind pairs of legs. Fleas have two pads of resilin associated with the two hindlegs, whereas snow fleas have four pads of resilin associated with each of the middle legs and hindlegs. The location of the resilin for each individual leg is the same in fleas and snow fleas. The question therefore arises as to whether snow fleas using two pairs of legs represent the ancestral mechanism and fleas have become specialised by the use of just one pair of legs. Alternatively, both mechanisms could be specialisations in different directions from an evolutionary common ancestral mechanism of locomotion that may not have involved jumping. The general principles and mechanisms of jumping are the same in snow fleas and fleas, even though Boreus uses four legs while fleas use only the two hindlegs. These functional and behavioural findings add clear further evidence to support a close relationship between the Boreidae and the Siphonaptera.

I am enormously grateful to John S. Edwards for sharing with me his unpublished data on snow fleas. I am also greatly indebted to Roger Northfield for introducing me to these insects, locating their habitats and catching them. I thank Greg Sutton and Steve Rogers for their many helpful suggestions during the course of this work and, together with Steve Shaw, for their incisive comments on the manuscript. Jo Riley provided tremendous support with experiments and analyses.