Walking and Surface Film Locomotion in Terrestrial and Semi-Aquatic Spiders

Despite the attention arthropod walking has received, there are relatively few studies devoted to the locomotion of spiders. Nevertheless, there are a number of recent studies on a broad taxonomic range of spiders, including mygalomorphs (Wilson, 1967; Seyfarth & Bohnenberger, 1980), wolf spiders (Moffett & Doell, 1980; Ward & Humphreys, 1981), sheet-web spiders (Frohlich, 1978), jumping spiders (Land, 1972) and crab spiders (Ferdinand, 1981). Bowerman (1975) has examined locomotion in the scorpion. The coordination used by most terrestrial arachnids and insects is similar, despite the difference in leg number. In general, each leg tends to alternate with its contralateral homologue and ipsilateral neighbours but moves in near synchrony with diagonally adjacent contralateral legs. In arachnids, the resulting gait is termed an alternating tetrapod, an analogue to the alternating tripod of insects. This stepping pattern provides a stable base of support throughout the locomotor cycle, an important feature of effective terrestrial walking.

Although spiders are best known for their accomplishments as terrestrial predators, certain species within the genus Dolomedes (Pisauridae) are among the largest and most adept semi-aquatic arthropods. These spiders may be found at the margins of ponds and streams awaiting prey at the water surface. When Dolomedes detects the vibrations of a potential victim, perhaps a struggling insect, the spider moves rapidly over the surface film to capture its prey and then returns to a solid resting place to consume it. Thus, the surface film serves as a ready-made ‘web’. Although diving, swimming (McAlister, 1959) and wind-assisted movements (Deshefy, 1981) of Dolomedes have been investigated, surface film locomotion, despite clear adaptive significance, has received little study. Ehlers (1939) briefly describes surface film locomotion in several species, including the European fishing spider Dolomedes fimbriatus. When on water, D.fimbriatus rests with its ventral body surface on the water and legs extended to increase the area in contact with the supporting surface film. During locomotion, the third and then second leg pairs retract in quick succession, often followed by retraction of the first pair. Each leg moves in synchrony with its contralateral homologue. The last pair of legs does not aid in propulsion but appears to function in stabilization. In some respects, rowing in Dolomedes is similar to that of its insect counterpart, the water strider (Hemiptera, Gerridae) (Anderson, 1976). However, the aquatic gait, characterized by synchrony of intrasegmental legs and near synchrony of adjacent ipsilateral legs, is quite different from the alternating pattern used by most spiders during terrestrial locomotion.

I have studied the control of spider walking and the evolution of rowing in Dolomedes by comparing terrestrial and surface film locomotion in a fishing spider, Dolomedes triton, and a wolf spider, Lycosa rabida (Lycosidae). D. triton is probably the most aquatic of the North American pisaurids (Carico, 1973). This species appears to use the same rowing gait as D.fimbriatus when moving on water and uses a gait similar to the alternating tetrapod during locomotion on a solid substrate. L. rabida is terrestrial, although many lycosids move over water facultatively (Ehlers, 1939), and is similar to D. triton in size and general proportions. The results of this analysis suggest that interleg coordination in spiders is governed by two distinct coupling mechanisms: intrasegmental and intersegmental mechanisms. These coupling units are identified by their independent effects upon coordination and may have served as the sites at which natural selection acted to produce the rowing gait used by Dolomedes.

All spiders used in this study were collected in Athens County, Ohio, USA. Dolomedes triton was found at the water surface among emergent vegetation growing along the margins of lakes and slowly moving streams. Adults of this species were collected from May to September. Adult Lycosa rabida were collected during late summer and were usually found well away from bodies of water. In both species, only adult females were used. (Males differ from females in being smaller and having relatively longer legs, factors which would have complicated species comparisons.) The average length of the carapace in females of both species is about 18 mm. Species were housed individually in plastic boxes (7× 12×7 cm) at an average temperature of 23 °C.

I filmed individuals of both species as they moved over a dry substrate (two D. triton, two L. rabida) and across the water surface (four D. triton, two L. rabida). A large arena was partitioned into a rectangular 0·25×1·0m runway. The floor consisted of a roughened plastic sheet marked with a 2-cm grid. To obtain sequences of surface film locomotion, I covered the arena floor with water to a depth of about 2 cm. The runway was illuminated by two 500-W slide projectors positioned about 2 m away. Light from the projectors was directed into a mirror which had been mounted above the runway at 45° to the arena floor. I aimed the camera lens at the mirror to capture the reflected dorsal image of the moving spider. Films were taken with a Ciné-8 high speed camera (Visual Instrumentation Corp., Model SP-1) at 100 frames s⁻¹ using Kodak 4-X reversal film (ASA 400). Sequences were filmed at an average air temperature of 23 °C and water temperature of 20°C.

I limited quantitative film analyses to sequences showing three or more complete stepping cycles during which turning did not exceed an angle of 30°. I examined all such sequences frame-by-frame using a stop-action Kodak Ektagraphic film projector. When analysing the stepping cycle of each leg, I considered retraction (the propulsive stroke) to begin when the leg contacted the substrate and then moved backwards. Protraction (the return stroke) began when the leg lifted from the substrate and moved forwards. If for some fraction of a stepping cycle a leg did not move relative to the spider’s body, it was included within the preceding phase. In films of walking spiders, these features were fairly easy to recognize, since the tips of the legs were usually stationary with respect to the substrate during retraction. In aquatic locomotion, the legs move relative to the substrate throughout the stepping cycle. Therefore, protraction and retraction could be recognized most readily as movements relative to the spider’s body. This was accomplished by projecting the film on a graphics tablet (Tektronix, Inc.; Model 4956) attached to a Tektronix graphic computer (Model 4051) and plotting the positions of the spider’s body and tarsi for each frame. I programmed the computer to calculate the angle between each tarsus and the spider’s midline for each frame. By comparing the magnitudes of the angles in consecutive frames, I was able to identify phases of the stepping cycle in each leg.

Since stepping is a cyclic process, one can in principle use any point as a reference when comparing coordination in two legs. To simplify analysis and allow comparisons with previous studies, I selected the onset of retraction as the point of reference. The time between the start of retraction in a reference leg and that of another is termed ‘lag’. The phase of the two legs is calculated by dividing lag by the stepping period of the reference leg. Phase values near 0·0 or 1·0 indicate that the legs are stepping in near synchrony and those near 0·5 indicate that the legs are moving in alternation. Circular statistics for analysis of phase, including calculation of mean and standard deviation, follow Batschelet (1965). All other statistical treatments follow Sokal & Rohlf (1981) and all tests were performed at the 0·05 level of significance.

The arrangement of segments in a spider’s leg is shown in Fig. 1.

When standing or moving on a solid surface, both L. rabida and D. triton adopt a posture typical of many terrestrial arthropods; the body is held close to the substrate suspended from a ring of legs which forms a broad base of support (Fig. 2A). Usually, only the tips of the tarsi bear the spider’s weight and convey propulsive forces to the substrate. When on water, the habitually terrestrial L. rabida places its ventral body surface on the surface film, but some elements of the terrestrial posture remain (Fig. 2B). The legs still provide some support and propulsive forces are transferred primarily through the tarsi, as indicated by the displacement and frequent penetration of the surface film by the tips of the legs. When on water, D. triton also places its sternum against the water surface, but the legs are extended horizontally, thus distributing the body weight over a larger area (Fig. 2C). Propulsive forces may therefore be generated along the length of the leg and not just at the tarsus.

During terrestrial locomotion, the leg movements of L. rabida and D. triton are similar and resemble those described in other spiders (see Ehlers, 1939). The first and fourth pairs of legs work in an essentially vertical plane; leg 1 pulls and leg 4 pushes. The movements of the second and third leg pairs are more complicated, but can be generally described as oar-like, having a large horizontal component.

When walking, both spiders use a gait which is approximately an alternating tetrapod; sets of diagonally adjacent legs (R1-L2-R3-L4 and L1-R2-L3-R4) tend to work in alternation. If one considers stepping of only ipsilateral appendages, retraction typically occurs in the order 4-2—3—1. While L. rabida uses leg 1 regularly during the stepping cycle, D. triton may go through several cycles in which one or both members of the first leg pair do not contact the substrate and remain extended in front of the animal. D. triton also differs from L. rabida in frequently dragging the tarsus of leg 4 following full extension. In these respects, terrestrial locomotion in D. triton resembles that of the burrowing wolf spider, Lycosa tarentula (Ward & Humphreys, 1981).

Intraleg movements used by L. rabida on water are very similar to those used during walking, although stepping movements of legs 1 and 4 tend to have larger horizontal components (Fig. 3). Interleg coordination differs from that seen during terrestrial locomotion. While each leg continues to alternate with its contralateral homologue, ipsilateral legs assume the metachronal stepping pattern 4—3-2—1. Although this spider can maintain a net linear course when on the water surface, it generates a great deal of turbulence and the body undergoes continuous yaw and roll.

Surface film locomotion in D. triton differs from walking and from aquatic locomotion in L. rabida. Movements producing retraction in legs 2 and 3 occur largely at the proximal joints. The tibiometatarsal joints (see Fig. 1) remain in an extended position, although some flexion may occur at the end of retraction (Fig. 4). Due to a rotation about the long axis of the coxotrochanteral articulation, the movements of the distal elements of leg 1 show a greater horizontal component than during terrestrial locomotion. During protraction, legs 1, 2 and 3 are lifted from the water surface. The fourth leg pair does not contribute to propulsion but remains extended behind the spider. Important directional control movements of leg 4 will be discussed below.

Interleg coordination during surface film locomotion in D. triton is quite different from that found in walking. As Ehlers (1939) described in D. fimbriatus, intrasegmental legs move in synchrony rather than in alternation. This behaviour may minimize the production of asymmetrical forces which generate yaw and roll. Likewise, ipsilateral legs assume the metachronal pattern 3–2–1. The nearly simultaneous retractions of legs 2 and 3 result in a period of rapid acceleration followed by deceleration during their protraction, a situation which seems energetically inefficient.

Directional control during rowing in D. triton can be divided into two types: turning (rotational movements made in order to establish a new course) and yaw correction (rotational movements made in order to compensate for deviations from an established course). Each form of directional control is associated with a distinctive pattern of leg movements.

Turning involves only the first three leg pairs, the fourth pair rests nearly motionless on the water surface. A left turn is usually initiated by the independent retraction of the third leg on the right side (R3). The rotation of the body resulting from this movement establishes the new direction of travel. Next, L3 and R2 retract simultaneously, followed by the synchronous retraction of L2 and Rl. LI remains extended in front of the spider throughout the turn. A right turn is the mirror image of a left turn. After the turning sequence is completed, the spider returns to the regular rowing pattern.

D. triton rides over the surface film on a layer of air held by many hydrofuge hairs covering the ventral body surface. This arrangement decreases friction with the water, probably reducing drag and thus the energy the animal must expend in moving from place to place. However, low friction increases the potential for unintended rotations (yaw) due to even slight asymmetry in the forces generated during the propulsive strokes. Transverse phase synchrony may have evolved to minimize yaw, but directional control remains an apparent problem.

The most obvious yaw corrective movements used by D. triton are rapid ‘protractions’ of leg 4. The kinematics of these sweeping lateral kicks are similar to those seen in normal protractions of the fourth leg pair during aquatic locomotion in L. rabida (Fig. 3). If a corrective rotation to the right is required, R4 is swept rapidly forward, while L4 remains motionless or undergoes slight ‘retraction’. Interestingly, these yaw corrective ‘protractions’ only occur just prior to, or simultaneously with, the protraction of leg 3, the period during which one would expect protraction in leg 4 if it participated in the metachronal sequence seen in the other legs. Legs 1, 2 and 3 may also participate in yaw correction by varying the timing of retraction across the body. This behaviour is especially evident in leg 1. Fig. 5 depicts a sequence of surface film locomotion in D. triton during which the spider makes several yaw corrective movements.

During surface film locomotion in L. rabida, intended and unintended rotational movements cannot be distinguished. It is interesting to note that leg pairs may move in transverse phase synchrony during rotational movements rather than the usual alternating pattern. The same behaviour has been observed in walking tarantulas (Wilson, 1967).

Durations of protraction and retraction for each leg were found to vary linearly with stepping period (Tables 1, 2). During terrestrial locomotion in both species, retraction changes more rapidly with stepping period than does protraction, but this relationship is reversed during aquatic locomotion in D. triton and in leg 3 of L. rabida (Fig. 6). The relationships for the other legs in L. rabida differ but, in general, retraction is shorter and protraction is longer during aquatic locomotion than during walking at similar stepping periods. Thus, the slope of the p/r ratio (y) versus stepping period (x) tends to be negative during terrestrial locomotion and positive during aquatic locomotion.

In terrestrial locomotion, mean phase values for each combination of adjacent ipsilateral legs approach 0·5 in both L. rabida and D. triton (Table 3). For surface film locomotion, means are consistently lower for all combinations of neighbouring legs within a species and are very similar between species. Alternate ipsilateral legs (i.e. R1 and R3, etc.) show no clear pattern.

Mean phase values for the first and second leg pairs are not statistically significant in walking D. triton (Rayleigh test, P> 0·05) (Table 4). Intrasegmental legs tend to alternate during surface film locomotion in L. rabida and during terrestrial locomotion in both species. Means of intrasegmental legs approach 0·0 in D. triton during aquatic locomotion, indicating that the members of each leg pair are moving in synchrony.

Phase values for diagonally adjacent contralateral legs are also presented in Table 4. Mean phase values for contralateral legs of segments 2 and 3 are very similar in walking L. rabida and both forms of locomotion in D. triton, while the values for the same combination of legs during aquatic locomotion in L. rabida are much higher. Diagonally adjacent legs of segments 3 and 4 tend to move in near synchrony in walking D. triton and during both forms of locomotion in L. rabida.

No significant relationship was found between phase and stepping period in any combination of legs.

Statistical measures of variation, such as circular standard deviation, are used to indicate the strength of coordination or coupling between legs; lower values indicate tighter coupling. The results are consistent with the respective habits of the two species. D. triton shows less variation for all leg combinations when rowing as compared to walking. L. rabida tends to have stronger coordination during terrestrial locomotion than during surface film movement. When interspecific comparisons are made, leg movements in D. triton tend to be more weakly coupled than those of L. rabida during terrestrial locomotion. The opposite is found during surface film locomotion.

Our understanding of the factors underlying coordination in arthropod walking has been aided by modelling the motor control apparatus as a system of interconnected subcomponents. The stepping cycle of each leg is considered to be governed by a single centrally driven oscillator, the timing of which may be altered by changes in sensory input. The relative timing of neighbouring oscillators is determined by a coupling mechanism (Stein, 1977). Such systems have predictive value, but it has yet to be shown whether these subcomponents represent real entities or are simply a convenient shorthand which captures certain properties of the actual mechanism (Delcomyn, 1985). The results reported here suggest that the subdivision of the motor control apparatus of spiders into separate functional components is justified.

Differences in interleg coordination during walking and surface film locomotion in the habitually terrestrial Lycosa rabida suggest that this behaviour is governed by at least two types of coupling mechanisms: intrasegmental and intersegmental mechanisms. When moving on a solid substrate, this spider uses the alternating tetrapod gait. During movement on the surface film, ipsilateral legs adopt a metachronal pattern of stepping while legs of the same segment continue to alternate as in the walking gait. The independent effects of substrate type on two elements of coordination seem most readily explained by the presence of functionally distinct units within the motor control apparatus of the spider.

Evidence from the comparison of locomotion in L. rabida and D. triton is also consistent with the notion of subcomponents within the motor control apparatus. These species show the same basic stepping pattern during terrestrial locomotion, implying the existence of similar motor control mechanisms. Although obvious interspecific differences occur during surface film locomotion (e.g. intrasegmental legs move in synchrony in D. triton), coordination of adjacent ipsilateral legs is essentially the same. This result suggests that natural selection has favoured the evolution of transverse phase synchrony in the amphibious D. triton, leaving the mechanism that controls intersegmental coordination unaltered. Thus, it appears that intra- and intersegmental coordination in spiders can be varied independently in both proximate and evolutionary time, implying that interleg coordination is regulated by two types of mechanisms which maintain a large degree of functional and genetic autonomy.

If the separation of the motor control apparatus into intra- and intersegmental components is valid, the question arises as to the orientation of the intersegmental mechanism. The typical walking gait of spiders (Fig. 7A) shows two internal patterns which serve as the basis of two qualitative models of pattern generation in walking arthropods. The alternating tetrapod model emphasizes the importance of synchronous stepping in sets of diagonally adjacent contralateral legs (R1-L2-R3-L4 and L1-R2-L3—R4) and suggests that intersegmental coupling is arranged diagonally (Fig. 7B). The metachronal model (Wilson, 1966) gives emphasis to the wave-like stepping pattern which travels along each side of the arthropod’s body. Although this model, in its strictest sense, fails to predict some stepping patterns used by walking mygalomorphs (Wilson, 1967), the assumption that intersegmental coupling is arranged longitudinally (Fig. 7B) is still useful.

The pattern of phase values observed during surface film locomotion in L. rabida and D. triton is consistent with the assumptions of the metachronal model. Fig. 7C represents the arrangement of the phase relationships between legs of any two adjacent segments, where a represents the intrasegmental phase value, b is the diagonal phase and c is the ipsilateral phase. The variables must show the relationship a+c = b measured on a circular scale. Since intrasegmental legs in D. triton move in synchrony (phase = 0·0), diagonal and ipsilateral phase values must be equivalent and should represent the effects of the intersegmental coupling mechanism alone. Changing intrasegmental phase to 0·5 (legs moving in alternation) and assuming that intersegmental coordination is governed by an ipsilateral mechanism, the phase values measured in D. triton accurately predict those found in L. rabida. When one assumes a diagonal mechanism and the procedure is repeated, the resulting phase values bear no resemblance to those measured in L. rabida (Table 5). That such an exercise can be performed at all points to the great similarity of the motor control mechanisms of the two species and the independent action of the intra- and intersegmental coupling mechanisms. This reasoning also suggests that, while diagonal phase synchrony may be of great functional significance to walking arthropods, one need not evoke a diagonally arranged coupling mechanism to account for this behaviour.

These results have been anticipated by the findings of previous research. Many of the spiders studied thus far show stronger coupling between adjacent ipsilateral legs than between intrasegmental legs (mygalomorphs, Wilson, 1967; Seyfarth & Bohnenberger, 1980; wolf spiders, Moffett & Doell, 1980; jumping spiders, Land, 1972). Seyfarth & Bohnenberger suggest that this indicates two different coupling mechanisms, one intrasegmental and one ipsilateral. Ferdinand (1981) also provides evidence to support the ipsilateral model of intersegmental coordination in his study of unusual stepping patterns in crab spiders.

In the transition from walking to surface film locomotion, L. rabida and D. triton show similar changes in coordination: the protraction/retraction ratio (p/r) increases in each leg and lag time between adjacent ipsilateral legs decreases for a given stepping period. Certain aquatic insects, such as nepid water bugs (Wendler, Teuber & Jander, 1985) and gerrid water striders (Bowdan, 1978), display stepping patterns similar to those of spiders when transferred from a solid substrate to water. Altered proprioceptive or exteroceptive input may produce such changes by triggering a shift from one specialized motor programme to another or by altering a single flexible system capable of producing a variety of patterns (Wendler et al. 1985). The demonstration of a gait intermediate between walking and swimming in nepid water bugs suggests that the latter mechanism operates in these insects. Sensory cues from strain (load) receptors located on the legs may provide the stimuli which modify motor output in the transition from walking to swimming (Wendler et al. 1985).

Gait shifts associated with changes in vertical loading or horizontal loading (resistance to retraction) have received considerable attention (Graham, 1985). Most of these studies concern the effects of increased loads on locomotor behaviour. However, when L. rabida and D. triton are on water, the legs do not support the entire body weight, and water would not be expected to resist leg movements to the same extent as a solid substrate. Therefore, the effects of reduced loads are probably more important in understanding the coordination used during surface film locomotion. Gait shifts similar to those found in L. rabida and D. triton have been observed in other arthropods under conditions of reduced load or reduced sensory input. For example, an inverted cockroach shows intrasegmental coordination similar to normal walking but the p/r ratio is higher; bursts from the levator muscles are proportionally longer and depressor activity is significantly reduced in comparison with walking at the same stepping period (Reingold & Camhi, 1977; Sherman, Novotny & Camhi, 1977). Also, if a cockroach is suspended over a low-friction substrate, reduction in horizontal and vertical loading may elicit a metachronal stepping pattern to replace the normal alternating tripod (Spirito & Mushrush, 1979). These changes appear to result from sensory modulation of a single flexible motor control system. Similar changes in p/r ratio and ipsilateral phase occur in lobsters when placed under conditions of reduced load (as when the legs are moving freely in the water) or when the legs are autotomized. However, this change appears to be a sensory-mediated transition from a walking programme to another programme that may be associated with gill ventilation (Pasztor & Clarac, 1983). In contrast to these examples, reduction of horizontal loading in a stick insect moving over a slippery surface does not significantly alter normal interleg coordination, while an increase in resistance brings about a metachronal pattern (Epstein & Graham, 1983).

Given this apparent diversity within motor control systems of arthropods, studies dealing with the effects of different substrates on coordination in spiders are needed before any firm conclusion can be made about the role of sensory input in producing the aquatic gaits of L. rabida and D. triton. Even then, the role of sensory input may differ between species. For instance, it seems unlikely that a terrestrial spider like L. rabida would have an alternative motor programme adapted for aquatic locomotion. I suggest that the coordination displayed by this spider when moving on the water surface is elicited largely (if not solely) by the effects of this novel sensory environment on a flexible motor, system adapted for effective locomotion on solid substrates. A single flexible system may also occur in D. triton (as in nepid water bugs) but the existence of two distinct motor programmes would not be surprising in this species.

Behavioural comparisons have provided some insight into how the evolution of the motor control strategy of Dolomedes is related to this spider’s lifestyle and phylogenetic history. D. triton occupies a heterogeneous habitat of emergent vegetation and open water. If a spider is to be successful in this habitat, its behaviour and morphology must be a compromise response to the different physical conditions of solid substrates and the water surface. Its potential for acquiring new behaviour is limited by the adaptability of the motor programme of its terrestrial ancestors; complex patterns of behaviour, like the coordination of legs during locomotion, cannot be expected to arise de novo. Therefore, it is not surprising to find evidence for the derivation of the rowing gait from the terrestrial motor programme. Dolomedes shows the same ipsilateral coordination when moving on the water surface as its terrestrial relative, L. rabida. The timing and kinematics of yaw corrective kicks used by the amphibious spider are like the protractions of the fourth leg pair observed in the terrestrial spider when it is forced to move on water. The two species also show certain similarities in the relative durations of protraction and retraction. Transverse phase synchrony used by Dolomedes is undoubtedly an adaptation to aquatic locomotion, but its evolutionary origin is obscure. If surface film locomotion had been examined without reference to a terrestrial species, all features of rowing in Dolomedes might have been assumed to be adaptations for aquatic movement, rather than a complex of primitive and specialized features.

I gratefully acknowledge the support and assistance of Dr Jerome S. Rovner throughout the course of this study. I thank Drs Ellengene Peterson, Ralph DiCaprio, George Lauder and two anonymous reviewers for their suggestions. Dr Sharon Simpson, Janet Easly and Susan Mason assisted in filming. This work was done in partial fulfilment of the requirements for Master of Science from the Department of Zoological and Biomedical Sciences, Ohio University.