Two distinct gait types in swimming frogs

J. Zool., Lond. (22) 258, 183±188 # 22 The Zoological Society of London Printed in the United Kingdom DOI:1.117/S952836921292 Two distinct gait types in swimming frogs Sandra Nauwelaerts and Peter Aerts Department of Biology, University of Antwerp (UIA), Universiteitsplein 1, B-261 Wilrijk (Antwerpen), Belgium (Accepted 15 October 21) Abstract During terrestrial locomotion, frogs use two distinct gaits: out-of-phase leg movements associated with slow crawling behaviour and in-phase leg movements during fast jumps. In Rana esculenta, crawling occurs during feeding, while jumping is used as an escape strategy. We examined whether a similar velocitydependent gait shift appears in swimming R. esculenta. Typically, swimming frogs propel themselves by kicking both hind limbs simultaneously. Observations of out-of-phase leg movements in swimming frogs have been reported, but were usually assumed to be associated only with directional changes. We demonstrate that alternating-leg swimming is used quite frequently and that it results in a signi cantly lower velocity to the one obtained by using in-phase leg movements. This difference is likely to be associated with energetic costs. Mathematical estimates of positive mechanical work required to move the centre of mass revealed that out-of-phase swimming is energetically less expensive than the in-phase gait at a comparable speed, but may not be ef cient at high speeds. Possible explanations for this phenomenon are higher inertial energy losses for in-phase swimming, but at high speeds jet propulsion and an interaction effect may gain importance. Key words: locomotion, gait transition, energetics, frogs, Anura, speed, Rana esculenta INTRODUCTION Two main types of terrestrial locomotion exist within the Anurans. Some species crawl, using out-of-phase leg movements, others jump, extending both legs simultaneously. Some species use both, e.g. Bufo woodhousii fowleri (Anderson, Feder & Full, 1991), Bufo bufo (pers. obs.) and Rana esculenta (this paper). These species change gait, from a walk at slow speeds to a hop at faster speeds (Walton & Anderson, 1988; Anderson et al., 1991). This gait change reduces the energetic cost of locomotion (Anderson et al., 1991). Out-of-phase leg movements during aquatic locomotion were ignored for a long time. Although Calow & Alexander (1973) mentioned them brie y, only the recent study of Abourachid & Green (1999) describes out-of-phase swimming in more detail. These authors link the swimming mode of frogs to their position in the phylogenetic tree. They stated that primitive frog species use an alternate-leg swimming style, while derived species swim with in-phase movements of hind limbs. In derived frogs, out-of-phase leg movements would be associated only with directional changes. All correspondence to: Sandra Nauwelaerts. E-mail: sandran@uia.ua.ac.be However, individuals of R. esculenta, a species that would be classi ed as a derived frog, were observed using both in-phase and out-of-phase limb movements during continuous straight swimming bouts. Does the selection of a swimming gait type re ect a velocitydependent relationship as observed during terrestrial locomotion? A widespread assumption is that animals change gaits to minimize metabolic energy consumption (e.g. Hoyt & Taylor, 1981; Hreljac, 1993; Alexander, 1999), although other hypotheses are formulated (e.g. Biewener & Taylor, 1986; Farley & Taylor, 1991). The present paper: (1) explores whether a similar velocity-dependence gait transition as in terrestrial locomotion exists in swimming frogs; (2) explains the co-existence of two gaits by comparing the energetic cost of transport. MATERIAL AND METHODS Animals and their maintenance Five male individuals of Rana esculenta were obtained from a commercial supplier (Bray-et-LuÃ, France). The mean ( sem) body mass and snout±vent length were 5.8 3.6 g and 79.3 3.3 mm, respectively. The animals

184 S. Nauwelaerts and P. Aerts were maintained together in a moist terrarium in a room at c.158c. They were fed crickets ad libitum. Filming Swimming Swimming animals were lmed from below using a PANASONIC F-15 video camera connected to a S-VHS recorder set at 5 elds s 71. A mirror was positioned at 458 beneath the bottom of a transparent swimming tunnel made from Plexiglas (height.1 m, width.15 m, length 2 m). The tunnel was closed and entirely lled with water, interconnecting 2 open aquaria. The temperature of the water in the tunnel was maintained at c. 28C. Animals were sometimes stimulated to swim through the tunnel by touch, resulting in a wide range of velocities. A total of 55 sequences from the 5 individuals was selected for further analysis. Crawling and jumping Animals were individually placed into a glass aquarium of 16.761 m, with several live crickets. The aquarium was in a room at 2 8C. The observer vacated the room while the foraging behaviour of the frogs was videotaped from above using the same equipment as previously described for lming swimming. A total of 1 sequences from the 5 individuals were selected for further analysis. Additionally, 1 (spontaneous) jumps were recorded. Performance and gait type Video images were digitized using an Iomega BUZZ frame-grabber and imported into the APAS software (Ariel Dynamics). Seven points on the image were followed through time: snout tip, right and left shoulder, right and left hand, and right and left ankle. The x- coordinates were used for further calculations as all animals were swimming in a straight line (y-coordinates around ). The mean velocity of the snout over several cycles was calculated for each sequence. The distances of the frog's snout to right and left ankle, right shoulder to right hand, and left shoulder to left hand were plotted against time. The relative phase between left and right ankle was used for gait type determination. The sequences in which the legs were moving simultaneously were de ned as `in-phase', while sequences in which legs were moving alternately were de ned as `out-of-phase'. Energetic costs of locomotion To estimate energy consumption, the ideal set-up involves respirometric measurements. Unfortunately, in a swimming frog, this procedure is dif cult to achieve because a breath-by-breath analysis is required for the short swimming bouts typical for anurans. Relative measures for energetic costs, however, can be obtained starting from the mechanical work. The mechanical energy of the movement is typically divided into external and internal components. External energy is spent to displace the centre of mass (CM), whereas internal energy is used for segmental movements with respect to the CM. In our study, internal work by frogs is likely to be identical in both gait types, as only the timing of the legs differs. Therefore, only external work was calculated and the amount of positive work used as a relative measure for the locomotor costs. Negative work was not considered as it was not delivered by the frog itself, but by drag forces working upon the frog's body. External, positive, mechanical work is obtained from the integration of the instantaneous positive mechanical power output over the total recording time. Instantaneous external power equals force times the instantaneous swimming speed. Estimations of the propulsive forces were based upon those to move an ellipsoid, with the dimensions of a frog body, moving through the water. Propulsion forces were calculated using the following formula (see also Nauwelaerts, Aerts & D'AouÃt, 21): F = (1 + AMC) m a (.5 r A C d v 2 ) where: AMC, added-mass coef cient =.2 (Daniel, 1984); m, mass of the frog during the experiment; a, measured acceleration; A, surface area =.4; v, velocity; r, kinematic density = 1; C d, drag coef cient =.5. The acceleration was obtained from rst ltering the displacement of the snout tip using a zero-phase shift low-pass Butterworth lter, and then deriving the displacement pro le twice to obtain smooth acceleration pro les. The drag coef cient was determined from the measured decelerations during the glide phases of swimming frogs and corresponds to the data presented by Gal & Blake (1988). Power pro les were obtained for each sequence by multiplying the calculated force with the derived velocity. By using numerical integration, it was possible to calculate mechanical work required to displace the frog's centre of mass. Only mass-speci c positive work per unit distance was considered in our analysis (see earlier). Statistical analysis The standard statistical procedure of MANOVA was used to examine variation in velocities obtained using in-phase and out-of-phase gait types. Variation between individual frogs was considered in this model. Analyses were performed using the software STATISTICA. To test statistical differences in energetic costs between the 2 gait types, an ANCOVA was performed, using velocity as a covariate, using only the data that referred to the overlap in the velocity range.

Gait transition in swimming frogs 185 Velocity in m/s.5.45.4.3.3.2.2.15.1.5. RESULTS ± SD ± SE Mean Min Max Velocities of gaits Out of phase In phase Gait type Fig. 1. Comparison of the mean of the observed velocities during out-of-phase (n = 19) and in-phase (n = 36) swimming in Rana esculenta. Shaded area, the transition velocity. Minimum and maximum velocities are also indicated. Thirty-six in-phase and 19 out-of-phase swimming sequences were used in the analysis. Frogs attained signi cantly different swimming velocities when using the two gait types (P <.1): they swam substantially slower when swimming out-of-phase. This result was consistent for all frogs (P =.771) and there was no signi cant interaction between individual and gait type (P =.998). The speed of the transition phase between gaits, based upon the overlap of standard deviations, is close to.13 m/s (Fig. 1). Description of gaits In-phase swimming involves a nearly synchronous timing of hind and front limbs with all limbs contributing to the propulsion simultaneously (Fig. 2). This synchrony results in a sharp increase of velocity, followed by a glide phase in which the velocity declines as a result of drag forces. In contrast, during out-of-phase swimming, the hind limbs move alternately and in synchrony with the diagonally opposing front limb (Fig. 3), i.e. the left front limb moves together with the right hind limb and vice versa. Frogs also use the out-of-phase gait when crawling on land when the timing of the limbs is similar to the outof-phase swimming pattern (Fig. 4). A jumping sequence, however, involves in-phase hind limb extension. The front limbs seem unimportant for propulsion during the fast jumping gait. Right ankle Left ankle ±±±± Right hand Left hand.14 Flexion Extension.12.1 Displacement in (m).8.6.4.2.2.2.4.6.8 Time 1 1.2 1.4 Fig. 2. Displacement of right (black) and left (white) ankle relative to the snout tip and of the right (black) and left (white) hand relative to the right and left shoulder in function of time during in-phase swimming in Rana esculenta. Note that extension and exion occur at approximately the same time for all limbs. 1.6 1.8

186 S. Nauwelaerts and P. Aerts Displacement (m) Right ankle Flexion 14 12 1 8 6 4 2 2 4.2.4.6.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4 Time (s) Energetic costs of locomotion When plotted against mean velocity, energy requirements for out-of-phase swimming are considerably lower then those for in-phase bouts at the same speed (Fig. 5; P =.4). DISCUSSION Left ankle Extension Right hand Left hand Fig. 3. Displacement of right (black) and left (white) ankle relative to the snout tip and of the right (black) and left (white) hand relative to the right and left shoulder in function of time during out-of-phase swimming in Rana esculenta. Note that extension and exion occur alternated. The left front limb moves in synchrony with the right hind limb and vice versa. Displacement in (m) Right ankle Flexion.14.12.1.8.6.4.2.2.4 Left ankle Extension Right hand Left hand.1.2.3.4.5.6.7.8.9 1 1.1 1.2 1.3 1.4 Time Fig. 4. Displacement of right (black) and left (white) ankle relative to the snout tip and of the right (black) and left (white) hand relative to the right and left shoulder in function of time during crawling on land in Rana esculenta. Note that extension and exion occur alternated. The left front limb moves in synchrony with the right hind limb and vice versa. Normalized work (L/kg.m) 1.8 1.6 1.4 1.2 1.8.6.4.2 In phase Out of phase.1.2.3.4.5.6 Velocity Fig. 5. External positive work as a function of the velocity during both swimming gait types. Swimming frogs typically propel themselves by kicking both hind limbs simultaneously (Calow & Alexander, 1973). Abourachid & Green (1999) stated that all derived anuran species, exempli ed by Rana sylvatica and Bombina variegata, swim using in-phase movements of hind limbs while holding the forelimbs at alongside the body. In R. esculenta, the front limbs are pulled away from the body before the propulsion phase of swimming and repositioned close to the body during actual propulsion. In this species, the same pattern holds for jumping: the front limbs are pulled forward in the rst phase of the movement, but held alongside the body during the airborne phase of the jump (pers. obs.). At least super cially, the kinematics of in-phase swimming and jumping in frogs are similar (Calow & Alexander, 1973; Gal & Blake, 1988). This is con rmed by EMG studies (Emerson & De Jongh, 198; Kamel, Peters & Bashor, 1996), which indicate that adaptations for jumping performance in frogs do not necessarily compromise swimming ability (Gal & Blake, 1987). Although mentioned by some authors (e.g. Calow & Alexander, 1973), swimming using alternating limb movements has been largely overlooked in previous work on anurans (Peters, Kamel & Bashor, 1996). Observations of out-of-phase leg movements in primitive frog species have been used as evidence: (1) to support the notion that swimming and jumping involve different kinematics; (2) to hypothesize separate evolutionary derivations of these two locomotor modes (Abourachid & Green, 1999). Abourachid & Green (1999) stated that out-of-phase aquatic movements of frogs seem at odds with their specializations for saltatory terrestrial locomotion. Although the frogs are good jumpers, they rarely use in-phase leg movements in the water. As Abourachid & Green (1999) claim that outof-phase movements are not related to terrestrial locomotion, it seems that both types of locomotor behaviour, used by the frogs, are independently derived characters. However, this study demonstrates that alternating leg movements are also used by derived frog species and that this type of movement results in a clearly different velocity to that obtained using in-phase leg movements. Speci cally, frogs alternate leg movements when swimming at slow speeds. Frogs kick their legs simultaneously at fast and slow swimming speeds (Fig. 1). Moreover, even primitive frog species as Leiopelma hochstetteri, L. archeyi, and Ascaphus truei sometimes

Gait transition in swimming frogs 187 use the in-phase gait type: for example, in an initial launch after having entered the water or as an escape behaviour (Abourachid & Green, 1999). Both these situations require that frogs rapidly attain a high swimming speed; it is suggested that out-of-phase and inphase leg movements by frogs are two distinct gait types used for swimming. The same velocity-dependent gait transition exists in terrestrial locomotion by frogs. That is, individuals use a crawling behaviour (walking-trot) when approaching a prey, while jumping is used when a fast escape is required. The species of primitive frogs examined by Abourachid & Greens (1999) may not have used in-phase leg movements because the frogs used in their experiments did not reach the higher velocities. Unfortunately, mean average speeds of the in-phase or out-of-phase movement sequences are not reported, although one example is given of an out-of-phase cycle resulting in speeds of 18 cm/s. This velocity lies in the lower range of the velocities that we obtained for R. esculenta. Either the frogs were not stimulated to swim at maximal speed or their lifestyle may not require adaptations for fast swimming performance. In the latter case, in-phase swimming might be lost in the behavioural repertoire of this species. Indeed, primitive aquatic frog species tend to dive to the bottom of streams and hide when faced with danger rather than eeing (Abourachid & Green, 1999). Moreover, one of the primitive frog species tested by Abourachid & Green (1999) is strictly terrestrial and therefore not expected to have adaptations for fast swimming. The question remains, why do the two gait types coexist? Why is a transition present? Although other hypotheses are formulated (e.g. Biewener & Taylor, 1986; Farley & Taylor, 1991), a widespread assumption is that animals change gaits to minimize metabolic energy consumption (e.g. Hoyt & Taylor, 1981; Hreljac, 1993; Alexander, 1999). Only the positive external mechanical energy was calculated and plotted against mean velocity. This shows that energy requirements for out-of-phase swimming are considerably lower than those for in-phase bouts at the same speed (Fig. 5). This result can be explained by considering the acceleration pro le of both behaviours. Since accelerations are much higher during in-phase swimming compared to the more steady swimming state of out-of-phase swimming, the former gait is characterized by higher inertial energy losses. Moreover, the drag of in-phase swimming is higher since a higher maximal velocity is attained. Therefore, the alternated swimming style seems to be more economical. However, it is possible that there are also potential drawbacks of swimming using the alternate gait, especially at high speeds. Swimming faster, implies that animals gain a higher thrust. Gal & Blake (1988) showed that a simple drag-based propulsion mechanism was not suf cient to describe the in-phase locomotor behaviour of frogs. They suggested jet propulsion and an interaction effect between the simultaneously kicking feet as additional sources of propulsive thrust. In out-of-phase swimming, these two effects are not applicable, since the legs move individually. Moreover, in-phase swimming (kick and glide swimming) has a glide phase in which no metabolic energy is required, since the frog's body stays rigid and decelerates due to drag forces. At a certain velocity the bene ts of swimming inphase will compensate for the costs associated with this behaviour and out-of-phase swimming becomes relatively less ef cient. It is hypothesized that inertial and drag arguments are important in the lower velocity range, while jet propulsion and the interactive effect are crucial at higher velocities. Besides this `whole body' energetic argument, the frequency in which the hind limbs are kicked, can be the limited factor as well. Kick frequencies must be higher in out-of-phase swimming since no jet propulsion or interactive effect is possible during alternated swimming. When the frequency of kicking becomes higher, the muscles possibly act in unfavourable force±velocity conditions. This would result in a lower ef ciency of the muscle work, which may prompt the frog to switch to another gait type with relatively higher ef ciency of muscle work. Several hypotheses are suggested that explain why two swimming gait types may co-exist in frogs. However, the question still remains why frogs sometimes use in-phase swimming at lower velocities. Perhaps the in-phase sequences displayed at slower speeds are portions of sequences where the intended locomotion objective is attaining a high velocity. Higher velocities in frogs would be attained more rapidly when using the in-phase swimming style. Similarly, human sprinters don't begin a fast dash by walking, but rather accelerate by running from the start. As the frogs do not show this behaviour frequently, only four sequences were obtained to test this hypothesis. Moreover, frogs use a kick-and-glide swimming that results in high uctuations in the velocity pro le. However, in three out of four sequences, peak velocities increase during the in-phase swimming at very low speeds; this indicates that there is acceleration. Based on our limited data set, however, the `acceleration' hypothesis can be neither con rmed nor rejected. Based on our results, it is suggested that a gait transition exists in frog locomotion during movement on land and in water. Additionally, parallels can be drawn between crawling and out-of-phase swimming and between jumping and in-phase swimming. The presence of this gait transition in both media encourages us to re-establish the hypothesis put by Gal & Blake (1987) that was subsequently questioned by Abourachid & Green (1999). This hypothesis states that adaptation for good jumping performance in frogs does not compromise swimming ability. This notion is extended by suggesting that terrestrial and aquatic modes might essentially involve the same gait transition and, therefore, are likely to have evolved together, with adaptations for one locomotor mode automatically being propitious for the other.

188 S. Nauwelaerts and P. Aerts Acknowledgements The authors are grateful to Anthony Herrel for helpful discussions. This work was supported by an IWT grant (99131) to SN and a GOA-BOF UIA-1999 grant to PA. PA is a research director of the FWO Fund for Scienti c Research, Flanders. REFERENCES Abourachid, A. & Green, D. M. (1999). Origins of the frog-kick? Alternate-leg swimming in primitive frogs, families Leiopelmatidae and Ascaphidae. J. Herpetol. 33: 657±663. Alexander, R. McN. (1999). Energy for animal life. Oxford: Oxford University Press. Anderson, B. D., Feder, M. E. & Full, R. J. (1991). Consequences of a gait change during locomotion in toads (Bufo woodhousii fowleri). J. exp. Biol. 158: 133±148. Biewener, A. A. & Taylor, C. R. (1986). Bone strain: a determinant of gait and speed? J. exp. Biol. 123: 383±4. Calow, L. J. & Alexander, R. McN. (1973). A mechanical analysis of a hind leg of a frog Rana temporaria. J. Zool. (Lond.) 171: 293±321. Daniel, T. L. (1984). Unsteady aspects of aquatic locomotion. Am. Zool. 24: 121±134. Emerson, S. B. & De Jongh, H. J. (198). Muscle activity at the ilio-sacral articulation of frogs. J. Morphol. 166: 129±144. Farley, C. T. & Taylor, C. R. (1991). A mechanical trigger for the trot±gallop transition in horses. Science 253: 36± 38. Gal, J. M. & Blake, R. W. (1987). Hydrodynamic drag of two frog species: Hymenochirus boettgeri and Rana pipiens. Can. J. Zool. 65: 185±19. Gal, J. M. & Blake, R. W. (1988). Biomechanics of frog swimming II. Mechanics of the limb-beat cycle in Hymenochirus boettgeri. J. exp. Biol. 138: 413±429. Hoyt, D. F. & Taylor, C. R. (1981). Gait and energetics of locomotion in horses. Nature (Lond.) 292: 239±24. Hreljac, A. (1993). Preferred and energetically optimal gait transition speed in human locomotion. Med. Sci. Sports Exerc. 25: 1158±1162. Kamel, L. T., Peters, S. E. & Bashor, D. P. (1996). Hopping and swimming in the leopard frog, Rana pipiens: II. A comparison of muscle activities. J. Morphol. 23: 17±31. Nauwelaerts, S., Aerts, P. & D'AouÃt, K. (21). Speed modulation in swimming frogs. J. Motor Behav. 33: 265±272. Peters, S. E., Kamel, L. T. & Bashor, D. P. (1996). Hopping and swimming in the leopard frog, Rana pipiens: I. Step cycles and kinematics. J. Morphol. 23: 1±16. Walton, M. & Anderson, B. D. (1988). The aerobic cost of saltatory locomotion in the Fowler's toad (Bufo woodhousei fowleri) J. exp. Biol. 136: 273±288.