Kinematic and Electromyographic Analysis of the Functional Role of the Body Axis During Terrestrial and Aquatic Locomotion in the Salamander Ambystoma Tigrinum

Aquatic and terrestrial environments place vastly different locomotor demands on animals. Whereas aquatic vertebrates primarily use axial undulations of the trunk and tail to propel themselves through the water, terrestrial vertebrates are perceived as using primarily limb action to propel themselves over the ground. Functional studies of aquatic locomotion in vertebrates (mainly in fishes), therefore, have focused primarily on the role of the body axis (Gray, 1933; Webb and Keyes, 1983; Williams, 1986), while similar studies of terrestrial locomotion have centered on the supportive role of the limbs (Hildebrand et al. 1985). Indeed, direct comparisons, of aquatic and terrestrial locomotion have largely neglected axial movements in walking animals (Goslow et al. 1989; Grillner and Wallén, 1985; Székeley, 1989).

Elongate ‘anguilliform’ and ‘carangiform’ fish exhibit traveling waves of lateral undulation that propel them through the water. ‘Lower’ tetrapods, in contrast, exhibit standing waves of lateral bending to increase their stride length in a sprawling gait (Edwards, 1976; Daan and Belterman, 1968). The patterns of axial muscle activity that are exhibited during these different locomotor modes, however, have not been previously investigated.

The urodele lissamphibians, or salamanders, present an ideal group to address aquatic/terrestrial differences in axial function during locomotion. These animals exhibit many primitive tetrapod features with their sprawling gait, elongate trunk and tail, and fore-and hindlimbs of similar size. Furthermore, they primitively (i.e. plesiomorphically) exhibit a biphasic life cycle involving a metamorphic transition from aquatic larvae to terrestrial adults. Finally, in certain species, such as A. tigrinum, an alternative neotenic life cycle involves foregoing metamorphosis into the terrestrial environment and reaching adulthood in a fully aquatic larva-like form (Fig. 1A). Whereas neotenic individuals of A. tigrinum are fully aquatic and never leave the water in nature, metamorphosed adults must return to the water to breed, and retain good swimming capabilities for this purpose. These two alternative developmental pathways make it possible to compare the locomotor features of the body axis of neotenic adults, which retain the aquatic adapted larval morphology, with those of terrestrially adapted metamorphosed adults.

A number of differences in axial morphology between neotenic and metamor-phosed individuals might be expected to influence their locomotion. The way in which axial muscle contractions initiate trunk bending may be affected by extensive repatteming of the dermal collagen fibers at metamorphosis (Frolich and Schmid, 1991) (the skin is a major site of axial muscle attachment), the ossification of major parts of the vertebrae at metamorphosis (Worthington, 1971) and the reduction in axial muscle cross section at metamorphosis (L. M. Frolich, in preparation). In addition, the hydrodynamics by which lateral undulations of the body axis are used to produce thrust may be affected by external morphological differences, such as the change in snout shape and the loss of the medial dorsal and ventral fins at metamorphosis.

The environmental and morphological differences associated with these two alternative life cycles, then, present a natural framework for addressing two fundamental questions regarding aquatic/terrestrial transitions in axial function during locomotion (Fig. 1B; Lauder and Shaffer, 1988). (1) What is the effect of morphological differences between neotenic and metamorphosed individuals on axial function during locomotion? (2) What is the effect of the transition to terrestriality on axial function during locomotion?

Three neotenic [15.5-16.9cm SVL (snout-vent length),x̄=16·0cm; 27.5-29.0 cm TL (total length), x=28.3] and three metamorphosed (10.2-11.5 cm SVL, x̄10·9cm; 18.5-21.3cm TL, x=19.8cm) individuals of Ambystoma tigrinum (Green, 1825) were obtained from a commercial supplier. All animals were kept at 8°C (neotenic individuals in aerated dechlorinated aquaria; metamorphosed individuals in moist boxes) and acclimated to 20°C for 3h before experimental recordings were carried out.

Following anesthetic sedation [MS222 (1:1000)], bipolar hook EMG electrodes (Gans and Loeb, 1986) constructed from fine (0.01mm) insulated silver wire (California Fine Wire Co.) were inserted percutaneously into the axial muscles of anesthetized animals using a hypodermic needle through which the electrode tips had been threaded. Insulation was removed from the tip of each wire (to a distance of 0.5 mm, measured under a dissecting microscope equipped with an ocular micrometer), and the two electrode tips were separated by 1.0mm along the length of the muscle fibers. The electrode wires were gathered into a bundle and sutured to the skin at the dorsal mid-trunk to prevent tension in the wires from disturbing the electrodes’ position in the muscle.

The epaxial muscles in Ambystoma tigrinurn form a single large dorsolateral epaxial bundle (M. dorsalis trunci after Francis, 1934). Along the length of the body, this bundle is partitioned by myosepta into myotomes, one corresponding to each body segment or vertebra. Fourteen or fifteen vertebrae are present in the trunk region (between the single cervical vertebra and the single sacral vertebra), three are present in the caudosacral region (between the sacrum and the base of the tail), and a variable number occur in the tail. Electrodes were inserted bilaterally at five sites in the trunk and caudosacral region with four or five segments between electrodes (Fig. 2). Because myosepta are angled anteriorly to attach to the neural spine forming a ‘medial anterior myotomal flexure’ (sensuNaylor, 1978), an electrode in the fourth trunk myotome (the myotome that inserts on the myoseptum of the fourth trunk vertebra) is lateral to the sixth trunk vertebra. The epaxial muscle situated dorsal and superficial to the vertebra in the region of the medial anterior flexure is differentiated into a separate bundle (M. interspinalis, which inserts on successive neural spines). Initial experiments revealed no EMG activity in these more medial fibers during swimming or trotting locomotion. Thus, in subsequent experiments, all electrodes were inserted into the larger lateral portion of the M. dorsalis trunci.

Small white marker beads were sutured to the skin at 16 locations along the mid-dorsal line (Fig. 2) to serve as kinematic reference points and the animals were allowed 2h to recover from anesthesia before kinematic and EMG recordings were started.

After recordings had been made, animals were killed with an overdose of MS222, and electrode positions were confirmed by radiography and dissection.

Animals were videotaped at 60 fields s⁻¹ with a 1/1000 s high-speed shutter (Panasonic PV330 video camera) while swimming in a flow tank (both neotenic and metamorphosed individuals) or trotting on a treadmill (metamorphosed individuals only). A white centimeter grid background was used for scale. Both dorsal and lateral views were captured in the same video field by suspending a mirror at 45° over the working section of the flow tank or treadmill. Videotapes were synchronized with EMG recordings by means of an electrical circuit that illuminates an LED in the video camera field and simultaneously emits a 5 V pulse that can be sampled, together with the EMG data, by a microcomputer (see below).

For each animal, at least ten swimming and ten trotting (metamorphosed individuals only) cycles were analyzed. In the water, both neotenic and metamor-phosed individuals exhibit low-speed paddling with the limbs and high-speed bursts with the limbs held back. Little axial EMG activity was observed at the sites that were sampled during the lower-speed paddling. We therefore analyzed only burst sequences, which were elicited by squeezing the base of the tail with forceps.

The first two cycles following the initiation of burst swimming, however, were not used. Only sequences in which the animal maintained a constant forward speed and direction and did not rotate from the plane of the video field were chosen for analysis. Edwards (1977) distinguishes a walking gait from a trotting gait in salamanders. In the walk, three limbs remain in contact with the ground at all times, with four limbs in contact during part of the cycle. In the trot, two contralateral limbs provide support for most of the cycle and contralateral limbs move approximately in phase. Metamorphosed A. tigrinum rarely exhibit a walking gait and preliminary experiments revealed no detectable epaxial EMG activity associated with a walking gait at the sites that were sampled. Thus, all data presented for terrestrial locomotion are for trotting sequences.

The kinematic approach we used to simplify analysis of complex waves involves measuring the magnitude and timing of lateral displacements of the body axis at specific points along the animal’s length. This analytical approach yields infor-mation on the pattern of wave propagation, wave amplitude along the body axis, wave velocity and phase lag.

Videotape recordings were digitized and analyzed using PEAK Performance motion analysis software run on an IBM-compatible microcomputer. The sixteen marker bead locations along the body axis, as well as the knees and elbows, were digitized for each frame from the sequences chosen for analysis. A single cycle of locomotion occupied 15-60 video fields with a time resolution of ±l/60s. Measurement error of maximal lateral displacement, based on five repeated digitizations of the same sequence, was ±1 mm at any given point along the body axis. To account for differences in size among individuals, lateral displacements at each marked point along the body axis were normalized by dividing by the animal’s SVL.

Raw EMG signals were amplified (5 x 10³-20 x 10³) and band-pass filtered (at 30 and 3000 Hz) using Grass P511K amplifiers. The amplified signal was then sampled at 1000 Hz by a 12-bit analog-to-digital converter and entered into a microcom-puter, effectively lowering the high-frequency resolution to 500 Hz. To verify that significant frequency components above 500Hz were not lost from the signal, EMGs were also sampled initially at 2000 Hz. No conspicuous attenuation in the frequency power spectrum of EMGs sampled at 1000 Hz versus 2000 Hz was observed.

Patterns of EMG activity were analyzed from the same sequences chosen for kinematic analysis. Onset and offset times (±l/100s, based on five repeated digitizations of the same sequence) of EMGs were digitized using a superimposed cursor directly on a computer screen graphic and entered into a data spread sheet for later analysis. EMG timing variables from all three neotenic or all three metamorphosed individuals were pooled for each locomotor mode. Means and standard deviations were then calculated for the pooled data (A=30) and compared using a Student’s t-test.

In the sequences chosen for analysis (see above criteria), swimming neotenic individuals have a mean forward velocity of 47.5 cms’¹ (range: 25.6-65.6cm s^-1), which is similar to that of swimming metamorphosed individuals (mean: 36.5cms^-1; range: 21.8-48.0cms^-1), but much faster than that of trotting metamorphosed individuals (mean: 8.1 cms^-1; range: 7.0-10.5cms^-1). To test whether velocity has a significant effect on measured kinematic and EMG variables within each group over the range of speeds analyzed, we carried out least-squares regressions of one kinematic variable (mid-trunk maximal lateral displacement) and one EMG variable (left mid-trunk burst duration) against velocity (independent variable). The slopes of these regression were not signifi-cantly different from zero (P>0.05 in all cases), indicating that velocity is not a significant source of within-group variation. On this basis, measurements of these and the other kinematic and EMG variables for the three neotenic or three metamorphosed individuals are pooled for each locomotor mode. Means for the pooled data (N=30 unless otherwise indicated) are compared using a Student’s t-test.

Cycle duration is the time period of a single locomotor cycle, defined here as the time between onsets of EMG activity in the left mid-trunk myotome. The distance that an animal travels per cycle equals the product of cycle duration and velocity. Within a given locomotor mode, distance per cycle varied less than 30% over at least a twofold range of cycle duration for the speeds that were observed (Fig. 3). Thus, for the speeds studied, the animals primarily increase velocity by increasing cycle frequency (Fig. 3). Because of the large variation in cycle duration, all absolute kinematic and EMG timing variables are normalized by conversion to a fraction of the locomotor cycle in which they occurred. Timing patterns of EMG activity and kinematics among individuals using each locomotor mode are consistent across the range of cycle durations observed within each group.

During swimming, both neotenic and metamorphosed individuals use a travel-ing wave of lateral undulation (Fig. 4A,B) that changes in amplitude as it passes along the length of the body. Maximal displacement of the snout (5-6 % SVL) is slightly greater than maximal displacement of the neck and pectoral region (3-4 % SVL) because of slight yaw of the head. Posterior to the pectoral girdle, however, lateral displacement gradually increases as the wave travels down the body axis, resulting in the largest displacements (25 % SVL) at the tip of the tail (Fig. 5A,B). Neotenes exhibit larger lateral displacements at all sites along the trunk; however, this difference is significantly different (N =30; P<0.01) only in the posterior trunk and caudosacral region (trunk 3 to tail 1).

The propagation of this traveling lateral undulatory wave can be clearly seen by the timing of maximal lateral displacements along the length of the trunk during a locomotor cycle (Fig. 5A,B). As the locomotor cycle begins (time 0.0 -arbitrarily defined as the onset of EMG activity in the left mid-trunk myotome), a wave of maximal lateral displacement is initiated on the right side anterior to the shoulder girdle. This wave then travels with a constant velocity (1.25 SVL cycle^-1 in neotenes; 1.23 SVL cycle^-1 in metamorphosed individuals; based on the slope of a least-squares regression on displacement timing versus fraction of SVL, r²>0.98 for both) down the right side of the animal (see Fig. 9). Halfway into the locomotor cycle (at relative time 0.5), this right-side wave reaches the anterior trunk (just posterior to the forelimbs) and a new wave is initiated on the left side. As the next locomotor cycle begins (time 1.0), the first right-side wave is at the base of the tail, the left-side wave has reached the anterior region of the trunk, and a new wave is initiated on the right side. This pattern is repeated with waves passing alternately down the right and left sides of the animal. As a result, the body axis is always thrown into an S-shaped curve that travels down the animal.

Corresponding to the traveling wave of lateral displacement, all animals exhibit a traveling wave of EMG activity during swimming (Fig. 6A,B). Because all animals showed extremely regular symmetrical patterns of alternating EMG activity between left and right sides of the body (Fig. 7), we concentrate on patterns along the left side only, as these are reflected by similar patterns on the right side. At any given location along the trunk, cessation of EMG activity in the left myotome occurs prior to maximal lateral displacement to the right side, except in the neck region of neotenes (Fig. 8). The duration of EMG activity does not differ along the length of the trunk; however, neotenes consistently exhibit longer burst durations (P<0.00l) than metamorphosed individuals at any given electrode site (Fig. 9). The EMG wave travels at a constant velocity down the animal’s trunk, but is slightly faster in neotenic individuals (Fig. 9; EMG onset travels at 1.88SVLcycle^-1; EMG offset travels at 1.75SVLcycle^-1; r²>0.97) than in metamorphosed individuals (EMG onset travels at 1.67 SVL cycle^-1; EMG offset travels at 1.48SVLcycle^-1; r²>0.96). These EMG wave velocities, however, are not statistically distinguishable (P>0.05).

Because EMG waves travel faster than kinematic waves (in both neotenes and metamorphosed individuals), the EMG-mechanical delay (or phase lag -see Grillner and Kashin, 1976; Williams et al. 1989) increases along the length of the trunk. This delay can be quantified by subtracting the time of EMG onset from the time at which maximum lateral displacement is reached (Fig. 10). This delay increases steadily along the length of the trunk (from 0.40-0.45 cycles at the neck to more than 0.65 cycles at the base of the tail), except in neotenic individuals at the sacral region (between sacral and caudosacral electrodes, where the delay actually decreases) and in metamorphosed individuals at the anterior trunk (between the neck and trunk 1 electrodes, where the delay remains constant). Neotenic individuals exhibit significantly longer mean EMG-mechanical delays than neotenic individuals at the mid-trunk and posterior trunk electrode sites (N=30; PcO.Ol).

Because the onset of EMG activity in the left mid-trunk myotome serves as a reference for all other EMG timing variables, it is always scored as time zero, preventing an assessment of its variability with respect to that of other myotomal EMGs. Histograms of burst durations in the left mid-trunk myotome for all three locomotor modes shown in Fig. 9 provide an indication of overall variability in left mid-trunk muscle activity. Burst duration is normally distributed with the means indicated in each graph. Neotenes exhibit a significantly longer burst duration than metamorphosed individuals (PcO.OOl) in the left mid-trunk (Fig. 11), as well as at other muscle recording sites along the trunk (Fig. 8). Sample sizes are not large enough to determine whether the longer muscle bursts of neotenic individuals are due to relatively earlier EMG onset or to relatively later EMG offset.

Both neotenes and metamorphosed animals exhibit traveling kinematic and muscle activity waves during swimming. Within the qualitatively similar traveling wave pattern, however, statistical analysis reveals subtle quantitative differences between the two groups in burst duration and EMG-mechanical delay. A key question, then, is how these similarities and differences, together with the important differences in axial morphology, affect swimming performance.

One measure of an animal’s mechanical efficiency during swimming is the ratio of its forward velocity relative to the posterior velocity of the traveling wave of flexion, often referred to as the ‘slip.’ Slip values can range from zero (where the animal has zero forward velocity) to one (where the animal’s forward velocity equals the velocity of the wave traveling down it body axis). In neotenic individuals, slip ranges from 0.56 to 0.75 (Fig. 12), and there is no significant correlation with velocity (at the 95 % confidence level based on a least-squares regression) over the range of speeds at which slip was measured (from 1.8 to 4.1SVLs^-1). In metamorphosed individuals, slip ranges from 0.42 to 0.63 and, again, there is no significant correlation with velocity (at the 95% confidence level) over the range of speeds at which slip was measured (2.1-4.2SVLs^-1). Given that slip is independent of velocity over the range of speeds studied for each group, a comparison of the pooled data shows that neotenes have significantly higher slip (x̄=0·±0·06,s,D.) than metamorphosed individuals (x̄=0·±0·06,s,D.) (P<0.001).

In contrast to the traveling wave used during swimming, a single standing wave (Figs 4C, 5C) with fixed nodes at the pectoral and pelvic girdles is employed during trotting. Maximal lateral displacements along the body axis during trotting are far less than those seen during swimming (Fig. 5). The smallest lateral displacements (3 % SVL) are at the pectoral and pelvic girdles (which serve as nodes for the standing wave of the trunk), while the greatest lateral displacements (6% SVL) occur at the snout, mid-trunk and tail (which are anti-nodes for the standing wave). Thus, whereas during swimming a wave of increasing amplitude travels down the trunk, peaking at the tail, during trotting a much smaller-amplitude standing wave oscillates between the pectoral and pelvic girdles with maximal displacements at mid-trunk. Trotting lateral displacements are significantly less than swimming displacements (N=30; P<0.01) in both neotenic and metamor-phosed individuals at all locations except at the snout, occiput and mid-trunk (trunk sites 2-4).

At the start of a trotting cycle, the trunk between the forelimb and hindlimb is maximally displaced to one side (left shown in Fig. 4C), while the regions anterior to the forelimb (neck and head) and posterior to the hindlimb (caudosacral and tail) are maximally displaced in the opposite direction (right shown in Fig. 4C). Halfway into the cycle (time 0.5), the trunk bends maximally in the opposite direction, forming a mirror image of the start of the cycle. Hence, during trotting, the body axis (between the head and base of the tail) forms a single standing arc that undergoes one full oscillation during a locomotor cycle.

Synchronous left-side EMG activity at recording sites between the forelimb and hindlimb (Fig. 6C) precedes simultaneous concave left flexion (coincident with maximal right displacement) of the standing kinematic wave used during trotting (Fig. 8C). EMG activity in the left neck and left caudosacral myotomes immedi-ately precedes maximal right displacement of the head and tail and is thus exactly out of phase with EMG activity in the trunk. The end of EMG activity precedes maximal lateral displacement by 10-20 % of a single locomotor cycle (Fig. 8C) and, as would be expected for a standing wave, this EMG-mechanical delay is constant along the length of the trunk (Fig. 8). Burst duration in the left mid-trunk myotome of trotting metamorphosed individuals does not differ significantly from that of swimming metamorphosed individuals.

Both neotenic and metamorphosed animals exhibit qualitatively similar travel-ing waves of epaxial muscle activity that precede the traveling waves of lateral undulation used during swimming. Statistical analysis reveals subtle differences between the two groups in the patterns of mechanical (i.e. kinematic) and EMG wave propagation, quantified in terms of wave amplitude, wave velocity and EMG-mechanical delay. These subtle EMG and kinematic differences may reflect how morphological differences alter the way in which (1) axial muscle forces are translated into lateral undulations of the trunk and (2) lateral undulations cause forward movement of the animal through the water.

A key unanswered question in studies of aquatic locomotion is how axial muscle contractions are translated into lateral undulatory movements of the body axis that propel the animal through the water. Neotenic individuals may rely more heavily on muscle activity to stiffen the trunk and thereby transmit axial bending forces to the tail, compared to metamorphosed individuals which exhibit briefer periods of muscle activity.

The only consistent difference in swimming patterns is that neotenic individuals exhibit longer burst durations than metamorphosed individuals at all electrode sites. In his general model for swimming movements in vertebrates, Blight (1977) views the body axis as a ‘hybrid oscillator’ wherein a stiffness-dominated trunk transmits force to a resistance-dominated tail. An extended period of muscle contraction in neotenes may serve to stiffen the trunk and thereby enhance the transmission of bending forces posteriorly to the expanded caudal fins. If the sole function of the epaxial myotomes of the trunk were to stiffen the body, continuous contractile activity might be expected during swimming. However, this function must be balanced against the need to produce lateral undulation of the trunk for thrust, which requires oscillatory EMG activity. To satisfy both of these require-ments, epaxial muscles should be active for only half of a locomotor cycle. To a good approximation (Fig. 8), this is what is observed in swimming neotenic A. tigrinum. The significantly shorter EMG burst durations of metamorphosed individuals, in contrast, suggest that muscular contraction to stiffen the trunk may be less important in these animals.

Another prediction of Blight’s (1977) model is that the amplitudes of lateral displacements in the tail should be much greater than in the trunk, since the tail is resistance-dominated and produces thrust to move the swimming animal forward. In both neotenic and metamorphosed individuals, wave amplitude increases as the wave travels the length of the body. However, neotenes produce slightly larger amplitudes in the tail, providing further evidence that muscular forces in the trunk are more effectively transmitted to the tail in neotenic individuals relative to metamorphosed individuals. In general, whereas the timing and displacement patterns of lateral undulatory movements may be very similar between neotenic and metamorphosed individuals, the way in which muscle force is transmitted along the length of the body axis to produce thrust may be quite different.

Williams et al. (1989) suggest that an increase in phase lag (EMG-mechanical delay) as a wave travels posteriorly reflects more effective transmission of propulsive forces along a stiffened body to the tail. By tensing progressively more-posterior epaxial myotomes before the traveling wave of flexion reaches that part of the trunk, a longer segment of the epaxial bundle is stiffened and bending forces from more-anterior myotomes are transmitted to the tail. Neotenic individuals exhibit longer EMG-mechanical delays at the mid-and posterior trunk. However, given that the basic mechanics by which axial muscle contractions produce bending of the trunk is not understood, it is not possible to evaluate how changes in EMG-mechanical delay affect the transmission of bending waves down the animal’s body. Experiments that address these questions are needed.

Enhanced transmission of bending forces along a stiffened trunk (indicated by the longer burst durations and slightly greater EMG-mechanical delay in the posterior trunk) in neotenes may be aided by the dermis, which is a major site of attachment for the axial musculature. The dermis of neotenic individuals is composed of a thick crossed array of collagen fibers that wind helically around the animal (Frolich and Schmid, 1991). This pattern is lost in favor of a thinner, less-organized mat of fibers in metamorphosed animals. Such helically wound fiber systems have been argued by Wainwright et al. (1978) to be a good mechanical design for transmitting muscular forces in the trunk to the tail during swimming in sharks. If the overlying dermis on which the muscle fibers insert (either directly or via the myosepta) is helically wound, as in sharks and in neotenic A. tigrinum, tension in the dermal fibers will be maintained during the myotomal shortening and area expansion that necessarily accompany muscle contraction (Wainwright et al. 1976, 1978). Maintenance of tension is necessary if the dermis is to function as a tendon to transmit force down the body axis. The helically wound organization of the neotenic dermis, thus, may work in conjuction with longer epaxial activation (as shown by increased EMG burst duration) and an increased delay between muscle activation and mechanical response to transmit muscle forces in the trunk more efficiently to the large caudal fins.

Compared to differing fish species (Williams et al. 1989) and different snake locomotor modes (Jayne, 1988), which exhibit phase lag (EMG-mechanical delay) differences of one-quarter to half a cycle at the same point along the animal’s body axis, neotenic and metamorphosed salamanders exhibit only small differences (less than 15 % of a cycle) in EMG-mechanical delay at any given point along the body axis. Jayne (1988) has shown that, whereas during aquatic locomotion snakes exhibit a significant increase in EMG-mechanical delay as a wave travels down the body, the same individual during terrestrial locomotion exhibits a consistent EMG-mechanical delay along the length of the body. In salamanders, despite the important axial remodeling and the radical transition in wave pattern that accompany metamorphosis into the terrestrial environment, both neotenic and metamorphosed individuals retain similar EMG-mechanical delay patterns. It would seem, then, that an increase in EMG-mechanical delay as a traveling wave moves down the body is an extremely important part of swimming locomotion that is always retained in those tetrapods that have been studied, even when terrestrial locomotion involves severe alterations of this pattern.

Based on our measurements of slip (which is clearly a limited indication of locomotor performance), neotenic individuals (x̄=0.66±0.66,S.D.) are better swimmers than metamorphosed individuals (x̄=0.53±0.66,S.D.). Larval A. tigrinurn exhibit similar maximal slip values (0.6-0.7) to neotenes (von Seckendorf Hoff et al. 1989). Metamorphic remodeling may result in a loss of aquatic adaptations, such as the medial dorsal and ventral fins, in order to accommodate the demands for support in the terrestrial environment (and other possible terrestrial complications such as water loss through a thin membrane).

Our general finding, that despite the overall similarities in kinematic and muscle activity patterns, neotenic individuals perform better in the aquatic environment than do metamorphosed individuals, mirrors the observations of Lauder and Shaffer (1986, 1988) in their studies of the functional morphology of feeding in neotenic and metamorphosed A. tigrinurn. Although extensive remodeling of the head and jaws occurs at metamorphosis, these workers found that the kinematic and EMG patterns of aquatic gape-and-suck feeding of larval and metamorphosed individuals are virtually identical. Larval individuals, however, did exhibit better aquatic feeding performance than metamorphosed animals.

Whereas metamorphosed A. tigrinurn exhibit traveling axial kinematic and electromyographic waves during swimming, the same individuals employ a standing wave pattern of axial movement and muscle activity when trotting on land. Thus, depending on environment, the same individual uses two distinct patterns of axial movement during locomotion. A key question is what controls which axial locomotor pattern is used. Is it afferent feedback from the environ-ment, or does integration of axial movement with movements of the limb require a standing axial wave? Most locomotor movements are thought to be generated in the spinal cord, with little or no peripheral feedback necessary (McMahon, 1984). This is clearly the case for limb movements in salamanders (Székeley, 1989; Székeley et al. 1969) and for axial movements in a variety of fishes (Williams, 1986). Evidence for whether swimming amphibians require afferent feedback to generate a traveling wave remains equivocal (see below) and virtually no neurophysiological data exist concerning the control of axial wave patterns during terrestrial locomotion in any laterally undulating tetrapod.

It is clear from Coghill’s work (1929) that both standing and traveling waves of lateral bending occur quite early in amphibian development, with standing wave patterns probably appearing first. However, the motor control of muscle activity that produces these two types of wave patterns is not well understood and has been previously studied only in embryos or very young larvae.

Salamander embryos (Triturs vulgaris) exhibit both traveling and standing waves of axial lateral bending (Soffe et al. 1983), axial muscle activity (Blight, 1977) and motoneuron activation (Soffe et al. 1983). It is not clear whether a rostral-caudal delay (i.e. a traveling wave) of motoneuron activation is preserved in deafferented amphibian embryos (Kahn and Roberts, 1982; Soffe et al. 1983; Stehouwer and Farel, 1980), leaving open the question of whether afferent feedback (and, by extension, environmental cues) is important for producing traveling waves in the aquatic medium. The clear message from experiments on neural control of axial movement in amphibian embryos is that both standing and traveling waves appear quite early in development, well before metamorphosis. This strongly suggests that metamorphosed individuals do not develop an entirely new motor control system to generate standing waves of axial lateral bending at metamorphosis but rather co-opt an existing embryonic program for use in terrestrial locomotion.

There is strong evidence that the earliest tetrapods (the ichthyostegids) were partially, if not fully, aquatic throughout their life cycle (Coates and Clack, 1990). Many later paleozoic tetrapod groups that exhibit definitively terrestrial adult forms had aquatic larvae. Thus, the original evolutionary transition from water to land in vertebrates probably occurred in species with biphasic life cycles involving metamorphosis from aquatic larvae to terrestrial adults. This is certainly the case for the dissorophid temnospondyls, which are commonly accepted as the closest sister group to the modern amphibians (Bolt, 1969, 1977; Bolt and Lombard, 1985). Thus, metamorphosis during the biphasic life cycle in Ambystoma tigrinum is certainly analogous to, if not homologous with, the earliest attainment of terrestriality by vertebrates. In this respect, salamanders make a good model for studying how this major aquatic/terrestrial transition in vertebrate evolution occurred. However, it is important to note that salamanders exhibit many derived features of their axial morphology relative to temnospondyls and other Paleozoic tetrapods (such as tiny reduced ribs) that must be carefully weighed in considering the evolution of terrestriality within this group.

Among those vertebrates that use lateral bending of the trunk during loco-motion, fish, swimming tetrapods (whether larval or adult) and limbless (or reduced-limb) terrestrial forms exhibit traveling waves (Blight, 1977; Daan and Belterman, 1968; Gray, 1933; von Seckendorf Hoff et al. 1989; Jayne, 1986, 1988; Manter, 1940; Wassersug and von Seckendorf Hoff, 1985; Webb and Keyes, 1983; Williams, 1986). Standing waves are seen in terrestrially locomoting limbed tetrapods, including salamanders (Gray, 1968; Roos, 1964; Sukhanov, 1974), lizards (Daan and Belterman, 1968; Schaeffer, 1941; Sukhanov, 1974) and crocodiles (Schaeffer, 1941). However, terrestrially locomoting salamanders of several species exhibit both standing and traveling waves (Edwards, 1976), and the present study demonstrates that metamorphosed A. tigrinum retain the ability to generate a traveling wave for aquatic locomotion.

Edwards (1977) suggests that the first terrestrial gait was a trot that employed a traveling wave of axial undulation, since this is the type of wave used by fish and amphibian larvae. However, the motor control system to produce a standing wave is present in amphibian embryos and larvae (see above). Many fish that use traveling waves during constant-velocity swimming probably have also standing wave capabilities for fast starts and escape responses (Webb and Blake, 1985). Thus, traveling waves of lateral undulation may not be primitive for terrestrial tetrapods, making it impossible on the basis of current data to support a statement as to which type of wave form was used in the earliest land vertebrates.

There is a dearth of information on how the vertebral column and associated axial structures are used during locomotion in all tetrapod groups. Unlike limbs, which are difficult to compare between fish and tetrapods (and often among distantly related tetrapod groups), axial structures are relatively easy to homolo-gize between major vertebrate groups. This, together with the fact that axial movement is an integral part of the locomotor system of all vertebrates, should make the study of axial locomotor systems a prime target for understanding major evolutionary transitions in locomotor mode.

This work was supported by NSF grant DCB 85-14899 to A.A.B. and a University of Chicago Hinds Fund grant to L.M.F. Dr Michael LaBarbera and Dr R. Eric Lombard provided helpful input throughout the course of the experiments and useful comments on earlier drafts of the manuscript. Videotapes were digitized using equipment kindly made available by Dr Brian Shea and Dr Sharon Swartz at Northwestern University.