Swimming in needlefish (Belonidae): anguilliform locomotion with fins

Anguilliform locomotion is widespread among aquatic animals and represents a convergent strategy for moving through water. Since Gray's pioneering work on a swimming juvenile Anguilla vulgaris (now Anguilla anguilla) (Gray, 1933),comprehensive and comparative analyses of anguilliform swimmers have revealed substantial variability in swimming kinematics. Although sources of this variability may include differences in neuromuscular control, ontogeny,swimming speed and phylogenetic history, a substantial component of the variation in kinematics may be due `to the external morphological differences in the shape of the trunk and the tail'(Gillis, 1996). Thus, a promising approach to understanding the causal basis for differences in anguilliform swimming kinematics is to make use of the morphological diversity found among anguilliform swimmers.

Elongate fishes in several phylogenetically and ecologically disparate families exhibit undulatory locomotion. In addition to the catadromous eels(Anguillidae), examples range from jawless fishes such as stream-dwelling lampreys (Petromyzontidae), to highly derived Perciformes such as rocky intertidal gunnels (Pholidae) and burrowing sand lances (Ammodytidae)(Nelson, 1994). In general,elongate fishes that swim using undulatory locomotion tend to live in benthic,structurally complex environments, are slow-to-moderate swimmers, and often have reduced or lost fins (Helfman et al.,1997).

An exception is the Atlantic needlefish (Strongylura marina), an elongate teleost related to the flying fishes (Exocoetidae) that lives in the surface waters of coastal marine environments. The behavior of S. marina leaping and skittering across the surface at high speeds when alarmed is well known (Collette,1977; Helfman et al.,1997) and has prevented them from being studied in captivity. Breder (1926) observed them more than 70 years ago, but thought them to be `rigid fishes...resembling esocids' whose swimming movements were not as close to the`anguilliform type of motion as might be expected judging from the form alone'. Needlefish possess a posterior arrangement of distinct, dorsal,anal and caudal fins that is unusual for anguilliform swimmers. Unlike in most elongate undulatory fishes, the bases of their relatively large pectoral fins are oriented closer to vertical than to horizontal(Collette, 1977; Helfman et al., 1997). No kinematic studies to date have described the axial kinematics of anguilliform locomotion in acanthopterygian fishes. In addition, there are no data on the fin kinematics of anguilliform swimming fishes, despite the fact that the median fins may contribute substantially to the lateral body profile.

Kinematic analyses of anguilliform swimmers have not emphasized the contribution of the fins since the principal force component is assumed to be generated predominantly by the body axis(Gillis, 1998). However,experimental work on several genera of fishes has shown that fins can alter the flow associated with the body as well as with other fins(Webb and Keyes, 1981; Jayne et al., 1996; Wolfgang et al., 1999; Nauen and Lauder, 2000; Hove et al., 2001). As we approach a more comprehensive understanding of fish locomotion, it is clear that kinematic analyses integrating both the body and fin movements are needed.

In this study, I examine the body and fin kinematics of steadily swimming needlefish and suggest some hydrodynamic consequences of having fins on an elongate body. The specific goals of this paper are (i) to describe the axial body kinematics of the Atlantic needlefish and highlight how they differ from other anguilliform swimmers, and (ii) to document the change in pectoral and median fin kinematics across speeds and discuss their possible functional roles.

Adult Atlantic needlefish, Strongylura marina Walbaum, were obtained from the New England Aquarium in Boston, MA, USA. Fish were housed in a 12001 circular polyvinyl chloride tank maintained at a temperature of 24±1°C and a salinity of 32-34‰. Fish were fed dried euphausiids, frozen silversides (Menidia spp.) and brine shrimp(Artemia sp.). Data were collected from four individuals (total body length, L=23.3±1.5 cm, mean ± S.E.M.).

Fish were acclimated to the flow tank for several hours before data were collected. Experiments were conducted in a 6001 aerated, recirculating flow tank (working section 28 cm×28 cm×80 cm) maintained at 24±1°C. Two electronically synchronized NAC HSV-500 video cameras filming at 250 frames s^-1 simultaneously recorded ventral and posterior views of swimming needlefish using two 45° front-surface mirrors placed below the flow tank and within the flow tank, respectively.

Up to five swimming speeds were chosen because they encompassed the widest range of speeds over which needlefish would swim steadily in the flow tank(0.25, 0.5, 1.0, 1.5 and 2.0 L s^-1). These speeds were selected because needlefish can swim steadily at each for at least 30 min without exhibiting burst-and-coast behavior. To ensure that swimming speed was equivalent to flow velocity, data were collected only for fish swimming steadily in the center of the flow tank at least 12 cm away from the side walls. Swimming speeds were tested in no particular order; however, needlefish could only swim steadily at high speeds if flow velocity was increased gradually. For most analyses, the four tail-beat cycles recorded for each fish at each swimming speed were consecutive.

For each tail-beat trial, at least 20 video frames were captured, separated in time by 12-20 ms, depending on the swimming speed of the fish. A customized software program was used to digitize 20 points on each side of the outline of the ventral silhouette of the fish (Fig. 1), for a total of 40 points for each image (note that the point placed on the tip of the jaw for left and right side overlap). A series of cubic spline functions were used to draw the best-fit line along these points(Jayne and Lauder, 1995; Gillis, 1997), and a midline was constructed and divided into 25 segments. The amplitudes relative to the midline for seven approximately equidistant points along the midline(Fig. 1) were calculated by dividing the total lateral excursion during one oscillatory cycle by two. The first location (Fig. 1, 0%L) coincided with the tip of the dentary, the second location (24%L) marked the body just posterior to the operculum, the third through to sixth locations (40-88% L) divided the body of the fish, and the seventh location (100% L) represented the tip of the tail. To characterize body amplitudes during gait transitions over the broadest range of swimming speeds, analysis of the seven body amplitudes focused on three speeds, including the lowest and highest speeds (0.25, 1.0 and 2.0 Ls^-1, N=16 wave cycles for each speed).

All five steady swimming speeds were analyzed to better examine the relationship between swimming speed and tail-beat frequency. Mean tail-beat frequencies for each speed were determined for each fish(Fig. 3, N=4). This was accomplished by tracking a digitized point on the tail from the ventral view over the course of one tail-beat cycle and dividing it by the elapsed time. The time required to complete one tail-beat cycle is the tail-beat period. Stride length is the distance traveled per tail beat, calculated by dividing the swimming speed by the tail-beat frequency.

Consistent with Gillis(1998), propulsive wavelength was measured directly from the reconstructed midlines as the distance between two successive peaks present on the body. Although wavelength can be calculated by dividing the mean wave speed by the mean tail-beat frequency(Webb et al., 1984), this method consistently underestimated the propulsive wavelength compared to direct measurements from the midline. Furthermore, this method assumes a constant wave speed along the body, which prohibits analysis on fast-swimming needlefish.

Mean propulsive wave speeds for three swimming speeds, (0.25, 1.0 and 2.0 L s^-1) were calculated for all fish. Wave speed was calculated by dividing the distance between the anteriormost point of the body exhibiting undulation and the tail tip by the time required for the crest of the wave to pass through these points. At high speeds, a larger portion of the body undulates, allowing the crest of the propulsive wave to be tracked from a more anterior position than would be possible at lower swimming speeds. Slip was calculated as U/V and Froude efficiency as 1-0.5[(V—U)/V](Lighthill, 1975), where U is the swimming speed and V is the propulsive wave speed.

To determine the possible effect of the median fins on the speed of the propulsive wave along the body, changes in wave speed were investigated for the two extreme swimming speeds. Differences in the propulsive wave speed were determined by halving the entire region of the body that was undulating and calculating the speed for each section. Thus, the anterior section of the propulsive wave (72-84% L) at 0.25 L s^-1 was different from the anterior section of the propulsive wave (24-60% L)at 2.0 L s^-1 (Fig. 1). Since undulation is restricted to the back of the body at 0.25 L s^-1, the anterior and posterior halves of the undulating region contain a similar portion of the median fins. In contrast, because a larger portion of the body undulates at 2.0 L s^-1, the anterior half of the propulsive wave is void of the median fins, which are now completely contained within the posterior half.

Needlefish fins are too delicate to be marked. However, images from the posterior view provided enough contrast to allow the apex of the dorsal and anal fins and the edge of the caudal and pectoral fins to be identified. To describe the phase relationships between the body and median fins, I analyzed four cycles for all fish swimming at 1.0 L s^-1, an intermediate speed at which needlefish swim steadily for the longest period of time. The longitudinal position of the dorsal and anal fin apices coincides with a point 80% down the body (Fig. 1), as measured on anesthetized individuals. Only a limited number of points that could be reliably identified using the outline of the fins were digitized. To perform a statistical analysis on median fin excursions,individual tail beats for all fish were aligned. Although this procedure resulted in depressed values for median fin amplitude compared to averaging maximum values (e.g. compare the tail-beat amplitude in Fig. 5 to that in Fig. 2), it accurately illustrates the average phasing of the fins relative to the body. Adduction and abduction speeds (mean of four trials for each of the four fish swimming at 0.25 L s^-1) for the pectoral fin were obtained by tracking the digitized tip of the fin. I measured the orientation and angle of insertion of the abducted pectoral fins using video images from the posterior view and lateral images of cleared and stained specimens, respectively. In addition, outline tracings of the fins at selected phases of adduction and abduction were produced by digitizing points around the edge of the fins using CorelDraw version 9.0 for the PC. The same technique was used to reconstruct the trailing edge of the caudal fin.

Means and standard errors (S.E.M.) were calculated for the amplitudes at each of the seven body locations, tail-beat frequencies and the propulsive wave speeds for the anterior and posterior regions of the body across swimming speeds. A two-sample t-test was used to determine if there were differences in the mean speed of the pectoral fin tips during abduction and adduction in the x and z directions. A paired-sample t-test was performed to detect significant differences in the anterior and posterior propulsive wave speeds for needlefish swimming at 0.25 and 2.0 L s^-1. Two-way, mixed-model analyses of variance(ANOVA) were performed separately for tail-beat frequency, period and stride. Significance levels were adjusted using the sequential Bonferroni technique(Rice, 1989).

To examine the effect of individual, swimming speed and longitudinal position on body wave amplitude, a three-way mixed-model (Model III) ANOVA was used. Swimming speed and longitudinal position were treated as the fixed effects and the individual was considered to be the random effect. A Bonferroni—Dunn post-hoc test was performed to determine whether there were significant differences among swimming speeds and longitudinal position (α=0.05). F-values for all ANOVAs were calculated according to Zar(1999). Statistical tests were performed by Statview (version 4.5) for the PC.

Needlefish generate thrust by passing waves down their body at all swimming speeds. The amplitude of these waves increases from anterior to posterior(Fig. 2), and the percentage of the body that undulates increases with swimming speed. At 0.25 Ls^-1, only the posterior 44% of the body undulates (including body point 4). At 1.0 L s^-1, all of the body posterior to the base of the cranium undulates (76% L). At 2.0 Ls^-1, the amplitude of undulation increases along the body and the elongate cranium exhibits yawing motions (0.3% L). Above 2.0 L s^-1, needlefish switch to burst-and-coast swimming. Propulsive wavelength increases from 49-73% L over the range of swimming speeds tested.

Over the range of swimming speeds investigated, tail-beat amplitude increases and does not plateau (Fig. 2). At 0.25 L s^-1, tail-beat amplitude is 2.6%L, at 1.0 L s^-1, tail-beat amplitude is 3.9%L, and at 2.0 L s^-1, tail-beat amplitude is 6.3%L. The relationship between tail-beat amplitude and swimming speed is best approximated by the equation a=0.49 U₁ +0.46 (r²=0.89, P<0.0001), where a is length-specific tail-beat amplitude and U₁ is length-specific swimming speed (L s^-1).

Tail-beat frequency also increases as a function of swimming speed, from 2.6±0.1 Hz at 0.25 L s^-1 to 5.1±0.3 Hz at 2.0 L s^-1 (mean ± S.E.M., N=16 cycles). This relationship is best described by the equation f=1.5 U₁+2.4 (r²=0.98, P<0.0001), where f is tail-beat frequency(Fig. 3A; Table 1). There is a significant difference in tail-beat frequency among individuals(P<0.001). Stride length increases with swimming speed(Fig. 3B). The effects of individual, swimming speed, and the interaction between individual and swimming speed on tail-beat period and stride are significant to at least P<0.01 (Table 1).

Absolute body wave speed (Fig. 4A) increases with swimming speed according to the equation V=30.9 U₁+18 (r²=0.88, P<0.0001, N=48), where the regression is performed on the raw values for the three swimming speeds. Means are 25.8±0.3 cm s^-1, 48.8±2.1 cm s^-1 and 77.7±5.4 cm s^-1, for fish swimming at 0.25 L s^-1, 1.0 L s^-1 and 2.0 L s^-1, respectively(mean ± S.E.M., N=16 wave cycles).

At 0.25 L s^-1, the propulsive wave speed does not change significantly along the portion of the body that undulates(α=0.05; Fig. 4B). However, at 2.0 L s^-1(Fig. 4C), the propulsive wave speed of the anterior body section (light gray column) is higher than that of the posterior section (dark gray column, P<0.001), indicating that at the highest swimming speed the propulsive wave decelerates as it travels towards the tail. Estimated mechanical efficiency increases with swimming speed, as reflected in the slip values, which increase from 0.23±0.01 at 0.25 L s^-1, to 0.48±0.01 at 1.0 Ls^-1 and to 0.62±0.02 at 2.0 L s^-1(N=16 wave cycles). Froude efficiency also increases with speed, from 0.61 at 0.25 L s^-1, to 0.74 at 1.0 Ls^-1 and to 0.80 at 2.0 L s^-1.

Results from a three-way ANOVA (Table 2) treating longitudinal position, swimming speed and individual as effects and body-wave amplitude as the dependent variable show that amplitudes vary significantly with longitudinal positions, swimming speeds and individuals (P<0.001). Bonferroni—Dunn post-hoctests indicate that the largest of the four fish (26 cm) has a significantly greater absolute bodywave amplitude than the other three fish(P<0.001). Post-hoc tests also reveal that the amplitudes corresponding to the yaw of the cranium did not change significantly with speed (P<0.001).

The dorsal and anal fins of needlefish taper abruptly from the anteriormost rays towards the tail, to form higher-aspect-ratio fins than otherwise found in anguilliform swimming fishes (Fig. 1B). Their combined surface area is greater than 116% of the caudal fin area, representing one of the most significant surfaces of the body interacting with water. Fully erect, the dorsal and anal fins provide a lateral profile that is three times the depth of the anterior region of the body. A phase relationship of slightly less than 180° between the anterior median fins and the body at the same longitudinal position (80% L) is maintained at a swimming speed of 1.0 L s^-1(Fig. 5). The lateral excursion of the dorsal and anal fins is approximately 1.7% L, similar to the lateral excursion of the body at 80% L. While the tips of the caudal fin show similar amplitudes to each other at 1.0 L s^-1,this is not true for the entire trailing edge, which adopts a complex dihedral conformation (Fig. 6).

At 0.25 L s^-1, the pectoral fins oscillate at a frequency of 3.41±0.46 Hz (mean ± S.E.M, N=16). Mean abduction speed (3.8±1.4 cm s^-1) along the xdirection (Fig. 7) relative to the body is less than the mean adduction speed (4.7±1.1 cm s^-1, P=0.01), where the x direction corresponds to the path of the downstream flow. During adduction, the fins are oriented along the transverse plane (y, z), where the y direction corresponds to the vertical axis with respect to the flow tank and the z direction indicates the direction across the flow tank. Because the adduction speed in the x direction relative to the body is less than the forward swimming speed (5.8 cm s^-1), the pectoral fins do not generate thrust. Along the z direction, abduction speed is 6.3±1.2 cm s^-1 and adduction speed is 8.4±1.4 cm s^-1 (mean ± S.E.M., N=16). As intermediate swimming speeds are approached, needlefish switch to holding their pectoral fins out as a positive dihedral (Fig. 8B),while at high speeds they fold them flush against the body surface(Fig. 8C). The insertion base of the pectoral fins in needlefish is positioned at approximately 35-40°relative to vertical.

The term anguilliform locomotion can be ambiguous. At times it refers generally to the undulatory swimming motion of elongate animals, and at other times it refers specifically to the proportion of the body that undulates or to the number of sine waves present on the body at any given instant. While the definitions of some authors include whole-body undulation(Webb, 1975; Lindsey, 1978; Blake, 1983), Breder's original definition of anguilliform locomotion(Breder, 1926) does not specify the proportion of the body that undulates. It is clear that these authors use these broad classifications to refer to swimming at `normal' or intermediate speeds. However, it is also clear that the proportion of the body that undulates changes with speed (Lighthill,1969; Gillis,1998). Breder's definition states that at least one half-wavelength and often more than one wavelength is present on the body at any time (Breder, 1926). According to this definition, needlefish are anguilliform swimmers, despite the fact that they do not undulate their entire body at low speeds, as has also been observed in eels (Anguilla rostrata)(Gillis, 1998). When needlefish undulate their entire body at higher speeds (1.0 Ls^-1 and above), they have approximately 1.5 waves along the length of their body, similar to anguilliform-swimming salamanders (Siren intermedia) (Gillis,1997). Unlike salamanders, which maintain a constant propulsive wavelength (64% L) across speeds, needlefish and eels increase their propulsive wavelength with speed (from 49 to 73% L for needlefish and from 44 to 54% L for eels).

Compared to eels (Gillis,1998), needlefish exhibit substantially different body kinematics. For example, even when swimming at lower relative speeds, needlefish have a longer propulsive wavelength. If a longer propulsive wave is passed down the body at a constant speed similar to that in eels, then overall, needlefish would pass fewer waves down their body and therefore have a lower tail-beat frequency than eels. In contrast, the data show that at similar swimming speeds needlefish have higher tail-beat frequencies. Several reasons may explain this finding. The length of the propulsive wave may decrease as it passes down the body. Higher tail-beat frequencies may also be reflected by lower slip values, resulting in the need to pass more wavelengths down the body to swim at the same speed. In this respect, needlefish are less efficient anguilliform swimmers than eels. Finally, the needlefish studied here were approximately half the length of the eels studied by Gillis(1998). Despite comparisons at similar length-specific swimming speeds, this size difference may contribute to higher tail-beat frequencies in needlefish, since intraspecific size-dependent variation has been demonstrated in other fishes(Webb et al., 1984). Although needlefish have higher tail-beat frequencies than eels at any given swimming speed, the slope describing the increase in tail-beat frequency as a function of swimming speed is 39% lower in needlefish(Gillis, 1997).

Previous studies on swimming fishes have shown that while tail-beat frequency continues to increase linearly with swimming speed, tail-beat amplitude increases linearly only over a certain range of swimming speeds,after which it plateaus (Bainbridge,1958; Webb, 1975; Blake, 1983). In needlefish,tail-beat amplitude does not plateau; it increases linearly over the range of swimming speeds investigated. This deviates from locomotor patterns of other fishes and may reflect the unique phylogenetic and ecological position of needlefish. Possessing an `acceleration specialist' morphology does not seem to be the cause of this behavior, since esocids show a non-linear relationship between tail-beat amplitude and swimming speed(Webb, 1988). The maximum length-specific steady swimming speed for needlefish is relatively low compared to that of carangiform and labriform swimmers(Collette, 1977; Helfman et al., 1997). Thus,reports of high speeds attained by these piscivorous predators in the field probably reflect short, unsustainable bursts of acceleration. Alternatively,needlefish may be capable of swimming steadily at much higher speeds in the field.

In needlefish, the significant increase in stride length(Table 1) at high speeds is a consequence of relatively high slip values. A high slip value indicates a larger contribution to rearward, thrust-producing forces than lateral forces. In salamanders, slip values, and therefore stride lengths, decrease with increasing swimming speeds (Gillis,1997), unlike eels which display a constant slip value and stride length across speeds (Gillis,1998; D'Aout and Aerts,1999). As typical of other swimmers, Froude efficiencies for needlefish increase with swimming speed. These values are higher than for tiger musky (Esox sp.), another fish with an `acceleration specialist' body type (Webb,1988). Although Froude efficiencies have not been reported for eels (Gillis, 1998; D'Aout and Aerts, 1999), their lower tail-beat frequencies would be likely to result in higher efficiencies compared to needlefish.

Gillis (1997) suggests that the maximum tail-beat amplitude of eels (8%L) is lower than that of sea snakes (Pelamis platurus) and salamanders (ranging from 11 to 19%L) due to the lateral compression of their body and to their continuous dorsal and anal fins (Graham and Lowell, 1987; Gillis,1998). The increase in lateral surface area provided by fins facilitates a greater transfer of momentum from the fish to the water,suggesting that the median fins in needlefish can accelerate a relatively larger volume of water than the median fins of eels because of their position and size. It is not surprising, then, that needlefish exhibit lower tail-beat amplitudes and body wave amplitudes than eels at comparable swimming speeds(Gillis, 1998; D'Aout and Aerts, 1999). Similarly, Webb (1988) found that the posterior location of the dorsal and anal fins in musky correlated with relatively smaller tail-beat amplitudes and higher tail-beat frequencies when compared to rainbow trout (Oncorhynchus mykiss).

Steadily swimming fishes typically possess a propulsive wave that either maintains a constant speed or accelerates as it passes down the body,depending in part on the change in wave amplitude. Gillis(1997) found that eels and slow-swimming salamanders maintain a constant propulsive wave speed along the body, while sea snakes and fast-swimming salamanders possess an accelerating propulsive wave. The decelerating propulsive wave seen in fast-swimming needlefish is a novel finding for an anguilliform swimmer and is probably correlated with the size and location of the anterior median fins (Figs 1, 4C). Wavelength equals propulsive wave speed divided by tail-beat frequency, and since tail-beat frequency is fixed, by definition a decelerating propulsive wave must shorten in length. Gillis's (1998)observation that propulsive wave speed remains constant in eels is probably due to the uniform lateral profile created by their continuous median fins. At high swimming speeds the tapering posterior region of sirenid salamanders,which do not have fins to extend their lateral profile, causes the propulsive wave to accelerate. Not only is the speed of the posterior propulsive wave higher than in the front of the body, it is also higher than the posterior region of similarly sized eels as well (Gillis, 1997, 1998). Interestingly, although sea snakes increase the depth of their posterior body region with a fin-like keel and a flat, paddle-shaped tail, they also exhibit an accelerating propulsive wave (Graham and Lowell,1987).

Discrete median fins can improve hydrodynamic efficiency and decrease drag in comparison to continuous median fins, especially when the gaps between fins are large (Lighthill, 1969; Webb, 1975). Theoretical and experimental work has shown that the wake shed by the dorsal and anal fins can be constructively utilized by the caudal fin(Weihs, 1973; Drucker and Lauder, 2001). Webb and Weihs have suggested a `double-tail hypothesis', in which the propulsive body wave causes the dorsal and anal fins (the first `tail') to be out of phase with the caudal fin, producing a relatively uniform thrust that is especially efficient for rapid acceleration(Weihs, 1973; Webb and Weihs, 1983). At 1.0 L s^-1, the caudal fin of the needlefish is indeed shifted out of phase with the anterior median fins, although not by 180°. Additional analyses show that dorsal and anal fin amplitudes remain relatively constant across swimming speeds, while body amplitudes increase. Initially, at the lowest speed the anterior median fins have a higher amplitude than the body, but at the highest swimming speed they both show similar magnitudes. In addition, as swimming speed increases, the tail-beat amplitude increases relative to the amplitude of the dorsal and anal fins. The apices of the dorsal and anal fins maintain a constant-phase relationship relative to the body across swimming speeds, suggesting that needlefish can modulate fin stiffness and height. Observations of needlefish suggest that at low speeds the dorsal and anal fins move actively, yet currently there are no data to evaluate median fin wakes and their interactions with the caudal fin.

The caudal fin of a needlefish is a flexible structure that changes shape with swimming speed. At slow speeds, the dorsal lobe of the caudal fin leads the stiffer ventral lobe during each lateral excursion of a tail beat(Fig. 6), a kinematic pattern that has been noted for homocercal tails in other teleosts(Lauder, 2000). At higher swimming speeds, the procurrent and middle rays of the caudal fin lead the tail as it sweeps to one side. This causes the more flexible rays of the dorsal and ventral lobe to lag behind, creating a W shape (Figs 6, 8). This tail shape may serve to direct or accelerate flow into the wake behind the fish and is believed to facilitate thrust generation in scombrids such as Pacific bonito (Sarda chiliensis) and wavyback skipjack (Euthynnus affinis)(Fierstine and Walters,1968).

At the lowest swimming speed, needlefish oscillate their pectoral fins 31%faster than their tail-beat frequency. Outlines of simultaneous ventral and posterior views of a fish swimming at 0.25 L s^-1(Fig. 7) indicate that the fin is feathered (tilted parallel to the frontal plane) during abduction and held to expose a large surface area along the transverse (y,z) plane during adduction. The speed of adduction is less variable and occurs faster than the speed of abduction. These lines of evidence typically suggest propulsive, drag-based locomotion (Gibb et al., 1994; Drucker and Jensen,1997; Walker and Westneat,1997). However, since the fin is brought back against the body at a slower speed than that which the water is moving past the body, no thrust can be produced (Webb and Weihs,1983). Why do needlefish oscillate their pectorals at high frequency, using a motion similar to drag-based propulsion, when they are not generating thrust with them? One possibility is that by continually oscillating their pectoral fins they may be able to react more quickly to generate forces over a range of directions. Using their pectoral fins may offer an additional degree of control over their forward swimming speed than if they were to swim with caudal fin propulsion alone. Observations in the field and in the laboratory support this idea; needlefish use their pectoral fins to decelerate body motions and facilitate directional changes of their prow-like cranium (Breder,1926). Selective drag production anterior to the center of mass,rather than lift generation, may play an important role in stability control and adjustment of their force balance(Drucker and Lauder, 1999; Wilga and Lauder, 1999; Liao and Lauder, 2000).

At higher swimming speeds (up to 1.5 L s^-1), needlefish stop oscillating their pectoral fins and instead hold them abducted from the body (Figs 7C, 8B). Maintaining abduction of a pectoral fin with a vertically oriented base creates an acute angle described by the medial surface of the fin and the body, creating a positive dihedral that is often utilized in man-made aircraft to provide stability(Smith, 1992). Rolling about the long axis of the body was observed at low to intermediate swimming speeds but was not measured. By orienting their pectoral fins as a dihedral rather than parallel to the frontal (x,z) plane, needlefish may decrease roll and side-slip, a condition that is made worse by head yaw(Fig. 2)(Smith, 1992). There is evidence that other anguilliform swimmers also use their pectoral fins during steady locomotion. In Gillis(1998), fig. 1 shows the pectoral fins of Anguilla rostrata being held abducted while swimming steadily at 0.4 L s^-1. Unfortunately, whether the fin is oriented in the horizontal plane or is held as a positive dihedral cannot be determined from the image.

Needlefish are the first fish in which a steady, positive dihedral conformation of the pectoral fins has been described (for a description of a negative dihedral, see Wilga and Lauder,2000). Above 1.5 L s^-1, the pectoral fins of needlefish are completely folded against the body. These speed-dependent behaviors define two gait transitions, and distinguish them from most fishes,which only show one pectoral fin gait transition over their entire range of swimming speeds (Gibb et al.,1994; Jayne and Lauder,1996; Drucker,1996; Walker and Westneat,1997).

My warm thanks to the staff at the New England Aquarium and especially S. Floyd for donating the needlefish and making this project possible. I would like to acknowledge L. Farrell, E. Freund, J. Lee, J. Nauen and J. Bertrand for graciously assisting in the data collection. This project has benefited greatly from the discussions and critical comments of G. Gillis, G. V. Lauder,M. Daley, J. Nauen, U. Muller, B. Collette, E. Freund, E. Drucker, E. Tytell and T. Hsieh. In addition, comments from three anonymous reviewers improved the quality of the manuscript substantially. Special thanks to A. Cullum and T. Hedrick for writing the digitizing software. Support was provided by a SICB GIAR and MCZ Putnam Grant to J.C.L. and a NSF grant (IBN-9807012) to G.V.L.