Tail Muscle Activity Patterns in Walking and Flying Pigeons (Columba Livia)

The demand for coordination of all flight surfaces has yielded a specialized integration of the avian neuromuscular system. Although the majority of avian flight studies have focused on the wings, the tail also appears to be crucial to the evolutionary success of birds as flying organisms. Despite recent advances in our understanding of wing muscle activity and function (e.g. Dial et al. 1987, 1988, 1991; Jenkins et al. 1988; Dial, 1992a,b), the motor control of the avian tail during locomotion has remained unexplored. Hypotheses of caudal muscle function are based on anatomical information (Fisher, 1946; Owre, 1967; Raikow, 1970, 1985; Baumel, 1988), qualitative descriptions of tail movement during flight (Brown, 1948; Pennycuick, 1968) or data from anesthetized (Baumel et al. 1990), restrained (Biederman-Thorson and Thorson, 1973; Bilo and Bilo, 1983) or surgically impaired (Fisher, 1957) birds. However, information is lacking on basic activity patterns of muscles controlling tail movement. This paper reports on the electrical activity of six major tail muscles of the pigeon during locomotion. We used electromyography (EMG) in conjunction with high-speed cinematography to record activity patterns of pigeon muscles during walking and four modes of symmetrical (i.e. non-turning) forward flight. These patterns are referenced to the activity in representative forelimb, hindlimb and trunk muscles.

Our results on freely locomoting pigeons are the first quantitative data available for the neural control of the tail in a vertebrate during flight. In view of recent studies of pigeon tail morphology (Baumel, 1988) and caudal muscle activity during ventilatory movements (Baumel et al. 1990), these EMG data provide a crucial test of predictions of tail muscle locomotor function. A basic description of this type is necessary for a more informed interpretation of results from reduced preparations (reflex studies, spinal transections, fictive locomotion, etc.) and data from more complex locomotor behaviors such as maneuvering flight. Finally, an understanding of caudal function is an essential step towards integrating both the wing and the tail into a more complete model of avian flight.

Ten adult wild type pigeons (Columba livia Gmelin) with a mean body mass of 334 g (±49g, S.D.) were used in thirteen electromyographic (EMG) experiments. Birds were collected from local populations, housed in stainless-steel cages (1 m×1m ×1.5m) and provided with commercial pigeon feed, vitamins and water ad libitum.

Birds were trained to fly onto an elevated perch-platform when either released from the hand or placed on the ground for unassisted takeoffs (Dial, 1992a,b). The flight corridor was 3.1m wide, 2.7m high and 50m long. In this study we present data from five modes of locomotion: (1) walking, (2) liftoff, (3) takeoff, (4) slow level flapping flight and (5) landing (Fig. 1). Walking birds moved freely upon the floor of the flight corridor. Liftoff is a transitional mode between terrestrial and aerial locomotion. Here, liftoff is defined as the first wingbeat and associated leg thrust of an ascent from the ground. Takeoff is designated as the first five wingbeats subsequent to liftoff, during which birds were gaining altitude and forward velocity. Slow level flapping flight (Brown, 1948, 1963) is achieved after birds have attained maximum speed and fly level (for a distance of 15–20m) down the corridor. Landing is defined as the last 5–7 wingbeats prior to touchdown on the perch-platform.

The morphology of the tail and pelvis was studied in fresh, frozen and preserved specimens (N=6) using a Wild M7-S binocular dissecting microscope. The epaxial musculature of the tail has been divided into two separate muscles by past workers (levator coccygis and levator caudae). In this paper we follow the newly revised terminology of Baumel et al. (1993), in which the caudal epaxials are considered as two parts (pars vertebralis and pars rectricalis) of a single muscle, the levator caudae. We recorded EMGs from the two parts of the levator caudae separately and treat them independently in our results.

Electromyographic data were gathered using surgically implanted bipolar electrodes similar to those previously described (Dial et al. 1988; Dial, 1992a,b). Birds were administered intramuscular doses of ketamine (25mgkg⁻¹) and xylazine (2mgkg⁻¹) to induce deep anesthesia during all surgical procedures. Injections were distributed around the body to avoid irritating muscles under study and supplementary amounts were given as needed. Feathers were plucked from four regions: (1) the middle of the back to accommodate a connector plug, (2) the left thoracic region to expose the pectoralis muscle, (3) the left side of the synsacrum to provide access to the hindlimb and trunk muscles, and (4) the base of the tail. The rectrices and most tail coverts were left intact. Electrodes ran subcutaneously from the female back plug (12-pole, consisting of two Microtech FG-6 miniature connectors), which was sutured to the thoracic interspinous ligaments. Target muscles were exposed by skin incisions and blunt dissection and then implanted with electrodes (twisted pairs of 100 μm diameter silver wire with 0.5mm insulation removed at each recording tip and offset 1mm) using a 23-gauge hypodermic needle. Electrodes were sutured to adjacent periosteum or fascia approximately 1–5mm from their exit from the muscle. A short loop of wire was left between the exit point and suture to allow the electrode tips to move freely with the contracting muscle, significantly reducing low-frequency movement artifact from EMG signals.

EMG recordings were made on the first postoperative day. A 25m, lightweight, 12-lead shielded cable (Cooner Wire Co., Part no. NMUF 4/30-4046SJ) with a male miniature connector transmitted the signals to the recording equipment. The cable was taped to the bird’s back and flank so that it passed over the right thigh beneath the tail, thus minimizing interference with locomotor movements. EMG signals were amplified and filtered (Grass P511J; 2000–20000×; 100–3000Hz bandpass) before being simultaneously recorded on a four-channel Hewlett-Packard 3694A tape recorder at 15inches s⁻¹ and to an eight-channel, 12-bit A/D converter (Keithley DAS 5000 series; 2040Hz sampling rate) for storage on a Zenith 386SX microcomputer. Data were printed on a Gould 2400 chart recorder and a Graphtec WR4000 thermal array recorder.

Eleven muscles were chosen for electromyographic recording. The pectoralis, the primary flight muscle, was implanted to monitor wing activity. The iliotrochantericus caudalis, a hip muscle, served as an indicator of hindlimb use. The pectoralis, iliotrochantericus caudalis and four additional muscles were sampled in most experiments. The caudofemoralis is a hindlimb muscle (based on its embryology; Romer, 1927), but originates from the tail and may be involved in caudal control. The longissimus dorsi provided information on epaxial muscle activity in the trunk. Two epaxials (levator caudae pars vertebralis and pars rectricalis) and five hypaxials (lateralis caudae, bulbi rectricium, depressor caudae, pubocaudalis externus and pubocaudalis internus) were recorded in the tail. Activity was recorded in the ventral fibers of the bulbi rectricium. The depressor caudae was implanted in its distal part, as exposed laterally, ventral to the caudal transverse processes; activity in the proximal and deep parts was not recorded. The pubocaudalis internus is a digastric muscle of which we sampled the pelvic head only. Electrode positions were verified by surgical removal or post-mortem dissection (in at least one bird for each muscle) following recording.

Simultaneous high-speed films were made of approximately half of the recorded EMG sequences. The camera (16mm Lo-Cam, Red Lakes Laboratories) ran at 150–200frames s⁻¹ for flight and at 50frames s⁻¹ for walking. An electrical pulse sychronized with each frame of film (Kodak 7250 Ektachrome) was recorded on tape and computer to permit correlation of kinematic and EMG data. Films were viewed using an L-W Motion Analyzer projector (model 224-S) to verify wing and leg position and to characterize tail movements qualitatively. Tracings were made from projected images of one bird during liftoff, takeoff, slow level flight and landing.

For analysis of the motor pattern of the recorded muscles, EMG signals were displayed from the stored digital data on a Tektronix 4109 graphics terminal. Signals were digitally rectified prior to measuring timing of activity using custom digitizing programs supplied by G. V. Lauder. Each muscle was analyzed to determine onset and offset times relative to a reference muscle (pectoralis during flight and iliotrochantericus caudalis during walking). EMG bursts were identified as changes in amplitude at least two times greater than baseline. Most EMGs were discrete bursts with amplitudes well above background activity (see Figs 7–9). Timing was measured at a resolution of 0.5ms. These data were then normalized to each wingbeat or stride in order to compare activity within a locomotor cycle between different birds. Mean cycle duration and frequencies are given in Table 1. Five cycles from each of two birds were averaged for each muscle within each mode of locomotion, with the exception of liftoff. Liftoff sample sizes were: pectoralis (25 liftoffs, 7 birds), iliotrochantericus caudalis (17, 4), caudofemoralis (6, 2), longissimus dorsi (7, 2), levator caudae pars vertebralis (12, 2), levator caudae pars rectricalis (7, 2), lateralis caudae (10, 3), bulbi rectricium (1, 1), depressor caudae (6, 2), pubocaudalis externus (7, 3) and pubocaudalis internus (9, 3).

Only a general description of caudal myology is presented here (Fig. 2; see Baumel, 1988, for a thorough study of pigeon tail anatomy). The tail skeleton is composed of five or six free caudal vertebrae and a terminal pygostyle. Six pairs of main tail feathers (rectrices) are rooted in a fibroadipose structure called the rectricial bulb (Baumel, 1988). Major movements of the tail feathers are controlled by muscles connecting the synsacrum, pelvis, femora and caudal vertebrae to the tail skeleton and rectricial bulbs.

One noteworthy characteristic of the avian tail is its relative isolation from surrounding trunk and hindlimb structures. Dorsally, the epaxial musculature of the trunk (longissimus dorsi) is not continuous with the adjacent epaxial musculature of the tail (levator caudae pars vertebralis). Between these muscles, dorsal rami of spinal nerves perforate the synsacrum but do not supply motor innervation. In the pigeon this hiatus spans at least four spinal levels. Laterally, the caudal musculature has almost no overlap with the muscles of the hind legs; both take origin from the fused pelvic bones. The one exception is the caudofemoralis muscle, which runs from the base of the pygostyle to the femur. Posteriorly, the eminence of the vent is connected to the tail by the septum supracloacale and the slender levator cloacae muscles. The influence of cloacal muscles on the tail during flight remains unknown.

Representative movements of the tail during liftoff and the three flight modes are shown in Figs 3 and 4. In the transition from terrestrial to aerial locomotion the tail undergoes dramatic changes in orientation (Fig. 3). When the bird is standing, the tail is typically furled (rectrices adducted) and slightly tented (concave down; Fisher, 1946; Baumel, 1988). As the wings are raised, the tail elevates, flares and arches (concave up). The tail is depressed during the first downstroke and remains highly flared. This general position is maintained into takeoff (Fig. 4A). In this mode, the tail is slightly adducted as the wings pass by the rectrices during the downstroke and then abducted again during the upstroke. In comparison to takeoff, both the body and tail are oriented closer to the horizontal in slow level flight (Fig. 4B). As birds approach the perch to land, the body axis once again become more vertical; the tail is highly flared during landing (Fig. 4C).

During walking (Figs 5A, 6; Table 2) the pectoralis and the majority of caudal muscles are inactive. The iliotrochantericus caudalis is active from late in the swing phase through most of the propulsive phase of the stride cycle. This activity is similar to that known for other avian species (Gatesy, 1989; Jacobson and Hollyday, 1982). During walking, the caudofemoralis (Fig. 6) shows either no activity (below approximately 2strides s⁻¹), low-amplitude activity in the mid–late propulsive phase (from about 2–3stridess⁻¹) or high-amplitude activity (above 3strides s⁻¹). The longissimus dorsi is activated continuously, but within the sustained activity its EMG exhibits a biphasic modulation of amplitude with each stride. It is clear from bilateral implants that this modulation of trunk epaxial EMG amplitude is relatively symmetrical. Within the tail, only the levator caudae pars vertebralis and pars rectricalis show a consistent EMG during terrestrial locomotion. These muscles fire one burst per stride, with pairs alternating their activity in phase with the striding legs. Caudal hypaxials are typically inactive during walking.

Birds move from terrestrial locomotion to flight during liftoff by initiating the first wingbeat and pushing off the ground with the legs (Figs 3, 5B, 6; Table 2). A preparatory period of leg and tail activity precedes the onset of activity in the pectoralis, but all muscles recorded exhibit pronounced activity during this explosive locomotor behavior. Activity normally appears first in the hypaxials; this continues through the first downstroke. Multiple bursts of EMG activity in the pubocaudalis externus and internus cannot be compared easily between trials and birds. The onset and offset of this collective activity are given in Table 2 and Fig. 5B. The depressor caudae has three EMG bursts per liftoff. The leg and epaxial muscles typically become active as the wing begins to elevate. The iliotrochantericus caudalis and caudofemoralis fire relatively simultaneously during pushoff and cease activity prior to the initiation of the first downstroke. The longissimus dorsi and the caudal epaxials continue their activity through the upstroke/downstroke transition. All but the levator caudae pars rectricalis have a second burst late in the downstroke. The tail is elevated and flared as the wings are raised and then depressed as the wings descend. The bulbi rectricium was activated just prior to the pectoralis in the one liftoff recorded for this muscle. The feet usually leave the ground during the second half of the downstroke, but the exact timing of last contact is somewhat variable.

During takeoff (Figs 4A, 5C; Table 2) and other flight modes the pectoralis is active to decelerate the elevating wing and provide the major power for the downstroke (Dial et al. 1988). The iliotrochantericus caudalis is usually inactive, but may show a brief, bilateral burst of EMG activity during the downstroke. Caudofemoralis EMG activity is also seen during the downstroke, but is of much lower intensity than in fast walking or liftoff (Fig. 6). Variation in leg muscle activity appears to depend on the orientation and movement of the legs during flight, which often differs among individuals. Most other muscles are active twice in each wingbeat. The longissimus dorsi is active at the upstroke/downstroke transition and again in late downstroke (Fig. 7). The levator caudae pars vertebralis (Fig. 7) and lateralis caudae show bursts of activity in late upstroke and late downstroke. The levator caudae pars rectricalis differs from these in that it has only a single (late upstroke) burst of activity (Fig. 8). Activity is continuous in the bulbi rectricium. The depressor caudae, pubocaudalis externus (Fig. 9) and pubocaudalis internus are active in early downstroke and during the downstroke/upstroke transition.

Once birds achieve slow level flapping flight, either from unassisted takeoff or after hand release, most muscles show only a single EMG burst per wingbeat (Figs 4B, 5D; Table 2). The pectoralis continues to power the wing, but compared to takeoff there is a pronounced reduction in its EMG amplitude (Dial et al. 1988). The iliotrochantericus caudalis is typically inactive. Caudofemoralis EMGs of low amplitude occur during the downstroke. Longissimus dorsi activity is reduced from its biphasic takeoff pattern to a single burst per wingbeat near the upstroke/downstroke transition (Fig. 7). Similarly, the levator caudae pars vertebralis and lateralis caudae show single bursts of activity, primarily during the upstroke. The levator caudae pars rectricalis, which has a single burst of activity during takeoff, becomes completely inactive during level flight (Fig. 8). The bulbi rectricium continues to be active throughout the wingbeat cycle. The dual bursts seen in the depressor caudae and pubocaudalis internus during takeoff merge into a single downstroke burst during each cycle of level flapping flight. Activity in the pubocaudalis externus is continuous rather than in discrete bursts, with some modulation of EMG amplitude (Fig. 9).

The pattern of muscle activity during landing (Figs 4C, 5E; Table 2) is similar but not identical to that during slow level flapping flight. Pectoralis EMGs increase in amplitude as birds slow down to land, while the caudofemoralis typically retains its level-flight activity. Within the tail, the levator caudae pars rectricalis, which is silent in the level flight mode, is active near the upstroke/downstroke transition during landing (Fig. 8). EMG activity in the bulbi rectricium remains continuous, but fluctuates in amplitude, being highest during the downstroke. Activity in the pubocaudalis externus (Fig. 9) is reduced from its continuous level flight pattern to a downstroke burst similar to that of the depressor caudae and pubocaudalis internus.

The limb muscles we sampled exhibit high-intensity co-activity only during liftoff, when the bird makes the transition between terrestrial and aerial locomotion. The pectoralis is silent during walking but active during flight. When walking, the iliotrochantericus caudalis and caudofemoralis burst with each stride. During flight, these hindlimb muscles have activity that, when present, is of lower amplitude than during terrestrial modes (walking and liftoff).

Within the epaxial musculature, activity often differs among the longissimus dorsi, levator caudae pars vertebralis and levator caudae pars rectricalis. All epaxials are active during walking. In this mode the longissimus dorsi exhibits continuous activity with bilaterally symmetrical fluctuations in EMG amplitude twice per stride. Both the levator caudae pars vertebralis and pars rectricalis have monophasic EMGs. The right and left levator caudae alternate in activity in phase with the ipsilateral striding hindlimbs. During liftoff and takeoff, the longissimus dorsi and levator caudae pars vertebralis fire twice per wingbeat, while the levator caudae pars rectricalis fires only once. In the transition to slow level flight, each epaxial loses one burst of EMG. Those that have a biphasic EMG during takeoff become monophasic (Fig. 7); the levator caudae pars rectricalis turns off completely (Fig. 8). Activity in the levator caudae pars vertebralis begins prior to that in the longissimus dorsi and pars rectricalis in most modes.

Many motor patterns are present among hypaxial muscles. The lateralis caudae, bulbi rectricium, depressor caudae, pubocaudalis externus and pubocaudalis internus are normally silent during walking. Once activated during liftoff, the bulbi rectricium remains active continuously during takeoff, level flight and landing. Other hypaxials exhibit at least two EMG bursts during liftoff and takeoff. As birds shift to slow level flight these dual takeoff bursts merge. The pubocaudalis externus is unusual in that it is continuously active (Fig. 9), whereas the lateralis caudae, depressor caudae and pubocaudalis internus burst phasically.

Baumel (1988) concludes his monograph on the pigeon tail by stating the need for more investigations of tail use and tail muscle activity during flight. The present study, combining electromyography and high-speed cinematography, provides initial experimental data on the neural control of caudal muscles during locomotion. Our data reveal that the pigeon’s major tail muscles exhibit phasic patterns of cyclical activity sychronized with either wing or leg movement. However, not all muscles are active in all modes, particularly walking. The most dramatic example of inactivity may be the levator caudae pars rectricalis, which is active during all locomotor modes studied except slow level flight. In the slow level flight mode, therefore, the levator caudae pars rectricalis does not normally contribute to tail control. The EMG activity pattern of this and other muscles is critical to understanding their function, but is not readily predictable from anatomical or kinematic information. The temporal patterns of EMG activity presented in this paper do not delineate muscle function conclusively, but set limits on the interpretation of each muscle’s contribution to the control of tail movement.

Our measurements of tail muscle activity differ significantly from those found for wing muscles of pigeons performing equivalent flight behaviors under identical conditions (Dial, 1992a). Fifteen of sixteen muscles of the wing showed only slight changes in timing of EMG activity between takeoff, slow level flapping flight, landing, ascending and descending (Dial, 1992a). All wing muscles studied were active in every mode and retained the same pattern of mono-or biphasic activity during takeoff, slow level flapping flight and landing. In contrast, the majority of tail muscles exhibit conspicuous changes in the timing of their activity during these same flight modes. The motor pattern of many muscles varies from no activity to a single burst, a double burst or a continuous burst of EMG activity for each cycle of limb movement in different locomotor modes.

We suggest that the birds have evolved different neuromuscular control patterns for their forelimb and tail flight surfaces (Fig. 10). The difference in the complexity of timing in the tail (modulated) and wing (stereotypic) may stem from the need to standardize control signals in the rapidly oscillating forelimb. Perhaps only a single temporal motor pattern evolved in the flapping wing as the degrees of freedom in the elbow and wrist were reduced during the origin of avian flight. This basic pattern, in conjunction with an automatic linkage system in the wing (Headley, 1895; Fisher, 1957; Dial, 1992b), forms the basis for steady-state flapping flight. Alterations of wing shape during non-steady flight modes appear to be controlled by varying the intensity of forelimb muscle activation (Dial, 1992a). In addition, complex changes in wing movements may be controlled proximally by the differential recruitment of subvolumes of the pectoralis (Dial et al. 1988; Dial, 1992b; D. F. Boggs and K. P. Dial, in preparation).

Unlike the wing, the tail does not undergo dramatic oscillations during flight and so may be free of a constraint to standardize temporal activation of its musculature. Although we did not quantify EMG intensity in this study, tail muscles clearly exhibit changes in both timing and intensity (S. M. Gatesy and K. P. Dial, unpublished data). This flexibility in caudal motor pattern between different modes of forward, non-turning flight may reflect the versatile role of the tail in locomotion. Certain types of flapping flight appear to require more contribution from the tail than others, but the wings are always in use and are therefore less likely to vary. One point that should be stressed is that tails of pigeons flying in our corridor were always at least slightly flared and depressed. This has been referred to as ‘slow flight’ by Brown (1948, 1963) and Simpson (1983). In fast flying pigeons in the wild, the tail is primarily adducted (furled) and held parallel to the body axis (Pennycuick, 1975; personal observation). We predict that during fast, non-turning level flight the tail is held by aerodynamic rather than muscular forces and that activity of caudal muscles is greatly reduced. This would be consistent with the activity of the levator caudae, which is inactive in slow level flight; other muscles may similarly be turned off as birds achieve higher speeds.

Most birds are considered to have two relatively independent locomotor systems: the wings for flight and the legs for walking, running and hopping (Pennycuick, 1986; Butler, 1991). This distinction between aerial and terrestrial locomotor systems is clearly illustrated in our simultaneous recordings from the pectoralis (wing) and iliotrochantericus caudalis (leg) during different locomotor modes (Fig. 6). The wing muscle is inactive when the bird is walking with alternating leg movements. Once airborne, the leg muscle is typically inactive as the bird flies with synchronous wing-beats. Much of the leg activity appears to be related to restraining the folded hindlimbs against the bird’s body during takeoff, when they may be experiencing more violent forces than those in level flight. Clearly, there must be some leg muscle activity (possibly in tonic fibers) to keep the legs flexed during flight, but we predict that this will be minor compared to that when the bird is moving on the ground. We cannot rule out the possibility of subthreshold excitation of motoneurons of muscles that we identify as inactive; this will have to be shown neurophysiologically.

The activity of caudal muscles supports the hypothesis that in the pigeon the tail has its primary allegiance to the wings rather than to the legs in this segregated locomotor system (Gatesy and Dial, 1991). Coordination of the tail and forelimbs during flapping and gliding (e.g. Tucker, 1992) is not surprising, but there is no evidence to suggest that this type of integration was present in the bipedal theropods that gave rise to birds. The tail’s primitive linkage is to the hindlimb and trunk, as found in living saurians such as lizards and crocodilians (Gatesy, 1990). What changes accompanied the tail’s unprecedented affiliation with the forelimbs?

There is anatomical and functional evidence that the tail has been decoupled during the evolution of birds from their theropod ancestors. Muscular connections have been eliminated or weakened so that the tail is almost completely independent from the trunk and hindlimbs (Fig. 2). This decoupling appears to have permitted the tail to specialize for the control of the tail feathers during flight and form a novel association with the wings. Verification of this decoupling can been found in each major muscle group of the tail. First, the epaxial musculature is no longer continuous along a bird’s dorsal surface. There is a distinct gap between the longissimus dorsi in the trunk and the levator caudae pars vertebralis in the tail. This anatomical segregation coincides with differences in motor pattern during locomotion; the caudal epaxials are not acting as simple segmental extensions of the trunk musculature. Second, the caudal hypaxials of walking pigeons are inactive so that their involvement in terrestrial locomotion, if any, must be passive. This differs from Alligator, in which there is both epaxial and hypaxial activity in the tail during walking (S. M. Gatesy, unpublished data). Third, pigeons walk at low speeds without using the caudofemoralis. The avian caudofemoralis is either reduced (relative to saurians) or completely lost (George and Berger, 1966) and is no longer a crucial element in powering the limb. In contrast, this muscle forms the basis of hindlimb retraction in crocodilians, lizards and, presumably, the dinosaurian ancestors of birds (Gatesy, 1990). In the pigeon, a small remnant of the tail’s primitive role can be seen in the alternating activity of the levator musculature during walking.

Past workers have often divided the caudal epaxial musculature into two muscles (Shufeldt, 1890; Fisher, 1946; George and Berger, 1966), but recently Baumel (1988) reunited them as two parts (pars proximalis and pars distalis) of a single muscle called the levator caudae. This nomenclature will be changed once again (pars vertebralis and pars rectricalis) to help make these names more informative (Baumel et al. 1993). Although the fascicles of these two parts are oriented in parallel, their insertions and apparent actions differ. Our EMG data confirm that these muscular subdivisions are not used in the same way during locomotion. For example, the levator caudae pars vertebralis fires two bursts per wingbeat during takeoff, whereas the pars rectricalis fires only one. In slow level flapping flight, the levator caudae pars vertebralis fires one burst per wingbeat, but the pars rectricalis is completely silent. It is hoped that the new terminology will reflect the heterogeneity within the levator caudae and not imply uniformity.

The EMG results presented here are a prerequisite for an understanding of tail function during locomotion. For example, EMG data help confirm the bulbi rectricium as an abductor of the rectrices, as predicted by Baumel (1988). The continuous activation of these paired muscles holds the tips of the calami of the rectrices together, thus maintaining the tail fan seen during flight (Figs 3, 4). The lateralis caudae musculature may assist the bulbi rectricium initially, but its phasic activation does not seem adequate to sustain the abducted position.

The relationship between the activity of caudal muscles and tail movement, however, is not always clear. Although we have not presented a quantitative kinematic analysis, our overall conclusion is that the tail is relatively stable during most wingbeats. Fluctuations in elevation/depression and abduction/adduction do occur, but these do not always have a one-to-one correspondence with the activity of caudal muscles. In takeoff, for example, the tail does not undergo observable pulsatile movements for each burst of biphasic activity. It is not known, however, to what extent caudal muscles continue to produce force after the cessation of EMG activity; this lag could act to smooth the kinematic output generated by punctuated activation.

We hypothesize that much of the EMG activity of caudal muscles is related to fixing tail position for appropriate aerodynamic presentation as the oscillating wings impose dramatic forces on the body. The coactivation of epaxials and hypaxials near the downstroke/upstroke transition during takeoff may stabilize the tail during this relatively violent movement. Unfortunately, the interaction of aerodynamic, inertial and muscular forces during takeoff and other flight modes is not understood. This makes the interpretation of caudal muscle function particularly difficult, since tail kinematics are likely to be the net result of all of these forces, not just of muscular activation.

Although we attribute the majority of tail muscle activity to stabilization, we cannot distinguish between modulated output from a central pattern generator and reflex reactions to body and tail movement. Both of these sources may be involved in the motor output we record from the tail musculature. Further research is needed to identify the nature of caudal muscle activity and its integration with the forelimb spinal locomotor networks.

We suggest that a better perception of caudal muscle function will come through studies of more complex flight behaviors. In turning flight, for example, the tail moves asymmetrically as the bird maneuvers (Fig. 10). This kinematic asymmetry is presumably initiated, maintained and terminated by differential activation of contralateral pairs of caudal and wing muscles. It is likely that the muscle activity during slow level turning flight will be similar to that during slow level non-turning flight, but with variations in timing and/or intensity creating asymmetries during the maneuver. Using non-turning EMGs as baseline signals may reveal the differences in activity that make the tail deviate during turning. Correlating these differences with tail movement may then lead to a clearer understanding of tail muscle function, the relative contributions of the wing and tail in turning, and the evolution of vertebrate flight.

We thank R. Trenary, N. Olsen, C. Bocker, D. Conway and L. Becker for their assistance during this project. B. Tobalske, J. J. Baumel and two anonymous reviewers made many helpful comments on earlier versions of this manuscript. This work was supported by National Science Foundation Grant BNS-89-08243 to K.P.D.