Compensation of Visually Simulated Wind Drift in the Swarming Flight of the Desert Locust (Schistocerca Gregaria)

For more than five decades, field research on the swarming flight of locusts has focused on the analysis of seasonal and geographical patterns of migration. External and internal factors affecting swarming behaviour, such as wind speed and direction, terrain, position of the sun, the locusts’ age and developmental stage and gregariousness have been examined (Kennedy, 1951; Waloff, 1972; Betts, 1976; Uvarov, 1977). These studies indicate that two major factors influence the orientation of day-flying swarming locusts. The first of these is visual orientation to other individuals or to the entire swarm, and the second is orientation to the visual effects of wind. As first pointed out by Kennedy (1951), the orientation to wind is mediated by optomotor responses to changes in the velocity of ground images over the ventral ommatidia (‘optomotor anemotaxis’; Kennedy, 1939). This velocity results from a vectorial addition of the locust’s own flight manoeuvres and wind-induced motion. Departure from a front-to-back movement of images and/or from a preferred moderate rate of retinal velocity is supposed to induce compensatory responses in air speed, orientation or flight altitude. The orientation of airborne locusts could thus be predictably related to wind speed. Locusts tend to fly upwind at wind speeds below their air speed. When the speed approximates or exceeds their air speed, they turn out of the wind or change their altitude.

A common orientation of groups of locusts within a swarm has been widely observed (Waloff, 1972; Riley, 1975; Schafer, 1976; Drake, 1983). Such a visual orientation to other locusts could also be explained by optomotor reactions to the relative movement of images of surrounding individuals (Kennedy, 1951). An inward orientation of locusts observed at the perimeter of the swarm was attributed to a strong optomotor effect of the swarm itself, which could be responsible for cohesion of swarms over long distances (Waloff, 1972).

During the same period, laboratory studies on locusts investigated more basic problems of insect flight. Much effort went into measuring accurately the aerodynamic forces produced by flying locusts tethered to a flight balance (Weis-Fogh and Jensen, 1956; Weis-Fogh, 1956) and in relating these forces to the kinematics of the beating wings (Zarnack, 1972, 1977; Koch, 1977; Waldmann and Zarnack, 1988). Other investigations concentrated on the neuronal pattern generator for flight (Delcomyn, 1985; Stevenson and Kutsch, 1987), its control by proprio-and exteroreceptive feedback (Wendler, 1985; Wolf and Pearson, 1988) and the influence of visual and mechanosensory organs of the head (Gewecke, 1982; Rowell, 1988).

Various aspects of flight behaviour have also been studied in the laboratory, but the results of this work have not been related to the sensory control and orientation of locusts’ flight during swarming, since visual stimuli were selected to investigate steering but not orientation. Rotational movements of striped patterns were used (Goodman, 1965; Thorson, 1966; Baker, 1979; Cooter, 1979; Taylor, 1981; Rowell et al. 1985; Thüring, 1986; Rowell and Reichert, 1986; Robert, 1988), but to study the visual control of ground speed, track direction and flight altitude visual input from translational movements presented underneath the flying locust is the relevant parameter.

The aim of the present paper is to begin to fill this gap between laboratory and field investigations by using translational stimuli to provide a phenomenological description of the locust’s orientation capabilities, a ‘flight ethogram’.

Desert locusts, Schistocerca gregaria Forsk., of both sexes from a crowded laboratory culture were used. Experimental animals were at least 10 days past their imaginai moult, when their body was no longer coloured pink. Experiments were performed at 29-32°C and were started 3 h after the onset of the photophase. A small metal pin was waxed to the locust’s dorsal pronotum, and the pin was then used to attach the locust to one of two friction-free force transducer measuring devices. One device monitored thrust and yaw-torque simultaneously, and the other monitored lift. The transducer with attached locust was located in a wind tunnel onto whose floor a movable ground pattern could be projected.

The thrust/yaw-torque meter consisted of two ‘linear variable differential transformer displacement transducers’ (Götz, 1968; Cobbold, 1974). Their ferrite cores were connected to a stiff rod, which was affixed above the midpoint of a vertical clock spring by means of a metal plate. The clock spring acted as both a torsion and a leaf spring and was kept tightened by a tension spring (Fig. 1A). The locust was attached along the long axis of the clock spring by means of a suspension rod. Yaw-torque produced by the flying locust led to torsion of the clock spring and displacement of the ferrite cores in opposite directions, whereas thrust produced displacements in the same direction. The thrust (yaw-torque) component of the response was obtained by electronic subtraction (addition) of the signals from the two transducers.

The lift meter consisted of a single transducer, mounted vertically (Preiss and Kramer, 1983). The locust was attached to the elongated ferrite core of the transducer. Lift produced by the locust was counteracted by horizontal leaf springs, supporting the ferrite core.

Both measuring devices gave linear responses over the range of forces exerted by the locusts and the counterforce of the springs was strong enough to allow open-loop conditions to be used during visual stimulation. Oscillations of the suspension rods were damped by ball-in-silicone dashpots and, in addition, high frequencies in the measured signals were electronically filtered (corner frequency 1Hz). Thrust and yaw-torque, or lift, and visual stimuli were registered on a chart recorder. The locust’s behaviour was observed from behind via a mirror. In some experiments, torque, thrust and the stimuli were digitized and stored in a computer for off-line averaging (Commodore PC40/20, combined with a CED 1401 signal-averaging programme, Cambridge Electronic Design, UK).

The locusts were flown in a continuous air current within a horizontal wind tunnel, which simulated their air speed (Fig. 1B). Turbulence was smoothed by passing the air through a honeycomb grating, a plastic meshwork and a reduction nozzle, located at the upwind end of the Perspex flight chamber (36cmx25 cmx25cm). To obtain sustained and regular flight, the air current was adjusted in steps of 0.5ms’¹. Following an erratic modulation of the flight parameters during the first minutes of flight, a smooth modulation of the parameters in correlation with the visual stimulus was generally achieved at an air current of 2-3 ms^-1. In air currents of less than 2ms^-1, flight was erratic and the locusts usually struggled with their legs and the abdomen was bent downwards. At 3.5 ms^-1 the wings were held in a gliding posture (Baker and Cooter, 1979) or the hindlegs were stretched upwards and flight was stopped intermittently.

The ceiling and sides of the flight chamber were lined with opaque white paper. The floor was covered with translucent plastic, which acted as a projection screen for visual stimuli, subtending an angle of 69° from side to side, and 98° fore and aft. A black/white periodic grating with a spatial wavelength of 22° and a contrast (I_max − I_min)/ (I_max + I_min) of 0.92 was projected onto this screen. The average pattern luminance was 1000 cd m^-2. The pattern was produced by moving a 35 mm film loop bearing equidistant black and clear stripes through the focal plane of a slide projector (Krüppel and Gewecke, 1985; Preiss and Kramer, 1986a). The film loop was moved by a d.c. motor combined with a reduction gear (548:1, Fa. Faulhaber) to stabilize lower speeds. The speed of the motor was measured by a coupled tachometer and was controlled by a function generator. Pattern motion, simulating ground speed of the locust (Fig. 1B), was either in line with, or transverse to, the animal’s long body axis. The polarity of the direction of motion was periodically reversed from progressive (anterior to posterior) to regressive (posterior to anterior) or from moving to the right to moving to the left. Pattern speed followed either a trapezoidal or a triangular time course. The maximal speed of a cycle of pattern movement varied between 18 and 0.15°s^-1.

Several hundred individuals were tested over a period of 2 years. The duration of an experiment was not held constant since the responsiveness of individual locusts was variable. Because the aim of this study was to produce an ethogram for visual orientation of locusts during flight, in most cases no attempt was made to estimate the frequency of individual responses. The primary interest was to describe the complete range of behavioural responses which occur repeatedly.

Locusts within the wind tunnel adopted either of two different flight postures (Krogh and Weis-Fogh, 1952; Weis-Fogh, 1956). In the typical long-term flight posture, the hindleg tibiae were pressed against the femora, the midlegs were pointed backwards and the forelegs were drawn up within the groove between head and pronotum; the abdomen was extended horizontally. In the other posture, the hindlegs were stretched out and, in most cases, bent downwards together with the abdomen. The midlegs were lowered and, occasionally, the foreleg tibiae were also extended. The former posture was frequently associated with high thrust and low yaw-torque, as expected during straight flight at high air speed. The latter posture was more often associated with low thrust, as would occur during straight flight at low air speed. High thrust with the hindlegs stretched out, however, also occurred. Changes in yaw-torque were, in most cases, accompanied by a typical turning posture of abdomen and legs (Dugard, 1967; Camhi, 1970; Arbas, 1986). The folded or extended hindlegs were bent sideways, usually together with the abdomen. In addition, the head was often rolled to the same side. In many experiments, however, the hindleg of only one side was stretched out and bent sideways, together with the abdomen, whereas the other leg was kept folded. The turning posture was always associated with high yaw-torque towards the side of bending. When in the long-term flight posture, however, some animals kept their abdomen and legs straight, though yaw-torque was high. We presume that, in these cases, yawing was produced solely by the beating wings, as is also observed in free flight (Baker et al. 1984). Changes between the two flight postures were usually correlated with a reversal of pattern motion when the pattern moved in line with the long body axis. With transverse motion only the direction of bending changed in most cases.

Pattern motion in the ventral visual field of the locust turned out to be an adequate stimulus to simulate passive changes in flight speed and flight direction. During both transverse and longitudinal pattern movement, a reversal of the direction of motion was correlated with modulation of thrust, lift and yaw-torque. Under the same experimental conditions, however, two opposing types of response were found for each measured flight parameter, and we could not predict which type would occur at the start of an experiment. The production of yaw-torque, thrust and lift was either increased or decreased under the same stimulus situation. The polarity of response was thus either positively or negatively correlated with the direction of pattern motion. While in most experiments the type of response did not change (up to 5 h), in several the animal switched back and forth from one type to the other at random during the experiment. The response did not diminish before the change, the animal switching instantaneously from one type to the other. This indicates that neither of these response types is due to fatigue of the locust but that they are distinct and equivalent behavioural responses. This was true for all three flight parameters studied. Fig. 2 illustrates such changes in the lift response of a locust with longitudinal pattern flow.

The time course of response differed markedly between experiments and even altered during an experiment. An identical triangular modulation of pattern speed evoked all transitions between a triangular and a trapeziform modulation of the flight parameters (Fig. 3). When the response was modulated in a triangular manner, the strength of response depended on pattern speed, as has been commonly found for optomotor responses in other insects. The amplitude of response was highly variable, often gradually changing into a trapeziform modulation. When modulated in a trapeziform manner, the strength of the response was independent of pattern speed. The amplitude did not increase further during the experiment, but often slightly decreased or gradually changed into a triangular modulation. Short pulses of yaw-torque were also recorded in some experiments. These pulses have a short rise time (about 1 s) and a somewhat longer decay time (Fig. 3, upper trace) and are probably equivalent to the ‘torque-spikes’ described in Drosophila (Heisenberg and Wolf, 1979). The torque signal usually returned to the baseline before another torque-spike began. Only in a few cases were they superimposed on a rather tonic modulation of yaw-torque (Fig. 3). Torque-spikes followed each other in an irregular sequence. The direction of pattern motion affected either the frequency of torque-spikes or their polarity. This flight mode was seldom found and only during the first 30 min of an experiment before yaw-torque changed to a triangular or trapeziform modulation, although in some cases single spikes might also then occur. In these cases, however, it was not possible to make a definite distinction from noisy fluctuations of the signal. Fixation of the head on the thorax did not increase the probability of this flight mode, as was found in Drosophila (Heisenberg and Wolf, 1979).

Commonly, thrust and yaw-torque were simultaneously affected by the stimulation. In several cases, at the beginning or intermittently during an experiment, either thrust or yaw-torque was correlated with the stimulus, while the other showed only weak fluctuations (noise). Furthermore, the response of thrust, yaw-torque or lift could cease for a few stimulus cycles while body posture maintained its modulation. During an experiment, the amplitude of the yaw-torque and thrust responses could gradually increase or decrease, either in combination or independently. The response could thus gradually change from yaw-torque to thrust or vice versa (Fig. 4). A temporal sequence of modulation of the flight parameters was also observed. In many cases, the locust first increased thrust and then, while thrust remained unchanged or decreased, increased yaw-torque, or vice versa. This sequential modulation could either occur intermittently with a combined modulation, or during the whole experiment. This variability suggests that a combined modulation of flight parameters is not the result of an artificial coupling of signals via the mechanics of the measuring system.

With transverse pattern motion simulating side wind, the polarity of yaw-torque responses was either with or against the direction of motion (Fig. 5A), both types occurring equally often. Both types of response were also observed in individual animals, as shown in Fig. 5B. In free flight, turning with the transverse component of pattern motion would result in upwind orientation and turning against it would result in downwind orientation (Fig. IB). Turning was accompanied either by constant thrust (either high or low) or it was modulated as a function of pattern speed, independently of the direction of motion (Fig. 5B). If thrust was constantly high, it was only reduced when pattern speed was reduced to nearly zero.

In many experiments, however, the flight posture indicated that only one direction of pattern motion actually induced turning (Fig. 6). Thrust was then also modulated asymmetrically, being always lower when the flight posture indicated turning. Neither leftward nor rightward motion was systematically preferred in this respect (either across all the group or within individuals), as would be expected if it were caused by an asymmmetry within the stimulus or tethering situation. The asymmetry, therefore, must be attributed to an internal shift of the set point of the response to transverse pattern motion.

Pattern motion in line with the long body axis (alternately simulating tail-and headwind when periodically reversing polarity) elicited a turning response in half of the locusts with progressive motion (head-to-tail; Fig. 7A) and in the other half with regressive motion (tail-to-head; Fig. 7B). Pattern motion in the respective opposite direction usually elicited straight flight, though sometimes a weak tendency to turn persisted, as was also indicated by a slightly bent abdomen. The polarity of yaw-torque was either maintained over many periods (Fig. 7) or was repeatedly inverted during a single period (Fig. 8). Thrust, measured simultaneously, was also modulated in correlation with this stimulus. In half of the experiments, thrust was reduced when the pattern moved progressively and increased when it moved regressively (Fig. 7). In the other half, the opposite occurred. Several experiments showed high thrust during both progressive and regressive pattern motion which was reduced only at the moment of reversal, i.e. when the pattern was stationary.

Table 1 shows paired combinations of yaw-torque and thrust modulations. Three out of the four possible combinations were found to occur equally often. The fourth combination, high yaw-torque with high thrust at progression and low yaw-torque with low thrust at regression, was never observed.

The two types of pattern orientation also induced a correlated modulation of lift. With transverse pattern flow, lift was modulated when the locust assumed an asymmetrical turning posture. Lift was high during straight flight and low during turning flight. With longitudinal pattern flow, lift was either high during progression and low during regression or vice versa (Fig. 2). At low lift, the amplitude of the fluctuations was often increased and gliding was observed. Since the lift response was measured in a separate series of experiments, we have no information on combinations of lift with thrust and yaw-torque responses.

The locust’s turning behaviour was not determined by pattern motion alone. When pattern motion intermittently stopped, some animals continued to show modulated yaw-torque (Fig. 9A). The amplitude of the modulation remained the same but the polarity changed either more or less frequently than during pattern flow. Restarting the pattern flow caused the modulation to become synchronized again, but changes in yaw, not correlated with the stimulus, sometimes occurred (Fig. 9A, arrows). In other experiments, the yaw response altered between two different mean levels (Fig. 9B). The behaviour then indicated either an alternation between turning and straight flight or between strong turning towards one side and weak turning to the same or the other side. This complex turning behaviour requires an additional, internal turning command that changes its polarity from time to time. These changes sometimes occurred rather frequently, as in Fig. 9B, or the polarity remained unchanged for longer periods. Furthermore, in some experiments the amplitude of the visually induced response decreased to zero over time, but the internal turning tendency was still apparent, as the polarity of yaw changed from time to time without any obvious correlation with the visual stimulus.

Wind-related orientation at high altitude requires a low threshold of the optomotor response. This threshold was investigated by a stepwise reduction of a trapezoidal speed modulation as soon as a correlated response of thrust and/or yaw-torque was observed. Owing to slackness between motor and film loop, the range of smooth pattern movement did not extend below 0.15°s^-1.

Pattern motion both in line with or transverse to the long body axis resulted in a modulation of flight parameters down to the minimum speed, indicating that the threshold of optomotor response is below 0.15° s^-1 (Fig. 10C). In previous experiments (Riley et al. 1988), the threshold of optomotor yaw responses to simulated yaw rotations was shown to be dependent on illumination and to be about 0.46°s^-1 at 0.16cdrrT². Roll responses of the head were induced down to 4X10^−3OS^-1 at 1600cd m⁻² (Thorson, 1966). The amplitude of modulation in our experiments usually decreased when pattern speed was reduced (Fig. 10A). In several cases, however, there were sequences of intermittent responses with greatly enlarged amplitude. This indicates that response amplitude depends not only on pattern speed but also on unknown internal factors. In some individuals, the amplitude of response was always maximal and independent of pattern speed down to ±0.27° s^-1. Though the response was delayed for up to 40 s at low pattern speed, when the stimulus frequency was changed the response frequency changed jn a similar manner (Fig. 10B).

The occurrence of two opposing types of response to the same stimulus is a striking feature of the orientation of tethered flying desert locusts to visually simulated wind drift. When lateral wind drift was simulated, the polarity of the yaw-torque response was such that in free flight either upwind or downwind orientation would result (Fig. 1B). This differs from results previously found in similar experiments with other insects. Male moths orienting within an odour plume, for example, always oriented upwind by turning in the direction of the transverse component of ground pattern motion (Marsh et al. 1978; Kennedy et al. 1981; David et al. 1983; David, 1986; Preiss and Kramer, 1986a,b). The adaptive sense of this moth behaviour is clear: the approach towards the pheromone source. We assume that the variability of visually induced behaviour seen here in the locust is a general property of insects, but only becomes apparent in the absence of adaptive necessities. Variability of response would, therefore, not be expected in previous optomotor experiments with locusts since only rotational changes of body posture were examined, which require compensation to maintain flight stability (Goodman, 1965; Thorson, 1966; Baker, 1979; Cooter, 1979; Taylor, 1981; Rowell et al. 1985; Thüring, 1986; Rowell and Reichert, 1986; Robert, 1988). In the case of translatory movements, the locust’s orientation behaviour is not always aimed towards a defined goal (Kennedy, 1951). The direction of swarm displacement over a long period is always downwind (swarms thereby approach the Inter-Tropical Convergence Zone where rainfall is most probable). When observed over a shorter period, such as a day, orientation of the whole swarm or of single locusts is variable, depending on wind speed. At low wind speed, upwind orientation prevails, and at speeds approaching or exceeding the locust’s maximal air speed, downwind orientation normally occurs (Kennedy, 1951; Waloff, 1972; Betts, 1976; Rainey, 1976). This was not the case in our experiments. The locusts oriented themselves either upwind or downwind in response to the same visual stimulus, and even changed orientation during an experiment. Though upwind and downwind turning could also be induced via the anemoreceptive system, this required different sensory inputs either from the cephalic mechanosensory hair plates or from the antennae (Arbas, 1986). In contrast, the polarity of the response in our experiments was determined by internal factors and changed in an unpredictable manner. A similar reversal of polarity was also found for thrust and lift responses, i.e. the same visual stimulus elicited speeding up or slowing down, climbing or falling.

In experiments with flies and moths, visual control of wind-related flight by means of the ground pattern was explained by two independent control circuits (David, 1986; Preiss and Kramer, 1986a). The first uses the transverse component of pattern motion for control of angular orientation to the wind by modulating yaw-torque. The second uses the longitudinal component for stabilization of ground speed by modulating thrust and lift (at least in cases where the lift response is not restricted by other stimuli such as pheromone; Preiss and Kramer, 1983; Preiss and Futschek, 1985). Surprisingly, this functional separation was not found in locusts. Pattern motion both in line with and transverse to the long body axis was capable of inducing a compensatory modulation of yaw-torque, thrust and lift. It is still unclear, however, whether the different responses of the same flight parameter to either of the two pattern orientations are due to separate control circuits and, if so, how these interact.

The ability to modulate yaw-torque and thrust independently would explain the high variability of response in our data. Yaw-torque and thrust were not correlated, with the exception that low yaw-torque during regression was never associated with low thrust. In free flight, the latter condition corresponds to a continuous, yet visually controlled, backward drift, which is never tolerated by locusts (Kennedy, 1951). The observation that locusts flying with high thrust turn over progressive patterns and fly straight over regressive ones, as found in the experiments, fits with those in the field.

A discussion of the role that these responses play during free flight must take two facts into account. First, our experiments did not examine whether the tethered locusts could modulate lift independently of thrust, as described by Weis-Fogh (1956) and Zarnack and Wortmann (1989). Second, interpretation of lift values remains ambiguous. Reduced lift could indicate sinking, maintenance of height or even slower climbing, since tethered locusts often do not compensate for their body weight (Gewecke, 1975; Dreher and Nachtigall, 1983; Krüppel and Gewecke, 1985), presumably because of the lack of visual feedback. When changing altitude, a centrifugal/centripetal flow of pattern elements occurs which has been shown to play a role in stabilization of flight altitude in other insects (Preiss and Kramer, 1983; R. Preiss, unpublished results with locusts). Lift production also varies with the inclination of the body axis (Gewecke, 1975; Zarnack and Wortmann, 1989), and in tethered flight experiments inadequate inclination could contribute to this deficiency. A similar ambiguity exists for thrust; low thrust could indicate that ground speed was either above or at its set point. Clarification of these ambiguities would require closed-loop experiments.

The relevance of responses to transverse pattern motion for free flight is apparent. Responses to longitudinal pattern motion, though, require explanation. Turning over a progressive pattern could indicate a tendency of the locust to orient upwind in the field. Apparently the tethered locust interpreted the strong optomotor signal as downwind orientation, in which the wind-induced signal sums with the signal caused by its own flight (they subtract from each other after turning upwind). Straight flight over a progressive pattern could indicate a tendency to orient either upwind or downwind in the field. When combined with slowing down, compensation of a strong optomotor signal above its set point would result, if it were interpreted by the tethered locust as caused by a decrease of wind speed during upwind orientation or an increase of wind speed during downwind orientation. Speeding up, however, would not result in such compensation unless it was combined with an increase in flight altitude. The decrease of the optomotor signal caused by the increase in altitude compensates for the increase of the signal caused by the increase of air speed. Clarification of this ambiguity would require closed-loop experiments.

Not only the polarity but also the amplitude of modulation of the three flight parameters varied greatly even when stimulus strength was unchanged. This variability, formerly also observed by Thüring (1986), suggests independent internal factors which modulate the gain of the response for each flight parameter. Gain modulations would also explain the variations in the time course of the response. Clipping as a result of extreme gain of a triangular modulation of pattern speed would cause a trapeziform modulation of the response even when pattern speed was close to threshold level, as found in some experiments (Fig. 10B). This clipping has been found in Drosophila for responses to a stimulus that saturates the visual pathway (Gdtz, 1968), whereas in our experiments this effect was due to spontaneous changes in gain. Gain modulations in the visual system have been previously shown in gypsy moths, in which the attractant pheromone enhanced optomotor responses, while they were attenuated by an inhibitory compound (Preiss and Kramer, 1983; Preiss and Futschek, 1985; Olberg, 1989). These internal gain modulations can be distinguished from habituation and sensitization (well known, for example, from the landing response of Drosophila-, Fischbach, 1981) in that the response strength is unpredictable. Gain modulations may be more common in insects than previously found since in earlier experiments responses were averaged, even across individuals.

The locust’s turning behaviour was not solely determined by visual stimuli. Modulation of yaw-torque continued in the absence of pattern motion and biasing of yaw-torque about zero occurred without any consistent relationship with the visual stimulus. This behaviour, however, was not due to the strong directional cue that might be provided by the striped pattern. When, in further experiments, a chequered pattern was used, the induced behaviour was the same (R. Preiss, in preparation). To explain this requires an internal turning tendency in addition to the visually induced turning tendency. In free flight, such an internal turning tendency might shift the point of balance (zero torque) to a point between upwind and downwind, resulting in an oblique course (‘menotactic orientation’). According to this theory (von Holst and Mittelstaedt, 1950; Mittelstaedt, 1963; modified by Kramer, 1975), the track angle should become less stable as it approaches ±90°, given that the visually induced torque follows the sine of the course angle, as found in gypsy moths (Preiss, 1991). Menotactic orientation of locusts has also been observed in field experiments (Waloff, 1972; Baker et al. 1984). Within a swarm, the heading of individuals in a cluster was always parallel, but the heading of clusters, sampled at 2 min intervals, varied widely (Waloff, 1972). Since wind direction had not changed, a variable menotactic orientation is indicated. Based on our experiments, this variability could be explained either by a variable internal turning tendency or by a modulation of gain in the visual pathway.

The experiments suggest that still other interactions between the visually induced and internal turning tendencies are possible. After pattern motion had been stopped, for example, the torque continued to be modulated with the same amplitude, although at a different frequency. In addition, fluctuations of yawtorque with a similar amplitude but not correlated with the stimulus also occurred during visually induced responses. It is therefore possible that a visually induced response can be intermittently suppressed by the internal system.

We are grateful to Dr E. Kramer of the Max-Planck-Institut für Verhaltensphysiologie, Seewiesen, FRG, for developing the measuring system and to Dr R. M. Olberg of Union College at Schenectady, USA, Dr E. Kramer and Dr B. Corrette for their most critical reading of the manuscript and for correcting the English. This work was supported by the Deutsche Forschungsgemeinschaft (Ge 249/10-1).